Seehausen 1996 Lake Victoria Rock Cichlids

Home , Chromis, Coptodon, Cyprichromis, Gnathochromis, Haplochromine, Haplochromis

1 4 Lake Victoria Rock Cichlids

— taxonomy, ecology and distribution —

by Ole Seehausen Institute of Evolutionary and Ecological Sciences, University of Leiden, Netherlands

3 Photo cover: Haplochromis nyererei, territorial male at Makobe Island at a depth of about 6 metres.

Photos on pages 132 bottom left, 156 top left, 208 bottom, 244 bottom, and 260 top and bottom by Frans Witte & Els Witte-Maas. All other illustrations by the author.

Printed in Germany

4 Contents

Preface by Les Kaufman ...... 7

Introduction ...... 9

Lake Victoria: climate, limnology and history ...... 13 Climate and Limnology ...... 13 Geological history of Lake Victoria ...... 16

The cichlid species flock ...... 20

The Mbipi, overlooked subflocks like the Mbuna of Lake Malawi? ...... 25

Taxonomy of rock-dwelling cichlids...... 29 Species recognition ...... 29 Generic taxonomy of Lake Victoria haplochromines ...... 31

Habitat and ecology of rock-dwelling cichlids — with co-authors A. Samwel-Terry & N. Bouton ...... 36

Keeping and breeding of rock-dwelling cichlids ...... 44 the aquarium ...... 46 water and food ...... 48 breeding ...... 50

Identification of rock-dwelling haplochromine cichlids ...... 58 key to (sub)genera and species complexes ...... 61

Introduction to the species discussions ...... 63 Vertical bar Mbipi ...... 66 Blue, black and OB, the Neochromis complex of algae scrapers ...... 66 Xystichromis-like epilithic algae scrapers ...... 91 The Haplochromis “carp” complex of Xystichromis-like algae scrapers97 In the colours of the rainbow - the Haplochromis nyererei complex ... 100 Large heads - the Haplochromis “deepwater” complex ...... 125 The Haplochromis “pseudonigricans” complex ...... 129

Chessboard Mbipi ...... 146 Tiny mouths - the H. (Paralabidochromis) “rockpicker” complex...... 146 The H. (Paralabidochromis) “rockkribensis” complex...... 158 Lobe-lipped cichlids, the H. (Paralabidochromis) chilotes complex .... 167 The H. (Paralabidochromis) chromogynos complex ...... 175

5 Rock-dwelling species of other lineages Shallow heads and narrow jaws - the Psammochromis lineage ...... 179 Hook-teeth - the Ptyochromis lineage of snail shellers ...... 190 A mill in the throat - snail crushing haplochromines ...... 201 “Mgobegobe” - fish eating haplochromines of rocky shores ...... 205 Egg and fry robbers of the Lipochromis lineage ...... 215 Silver arrows of the “Double stripe” complex ...... 227 Teeth like a scalpel - algae scrapers of the Haplochromis lineage ...... 229 Anatomically unspecialized haplochromines of Astatotilapia type ...... 235 Insect eating rock cichlids that cannot be grouped ...... 244

Frequent intruders from other habitats ...... 250

Astatoreochromis, Oreochromis and Tilapia at rocky shores ...... 258

Non-cichlid rock fishes ...... 263

Evolution of rock-dwelling cichlids ...... 269 Evolution of habitat specificity and trophic primary radiation ...... 269 A role for sympatric speciation? ...... 271 Secondary radiations by allopatric speciation ...... 274 Factors driving divergent evolution in allopatry ...... 275 Matrix species ...... 277

Conservation of rocky shores and their fish ...... 278

Distribution maps ...... 281

Tables with taxonomic measurements ...... 291

References ...... 296

Species Index ...... 302

6 Preface

Professionals and amateur students of economically almost worthless little fishes aquarium fishes both begin with joyous, that started it all by checking out en masse. boundless enthusiasm, but things have a If the shattered remnants of bombed-out way of getting too serious in the pro world, Nature would just fade quietly into the and the fun starts to slip away. For those of realm of paleontology, biologists would be us who work on cichlids from East Africa, able to savor their grief without ever hav- the fun really started wearing thin some- ing to worry about social responsibility. That time in the late 1980s, when it became ap- isn’t how it happened in Lake Victoria. parent that several hundred species of Large chunks of the native fauna just hung cichlid fishes from Lake Victoria that hardly on. A few species are even resurging in anyone had ever seen, might never be abundance now that the famed Nile perch seen again. The world’s second richest is petering out a bit. Included among these lacustrine fish fauna had been destroyed by are the magnificent fauna described in this humans mucking about, reorganizing things book. These may be the most beautiful, so that the lake could produce fish more abundant, and easily located vertebrates to marketable overseas. Hundreds of fishes have escaped the notice of people for this extinct, but then again, hundreds of mil- long. They’ve got the Vietnamese bovids lions of dollars per year pumped into one beat hands down. of the neediest regional economies in the How could more than 200 species of world. A Faustian compromise if ever conspicuous aquatic vertebrates crowd there was one. the shallow waters of the earths’ second Someday there could be an inspiring largest lake, yet go unnoticed through movie or television show made about the most of the 20th century? Perhaps the group of scientists from Africa, who along same way that an equal number could with their colleagues from Europe, the vanish completely from these same wa- Middle East, and North America, got to- ters, within a period of only five or six gether to save their beloved lake from years, yet inspire barely an epitaph save Nile perch, pollution, water hyacinth, and for the sorrow and disappointment of a all the other symptoms of the suffocat- handful who knew and understood. Other ing pressure of humanity. They got the than the African and visiting scientists data, blew the whistle, got into trouble who have borne witness to these events with authorities, worked hard, raised firsthand, nobody could feel the full loss. money, fought over money, sued each But if you are one of these scientists or a other, died, had kids, and generally knowledgeable hobbyist, you should be shared two decades of the joys and trag- seething with anger – anger craving a use- edies of East Africa. Through it all they ful and rewarding outlet. learned a whole lot about each other, and You can’t miss something that you never about life. Meanwhile, the East African even knew was there. But you cannot save Community, left in tatters in the wake of (from our own brutal hand) something that Idi Amin, was awoken from its deep you do not know and love. So, to know and slumber, in good part by embarrassment to love, to celebrate and to share, to gener- (and an embarrassment of international ate interest and funds to where they are funding and publicity) over the state of most needed in conservation: that is what Lake Victoria. Yet, almost lost in the shuf- aquarists can do to save faraway species fle were the haplochromines them- from extinction. But to do so you need a selves... those beautiful, fascinating, and primer, a guide. You need the information.

7 sipping fresh beer through long, long straws, and telling rowdy stories. It is maybe eleven at night and dinner is still on the fire. You hear a hy- ena call, not far away, and then somebody’s dog. And what are people talking about? After the stories about who fell out of the boat, or caught the best fish, or made the villagers clutch their sides the laughter was so painful, the conversation shifts to the price of haplo- A male Haplochromis "red back scraper" in the aquarium. chromines on the European marketplace. Or the possibility that a school might be built with You need the direct contact with your sha- the proceeds of aquarium fish sales, right on man, the aquarist-turned-field biologist. This your campsite, to help people, and to help book is the first specialized guide on any them know about value, and safeguard their of Lake Victoria’s endemic ‘treasures writ- aquatic heritage. About how after a whole day, ten especially for fish knowers and lovers. you finally succeeded in explaining to the lo- And Ole Seehausen is a shaman par excel- cal fish collectors (mostly little children) that lence. So do it. You are the troops, and here they should return the young, unsalable fishes are your orders: get to know these little alive to the lake instead of tossing them to fishes. Keep them in your aquariums, breed die up on the grass. How are the kids in Bos- them, photograph them, watch them, wor- ton, how’s so-and-so’s health in Leiden, is that ship them, in the funny little way we do that fellow’s father down in Masaka feeling better? nobody else seems to understand. Write Somebody says to your friend, who is lifting a about them, talk about them, spread enthu- haplochromine appetizer from out of the siasm and appreciation, celebrate the coals, “You should have preserved that as a fishes, the fauna, the lake, explain your dedi- specimen – you cooked a rare fish!” and he cation to non-aquarists. Do not let them die answers, “Yes, but I cooked it only briefly... and – in the wild, or in your hearts and minds. I will study it internally”. During dinner every- This is your revenge. Do you doubt that one talks with animation about what was what you do can make a difference in the learned about the fishes or the lake that day: context of a distant, different culture? But why it matters, and how best to tell the rest then for you, which culture is truly the al- of the world about it. These are your soul ien one? The culture of Kampala/Munich/ mates. This book is, through Seehausen’s art, London/New York that exists apart from, a gift from them to you. It is also your initia- and in disdain of, all nature and things wild? tion into our circle. Enjoy this book, enjoy the Or the culture of the woods and waters eve- fishes, and then, enjoin the battle to save all rywhere remote, where the last proud rem- living things beautiful, and richly varied, and nants of the planet’s living wonders quietly worthless in the eyes of the ignorant. await their demise? If you are reading this book, you needn’t Les Kaufman even think about your choice. You are one 8 Sept.1996 of us. This is your adventure. You are in East Boston University Marine Program, Africa, at lake side. It is a dark night, and you Kenya Marine and Fisheries are the only foreign visitor amongst your Research Institute, and Fisheries many African friends around the campfire, Research Institute of Uganda

8 Introduction

A number of books have been published cies assemblage of Lake Victoria cichlids, during the past years, that dealt with the that was explicitly believed not to exist! A biology of East African lake cichlids in their huge group of probably more than 200 natural environment. In this book, I want to haplochromine cichlid species had simply pick up that thread. Nevertheless, the book been overlooked. is by necessity quite different from the oth- All around the world the cichlid faunas of ers. Not only that it is the first popular book the East African Great Lakes fascinate stu- about Lake Victoria cichlids, but it reports dents of evolution as much as aquarium about a species assemblage that is largely hobbyists, enjoying and observing the prod- new to aquarists and scientists alike. Of the ucts of evolution. A continuous flow of pub- 173 species discussed in it, merely 34 have lications of various kinds results from this been scientifically described yet. The ma- fascination. Apart from their enormous di- jority of the "new" species were discovered versity in form and coloration, a number of between 1991 and 1996 during the first unique features made the cichlids of Lakes extensive ecological and taxonomical sur- Malawi (Nyasa) and Tanganyika famous: vey of rocky shores in south-eastern Lake One is the highly developed habitat Victoria. They constitute a Mbuna-like spe- specificity and site fidelity of rock-dwelling

The north-west corner of Gana Island, the most remote locality visited by the author.

9 cichlids, resulting in unusually high levels to give a first introduction to this over- of regional and local within-lake endemism looked fauna, and to make researchers, and extraordinarily high numbers of spe- hobbyists, and conservationists aware of its cies. Another one are the fascinating simi- riches. larities that evolved among individual rock- Most emphasis is laid upon the presen- dwelling species of Lakes Malawi and Tan- tation of the various species complexes ganyika as well as among whole commu- and species, with much new information nities (Lowe-McConnell 1993). about distribution, ecology, and behaviour It has long been believed that highly de- in their natural environment. For this pur- veloped habitat specificity, site fidelity, and pose underwater work by Scuba was restricted distribution within the lake are done for the first time in Lake Victoria. much more muted, if at all developed The lake, widely known as a turbid one, among the cichlids of Lake Victoria, and has never attracted diving researchers or that a species rich rock-dwelling cichlid as- cichlid enthusiasts. This is certainly one semblage would not exist in this lake. Re- reason for that its rock-dwelling cichlids sults of more recent research had indicated were discovered that late. Lake Victoria since the early eighties, that these assump- is indeed turbid compared to Lakes Ma- tions might be wrong. The existence of a lawi and Tanganyika. Nevertheless, it is at group of cichlids, very similar to the some places clear enough to study the be- Malawian Mbuna, inhabiting the rocky cliffs haviour and ecology of the cichlids. Much in Lake Victoria, was first mentioned by van of the new ecological information brought Oijen et al. (1981). The first cichlid collec- together in this book has been gathered tions at rocky islands in southern Lake Vic- by Scuba. Unfortunately, however, the low toria were done in 1978/79 by Els Witte- water transparency makes photographing Maas and Frans Witte, in 1986 by Jan much more difficult and less attractive Sevenster and in 1989 by me. In 1990 long than it is in the clear Lakes Tanganyika and term field research on rock-dwelling cichlids Malawi. Consequently most of the fishes was taken up by Niels Bouton and Yves for this book had to be photographed in Fermon, and in 1991 by me. Over those a photo cuvette immediately after cap- years, and in particular during the last five ture. years, a large variety of new species has Apart from harbouring the large assem- been discovered. I hope the results of this blage of "new" Mbuna-like rock-restricted most recent work on rocky shores will con- cichlids, rocky shores and islands are the vince most of those authors that until now most important refugium for a number of kept to the old assumptions, when compar- cichlid species that were formerly not re- ing the cichlid flocks of the East African stricted to rocky substrates, but survived Great Lakes. only there after the Nile perch upsurge In species richness, ecology, morphology ("rock-refugees"). Many readers may be fa- and behaviour of the species, the assem- miliar with what happened in Lake Victo- blage of rock-dwelling cichlids in Lake Vic- ria. The Nile perch (Lates sp.), a big preda- toria closely resembles that in Lake Malawi. tor, was introduced by man in the 1950s. At the time of writing more than 170 haplo- When its numbers increased explosively chromine species are known from rocky in the early 1980s, the populations of substrates in the south-eastern lake region. most open and deep water dwelling This has far reaching implications for the in- cichlid species crumbled. An estimated terpretation of evolution in Lake Victoria. number of 150-200 endemic cichlid spe- The "new" cichlid assemblage carries a cies disappeared, many of which probably great potential both for biological research went extinct (Witte et al. 1992). The and for the aquarium hobby. Apart from that cichlids that inhabit rocky shores and its role in the ecosystem of the lake needs some other littoral habitats were less than to be studied. The purpose of this book is others affected by these events. Rocky

10 habitats of Lake Victoria currently prove and is probably one of the most skilled to be of major importance to the survival boat drivers in the Tanzanian part of Lake of at least some of the endangered spe- Victoria. Apart from his great skills he also cies. At present they harbour the largest contributed to the work a great lot of en- anatomical, ecological, and possibly ge- thusiasm for the cichlids. Ruben Enoka netical diversity in Lake Victoria. This has always been the chief fisherman in refugium which is dispersed all around our team. I thank him for many years of and across the lake, may come to play a excellent collaboration and particularly for central role in the further evolution of solving various kinds of difficulties with Lake Victoria cichlids. great discipline. Anna Samwel-Terry has The first rock-dwelling Lake Victoria as field and laboratory assistant been con- cichlids that became available to aquarists tributing a lot to various kinds of work on were exported from Uganda in the 1980s. the lake and in the laboratory, and con- Recently the utilization of rock cichlid cerning this book in particular. She did populations as source for aquarium fishes most of the stomach content analyses of is becoming more permanent. Currently the new species. Hence her contribution at least about 50 species are kept by to the book goes beyond her co-author- aquarists. It is likely that in the near fu- ship in the chapter on ecology, and all ture they will become the major source across the species chapters. I am very for aquarium fishes from Lake Victoria. much indebted to Mohamed Haluna, Ali Rock-dwelling Lake Victoria haplochro- Marwa, Masoud Ilomo and Aloys Peter mines, if given the publicity, they de- for their expertise as fishermen in cichlid serve, are likely to become as attractive collecting. Radhmina and Gonza Mbilinyi- to aquarists as Mbuna are. Kitery, Mr. Mapunda, Kassim Mgady High time therefore, to look at Lake Vic- Hamissi, Sylvester Wandera, and John toria's rock cichlid assemblage more care- Balirwa are thanked for much logistic and fully. I wish this book can make a start in other support. Niels Bouton and Yves documenting its extraordinary diversity, Fermon are acknowledged for collabora- and lay the ground for further documen- tion and logistical support during my first tation in the natural environment, as well stays in Tanzania. Both are researching as for identification and documentation of about evolution and ecology of Lake Vic- fishes that are being imported to America toria rock cichlids, and the reader will and Europe in increasing numbers. Be- come across their names here and there cause more than 120 species are pre- in the book. Finally, Frans Witte, Jacques sented here for the first time in literature, van Alphen, and Kees Barel are greatly ac- and no other references exist to them, knowledged for a lot of encouragement more taxonomical details had to be given and scientific advice in our work. Frans in this book than would otherwise be Witte read and improved parts of the needed in a popular book. It is my hope book manuscript, discussed with me spe- that the book, nevertheless, makes inter- cies identifications and made available esting reading for hobbyists and biolo- some of his and Els Witte-Maas' data and gists alike. photographs. I thank Rob Hoogerhoud for It needs to be stressed that the work, making available his drawings of cichlid the book is based on, is not the work of jaws, my grandfather Helmut Seehausen one person. Rather it is the work of a for his paintings of cichlids, and Peter team of varying composition. I am very Snelderwaard for X-ray photographs and much indebted to all members. Those to other support. be mentioned in particular are our boat The book would not have been written captain and chief fisherman Mhoja Kayeba without the support of the Tanzanian gov- who worked with the research teams of ernment. I am grateful to its Commission Leiden University for more than 15 years, for Science and Technology for the re-

11 search permit that allowed me to carry out the work, and to the Tanzania Fisheries Re- search Institute (Prof. Philip O.J. Bwathondi) for accommodating my research. Egid F.B. Katunzi, as the director of the Mwanza centre of TAFIRI deserves particular mentioning for his and his staffs' hospitality and support during my ongoing studies. I am grateful also to the Ugandan Fisheries Re- search Institute (FIRI) in Jinja for their hospitality and support during my short visit in 1989. The research, of which this book is a spin-off, was financed by grants from the Netherlands Organization for the Advance- ment of Tropical Research (WOTRO), the TETRA company (Germany), the EHEIM company (Germany), and the German Cichlid Association. The new information about behaviour of rock-dwelling cichlids in their natural environment would not have become available without the support of the companies SCUBAPRO (Germany) and SEAWAY (Germany). The efforts and enthusiasm of Ludwig Schadhauser (journal "Wild about Animals") was instrumental in secur- ing much of the financial support. Last not least I want to thank the publisher Dirk Verduijn for his pleasant co-operation in the production of this book. I use in this book "we" and "I" at different places. When it concerns general observations, discoveries of "new" species etc., I use "we". When conclusions are drawn from our results, and decisions made that are mine, and do not necessarily reflect the opinion of all other team members, I use "I".

12 Lake Victoria: climate, limnology and history

Climate and limnology Northeast, drain the forested mountains of western Kenya. The third big river, the Lake Victoria, the biggest tropical lake, Mara, also has its origin on these moun- and the third biggest lake on Earth, is situ- tain slopes. However, it also drains, to- ated on the East African plateau at 1134 gether with smaller and semipermanent m above sea level, between the Central rivers, the relatively dry Serengeti plains African Rift in the West and the East Afri- in the Southeast. can Rift in the East. It stretches across the The lake is a huge sink with a single equator, extending from 0.2° North to outlet, the river Nile, breaking out from 3.0° South and is surrounded by three the lake at its northernmost point and countries, Tanzania (49 % of lake surface, connecting Lake Victoria to Lake Kyoga, roughly 60% of shoreline), Uganda (45% a shallow, extremely denticulated exten- of surface, roughly 30% of shoreline) and sion of the Nile. Downstream of Lake Kenya (6% of surface, roughly 10% of Kyoga, the waters of the Nile plunge shoreline). The lake has a large catchment down Murchison Falls, which are a formi- area (194,200 km²; Piper et al. 1986) with dable barrier to migration of fish. It is several bigger rivers draining into it. The Murchison Falls that isolate the Victoria largest one, the Kagera river in the West, basin from the ichthyofauna of the rest drains the mountains of Rwanda and Bu- of the Nile and its associated Lake Albert. rundi and hilly rainforest areas in western The climate is equatorial with two wet Tanzania. The second largest, the Nzoia seasons, in the southern parts one be- river and several smaller ones in the tween October and December, the other

Cichlids at the rock-sand interface at a depth of 6 metres at Makobe Island.

13 Figure 1. The catchment area of Lake Victoria (surrounded by stippled line), and adjacent catchment areas of (clockwise from top) Lakes Kyoga, Eyasi, Tanganyika/Kivu, Edward/George, and Albert. Arrows indicate the direction of water flow in the rivers. Lake Victoria is connected to Lakes Edward/ George through the Katonga River. However, the area in which the river undergoes reversal of its flow direction is difficult to penetrate for cichlids. It is a papyrus swamp in which waters have very low oxygen concentrations.

14 one between February and April. Annual 21°C. Heavy SE winds during the dry sea- rainfall averages about 1000-1500 mm, the son (June to September) stir up the lake northwest being more humid than the and cause mixis of the water body. The wa- southeast and east. Differences in rainfall ter of Lake Victoria has a neutral to slightly are reflected in the natural vegetation sur- alkaline pH. Fryer & Iles (1972) give a range rounding the lake. Mighty rainforests grow from 7.1 to 9.0. We measured a similar along the northwestern shores and on the range of pH (6.9 to 9.0), and a hardness of huge island archipelagos in the Ugandan between 2 and 8°. The conductivity lies, ac- part of the lake. Less luxurious evergreen cording to Fryer & Iles (1972) between 98 forests are found at many places along the and 145 mho./cm. The water transparency west and on islands in the southwest (e.g. is low compared to that in Lakes Malawi Rubondo Island), south (e.g. Kome Island) and Tanganyika but not everywhere as low and southeast (e.g. Ukerewe and Bwiru Is- as frequently assumed. In the early days of lands). The mainland shore in the south is this century, visibility reached up to 8 m dominated by shrub vegetation and grassy (Graham 1929). Since then it has decreased hills, while very hilly areas in the southeast considerably. The maximum measured by (e.g. Mwanza Gulf) with heavily eroded gran- us at offshore stations during the 1990s ite rocks are dominated by deciduous and was 4.2 m, however, at many places trans- semi-deciduous shrubs and euphorbias. parency is below 1 m. The half-evergreen vegetation on many Limnological studies have been carried small rocky islands in the Southeast is domi- out on Lake Victoria in the early years of nated by fig trees and euphorbias. In the this century by Graham (1929) and were East the grassy Serengeti plains reach the taken up again much later (Talling 1957, lake shores. This part of the lake region re- 1966, Akiyama et al. 1977). Recently limno- ceives the lowest rainfall. logical work has been intensified, with data The lake basin morphology of Victoria collecting, however, largely restricted to is very different from that of the rift lakes. northern parts of the lake. This work shows It is basically saucer shaped, with a large that dramatic changes have been happen- open water body of between 60 and 100 ing in the nutrient household of Lake Vic- m depth, a greatest length of 400 km and toria. Concentrations of nitrogen are increas- greatest width of 320 km. The compara- ing since the 1920s, those of phosphate tively little depth, combined with the since the 1950s, induced through precipi- wide surface, allowing annual total mixis tation and human activity in the catchment of the water mass, makes it that until the area, such as agricultural and urban devel- recent changes took place, the entire wa- opments and deforestation (Hecky 1993, ter body was inhabitable for fish. Apart Gophen et al. 1995). Increasing phyto- from the open waters and the wind and plankton densities and frequent algae wave exposed western and eastern shore blooms are the consequence (Ochumba lines, the lake has relatively more shel- & Kibaara 1989, Lehman & Branstrator tered shore lines along the large gulfs 1993, Mugidde 1993). Indications that the (Napoleon Gulf, Kavirondo Gulf, Speke lake is undergoing eutrophication have Gulf, Mwanza Gulf, Emin Pascha Gulf) and been recorded since the 1960s (Hecky within the island archipelagos in the north, 1993), and have become more apparent southwest and southeast. after the population upsurge of the intro- Due to the high altitude, temperatures duced Nile perch, and the concomitant de- stay moderate, with daily mean maxima cline of the sublittoral haplochromines of about 28°C and mean minima of about (Goldschmidt et al. 1993). Massive fish 16°C in Mwanza (based on data from the kills have been observed (Ochumba 1987) meteorological station Mwanza). Water and the recent investigations discovered temperatures are likewise moderate. We that most of the deep water in the for- measured maxima of 27°C and minima of merly entirely oxygenated Lake Victoria,

15 is now consistently anoxic (Hecky et al. 1994, Gophen et al. 1995). There are indications that annual vertical mixing of the water body does not reach the deep layers any more (Gophen et al. 1995). With these recent alterations the deeper waters of Lake Victoria become probably as hostile to fish life as are those in Lakes Malawi and Tangan- The sources of the Nile at Jinja in Uganda. yika. Upwelling of hypoxic water has been observed also at mass, that after a long period of relative tec- steeply sloping rock shores (N. Bouton tonic stability, began to experience earth pers. comm., pers. obs,), where it poses movements and volcanic activity in the a threat to rock-dwelling cichlids. Miocene and Pliocene epochs (roughly about 20 to 5 Myrs ago). In the course of Geological history of Lake Victoria these, the rifts formed and at several places started to fill up with water, as a conse- The African continent is an ancient land quence of river severance. The position of modern Lake Victo- ria was occupied by a highland ridge, constituting, near what is now the eastern shore of the lake, a watershed between rivers that flew westward via modern Zaire river system into the At- lantic ocean, and others that flew eastward to the Indian Ocean (Cooke 1958). Smaller lakes existed by that time and fish fossils from lacustrine deposits were found. The fossil ichthyofauna had little in common with the modern Victorian fauna. It contained Zinga Island was separated from the mainland (below) a few Oreochromis of the O. mos- decades ago. If the water level rises further, the island will be sambicus-type (today con- split into two. fined to rivers flowing into

16 UGANDA Ninja

Nyanza Gulf KENIA

Ukerewe Island

Speke Gulf Mwanza Emin Kome TANZANIA Pasha Island Gulf Mwanza Gulf 0 100km

Figure 2. Top right: Lake Victoria with depth contures and important localities. Large drawing: Lake Victoria, the area surveyed for this book, and the distribution of rocky habitat (red) in the surveyed area. In green areas the exact distribution of rocks is not known. Photo top left: Senga Point and the Speke Gulf.

17 the Indian Ocean), other cichlids that ago the lake was much bigger, extending could not be identified, Polypterus and westwards into modern Kagera valley. Lates (Fryer & Iles 1972). Thus, it con- Several satellite lakes and semi-lacustrine sisted mostly of fishes that do not natu- water bodies have remained there (e.g. rally occur in modern Lake Victoria but are Lakes Ihema, Twamwala and Rushwa). characteristic either for the East African, The maximum extension of the lake was or for the nilotic fauna. The extinction of followed by several phases of lake level these taxa is considered evidence for recessions, which have been suggested complete drying-up of the lakes before to correlate with phases of increasingly modern Lake Victoria came into being. deep incision of the Victoria Nile outlet. The latter process cannot yet be accu- Not only tectonic activities, but also cli- rately dated but there are indications that matic fluctuations contributed consider- the region west of the modern lake got ably to the instability of the lake shore to- elevated about 500 000 years ago pography. During the last glacial period through warping. As a result, the west- (about 20,000 years ago) temperatures in ward flowing rivers (proto-Nzoia/Katonga equatorial Africa were considerably lower and proto-Mara/Kagera) were ponded than today. When the temperature rose back, and an internal drainage system was (about 15,000 years ago), the ratio of pre- formed. The former river beds broadened, cipitation to evaporation must have and riverine environments got trans- changed in favour of evaporation, which formed into lacustrine environments, has sometimes been assumed to have edged by extensive swamps. Based on caused a drop in lake level (Fryer & Iles the topography of the present day drain- 1972). This has largely been confirmed by age system, it has been suggested that recent micro-fossil (e.g. unicellular algae, at least three such lakes have existed sponge needles, tree pollen) analysis of (Fryer & Iles 1972). Relatively small local sediment cores from Lake Victoria, that earth movements may have been suffi- provided important new insights into the cient to break and create watersheds lake's recent history. among the lake-like river arms, and to cut The micro-fossil record currently ex- off bays from the major body. It has been tends back to 25,000 years BP, and shows suggested that modern Lake Kyoga, a that the lake level has been fluctuating semi-lacustrine extension of the Victoria very much within this period (Stager et al. Nile, is representative for this riverine- 1986). Maximum aridity occurred be- lacustrine transitional state (Fryer & Iles tween 15,000 and 13,000 BP. Since the 1972). It is neither known how long the deepest sediment core has been taken at transformation of the river system into a a water depth of 66 m, it cannot yet be lacustrine environment took, nor how said, whether Lake Victoria dried up com- many isolated lakes were formed, and pletely, though the concomitant reduc- whether they stayed sufficiently long iso- tions in the levels of Lakes Tanganyika lated to be of interest in the reconstruc- and Malawi would suggest so (Owen et tion of the evolutionary history of the Vic- al. 1990). However, it seems certain, that torian cichlid flock. Nevertheless, cichlid between 14,750 and 13,700 BP the wa- researchers till today draw heavily on this ter level fell below the level of the cor- transitional phase to explain speciation ing site. This is about 70 m deeper than (see the chapter about the species flock). now, and means that the lake was re- As the saucer was eventually filled, the duced to 20% of its present surface area, water broke through the northern water- and that its circumference was reduced shed, forming the Victoria Nile. by more than 45%. Maximum depth was There is good evidence for that Lake merely 26 m (see text figure 10 in the Victoria's shoreline was instable ever chapter on evolution). This is of major in- since its formation. About 60 000 years terest in the discussion about the evolu-

18 tion of Lake Victoria's species flock, and the rock-dwelling cichlids in particular. The south-eastern part of the lake, the rocky shores of which were surveyed by us as reported in this book, was entirely dry! The lake, however, did not separate into disjunct water bodies. After 13,700 BP the lake level rose and was 8,000 BP considerably higher than today. A second overflow may have existed south of present days Mwanza Gulf. Since then, the water level is recessing again (Stager et al. 1986). The most recent major recession has been dated to about 3700 years ago (Temple 1967), another date of interest in the discussion about the evolution of rock-dwelling cichlids. Interesting are also changes that happen on very short term: A comparison of the present shoreline topography (see photo 16) with a detailed map from 1908 (revised reproduced on page 65) shows, that the water level is currently about two metres higher than it was by 1908. Several rocky areas that were peninsulas by then are now separated from the mainland by shallow water (Ndurwa Point = Ndurwa Island, Kawangazi Point = Igombe Island, Zinga Point = Zinga Island).

Postscript: Latest research results published in the journal Science of 23 August 1996 make most likely that Lake Victoria had entirely dried up 12,400 years before present (Johnson et al. 1996).

19 The cichlid species flock

The species flock of haplochromine cichlids it may have been close to a group of river- in Lake Victoria has been considered the ine species, usually referred to as the most striking example of explosive evolu- Astatotilapia bloyeti-complex. Populations tion and adaptive radiation among verte- of this complex inhabit almost every East brate animals (Wilson 1994). Its complex African river that drains to the Indian ecology and evolutionary history captivates Ocean, but have not unambiguously been the minds of biologists since more than a recorded from river systems connected to century ago. Though the picture of the Lake Victoria. Molecular data confirm that flock, and of its causal and temporal con- A. bloyeti is closely related to the Victo- text, has become refined over several gen- rian flock (Meyer 1993). However, since erations of research, many basic questions the rivers that were "back-ponded" to form are still unresolved. Among them the ques- Lake Victoria, drained to the West and not tions about the ancestry of the species to the Indian Ocean, it seems likely that flock, and why and how the species be- the closest relatives of the flock's ances- came that many (modes of speciation). tor are found in some of the smaller west- The question, whether the haplochro- ward flowing rivers, that are now cap- mine cichlids of Lake Victoria form a spe- tured between the ridges of the eastern cies flock in the strict sense (sensu Green- and central African rifts. More recent mo- wood 1984), that is, a monophyletic group, lecular work supports this (Meyer at al. derived from one common ancestor, has 1994). long been debated. An ancestral species Although anatomical studies so far failed would likely be a riverine (river-dwelling) to produce any evidence for a monophyletic cichlid, similar to those haplochromines origin of the Lake Victoria cichlid species that inhabit the rivers of modern East Af- flock, the now available molecular evidence rica. Frequently it has been assumed that is positive (Meyer et al. 1990). Neverthe-

Astatotilapia bloyeti from a river flowing into the Indian Ocean.

20 less, because not only one, but several riv- 1984, Meyer et al. 1990) and 14,000 years ers were back-ponded to form Lake Victo- (Owen et al. 1991). Despite its very recent ria, it is well possible that more than one origin which is reflected in very little ge- riverine population founded the flock. netic differentiation (Sage et al. 1984, Meyer Whether the flock is of mono- or et al. 1990), the Victorian cichlid species oligophyletic origin, strictly speaking, de- flock is ecologically not less diverse than pends then on whether the founding popu- those of the older Lakes Malawi and Tan- lations had, before the lake was formed, ganyika. The modern haplochromine been long enough isolated in their river ba- cichlids of Lake Victoria can be grouped sins, to behave as separate species on re- into at least 16 different feeding groups, unification, or whether they merged into so called trophic groups (Greenwood 1974, one. However, I am not sure whether this Witte & van Oijen 1990), from small inver-

View on Anchor Island from Saa Nane Island. would ever be resolved, and the difference tebrate and fish fry predators, similar to does little to the amazing speed with which those living in today's rivers, to tiny plank- the founder population(s) evolved into an ton feeders, large fish hunters, snail eaters assemblage of hundreds of species that, as and algae scrapers. They inhabit every avail- E.O. Wilson has put it, fill almost all major able lacustrine habitat from narrow crevices niches available to freshwater fishes as a among rocks, water lily beds, and floating whole (Wilson 1994). papyrus islands, to the entirely un- Recent estimates of the age of the flock sheltered pelagic zone of the open wa- vary between 200,000 years (Sage et al. ter body, and to the deepest parts of the

21 lake that are not reached by daylight. adaptive (e.g. many rows of densely set With the latest update presented here, teeth in the mouth of an algae scraper), ac- more than 500 different species of haplo- tual adaptiveness (= possession of the char- chromine cichlids are known from Lake Vic- acter gives competitive advantage over in- toria. The true species number is almost dividuals that do not possess it) has rarely certainly even higher. Several writers that been demonstrated. This gap in the argu- compared the ecological composition of ment of adaptive radiation has sometimes the three large cichlid "flocks" (note: the been hotly debated (see for instance Liem Lake Tanganyika "flock" is not a species flock 1980 and Barel 1983). Some indirect evi- in the strict sense because it has been dence for the adaptivity of interspecific dif- shown that it is polyphyletic (Sturmbauer & ferences in the morphology of the feeding Meyer 1993)) noted two apparent peculiari- apparatus in Lake Victoria rock cichlids has ties of the Lake Victoria flock: a very high recently been produced (Seehausen et al. proportion of piscivorous cichlid species in press [a]). Much work remains to be done (Fryer & Iles 1972, van Oijen 1982, Green- to understand the phenomenon of adaptive wood 1984) and a very low proportion of radiation. It is often assumed that competi- rock-dwelling algae scrapers and other rock- tion between species is driving adaptive ra- restricted cichlids (Fryer & Iles 1972, Green- diation, but again, well demonstrated evi- wood 1984, Ribbink 1991, Goldschmidt & dence for interspecific competition in Witte 1992). This picture of the overall eco- cichlid assemblages is very rare, and noth- logical shape of the flock is changing con- ing is known about its evolutionary signifi- siderably with the discovery of the rock- cance. dwelling cichlid assemblage described in The term adaptive radiation is mostly this book, and is becoming more similar to used to describe the enormous diversifica- that of the "flocks" in the other lakes tion of the feeding apparatus and feeding (Seehausen 1996a). It is likely that the re- behaviour in cichlids (Fryer & Iles 1972, semblance will become more evident, as Greenwood 1974, 1981). Lacustrine more of the littoral habitats along the lake cichlids, however, radiated in a number of shore will get surveyed. However, a high different, though not independent traits, proportion of piscivorous cichlid species such as substrate affinities and depth dis- may remain a peculiarity of the original Vic- tribution. This aspect of adaptive radiation torian flock. was long overlooked by scientists and The term adaptive radiation stands for aquarists working with Victorian cichlids ecological diversification of a founder stock alike. An apparent eurytopy of Lake Victo- (the ancestral species), going hand in hand ria cichlids, as expressed by generally with splitting of its gene pool into a multi- weakly developed habitat specificity (Fryer tude of isolated gene pools (species), each & Iles 1972), and absence of geographically of which occupies only a small part of the restricted distributions within the lake (Fryer total ecological space. Adapted to the dif- & Iles 1972, Greenwood 1974), was, and ferent niches, these species can utilize the by some people still is considered a strik- particular resources of any niche more effi- ing difference between Victorian and ciently than the non-specialized ancestor Malawian/Tanganyikan cichlids. Evidence could, but are less efficient in utilizing other against these assumptions has been niches which have their own specialists. brought forward by several authors (Witte Consequently each species is rather nar- 1984, Goldschmidt et al. 1990, van Oijen rowly restricted to its niche. This concept 1991, Seehausen & Bouton 1996a, in press) appears simple and is often regarded an and much new evidence for habitat obvious fact, particularly in aquarium litera- specificity and intralacustrine endemism is ture about cichlid evolution. However, real- provided in this book. Less dramatic, but ity seems much more complicated. For nevertheless existent among Victorian characters which common sense considers cichlids, are radiations in social behaviour

22 and in reproductive strategies (table 2, tional case (L. Kaufman pers. comm., Goldschmidt & Goudswaard 1989), for Seehausen et al. in press [b]). Kaufman and which adaptiveness could be assumed as Ochumba (1993) phrased the term "species well. A dramatic radiation has finally oc- pump" for the process by which new spe- curred in male coloration, adaptiveness of cies are added to the flock of Lake Victoria which is not quite apparent on the first when periods of high water level alternate glance. with such of low water level and speciation The chief theoretical puzzle created by in lake side lagoons. However, species en- species flocks is the process by which they demic to satellite lakes, usually differ from grow (Wilson 1994: 102). Neither the their nearest relatives in Lake Victoria pre- mechanisms of growth (i.e. speciation) nor dominantly in coloration and not much is the mechanisms by which growth is con- known yet about whether these differ- trolled (how many species can coexist) are ences would be sufficient to prevent inter- yet fully understood by biologists. Several breeding when the geographically isolated potential ways of speciation in cichlid lakes siblings would be reunited by a rise of wa- have been suggested. Reduced to the es- ter level. Thus, the speed with which dif- sential message, these are, (1) classical ferences between satellite populations and allopatric speciation by geographical isola- the main lake populations evolve, support tion, (2) micro-allopatric intralacustrine the hypothesis of speciation in separated speciation by ecological isolation, (3) sym- lake basins, but do not prove it. patric speciation. For several reasons speciation by geo- The evolutionary significance of graphical isolation in isolated lake basins is speciation in sympatry, in the absence of most unlikely to be the only mechanism geographical and ecological barriers that that accounts for the species flock forma- prevent interbreeding among diverging tion in Lake Victoria. forms, has long been disputed. Recently, (1) it is difficult to perceive how repeated however, strong supportive evidence for its fragmentation and re-unification of the lake role in cichlid species flock formation is ac- could have led to the evolution of more than cumulating from both empirical (e.g. 500 species. Schliewen et al. 1994), and theoretical work (2) splitting of gene pools in geographi- (e.g. Turner et al. 1995). The potential role cal isolation cannot explain the radiation of sympatric speciation should no longer be into various feeding and habitat niches. The neglected in attempts to explain evolution pattern of major niche complementarity is in Lake Victoria. better explained by intralacustrine diversi- As a logical consequence of the assump- fication of the kind recently demonstrated tion that Victorian cichlids display little habi- for several fish groups in geologically young tat specificity, micro-allopatric speciation post-glacial lakes of the northern hemisphere has not been considered much either in (Bentzen & McPhail 1984, Bentzen et al. 1984, the attempts to explain the explosive evo- Skúlason et al. 1989, Taylor & Bentzen 1993, lution of Lake Victoria cichlids (but see and references herein). Ribbink 1991). Instead the focus has long (3) satellite lakes usually have only a few been on allopatric speciation in isolated lake habitat types, they are mostly shallow, sur- basins. This idea was predominantly enter- rounded by swamps, have reed and papy- tained by P.H. Greenwood who used Lake rus edges, muddy bottoms, some shallow Nabugabo as a striking example of rapid open water and maybe some sand bottom. speciation in geographical isolation. This It is highly unlikely that rock-restricted former lagoon is cut off from the main lake cichlids, deep water-restricted cichlids or since no more than 4000 years ago, but is pelagic open water dwellers could ever inhabited by 5 endemic cichlid species. have evolved under such conditions. Line- Ongoing work on other satellite lakes ages with specific habitat demands such shows that Nabugabo is certainly no excep- as pelagic Yssichromis and rock-dwelling

23 Neochromis are absent from the small tor reshaped the species flock dramatically present-day satellite lakes, and poorly rep- but is unlikely to cause the destruction of resented in the biggest satellite lakes. the entire flock (Seehausen 1996a). It is the (4) the strongest argument for speciation species with largest habitat overlap with modes other than classical allopatric, is the Nile perch, and the larger species, in par- high degree to which regional endemism ticular the fish eating cichlids, that were ex- is developed among the rock-restricted terminated almost entirely. Some small de- cichlids, that cannot be correlated with pos- tritus and zooplankton feeders apparently sible distribution of former satellite lakes manage to coexist with Nile perch (Witte (see the chapter on evolution). et al. 1995) and larger species diversity cur- It seems therefore certain that modes of rently survives in a number of refugia speciation, other than classical allopatric, are (Kaufman & Ochumba 1993, Seehausen & involved in the formation of the Lake Victo- Witte 1995). These are habitats, not well ac- ria cichlid species flock. cessible to Nile perch, or in which its hunt- As much as the cichlid flock of Lake Vic- ing efficiency is reduced. The most species toria is an example of explosive speciation, rich of these refugia are rocky shores, is- it is one also of mass extinction. Nile perch, lands, and small, isolated rocky reefs Lates spec., which since its extinction (Seehausen et al. in press [b]). The preda- about 10 to 20 Myrs ago (long before the tor-reshaped cichlid species flock of Lake cichlid flock came into being), was absent Victoria is in its ecological composition from the Victoria drainage, was brought into rather similar to that of Lake Tanganyika, a the system by man in the 1950s. It had flock that evolved in the presence of Nile been kept in aquaculture in Uganda for perch. Rocky habitats may play a particularly some years, and an introduction into the important role in the further evolution of lake was discussed and disputed, when in Victorian haplochromines. the late 1950s some individuals somehow got into the lake. How this happened is not Postscript: quite clear, however, subsequently the lake While this book went into press, work was stocked several times deliberately in was published by Johnson et al. (1996) that the early 1960s with Nile perch from at makes most likely that Lake Victoria had least two sources, Lake Albert and Lake entirely dried up as recently as 12,400 years Turkana (Harrison 1991). During the first 25 ago. The authors studied lake bottom years after its introduction, Nile perch did sediments by seismic reflection profiles not exert any apparent impact on the and piston cores. They found a terrestrial cichlid fauna. Nile perch catches were low, soil (vertisol) horizon with grass roots in the until, in the early 1980s, they increased ex- deepest part of the lake under an only 7 m plosively. Simultaneously, many other fish thick layer of lacustrine deposits that had species declined, and most dramatically the sedimented since the area is flooded again. haplochromine cichlids. Already in the sec- According to the authors, during the dry ond half of the 1980s Nile perch comprised period the lake was neither split into sev- more than 90% of the total catch weight eral basins, nor would the climate have al- of trawlers, and cichlids had basically dis- lowed the persistence of satellite lakes. The appeared from the catches (Oguto-Ohwayo authors conclude that most of the en- 1990, Barel et al. 1991). Within five to ten demic cichlid species of the lake have years between 150 and 200 endemic haplo- evolved during the past 12,400 years. Spe- chromine species disappeared, many of cies formation in Lake Victoria must have which probably became extinct (Witte et al. been largely intralacustrine, and was more 1992). explosive than most evolutionary biologists Since Nile perch does not inhabit all habi- ever imagined. tats alike, and does not forage on all kind of cichlids alike (Witte et al. 1992), the preda-

24 The Mbipi, overlooked subflocks like the Mbuna of Lake Malawi?

Among more than 100 haplochromine spects closely resembles the Mbuna, the cichlid species, that Greenwood described rock-dwelling cichlids of Lake Malawi and redescribed from Lake Victoria, was (Nyasa), and is for cichlid hobbyists certainly one strictly lithophilic (rock-restricted) algae no less an attraction, and one of the most scraper, which later, in Greenwood's (1980) interesting components of the Victorian generic revision became the sole Victorian flock. representative of the genus Neochromis. The majority of the rock-restricted Greenwood seemed convinced that strictly cichlids of Lake Malawi were considered, lithophilic cichlids were uncommon in Lake by fishermen, taxonomists and aquarists, Victoria, and was surprised when the Hap- a group on its own within the Lake Malawi lochromis Ecology Survey Team discovered a group of such fishes 30 years after he had started working on the lake (van Oijen et al. 1981, Greenwood 1994). This group consisted of eleven species (Witte et al. 1992), one of which has in the meantime been described (Witte-Maas & Witte 1985). It took another decade until about 10 further rock-restricted species were discovered between 1986 and 1991, when J. Sevenster, N. Bou- ton, Y. Fermon and I visited some rocky islands that had not been sampled previously. Based on these results, it was predicted that the number of rock-dwelling cichlids in the lake is likely to be even much higher (Seehausen 1992). Irrespective of this, it was a great surprise again to ichthyolo- gists, who had been working on Lake Victoria before, when the survey of rocky shores in southeastern Lake Victoria that we conducted between 1991 and 1296, produced about 90 more, previously unknown species of strictly lithophilic cichlids. It is just now becoming apparent that a large and taxonomically diverse part of the Victorian cichlid species flock had simply been overlooked in the past. This newly discovered assemblage of rock cichlids in many re- Steep rocky coast are rare along the Victorian shoreline.

25 cichlid species flock long before molecular of other habitats, share a common facies biologists largely confirmed their status as that is recognized by the biologists work- a subflock — i.e. a group of species derived ing with them. Though difficult to quan- from one common ancestor within the tify, it may indicate that they form a mono- larger flock. The Tumbuku term Mbuna with or oligophyletic group within the Victo- which fishermen at Lake Malawi collectively rian flock. denote the either dark or brightly coloured On the basis of melanin pattern, two rock cichlids, was adopted by biologists and distinct groups within the rock cichlids aquarists. Similarly do fishermen at south- can be discerned. The "Vertical bar Mbipi" ern Lake Victoria have a collective name for are characterized by a number of simple Victorian rock cichlids: Mbipi. In the vertical bars on the flanks which can be Kisukuma language, which is the most crossed by a faint mid lateral band, as ex- commonly spoken language among fisher- emplified by H. (Neochromis) nigricans men in the southwestern lake region, Mbipi and H. (?) nyererei. (photos on page 28; means as much as "the dark ones". for explanation of the "(?)" see page 35). Fishermen were aware of the exist- The "Chessboard Mbipi" are characterized ence of the Mbipi long before biologists by a mid and a dorsal lateral band cross- happened to "discover" them. Fishermen ing the vertical bars, reminiscent of the are frequently rather good naturalists and black and white pattern on a chessboard, parallels between their popular classifica- as exemplified by H. (Paralabidochromis) tions of fishes, and the scientific classifi- chilotes. The melanin pattern is most eas- cations made by biologists, are common. ily seen in juveniles and females. Males, With names like Mbuna and Utaka, fish- and less frequently females of chess- ermen of Lake Malawi defined groups board species, when sexually active, may among the cichlid species, that are very lose the longitudinal bands while retain- similar to those, defined and delimited by ing vertical bars. Though trophic (= feed- biologists. The cichlids of Lake Victoria are ing related) diversity is lower among divided by fishermen into three major "Chessboard Mbipi" than among "Vertical groups, Furu (Haplochromis), Komaga bar Mbipi", I did not observe any obvious (Astatoreochromis) and Sato (Oreochromis). correlation between melanin pattern and Within the Furu three subgroups are com- behaviour or habitat. In both groups exist monly made: (1) Mgobe-gobe, the large- more and less aggressive species, inhab- mouthed piscivorous species of the itants of the rock surface, and those in- Harpagochromis and Prognathochromis lin- habiting crevices and caves, as well as dif- eages; (2) Mbipi, the rock restricted, mostly ferent feeding strategists (compare also dark or brightly coloured cichlids; (3) all oth- text figure 10a), though a group of ers. With the distinction of Oreochromis, planktivores has apparently evolved only Astatoreochromis and Haplochromis, the among the "Vertical bar Mbipi". folk taxonomy identifies the same major The melanin patterns may therefore bear groups of cichlid species in Lake Victoria more relevance for the reconstruction of that can be readily identified by molecu- phylogeny (= line of descent), than anatomi- lar biology as the major cichlid lineages cal characters that are often directly associ- in Lake Victoria. With the recognition of ated with ecology (see also Eccles & piscivores and rock restricted haplo- Trewavas 1989 for Lake Malawi haplochro- chromines as subgroups within the Furu, mines). My hypothesis is that the two mela- the fishermen may prove to be ahead of nin pattern groups represent two phylo- biologists. Particular vernacular names for genetically distinct lineages within the rock- weed bed- or sand-dwelling cichlids do dwelling cichlids of Lake Victoria. Apart from apparently not exist or are not much in melanin pattern, the shape of the dentary use. Indeed the majority of the rock-re- (tooth carrying lower jaw bone) differs be- stricted Victoria cichlids, more than those tween the two groups. In the "Vertical bar

26 Mipi" the tooth carrying surface of the trophically aberrant members of the verti- dentary runs slightly concave, and the cal bar Mbipi. outer teeth are implanted more or less Based on the similarity of melanin pat- upright (text figure 4). In species of the terns and on jaw and dental morphology, a "Chessboard Mbipi", the tooth carrying close relationship between the "Vertical bar surface of the dentary describes anteriorly Mbipi" and the Haplochromis lineage (sensu a ventral inflection. In lateral view it runs stricto) on the one hand, and between the somewhat convex. Consequently the an- "Chessboard Mbipi" and the Psammochro- terior outer teeth are implanted slightly mis lineage on the other hand seem likely. to strongly procumbently (text figure 4). The first of these assumed relationships Within each melanin pattern group, a would be congruent with Greenwood's number of species complexes can read- (1980) hypothesis of a superlineage consist- ily be recognized, based on anatomy. The ing of Neochromis, Xystichromis and Haplo- anatomical diversity within each complex chromis. It may have to be extended to in- is only partly reflective of interspecific dif- clude all species complexes of "Vertical bar ferences in ecology, and I am moderately Mbipi". Similarly the assumed relationship confident that the majority of the species between Paralabidochromis and Psammo- in each complex represent a mono- chromis would be congruent with Green- phyletic group. In the following account wood's hypothesis of a Psammochromis- of the rock-dwelling cichlids, I present the Macropleurodus superlineage, which con- species in these complexes. About 60 tains Paralabidochromis. This superlineage species are currently known in the "Verti- does not only overlap in species composi- cal bar" group, slightly more than 20 in the tion with the "Chessboard Mbipi" but shares "Chessboard" group. anatomical characters with both Mbipi Apart from the Mbipi, we found among superlineages. These are in particular as- rock-dwelling Lake Victoria cichlids several pects of the general head anatomy: a rela- of the lineages described and defined by tively short lower jaw, supported by mus- Greenwood (1980): Psammochromis, cles that allow powerful biting, and the ten- Ptyochromis, Labrochromis, Lipochromis, dency to increase the number of inner tooth Harpagochromis, Haplochromis, lacustrine rows. These characters appear to be adap- "Astatotilapia" and fluviatile Astatotilapia. All tations to feeding upon benthic organisms, of these non-rock restricted lineages, ex- that need to be removed from a firm cept fluviatile Astatotilapia, have produced substrate by biting, and may well be inde- a few strictly lithophilic species, but none pendently derived. In that case they would of them comes even close, in terms of spe- not indicate a recent shared ancestry of the ciosity at rocky shores, to the "Vertical bar" Psammochromis-Macropleurodus lineage, and "Chessboard Mbipi". If my phylogenetic including the "Chessboard Mbipi", with the hypothesis is correct, the rocky habitats of "Vertical bar Mbipi". However, it is equally Lake Victoria are dominated by two phylo- well possible that an ancestral species, af- genetic superlineages, which both have un- ter having achieved the ability for forceful dergone considerable secondary radiations. biting, invaded all feeding niches that de- These radiations are particularly apparent in mand for it, and consequently split up into the Vertical bar superlineage. It contains lineages with different feeding ecology. In large groups of algae scrapers, insect eat- that case the Psammochromis-Macro- ers, and plankton eaters, and thus, has ra- pleurodus superlineage could perceivably diated to tap the three major food sources be the sister taxon to the "Vertical bar found in rocky habitats. It is not impossible Mbipi". I hope that developments in mo- that certain anatomical and ecological spe- lecular genetics and other phyloge- cialists among the rock-cichlids, that I cur- netic techniques will soon make such hy- rently assign to other lineages on basis of potheses testable. their anatomy (e.g. Lipochromis), are Among the now over 130 species of

27 rock-restricted and other rock-dwelling several decades of field research, is per- cichlids, we found high degrees of re- haps due to the murky waters of Lake Vic- gional endemism and distinct zoo- toria that make faunistic and ecological geographical patterns in distribution. Be- work much more difficult than it is in the cause to date we have surveyed merely clear waters of Lakes Malawi and Tangan- about 15% of the rocky shores of Lake yika, but not less fascinating. Victoria, I expect many (probably over 200) more rock cichlid species to await scientific discovery. That their extreme diversity was simply overlooked during

H. nyererei from Ruti Island; an F1 male showing a typical vertical bar pattern.

A male H. "rockkribensis" in the aquarium, exhibiting a typical chessboard pattern.

28 Taxonomy of rock-dwelling cichlids

Species recognition taxonomist must therefore be to identify groups among populations that are closer Because stenotopic rock-dwelling related to each other than to others, and to cichlids form many small isolated popu- define and delimit species among them. lations, species recognition and delimita- Ribbink et al. (1983), when confronted tion among them is more difficult than with the same problem in Lake Malawi among cichlids of the sublittoral and open (Nyasa), decided to apply the Specific Mate waters. In the biological species concept Recognition System (SMRS) described by (defined by Mayr 1942) a species is a Paterson (1978) to define and delimit spe- population or group of populations whose cies. This system stresses the importance members have the potential to interbreed of mutual recognition among individuals of with one another in nature, producing fer- a species as partners in reproduction. Sym- tile offspring. A species is isolated by in- patric forms that do not recognize each trinsic ecological, behavioural or other other, and thus do not regularly interbreed, barriers from other species with which it are considered different species, no mat- does not successfully interbreed in na- ter how similar they are to the human eye. ture. This concept cannot be strictly ap- In a number of field studies of sympatric plied to rock-dwelling haplochromines. forms of Lake Malawi Mbuna, that were Because populations, living at different considered colour morphs of one species, islands (or habitat islands), are effectively it has been shown, that individuals of cer- isolated from each other by stretches of tain colour morphs that differ also in behav- sand bottom and open water, it is often iour and microhabitat, do not mate with impossible to judge whether members of each other, and represent different species two would interbreed if they could meet. (Holzberg 1978, Marsh et al. 1981). Other We found that some anatomical and/or studies had demonstrated earlier, that col- coloration differences exist among any oration and behaviour play important roles two island populations in most of our re- in cichlid communication during reproduc- search area, and it has been shown that tion (e.g. Baerends & Baerends-van Roon gene flow is restricted even among those 1950). that are topographically not far from away To categorize allopatric populations of each other (Dorit 1990). Thus, one could with Mbuna, Ribbink et al. (1983) assumed that some justification declare most populations if the characters, known to be important in different species. This is not only likely to species recognition (body shape, coloration, be wrong in most cases, but would veil melanin pattern, behaviour, microhabitat), morphological and ecological diversity by are the same among two geographically giving the same taxonomical rank to un- isolated populations, individuals would rec- countable very similar forms (indeed thou- ognize each other if they would chance to sands) and much fewer more different meet. Individuals of other populations, how- forms. Such a system would furthermore ever, that differ in one or more of these char- not provide any information about phyloge- acters, so they assumed, would not recog- netic relationships among populations. nize each other. Applying this concept Such information, however, is required to strictly on some groups of Mbuna, Ribbink understand evolution of rock-dwelling et al. emphasized differences rather than cichlids, and to interpret ecological patterns similarities, and considered populations, in an evolutionary context. The task of the that differed distinctly in one of above

29 Figure 3: Jaws of a typical “Vertical bar Mbipi”: Figure 4: Jaws of a typical “Chess-board Mbipi”: H. (Neochromis) “velvet black”. The outer teeth H. (Paralabidochromis) “rockpicker”. The outer in both jaws are implanted more or less upright. teeth in the lower jaw are implanted procumbently. Drawings made by R.J.C. Hoogerhoud from Drawings made by R.J.C. Hoogerhoud from material of F. Witte and E.L.M. Witte-Maas. material of F. Witte and E.L.M. Witte-Maas.

mentioned characters, as being poten- species complex, has only very recently tially non-interbreeding, i.e. different spe- been shown in laboratory experiments (See- cies. hausen 1996). While the results of that For Lake Victoria cichlids, assortative study support the applicability of the SMRS mate choice among anatomically similar, concept also to Lake Victoria Mbipi, there but in coloration different forms of one are also a number of problems involved

30 with its application. that the two forms are reproductively iso- (1) The result of species delimitation lated, or would be so if they chanced to among allopatric populations can depend meet. Nevertheless, final confirmation can on the geographical scale of the study. If only be obtained by studying mating behav- between two populations only minor differ- iour. ences exist, and non of the SMRS charac- As a consequence of these difficulties, I ters is fundamentally changed, they will be consider differences in one of the SMRS considered conspecific. This is commonly characters not as sufficient to separate spe- the case among populations from nearby cies within complexes of closely related patches of rocky habitat. On the other hand, allopatric populations. I suggest to regard two allopatric populations that differ in two forms of rock-dwelling cichlids as dif- anatomy and microhabitat distribution from ferent species only if they differ in male col- each other as much as each of them dif- oration plus at least one other character. fers from other sympatric species, would Furthermore I refer to distribution patterns be considered two different species. How- wherever the available data allow. Thus, I ever, they can turn out to represent merely consider two distinct forms that are con- extreme end points of a cline of gradual nected by a cline of intermediate popula- change, once the populations are known tions, as belonging to the same species, that live at localities lying topographically unless the extreme forms coexist sympat- between those of the extreme populations. rically in parts of their range. In the latter Such a case is known from H. (Neochromis) case, we are dealing with chain species "velvet black". Western Speke Gulf (see also the chapter on evolution). In such, populations, living in clear waters, are ana- and other cases, occasional gene flow may tomically highly specialized epilithic algae exist between two species, as long as both scrapers of Neochromis type, that live pre- phenotypes remain distinct. dominantly on the rock surface. Those of the Mwanza Gulf and eastern Speke Gulf, Generic taxonomy of Lake Victoria living in less clear waters, are much less haplochromines specialized as algae scrapers. They have less closely set teeth, fewer tooth rows, Haplochromis is probably one of the and a less steep dorsal head profile. At the most broadly applied, and most inconsist- same time they live more cryptically among ently used cichlid generic names in popu- rocks. If only the extreme populations lar and scientific literature. The genus would be known, they would probably be Haplochromis was first described by considered specifically distinct, though the Hilgendorf (1888), as a subgenus of the two forms are connected by intermediate marine genus Chromis (damselfishes), for populations. Consequently, geographical a Lake Victoria cichlid with a peculiar den- and ecological patterns of distribution have tition (H. obliquidens, see page 230). In to be considered in species delimitation. the course of the 20th century (2) Different from studies on sublittoral- Haplochromis, given generic rank by and open water-dwelling Lake Victoria Boulenger (1906), grew rapidly with the cichlids, our studies on rock-dwelling spe- discovery of many "new" species from cies revealed a good number of cases of the Lake Victoria basin and Lake Malawi male colour polymorphism (Seehausen & (Nyasa). In the late 1970s it contained Bouton 1996b). Thus, different male colora- more than 300 described and valid spe- tion alone is not sufficient evidence for two cies, and was the most speciose taxon forms to represent different species. If, within the family Cichlidae. With the grow- however, differences in coloration coincide ing species numbers, also the generic with differences in ecology, as may be mani- limits grew far beyond the original diag- fest in details of dentition or microhabitat nosis of Hilgendorf, and accommodated distribution, the probability is much bigger species with various tooth shapes. By that

31 time Haplochromis contained many Ma- ers, which would indicate common ances- lawian, and all but four of the endemic try to those that share them. In this way Victorian haplochromines. Those four he divided the former Haplochromis spe- represented four monotypic genera, dif- cies of the Lake Victoria basin over 16 fering from the other Victorian new and resurrected genera and one of haplochromines in dentition characters, the monotypic genera that had already but obviously being closely related to been nominal before. A similar revision them (Regan 1922). was later done also for the Lake Malawi

The outer teeth of H. "velvet black". The outer teeth of H. "blue scraper".

Considering this classification unsatis- haplochromines by Eccles & Trewavas factory, particularly from a viewpoint of (1989). Haplochromis became restricted understanding the evolutionary history to species with a dentition resembling and phylogeny of this species rich as- that described by Hilgendorf (1888). semblage, Greenwood (1979) split up the By the time of Greenwood's revision, genus Haplochromis. Attempting to ar- slightly above 100 haplochromine spe- range the species in monophyletic cies were known from Lake Victoria. In groups (groups consisting of species de- the years after the revision many "new" rived from one common ancestor that is species were discovered, predominantly not shared with other groups) as genera, by the Haplochromis Ecology Survey he tried to identify derived characters, Team (HEST), and researchers were con- shared by some species and not by oth- fronted with difficulty when attempting

32 to use the new genera. The problems piscivores into the two genera. In each of were mainly the following: (1) New spe- the characters he found greater overlap cies could not be assigned to Green- between the genera than Greenwood wood's genera without extending generic had found and concluded that the char- limits (Snoeks 1988, as an example from acters seem to form continuous Lake Kivu). (2) New species were found morphoclines in which any division would that bridged anatomical gaps between be arbitrary. Thus, he considered a split- several genera (Labrochromis and ting of the piscivores into two genera not Gaurochromis, Hoogerhoud 1984; desirable. Yssichromis and "Astatotilapia", Witte & Among the many new rock-dwelling Witte-Maas 1987; Prognathochromis and cichlids, one encounters some similarly Harpagochromis, van Oijen 1991). (3) New problematic species, but the majority can data and a re-evaluation of Greenwood's quite unambiguously be assigned to data showed considerable overlap in sup- Greenwood's genera and some new lin- posedly diagnostic characters among eages that were not yet known by the species that Greenwood had assigned to time of the revision. That generic limits different genera (Hoogerhoud 1984, van have to be extended after discovery of Oijen 1991). new species, is not unusual in a fauna Consequently these researchers opted that is still poorly investigated. My expe- to use the broad definition of the genus rience with the classification proposed by Haplochromis, that was in use prior to Greenwood, is more positive than not, Greenwood's revision, rather than the and I would like to call a few points into new classification. Recently van Oijen consideration, that may partly explain the proposed to exclude only the riverine negative experiences, other researchers Astatotilapia and to restrict Haplochromis had with it. Hoogerhoud and van Oijen to cichlids of the Lake Victoria basin, in- based their rejection of Greenwood's ge- cluding the monotypic genera (van Oijen neric division on intergeneric overlap in 1995). individual quantitative morphometric char- When Hoogerhoud attempted to test acters because these were (part of) the Greenwood's delimitation of the genera characters Greenwood used to delimit Gaurochromis and Labrochromis, he took the genera. However, in clades (groups of relative measurements of the oral and related species) that underwent rapid pharyngeal jaws, that should be different speciation and adaptive radiation in an among the genera according to Green- unstable environment, it seems to me wood. He demonstrated considerable unlikely that individual morphometric char- overlap among the genera, found one acters can ever accurately reflect species bridging them, and concluded phylogeny. Too large is the susceptibility that intraspecific variability is as great as of any one such character to the impact the intergeneric differences, and the clas- of temporal and spatial instabilities in the sification of the species into two genera environment, to which it was exposed in artificial and arbitrary. the course of evolution. Because the sus- In a similar study, van Oijen (1991) ana- ceptibility to any factor in the environ- lysed four quantitative characters of the ment differs between characters, the skull (neurocranium; see for details combination of several characters is likely Greenwood 1980, van Oijen 1991) that, to be a better indicator of phylogeny than according to Greenwood, are important individual characters, since the impact of for separating the two fish eating genera small scale adaptations is buffered Prognathochromis and Harpagochromis. against. Combinations of characters make He re-measured the very 66 skulls of 43 up the "facies" of species that taxonomists piscivorous species upon which Green- perceive, and according to which they, wood had based his decision to split the sometimes rather subjectively, group

33 them. It is often difficult to translate these genus. "general appearances" into objectively Thus, in both cases multivariate analy- quantifiable characters. However, multi- sis of critical characters largely supports variate statistics such as cluster analysis Greenwood's classification. The neces- may be very helpful to objectively quan- sity to reassign some species should be tify combinations of several characters. expected after the first attempt of classi- Though Hoogerhoud (1984) stated that fying a complex fauna. Since the charac- the overlap that he found among ters measured in above cited publications Gaurochromis and Labrochromis could were only part of Greenwood's diagnoses probably be reduced by using multivariate of the genera, the sorting of species into statistics, he did not discuss this further. his genera by cluster analysis might even I performed a cluster analysis (UPGMA) of become more complete, if all diagnostic his morphometric measurements of characters would be quantified and used. three nominal Gaurochromis and four I do not see evidence for the invalidity of nominal Labrochromis species. The result the genera proposed by Greenwood, were two distinct species groups. One apart from Astatotilapia (see page 235ff). contained three Labrochromis, the other Recently another line of evidence is com- one three Gaurochromis plus one ing up, that in large produces support for Labrochromis that was considered as Greenwood's classification: A detailed bridging the genera by Hoogerhoud. study of squamation patterns comes to Thus, though one species may have to be results that are similar to those of Green- reassigned between them, Greenwood's wood (E. Lippitsch, in manuscript). distinction of the two genera is defend- All currently available evidence sug- able on the measured characters, if they gests that Greenwood was more right are combined. I performed the same type than not in recognizing phylogenetical lin- of cluster analysis also with the eages among Victorian haplochromines. neurocranial measurements of van Oijen He may in part have failed to decode his (1991) and again obtained two major integrated picture of the "general appear- groups which are quite similar to Green- ance" or "facies" of a lineage, and to quan- wood's genera in their species composi- tify its essential components. This can be tions. Seventy one per cent of the Harpa- done now with various multivariate statis- gochromis species clustered into one tical procedures. To understand the evo- group, 91% of the Prognathochromis spe- lution of the unique ecological and mor- cies into the other one. Thus, though phological diversity of Victorian cichlids, again some species may have to be reas- we need a picture of phylogenetic pat- signed between them, the existence of terns amongst them. Greenwood made two major lineages among the pisci- a first step into this direction. His generic vorous Lake Victoria haplochromines subdivision of the Lake Victoria haplo- seems defendable on skull anatomy chromines seems not less justified than when the combination of characters is the generic subdivision of Lake Malawi quantified. Psammochromis acidens, a cichlids and some other cichlid groups. species that Greenwood assigned to an- None of these have ever been perfect, other lineage despite overlap with but they provide hypotheses that can be Harpagochromis and Prognathochromis in tested and altered where needed, to bring individual characters, came out as the sis- forward a more complete phylogenetic ter group to Harpargochromis and the picture of cichlid explosive evolution. Per- two together as the sister group to Pro- sonally I feel, to reallocate all species of gnathochromis. One nominal species of Victorian cichlids back to one broadly de- Prognathochromis (P. arcanus) came out fined genus Haplochromis, means a great as the sister group of all the others, and loss of information. Without giving the should probably be removed from that doubtlessly existing species complexes

34 formal taxonomic status, they would have remained known only to "insiders". For the majority of other biologists and cichlid hobbyists, however, the Victorian cichlid assemblage would on the paper remain an amorphous collection of species. For these reasons I am in general supportive of Greenwood's classification. Many of the new species of Mbipi can be assigned to the genera Neochromis and Paralabidochromis. Many others can be assigned to new lineages that were not known by the time of Greenwood's revision. Some of these deserve the same taxonomic rank as the described lineages. In order to give genera similar diagnostic width, the genus Paralabido- chromis, which was rather broadly defined by Greenwood, and now consists of several large species complexes, may have to be split up. Alternatively, the definition of the genera Neochromis and Xystichromis, which were narrowly defined by Greenwood, would have to be considerably broadened, and possibly joined, to incorporate several of the new "vertical bar Mbipi" lineages. Since this book is not a taxonomical revision, I will present just informal diagnoses of new species complexes, to enable the reader to gain an overview over the diversity of rock-dwelling Lake Victoria cichlids. Pend- ing a formal taxonomical revision of the rock-dwelling cichlids, I opt in this book for a nomenclatorial compromise. In order to provide phylogenetical hypotheses without causing nomenclatorial confusion, particularly among cichlid hobbyists and aquarium fish traders, I use the new generic names as subgenera of Haplo- chromis until a more definite generic assignment of the new species will be done.

35 Habitat and ecology of rock-dwelling cichlids

by Ole Seehausen, Anna Samwel-Terry & Niels Bouton

It has not only frequently been assumed land and rocky islands constitute about a that Lake Victoria has few lithophilic cichlids third of the shoreline in southern Lake Vic- but also that rocky habitats are uncommon toria (text figure 2 on page 17). Rocky in Lake Victoria (Greenwood 1994 as the shores are very common on the large island most recent reference). This view has archipelagos in the northwestern part of the largely been adopted by aquarists, many of lake (J.J.M. van Alphen & F. Galis pers. which consider rocks a most typical cichlid comm.) and in the northeast (L. Kaufman habitat in Lakes Malawi and Tanganyika, but pers. comm.). Even Greenwood himself muddy and swampy areas as the typical once wrote that rocky habitats are "not in- habitats for Lake Victoria cichlids. Certainly frequent" along the northern shores (Green- are mud bottoms and marginal swamps wood 1956a). common habitats in Lake Victoria, and Alternating with stretches of sand and maybe more common than in the other small patches of vegetation in sheltered lakes. However, this does not mean that situations, rocky shores form, in the south,

A rocky shelf interrupts the swampy H. "blue obliqui- shoreline at Bunyago (Mwanza Gulf). dens" grazing algae from a rock. rocky habitats are less common. a habitat mosaic, They are, at least in our research very much like area, in fact just as common as in that described for Lake Malawi. Rocks make up Lake Malawi roughly one half of Lake Tanganyi- (McKaye & Gray ka's shoreline (Konings 1988) and 1984). To some- about a quarter or a third of the body travelling by shoreline of Lake Malawi (Ribbink boat along the & Eccles 1988, Konings 1995 re- lake shores, it be- spectively). Though not yet investi- comes apparent gated on its entire length, all avail- that the shore- able evidence suggests that the line of Lake Victo- shoreline of Lake Victoria is made ria is a drowned up by rocks to not less than a quar- Small rocks in shallow habitats landscape. There ter either. Stretches of rocky main- provide shelter and feeding sub- are old pleisto- strate for small fishes.

36 A territorial male H. nyererei at Makobe Island at a depth of about 6 metres.

A mouthbrooding OB-female H. "blue scraper"

A shoal of plankton-feeding H. nyererei at a Algae-grazing H. "zebra nyererei" (left) and H. depth of about 6 metres. "blue scraper".

37 cene river valleys, cut deep into the rocks (Neochromis) "giant scraper" have never on one (Speke Gulf) or on both banks been observed or caught at water depths (Mwanza Gulf), shallow flooded valleys, hill- beyond 5 m (Seehausen & Bouton in press). top islands like the Vesi Archipelago, and Due to this extreme stenotopy of the highland ridges like Ukerewe. The sub- Mbipi, stretches of sandy bottom or deep strate at the shore, whether sand, shingle water are effective ecological and geo- and boulder fields, or steep rock walls, de- graphical barriers to their dispersal. Conse- pends on the topography of the drowned quently, populations of one species that in- environment. The extension of uninter- habit different rocky islands are effectively rupted rock shores is very variable. It ranges isolated from each other by deep water. from less than 50 m to 10 km, but continu- Similarly populations are isolated from each ous rock shores of more than a kilometre other that inhabit rocky mainland shores, length are rare. There are considerable re- separated by sandy beaches. Migration and gional differences in the abundance of therefore gene flow among them is very lit- rocky habitat. In regions with steep shores tle and each population evolves largely in- like the Mwanza Gulf, the northern Speke dependently, in response to local regimes Gulf, Ukerewe and Kome Islands, rocks are of natural selection. If gene flow between the dominant shore substrate, frequently them is little enough or completely absent, interrupted only by small bays with sand the populations may evolve into two differ- bottom, and make up about 50 % of the ent species, provided that the external iso- shoreline. In other regions, like the south- lation persists long enough. The speed of ern Speke Gulf, sand beaches dominate this process depends on population size, on the shore, and are interrupted merely by how different the selection regimes are in rocky peninsulas, so that rocks make up the respective environments of the isolated only about 10 % of the shoreline. populations, and on incidence. Thus, the The cichlid species found along rocky combination of extreme habitat stenotopy shores of Lake Victoria have in their major- and patchy distribution of the required ity very specific habitat demands, just like substrate, make the Mbipi good candidates those from Lakes Malawi and Tanganyika. for micro-allopatric (intralacustrine) Less than ten of about 80 currently known speciation (see the chapter on evolution). Mbipi species have ever been seen over a The ecology of rock-dwelling Lake Victo- substrate other than rock. H. (Neochromis) ria cichlids is further complicated by the cir- nigricans and H. (Paralabidochromis) "rock cumstance that at any rocky place between kribensis" have a few times, H. (Paralabido- 7 and more than 20 different species coex- chromis) chilotes and H. (Paralabidochromis) ist and share the resources food and space. chromogynos more frequently been caught Before I give an impression of how this is over sand, but hardly ever more than 50 organized, first another look at the rocky metres away from rocks. A species of the habitat. The landscape drowned by the H. (?) "pseudonigricans" complex (H. (?) "grey huge water masses of Lake Victoria, is a very pseudonigricans" according to Witte et al. ancient one. The dominant rock formation 1992) had been observed to leave the is therefore strongly eroded water- and rocks by night to forage pelagically on weather-shaped granite, that exhibits sharp phytoplankton under the surface of open edges only at recent cracks, and is light to waters 100 to 150 m away from the shore medium dark grey in colour. However, at (Witte et al. 1992, Witte pers. comm.). H. some places black rocks occur, that are (Xystichromis) "copper black" can be ob- more sharp edged and appear less strongly served in rock-sand mixed habitats. The eroded. A third type is a very rough kind of other Mbipi seem entirely confined to rocky rock, brown-grey in colour and consisting habitats. Similarly narrowly restricted are of huge almost continuous units, eroding the depth ranges of many species. For in- into boulders of various sizes, that are stance H. (Neochromis) nigricans and H. standing like mushrooms, many at their ba-

38 sis still connected to the "mother rock". We ice- and cave-dwelling cichlids, like the "crev- know this type of rock only from the hilltop ice hunters" of the H. (Harpagochromis) Mabibi Islands in the central Speke Gulf. howesi complex, and crevice-dwelling spe- Depending on qualities of the underlying cies of the H. (?) nyererei complex. Further- rock material, time of its exposure to wind more there are no vertical walls along which and weather, and topography, rock habitats plankton-rich currents could pass. Plankton vary in terms of boulder size and slope. eating cichlids therefore do not find suitable There are huge compact cliffs, as well as conditions either. big boulders, slabs, heaps and fields of The most diverse cichlid communities small boulders, stones and coarse pebble, frequently exist in habitats with medium dominating the rocky habitat at different sized rock boulders. This in especially, when places, and often sharply alternating with the rocks cover a depth range of at least each other. At some places extensive hori- six meters. Such habitats mostly occur at zontal surfaces of very uneven rocky moderately steep slopes and a representa- substrate are covered by shallow water, so tive community has been described by that outcrops of various sizes emerge from Seehausen and Bouton (1996). The role the water surface. Such areas will be re- that the microstructure of the habitat plays ferred to in this book as "rock gardens". in rock cichlid communities, is complex. Microhabitat characteristics again depend The availability of territorial space and shel- largely on slope and boulder size. A steep ter, and the topographical relationship be- precipice of a compact rock wall has little tween these two factors depends on it, but to offer for cichlids. Its surface, as substrate also the composition, diversity, abundance for algae and animal Aufwuchs, is much and availability of food. There are two ma- smaller than that of piles of boulders. On jor food resource bases for cichlids at rocky top of that, a vertical surface receives less shores: epilithic Aufwuchs and plankton. sun light. Thus Aufwuchs is less abundant Epilithic (= on rocks growing) Aufwuchs than in habitats with smaller boulders consists of a variety of unicellular and fila- whose surfaces have various inclinations to mentous green and blue-green algae, grow- sun light. Probably even more important, an ing directly on the rocks, diatoms that unstructured rock wall does not offer hide- mostly grow on the filamentous algae, outs for cichlids. Shoals of females and moss animals (Bryozoa), Ciliata colonies and nonbreeding males of plankton eating sponges. Associated insect larvae and mi- Mbipi, e.g. H. (?) nyererei can sometimes cro-crustaceans living between the algae be observed in such habitats, but territorial filaments are often considered part of the males close to never. On the other end of Aufwuchs as well. Furthermore, different the range of rocky habitat types are gentle species of snails live on the rocks, grazing slopes, composed of slabs, small boulders, the Aufwuchs. The zooplankton is domi- or pebbles. Here inclination to the sun light nated by different micro-crustaceans, pre- is ideal and algae Aufwuchs usually plenti- dominantly species of copepods at the ful. However, slabs again do not provide more offshore islands, and species of hideouts, and are sparsely populated. phyllopods (Cladocera: Daphnia) in shel- Among small boulders are many small hol- tered shallow water areas. The phytoplank- low spaces that serve cichlids as hideouts ton is frequently dominated by diatoms like in case of danger, and that form the centre Melosira, and blue-green algae like the ball- of every male's territory. In such habitats shaped colonies of Nostoc. Apart from cichlids occur in large numbers but their plankton and Aufwuchs, the detritus eating ecological diversity is limited because the prawns of the genus Caridina that have re- habitat is rather uniform, the microhabitat cently increased in Lake Victoria can be a diversity is low. Because of the small boul- rather abundant source of cichlid food at ders, there are no large caves, ledges and certain places and seasons (Bouton et al. rocky interstitial spaces, the habitats of crev- in press). Other food sources, tapped by

39 rock-dwelling cichlids are snails, Potamo- microhabitat distribution or microdistri- nautes crabs, eggs, fry and subadults of bution) and diet. Physical parameters that haplochromines. All these depend directly seem to determine specific microdistri- or indirectly on Aufwuchs and/or plankton. bution, are water depth, the amount of shel- The question about the mechanisms of ter in form of caves and crevices, the expo- coexistence and resource sharing in com- sure to wind and waves, and the distance munities of many, closely related species, from the shore. Spatial niche partitioning is a long discussed unresolved problem in among the members of a rock cichlid com- ecology. It is a widely accepted rule, known munity in Lake Victoria exists on a level that as Gause's principle, that two or more spe- is not less refined than that observed cies with identical ecological demands, can- among the Mbuna of Lake Malawi (Ribbink not permanently coexist, because one et al. 1983). It shall be described here ex- would in the long run be slightly more suc- emplary on one island. Species composi- cessful, and outcompete the other one. In tion, as well as niches of particular species, order to coexist in the long run, two spe- can differ between localities. Thus, gener- cies must ecologically differ to an extend alizations beyond those that we make, that allows them to share resources by should not be deduced from the following slightly different specialisations. Thus, they description. must subdivide an ecological niche into two Descending slowly, as a diver, along the parts, each utilized by one species more ef- moderately steep rocky slope of Makobe ficiently than by the other one. This phe- Island in the Speke Gulf, one meets first nomenon is known as niche partitioning. In with several species of smaller and bigger the early days of ecological research on algae scraping haplochromines of the Neo- Lake Victoria, scientists got the impression chromis lineage. The highest density of that Gause's principle was violated in cichlid Neochromis is found at about 1 to 3 m wa- communities (Fryer & Iles 1972, Green- ter depth, but some species penetrate into wood 1974). Too extreme appeared the waters as shallow as 0.5 m (the quieter the anatomical and ecological similarity among water, the further do they penetrate into the closely related cichlid species, as that one shallows). The anatomically very similar would have thought of ecological niche par- species of Neochromis have staggered titioning among them. However, since then depth ranges, with a trend for bigger spe- it has been demonstrated on cichlids from cies to live in shallower waters. The maxima habitats other than rocks, that even among of the population densities of sympatric anatomically extremely similar species, fine Neochromis species (e.g. H. (N.) "giant ecological differences exist, that may suf- scraper" [1], H. (N.) nigricans [2], H. (N.) "blue fice for successful coexistence (van Oijen scraper" [3], numbers refer to the picture on 1982, Hoogerhoud et al. 1983, Witte 1984, pages 41) are usually separated from each Goldschmidt et al. 1990). other by a difference in water depth of not More recently similar studies have been more than one metre. The depth ranges of carried out on communities at rocky shores the species vary between about 3 m and (Seehausen & Bouton 1996 in press, about 6 m. How far the "deepest" Neo- Seehausen et al. in press [a], Bouton et al. chromis penetrates into deep water, de- submitted). Rocky areas support more spe- pends on the clearness of the water. Even cies on less space (higher "species pack- at clear water islands, however, they hardly ing"), than do most other habitat types. It go deeper than 6 m. was found that each of the different haplo- The fine interspecific differences in depth chromine cichlids, that coexist at a rocky distribution are accompanied by slight dif- shore, has its specific and unique niche ferences in diet, i.e. the proportions of al- within the ecological space available to the gae and other items, and of different algae community as a whole, consisting of spe- types. Though all species the feeding be- cific distribution within the habitat (= haviour of which we studied in the labora-

40 tory (Neochromis and other Mbipi) have the same broad repertoire of feeding techniques, they differ considerably in how often they use the different techniques, even if interspecific compe- Figure 5: An idealized picture of tition is experimentally ex- the cichlid community at Makobe cluded. Thus interspecific Island (numbers refer to the differences in diet do not explanations in the text). merely reflect those in microhabitat distribution, but are caused by differences in feeding behaviour as well. All Neochromis at Makobe Island feed predominantly on filamentous algae. In the lower part of the depth range of the "deepest" Neochromis algae scraper, lives a rare algae scraper of another lineage (H. (Paralabidochromis) "short snout scraper" [4]), that feeds predominantly on unicellular diatoms, growing in the less well illuminated parts of the rocky shelf. A Cladophora belt in the wave zone. A different assemblage of shallow water algae scrapers is found at places where offshore rocks break the waves, so that the (H.) "blue obliquidens" [5] and H. (H.) "pur- shallow water is very quiet. Here species ple yellow"), species of the Xystichromis lin- of the Haplochromis (s.str.) lineage (e.g. H. eage (e.g. H. (X.) "carp" [6]), and Aufwuchs

41 eating stages of Oreochromis niloticus, O. rocks. Finally H. (P.) "rockkribensis" [16] leucostictus, and O. variabilis [7] mix with feeds upon larvae of midges and caddis flies the Neochromis. (Trichoptera) and has a wider depth distri- Most Neochromis species spend most of bution than most others. their time on the surface of the rocks from In very low densities, and without appar- which they scrape algae. Another Aufwuchs ent depth preferences, several predatory scraper that is more omnivorous lives in the haplochromines occur in the community. same depth range as those, but more in the Among them the large H. (Harpagochromis) interstices between the rocks (H. (Xysti- serranus which preys upon small haplochromis) "copper black" [8]). However, it is chromines, and the paedophage H. (Lipo- predominantly the male that spends much chromis) melanopterus [17], preying upon time in interstices, while the females live haplochromine eggs and buccal stage lar- like Neochromis on the rock surface. De- vae, obtained from brooding females. Par- scending to greater depths, one meets at ticularly the species of the deeper parts 5 m depth bright red territorial males of H. of the rocky shelf overlap considerably with (?) nyererei [9]. Their territories are usually Nile perch [18] of fish eating size and are a closely spaced and the fishes are fre- common diet of this predator. Predators quently involved in some small disputes that hunt for rock cichlids in shallow water, along territorial boundaries. Only on the sec- are little egrets (Egretta garzetta), pink- ond glance the presence of the cryptically backed pelicans (Pelecanus rufescens) and coloured females [10] is noticed. They pied king fishers (Ceryle rudis). The catfish move in shoals across the habitat and re- Bagrus docmac, the spotted neck otter main for some time at spots with high plank- (Lutra maculicollis), and cormorants ton densities. Sympatric with H. (?) nyererei (Phalacrocorax carbo, P. africanus) do hunt live two more planktivorous species, H. (?) cichlids at various depths. Brooding fe- "pink anal" [11], which has its maximum males of most Mbipi, unlike females of population density slightly deeper, and H. many Mbuna, guard their fry for a week or (?) "yellow chin pseudonigricans" which pre- more after releasing them for the first time. fers places with somewhat larger rock boul- For this purpose they defend very small ders. Shoals of planktivorous Mbipi usually territories around a hideout. tend to be composed of fishes of relatively In cichlid literature interspecific competi- uniform size. tion for food resources is sometimes in- Most species in any community occur in voked as a mechanism organizing cichlid much lower densities than the algae scrap- assemblages, and driving the evolution of ers, omnivores and planktivores. These are apparently highly specialized anatomical a number of insect eaters, among which forms. The problem with this argument is, again fine ecological differences exist. The that interspecific competition in the African crevice dwelling H. (?) "zebra nyererei" [12] cichlid assemblages has never been un- and H. (Paralabidochromis) chilotes [13] with equivocally demonstrated. Some patterns their shallow head profiles and pointed that we found in the spatial and trophic snouts, feed predominantly on the very (food related) organization of Lake Victoria mobile larvae of may flies (Ephemeroptera: rock cichlid communities may indicate that Baetidae), living on the underside of rocks. competition among species exists. We In contrast, the steep-headed and short- have not been able, however, to produce snouted H. (Paralabidochromis) "blue hard evidence. rockpicker" [14] and H. (P.) "yellow Evidence for competition for space may rockpicker" [15] live exclusively on the rock be derived from our studies on reproduc- surface, and feed almost exclusively on the tion. The reproductive ecology of many smaller, and less mobile larvae of midges Mbipi is very similar to that of many Lake (Diptera: Chironomidae) which live among Malawi Mbuna. Reproductively active filamentous algae on the upper side of males occupy territories of between one

42 and several metres diameter. The depth ranges, across which territories are distributed, are quite similar to the usual foraging depth ranges of the species. Thus, in any given depth, several species reproduce in sympatry (Seehausen 1996d, and unpubl. data). Territorial densities can be extremely high. Males often defend their territories not only against males of their own species but also against those of other species, so that interspecific competition for territory sites may exist among them. Patterns of geographical and microdistribution in some species pairs are suitable to lend support to this hypothesis. So does the microdistribution of territories of H. (?) "blue nyererei" at different islands in the Speke Gulf correlate with the absence or presence of H. (?) nyererei (see page 119). Apart from showing that each species in the complex communities of rock-dwelling cichlids has its unique combination of microhabitat, behaviour and diet, our studies show that these communities are relatively stable but not static assemblages. Environmental fluctuations have considerable impact on the degree of ecological niche partitioning, thus on the magnitude of ecological difference among species. Niche differences can be largely reduced at times of high abundance of one food item, when most species prey heavily on it (Bouton et al. submitted), and at times of low water levels, when spatial overlap among species increases (Seehausen & Bouton in press). This implies that the complex communities of rock cichlids are very sensitive to fluctuations in the environment.

43 Keeping and breeding rock-dwelling cichlids

Nothing is more beautiful than a large (Neochromis) nigricans and H. (?) nyererei rock aquarium with a group of Haplochro- have been in the hobby for quite some mis nyererei whose fire red males dance years now. However, a few species, “iso- like butterflies around the females, with lated” from their taxonomical, ecological sky blue male H. (Neochromis) “blue and evolutionary context, are not enough scraper” and yellow, orange blotched and to establish and keep busy a lobby piebald females, busy scraping algae among cichlid enthusiasts. This in particu- from rocks, with a few deep yellow fe- lar, since information about Lake Victoria males and orange-red males of H. rock cichlids was not available, neither in (Paralabidochromis) “rockkribensis”, and aquarium literature, nor in scientific litera- with some H. (Paralabidochromis) chilo- ture. With the many new discoveries tes, digging with their over-dimensional along rocky shores, and the upcoming lips in the sand among the rocks. trade with Victorian cichlids, this picture In spite of their beauty, very little has is now changing. About 50 rock-dwelling been published about aquarium mainte- species are already maintained in aquaria nance of rock cichlids from Lake Victoria, (table 1). The first articles about rock compared to the extensive literature that cichlids are appearing in aquarium jour- is available about those from Lakes Ma- nals, though those among them that de- lawi (Nyasa) and Tanganyika. Victorian scribe first-hand experience with keeping rock cichlids, with few exceptions, have and breeding, are still the exception (e.g. just recently become available to Schraml 1990, Raymond 1995, Boehme hobbyists. A few species, such as H. 1996). Unfortunately some pioneering au-

A male Haplochromis "crimson tide" in the aquarium.

44 A pair of Haplochromis chilotes in the aquarium. Fighting males of Haplochromis chilotes.

An OB-female of H. "blue scraper" from Makobe (left) and a male H. "red pseudonigricans".

A male Haplochromis chilotes.

45 thors do explicitly or implicitly assume mis lineages, Nyererei-, “Deepwater”-, that their experiences with some Victo- and “Pseudonigricans”-complexes, Para- rian cichlids are applicable to most. Such labidochromis lineage). However, their ag- wrong generalizations were made also gression is the result of keeping mis- about Malawi and Tanganyika cichlids takes, and can be overcome by some when the first species entered the hobby, tricks. implying that Victorian aquaristics is still To successfully maintain rock cichlids, in its infancy. aquaria should be 1.5 m long or longer Understanding of taxonomy and ecol- and contain 250 litres water or more. ogy of Lake Malawi and Tanganyika With some experience, groups of one cichlids would certainly be far less ad- male and several females can be success- vanced without the many valuable obser- fully kept in aquaria of 1 m length and 160 vations made by cichlid hobbyists, which litres volume, but this deserves careful consequently fuelled scientific research. observation, in order to be in time to in- What is needed to achieve a similar par- tervene if the social balance in the group ticipation of aquarists in the accumulation tilts. Anybody who has observed Victorian of knowledge about Lake Victoria rock cichlids in their natural environment cichlids, is literature to provide the or in big aquaria, knows that these fishes upcoming hobbyist lobby with back- are territorial but not really aggressive. ground information about identification, Serious territorial disputes are rare in na- taxonomy, ecology and the natural envi- ture, and I have never seen real fights be- ronment of their new clientele. Though tween males. Neither do males in nature this book is mainly aiming at giving this engage in aggressive behaviour against information for rock cichlids, I want to give brooding females. Territories of Mbipi a brief introduction to aquarium keeping males have between less than 1 and sev- and breeding, which will at places expand eral metres diameter, depending on the beyond the species that live at rocky species and the size of a male. It is sim- shores. ply the space limitation in small aquaria that makes them aggressive. It does not The aquarium allow rock cichlids to establish territories of adequate size, and worse, it does not Aquariums for Victorian cichlids can be allow submissive individuals to leave the as diverse as the cichlids. The minimum territory of a dominant one. Resistance of aquarium size, however, should not con- an intruder against leaving somebody's tain less than 80 litres water. Such small territory, would in nature most likely indi- aquaria, if equipped with a good filter, are cate that it does not accept the domi- perfectly suitable to keep and breed nance of the territory owner. Conse- small sublittoral and pelagic zooplankti- quently the latter would continue to dis- vores, such as Haplochromis (Yssichro- play its dominance until the intruder mis) pyrrhocephalus, H. (?) piceatus and leaves. In an aquarium that is too small H. (?) “argens”. The wide spread belief to provide extraterritorial space for sub- that Victorian cichlids are very aggressive, dominant individuals, this must lead to a is based on one of the many wrong gen- catastrophe. Therefore, to begin with rock eralizations. These pelagics, that are in cichlids, I would strongly advice an some ways similar to Cyprichromis from aquarium of not less than 250 litres vol- Lake Tanganyika, exhibit very little intra- ume. and interspecific aggression. Many (not The interior of the aquarium can be all) rock-dwelling cichlids, however, are very important to ensure peace among potentially quite aggressive. This holds the fishes. Mistakes in this regard are mainly for the “Vertical bar” and “Chess- one of the most common reason for loss board Mbipi” (Neochromis and Xystichro- of fishes. A few “hideouts”, say five or

46 ten rock caves or flower pots, usually help tion in relatively small aquaria. However, nothing to protect submissive individuals it does not function if group size is small against the attacks of a dominant tank because the aggression cannot be spread mate. As long as the latter can enter over many enough individuals, and abuse “hideouts”, it will be very consistent in of the weaker fish will be the conse- searching each, until it has found the sub- quence. Even if group size is large, the missive fish, that apparently does not behaviour of the fishes cannot be reliably “want” to leave the territory. The territory predicted, and careful observation is owner will soon be so experienced in pa- needed, in particular during the first days trolling each “hideout” that the submis- after setting up a group. sive fish will get no rest anywhere in the Nevertheless, also small groups of one aquarium. There are a number of possi- male and two or three females, even bilities to avoid this. In cases where the pairs, can be kept successfully if the aggressor is much bigger (deeper bodied) “dither fish method” is applied. Dither than the inferior individuals, enclosures fish are fishes of other species that are can be provided with small entrances introduced to help divert the attention of that allow only the smaller fishes to pass, the male, and to disperse its aggression. and make them real hideouts for them. Dither fish can be any fish that are tough Among rock cichlids this can often be enough to receive attacks of a helpful to protect females against male haplochromine male, and not strong attacks. In most rock cichlids, and particu- enough to chase him away from his terri- larly in the strongly territorial algae scrap- tory. I made good experience with small ers of the Neochromis and Xystichromis groups of subadult tilapias or females of lineages, males grow distinctly bigger other haplochromines. In the latter case than females. Such real hideouts can be only such species must be used that can- PVC tubes or broken flower pots that are not be confused later with the females pressed into the sand so that one or (bet- of the species to be bred. ter) several openings exist that allow only It should be stressed, that not all rock- the females to enter. Better of course are cichlids are equally aggressive. I am keep- more spacious hideouts if provided with ing and spawning a group of two full the respective small openings. grown males and three females of the If the females are not smaller than the paedophage H. (Lipochromis) melano- male, or if more than one male are to be pterus since more than a year in a 160 kept in one tank, the most promising so- litres aquarium without dither fish. The lution to overcome problems with ag- dominant male is that little aggressive gression, depends on the relationship be- that I never had to remove brooding fe- tween aquarium size and fish group size. males, hardly ever saw them being If the group consists of ten or more indi- chased away even from the close vicin- viduals, either large parts of the aquarium ity of the male, and never found their fins should be covered by complex structures or scales damaged, nor those of the sec- of rocks, creating tunnels and gaps of vari- ond male. Also H. (Lipochromis) ous sizes, and separating the aquarium “matumbi hunter” and H. (Harpagochro- into several compartments by visual bar- mis) “orange rock hunter” are very little riers, or the aquarium should have no hid- aggressive, and it seems that predatory ing places at all. If a larger number of cichlids are on average less aggressive fishes in an unstructured aquarium con- than algae scrapers and rock-dwelling in- tinuously see each other, the aggression sect eaters. of dominant individuals will be spread If aquarium size and cichlid group com- more evenly over the others, and males position allow to structure the aquarium, become in such situations not strongly rocks of different sizes should be piled up territorial. This is usually the safer solu- against the back side and at least one lat-

47 eral side of the aquarium. The bottom can dig much, such as H. (Paralabidochromis) be filled up with gravel. However, if one chilotes. One can simply fix them on, or wishes to observe male haplochromines among rocks, and after a few weeks they constructing their spawning pits, sand will get with their roots attached to the should be used instead of gravel. Never- rocks. In aquaria with smaller, and not theless, caution is needed that the rock much digging haplochromines, I made piles are very stable, and cannot be un- good experience also with Cryptocoryne dermined by digging. It is advisable to fix affinis. Some Victorian haplochromines at least the lower layer of rocks with sili- eat plants. Among those discussed in con on the aquarium ground. Spawning this book, it is mainly H. (Haplochromis) pits are usually dug out under overhang- “purple yellow” and maybe other mem- ing rocks! In their natural environment, bers of the Haplochromis lineage sensu territorial rock cichlid males have several stricto, Astatotilapia nubila and probably hideouts in their territory which they some species of the Neochromis and know very well, and into which they flee Xystichromis lineages. However, also in case of disturbance. These hideouts these do not eat all plants. They hardly generally are crevices among the rocks. ever touch Bolbitis and Microsorium, Males of Victorian rock cichlids have a while they can finish Vallisneria down to considerable demand for “security”, and the roots. are in the aquarium more shy than females, just like Konings (1995) observed Water and food among rock cichlids from Lake Malawi. To meet these demands, it is very important Concerning water chemistry, Lake Vic- to paint the background, and maybe one toria cichlids do not have particular de- or both sides of the aquarium with dark mands. The water in their natural environ- colour. If light falls into the aquarium from ment is usually rather soft, the pH neu- behind, caves among the rocks will not tral to alkaline. In the aquarium it is diffi- be dark, will not be considered “caves” cult to keep soft (bicarbonate poor) wa- by the fish, and will not contribute to their ter at a neutral to alkaline pH, because of feeling of security. Colourless fishes will its low buffer capacity. Most Victorian be the consequence. This should be cichlids thrive and breed perfectly in taken very serious because Lake Victoria somewhat harder water. In regions with haplochromines, more than those from soft and acidic tap water, it is advisable Lake Malawi, can lose all their colours to improve buffer capacity by putting a within minutes or hours. They become limestone into the aquarium and adding then totally unsightly. This is one reason sodium bicarbonate, as described by why Victorian cichlids rarely look attrac- Forsberg (1993). In recently set up tive in the tanks of traders and pet shops. aquaria that do not yet have a function- Plants rarely grow in the natural habitat ing flora of nitrifying bacteria, as well as of most rock-dwelling cichlids. Neverthe- in overcrowded, mechanically filtered rear- less, some hardy species are suitable to ing tanks that anyway hardly get such make a rock cichlid aquarium more attrac- flora, problems with nitrite accumulation tive. Vallisneria, Ceratophyllum and Najas are commonly encountered. Some Victo- are frequently found in sandy and muddy rian cichlids are more sensitive to it than littoral habitats of Lake Victoria. At least others. Open water dwellers and rock the first two are suitable for a rock cichlid cichlids are generally more sensitive than aquarium. I also have very good experi- inhabitants of muddy bays and sub- ence with Bolbitis and Microsorium ferns merged vegetation. Nitrite at higher con- and with Anubias. None of them grows centrations blocks the oxygen uptake of in Lake Victoria but they are particularly fish. Fish suffering from nitrite intoxica- suitable in an aquarium with cichlids that tion breath heavily without ascending to

48 A male Astatoreochromis alluaudi from an aquarium population. the surface as they would do in case of is very useful for aquarists. In aquaria, Vic- suffocation. Immediate change of 50% of torian cichlids are usually not choosy. the aquarium water usually helps. Addi- Coming from the wild, many species tionally it can be useful to stop feeding have initially some difficulties in accept- for a day or two to interrupt the supply of ing flake food. With white and black mos- nitrogen. We experienced very different quito larvae I always succeeded to get levels of sensitivity to organic water pol- them start eating. Later they usually ac- lution, when, at the Tanzanian fisheries cept also flakes. In composing the research institute in Mwanza, we failed to aquarium diet of Victorian cichlids one maintain cichlids in water from the should, to some extend, consider the Mwanza Gulf, that were caught in the rela- natural diet of the different species. The tively clean waters of the open lake. De- most commonly made mistake in feeding tailed information about filter systems for haplochromine cichlids is to offer minced cichlid aquaria is given by Fohrman meat or cattle heart. Cichlids should never (1993a) and Bailey (1993). be fed with mammal products because In a recommended article about keep- they cannot assimilate the fat of warm ing and breeding of Victorian cichlids, blooded animals. The consequence is Raymond (1995) rightly complained that liver adiposis which earlier or later leads no popular text existed that would pro- to death. A recommended text on cichlid vide information about what the indi- feeding has been written again by vidual species of Lake Victoria cichlids eat Fohrman (1993b). In the keeping and in the wild. Information about the diet of breeding system of the Institute of Evo- various rock-dwelling cichlids is provided lutionary and Ecological Sciences in in other chapters of this book. Though not Leiden, we have good experience with actually a popular text, a publication of keeping various species of Victorian Witte & van Oijen (1990) contains easily cichlids on minced prawns as the staple. understood information on the diet of Vic- It is important that the chitin shells are not torian cichlids from other habitats, which removed before mincing the prawns.

49 They are important as bulkage in the di- to the food of snail eaters. Meet of gestive tracts of the fishes. Mytilus and Edulis is equally well eaten, For herbivorous species, like Neo- though most fishes need to get accus- chromis, Xystichromis and Haplochromis tomed to its taste and rather strong fla- (sensu stricto), vegetative matter should vour. All snail eaters can be fed with form part of the diet. In a well illuminated minced prawns as a staple. and sparsely populated tank, green algae Fish eating predators and large insect growing on the rocks can do. In most situ- eaters have successfully been kept on a ations, however, additional vegetative mixed diet of prawns, bivalve meet and food, e.g. spinach or peas will be re- trout pellets (make sure they contain nei- quired. I made good experience with al- ther meet nor fat of mammals or birds). gae covered stones from local lakes that From time to time I feed piscivores with I exchange every three to four weeks. juvenile cyprinids (bream and rouche). If However, one should be aware of the no live fish is available, small pieces of possibility to introduce fish parasites in cod are also eaten. Copepods are a very this way. Alternatively one can, at least good food for all smaller haplochromines. during the summer months, keep some Feeding the anatomically specialized stones in a shallow water basin in the gar- pelagic zooplankton eaters with living den until they are overgrown by green al- copepods will allow the observer to ad- gae. This can take a while, and the mire the widely protractile sucking mouth Aufwuchs will never be as diverse as that of these tiny cichlids. Finally the usual ad- on stones taken from a lake, but the dan- vice to any fish keeper: Do not overfeed. ger of introducing fish parasites can be Do not give more than can be eaten avoided this way. Personally I prefer to within two minutes, and do this not more feed algae scraping cichlids with algae often than twice a day. covered stones, rather than with peas and spinach only, because it offers a possibil- Breeding ity to observe their natural feeding behaviour. Algae scrapers, more than others, Breeding of Lake Victoria cichlids is will not prosper on a diet that consists usually, but not always, easy. As far as exclusively of easily digestible, protein- known, all species are female mouth- rich food, such as mosquito larvae or brine brooders without pair bond. Neverthe- shrimps. They may soon suffer from in- less, a potential for male mouthbrooding testinal bloat. exists at least in some species. I observ- To observe their natural way of feeding, ed a complete sequence of mouth- it is most rewarding to offer living snails brooding over six weeks in a male H. to snail eating haplochromines. Most spe- (Harpagochromis) “orange rock hunter”. cies of Ptyochromis, as well as Macro- To breed rock cichlids, one should ideally pleurodus and Platytaeniodus, even if start with a group of one male and five well fed on other food, readily take to or more females. Regrettably this will snails as soon as they are thrown into the rarely be possible because haplochro- aquarium. Though I have not tried it, I mines are mostly still sold in pairs. Not think that a few individuals of any of these many people will be able, or even wish cichlids are perfectly suitable to control a to afford purchasing five pairs, in order to snail plague in an aquarium with other keep four of the males in a separate tank. fishes. However, they prefer planorbids, In such cases, I suggest to start out with and I never saw them eating Melanoides a pair in the following way: Do not put snails. The latter are eaten by some snail them straight pairwise into the spawning crushers, e.g. by Astatoreochromis tank. Instead keep them separate (the fe- alluaudi. When my stock of snails is ex- male in the spawning tank) until the fe- hausted, I use to add marine bivalve meet male’s abdomen is swollen of eggs, and

50 her genital papilla well visible in lateral time. If this method does not work, you view. Bring then the male, together with can try to keep your pair permanently with some dither fish, into the spawning tank a group of females of other haplochro- (if you do it the other way round, and mines. The older the fishes become, the bring the female to the male, the latter is more difficult will it become to spawn usually more aggressive because it con- them. Once breeding with a pair was suc- siders the female a territorial intruder as cessful, a larger breeding group should be soon as she is not ready to spawn). Ob- assembled from the offspring. From then serve the pair. If the female follows the on things usually go much easier. How- courting male on his lead swimming (see ever, if you can, try to spawn the pair sev- below), things may go well. However, eral times, and compose a breeding continue to keep an eye on them and be group from several nests, to keep up ge- ready to separate them again if you see netic diversity. If enough fishes are avail- that the female is not yet ready to spawn able Victorian cichlids should never be and is being chased aggressively. The kept in pairs or small groups for spawn- aquarium should contain some hideouts ing. The bigger the group, the better. for the female. Dense stands of Bolbitis Larger groups of most rock cichlids or Cryptocoryne sometimes do wonders spawn without any particular action on and can make it possible to leave pairs the part of the care taker. together even if the female is not yet ready to spawn. Never use male Table 1: Rock-dwelling Lake Victoria cichlids haplochromines of other of which aquarium populations exist species as dither fish. They may fertilize eggs of your Mbipi species other rock-dwelling species female even if they are H. (Neochromis) “black nigricans” Astatoreochromis alluaudi dominated by the male of H. (Neochromis) “blue scraper” Astatotilapia nubila H. (Neochromis) “cross dresser” H. (“Astatotilapia”) brownae ? the breeding pair (see be- H. (Neochromis) “giant scraper” H. (Haplo.) “blue obliquidens” low). In my aquaria even H. (Neochromis) nigricans H. (Haplochromis) lividus ? such different species as H. (Neochromis) “red anal nigricans” H. (Haplo.) obliquidens ? the pelagic zooplankton H. (Neochromis) “unicuspid scraper” H. (Haplo.) “red back scraper” eating H. “argens” and the H. (Neochromis) “velvet black” H. (Harpago.) “orange rock hunter” bottom dwelling snail H. (Xystichromis) “carp” H. (Harpagochromis) serranus sheller H. (Ptyochromis) H. (Xystichromis) “copper black” H. (Labrochromis) “stone” sauvagei hybridized with- H. (Xystichromis) “Jinja blue scraper” H. (Lipochromis) “matumbi hunter” H. (?) “big blue” (blue) (Nye) H. (Lipochromis) melanopterus out difficulty. After spawn- H. (?) “black & orange nyererei” (Nye) H. (Psammochromis) riponianus ing has happened, the H. (?) “crimson tide” (Nye) H. (Psammochromis) saxicola male and the dither fish H. (?) nyererei (Nye) H. (Ptyochromis) sauvagei should be removed. H. (?) “red head nyererei” (Nye) H. (Ptyochromis) xenognathus If the pair does not H. (?) “zebra nyererei” (Nye) H. (?) “double stripe” Uganda spawn within 24 hours, H. (?) “deepwater” (Dwa) you should remove the H. (?) “blue pseudonigricans” (Png) occasional intruders from other habitats male and try it again after H. (?) “pseudonigricans” (Png) other habitats H. (?) “Hippo Point blue bar” (?) H. (Macropleurodus) bicolor a couple of days. If then H. (Paralabidochromis) chilotes H. (Prognathochromis) perrieri they still do not spawn, H. (Paralabidochromis) plagiodon H. (Yssichromis ?) piceatus most likely the eggs of the H. (Paralabido.) “rockkribensis” H. (?) “argens” female are not ripe yet. You H. (Paralabido.) “blue rockpicker” H. (?) “citrus” ? have to separate the pair H. (?) “thick skin” again, keep particularly the female on optimal diet, (Nye) = H. nyererei complex, (Dwa) = H. “deepwater” complex, and try it again after some (Png) = H. “pseudonigricans” complex

51 52 Photos 1 to 9: A spawning sequence of Astatotilapia nubila in the aquarium. See text for explanation. Photo 10 to 12: A male "exchange" during spawning. While a pair A. nubila spawns (10) a male Thoracochromis sp. attacks the A. nubila male (11), chases it away and continues spawning with the A. nubila female (12).

53 The mating of Lake Victoria cichlids in parently stimulating her to spawn more gross follows the ritual described for other eggs. This sequence of acts is repeated haplochromine cichlids with fertilization until the last eggs have been spawned of eggs in the mouth of the female. How- and fertilized. Spawning is in nature fre- ever, differences in details exist between quently interrupted by disturbance. When different ecological groups. In all species the male has to chase away intruders, the the male attracts a ripe female by lateral female usually remains in the nest, wait- display. For this purpose he usually ap- ing for him. However, I have observed fe- proaches or overtakes the female and males leaving the nest in such moments, presents his beauty laterally in front of to continuing spawning with another her, stretching all his fins. Some cave male elsewhere! dwelling rock cichlids hardly leave the en- In some pelagic Victoria haplochro- trance to their cave while performing lat- mines, the ritual that leads to spawning, eral display, trying to attract females on can deviate slightly from the above de- larger distance. If the female does not es- scribed one. Lead swimming to the bot- cape, or if she approaches the male in tom rarely occurs. Instead the pair can as- case of “cave entrance courtship”, the sume T-position in the water column and male’s lateral display merges into quiver. sink down to the bottom in this position. A shaking runs across the male’s body, Once a female of the pelagic H. (Yssichro- while the spinous part of the dorsal fin is mis) pyrrhocephalus was seen spawning folded, the anal is stretched and the cau- some eggs in the water column (F. dal fin bent away from the female. Quiv- Liczkowski pers. comm.). However, only ering merges into lead swimming if the aquarium observations are available yet female has not moved away yet. While on courtship and spawning of pelagics, leading her to the nest, the male per- and even these are few. forms waving tail beats. On the nest the As stated above, spawning of Victorian male performs a loop and returns to the rock cichlids is most successful in a larger female to repeat the procedure. The fe- group. Since it will often not be possible male follows usually only stepwise. Once to remove all other fishes after a pair has the female reaches the nest, the male at- spawned in a large group, brooding fe- tains T-position, placing himself horizon- males at some point have to be removed tally in front of her head, and presents his from the community aquarium. This can anal with the egg dummies pressed be a most difficult task, particularly with against the bottom of the spawning pit. rock cichlids, spawned in a well structured The T-position can be followed by some rock aquarium. Brooding females do usu- rounds of circling, during which the part- ally not respond to bait and cannot be ners turn around each other with their caught in baited traps. In most cases vents on the bottom of the pit, and dur- there will be no way to avoid disassem- ing which no eggs are laid yet. bling the rocks to get the female. How- If up to here everything went well, the ever, before this is done, it is sometimes female at some point drops a few (1-5 but useful to try to catch her by surprise. A usually 2 or 3) eggs and immediately brooding female that moves around with turns to pick them up. While she picks the other fishes and does not hide, can them up, the male resumes T-position, sometimes be caught by one fast and presenting his egg dummies just next to well targeted movement with the hand the real eggs. After having collected her net, e.g. during feeding of her tank real eggs, the female picks at these dum- mates. If this goes wrong and the rock mies, the male releases milt, and the piles have to be disassembled, it proved eggs are fertilized in the female’s mouth. very helpful to manipulate the brooding The male then turns, and touches with his female into one corner of the aquarium, mouth the anal region of the female, ap- and block that part of the aquarium with

54 a PVC board. The remainder of the aqua- the fry can sometimes lead to a panic re- rium can then stay undisturbed. lease before the female is caught. An important question is when to re- As nursery, a 50-litres aquarium is suf- move a brooding female from the com- ficient, furnished with a heater, a sponge munity aquarium. I have negative experi- filter and a hideout for the mother. A hide- ence with doing this shortly after spawn- out in a nursery should not be too small ing. Females frequently jettison (rather and should be easy to look into. In nature than swallow) their eggs in the net or af- brooding females of rock cichlids defend ter being released in the new environ- small territories around a rock hole. Never ment of the brooding tank. Females that put more than one female into one nurs- have already been brooding for some ery tank. They can become very aggres- days, are less likely to jettison their eggs sive, particularly after the first has re- or fry. Females of most rock cichlids generally hold their clutch better than females of, Table 2: Approximate duration of parental care of some Lake for instance, pelagic zooplank- Victoria haplochromines. tivores. To ensure a good tim- Determined at temperatures between 24 and 26°C in aquaria. ing of the removal of brooding females, knowledge about the days until first additional days incubation time is needed. Re- release of fry until last mou- markably little, not to say noth- (mean) thing * ing, has been published about Vertical bar Mbipi this aspect of reproduction bi- H. (Neochromis) nigricans 21 up to 21 ology in Victorian haplochro- H. (Neochromis) “unicuspid scraper” 20 up to 20 mines. My data are not many H. (Xystichromis) “copper black” 21 up to 21 but they are sufficient to show H. (Xystichromis) “carp” 21 up to 17 that incubation period (defined H. (Xystichr.) “Uganda blue scraper” 21** ? here as time between spawn- H. (?) nyererei 21 7 ing and the first release of the H. (?) “zebra nyererei” 21 7 fry) and post buccal care dura- Chessboard Mbipi tion (defined here as time be- H. (Paralabidochromis) chilotes 15 up to 20 tween first release and last H. (Paralabidochr.) “rockkribensis” 18 2 guarding of the fry) are worth H. (Paralabidochr.) “blue rockpicker” 15 up to 15 looking at more carefully. Though incubation period de- rock-dwelling fish predators pends also on water tempera- H. (Lipochromis) melanopterus 20 up to 38 ture, both incubation period H. (Lipochromis) “Matumbi hunter” ? 20 and post-buccal care duration H. (Harpagochr.) “orange rock hunter”21 21 are, at equal water tempera- others ture, very different among dif- H. (Ptyochromis) sauvagei 21 7 ferent species. Some exam- H. (Haplochromis) “blue obliquidens” 22 up to 6 ples are given in table 2. My Astatotilapia nubila 18 up to 28 experience is that the best time to shift a brooding female pelagics into a nursery tank, lies be- H. (Yssichromis) pyrrhocephalus 13.5 1-2 tween the beginning of the H. (“Astatotilapia”) piceatus 15 1-2 H. (?) “argens” 22 1-2 last third of the incubation period and three days before the expected date of release. At- * maximum values are given; up to n, indicates large tempts to shift a female too variability shortly before the release of ** from Selbrink 1985a

55 leased her fry. Females of a few species breed them in community tanks with (e.g. H. (Harpagochromis) “orange rock males of more than one haplochromine hunter”) continue to eat during mouth- species. Females are usually choosy dur- brooding. They suck the food into the ing courtship, and select their own males, mouth, which they open just a tiny gap if these are available (if not, they spawn wide, while the eggs or fry are kept far with any male). However, once a female behind in the buccal cavity. However, in is in the nest and has started spawning, most species brooding females are appar- she can loose her choosiness and con- ently unable to suck in food while brood- tinue spawning even if a male of another ing ( I have not tried with plankton yet) species has chased away her male, and and should during this time not be of- has taken over his position (see photo se- fered food to prevent water pollution. ries). This can happen, if the owner of a Feeding must be resumed as soon as the territory was just slightly dominant over fry is released for the first time. another male, and if this instable domi- The size of the fry at the time of first nance relationship tilts during spawning release is not uniform among species, but when the other male starts to fight for the the fry of most species are large enough nest with the female in it. However, even to eat brine shrimp nauplii, microworms a clearly dominant male can often not and crushed flake food. Those of pelagic prevent others from fertilizing some eggs zooplanktivores are particularly big, in of “his” female by sneaking into his nest spite of the short incubation times of during spawning. Sneaking is most suc- those species. The fry of vertical bar cessful if sneakers are much smaller than Mbipi is usually also rather big, those of the territorial male. chessboard Mbipi tends to be smaller. The smallest fry I have seen, were those of the snail sheller H. (Platytaeniodus) degeni, which were unable to handle crushed flake food and starved when no brine shrimp nauplii were fed. After re- fraining from brood care, a female should be removed from the nursery and get the opportunity to rest and recondition before she is brought back into a tank with a male. Clutch sizes among Victorian haplochromines vary in nature between as few as 13 eggs in some pelagic zooplanktivores and 170 in Astatoreochromis alluaudi (Goldschmidt & Goudswaard 1989). Juve- niles grow slowly and reach sexual ma- turity at an age of between six months and a year. Already somewhat earlier one can often tell females and males apart. However, in most species some individuals develop their secondary sexual characters very late. The maximum life ex- pectancy of Victorian haplochromines under good aquarium care is about three years. If you want to reproduce your fish and maintain the different species, do not

56 Astatotilapia nubila

APPROACH

LATERAL DISPLAY

QUIVER NEST LOOP FAST LEAD

FERTILIZATION CIRCLE WITH PRESENTATION OF SPAWN EGGDUMMIES

Haplochromis pyrrhocephalus

APPROACH LATERAL DISPLAY

QUIVER

T-POSITION

HEAD TO HEAD POSITION (CIRCLE)

FERTILIZATION

SPAWN

Figure 6: Elements of courtship behaviour and spawning in two Lake Victoria haplochromines: the benthically living Astatotilapia nubila (above) and the pelagically living H. (Yssichromis) pyrrhocephalus (below). Drawings made after photographs. Most rock-cichlids court and spawn similar to A. nubila.

57 Identification of rock-dwelling haplochromines

That identification of haplochromine task. However, it was also for other rea- cichlids from Lake Victoria can be quite sons not desirable at this point. Some complicated is becoming apparent if one “new” species of rock-dwelling cichlids considers that more than 70 % of the are currently being discovered on every species, featured in European and Ameri- expedition into an area, that was not sam- can aquarium literature, were shown un- pled previously. This would render a spe- der wrong names. This rate lies certainly cies level key an only locally useful tool. much above the average for cichlids, and Instead, I provide here merely some help is probably partly due to the lack of popu- for the identification of the major species lar identification literature, but partly to complexes and (sub)genera that make up the misunderstanding of the importance the cichlid communities of rocky shores of coloration in species identification. Col- in Lake Victoria. Most of the “new” spe- oration is very important for the identifi- cies that are to be expected within the cation of Victorian cichlids, but only in coming years, will probably belong to combination with other morphological these species complexes. Thus, the key characters. Attempts to identify species may retain some relevance, though the only by coloration are frequently bound discovery of more species will certainly to fail. Very similar colour patterns occur quantitatively, and maybe also qualita- repeatedly in Lake Victoria among species tively, change the group characters and of different taxonomical and ecological their ranges (see the chapter on tax- groups. A typical example is black body onomy). with red anal and caudal fin and dorsal fin The key, addressing cichlid keepers and rim, found in Astatotilapia, “Astatotilapia”, breeders as well as field workers, uses Xystichromis, Neochromis, Haplochromis only characters that can be measured, and the “Deepwater” complex. Most counted, or observed on a living, anaes- fishes with such coloration are errone- thetized fish. However, a binocular micro- ously “identified” as A. nubila in popular scope will be required, and at one place literature. For any reliable identification, in the key an otoscope, to look into the some anatomical characters have to be throat of a fish. This key differs from the investigated. When this has resulted in simple dichotomous one that taxono- the identification of the ecological and/or mists usually wish to have. Like other au- taxonomical group a species belongs to, thors before (Greenwood 1973, 1980, coloration may be of great value to com- Snoeks 1994), I found it impossible to plete the task to identify it to species compile such a key. Differences among level. haplochromine species and species com- To investigate some taxonomical char- plexes can often not be pinned down to acters is not as difficult as some cichlid a single character. Rather it is in many hobbyists fear, and quite something can cases the combination of characters that be done on the living, anaesthetized fish sets one group apart from another one. without causing it much harm. I hope Each of these characters exhibits consid- that the identification key presented be- erable individual and intragroup variation, low, will aid in this. The key does not go so that often more than one character down to species level. Compiling a spe- have to be consulted to separate two cies level key to haplochromine cichlids groups. of Lake Victoria, is a basically infeasible Ranges of morphometric measure-

58 a. dorsal fin ments used in this key, lateral line upper part are ranges of mean gill caudal cover pectoral lower part fin values of the species fin anal fin that make up the cheek pelvic fin keyed out species complexes. Ranges of individual variation are b. larger, are strongly in- BD fluenced by extreme HL SL individuals that exist in any larger sample of Victorian haplochromines, and depend therefore much more than means on inciden- c. d. e. tal sample composi- EyL tion. The consequence LJW IOW is that this key is not suitable to identify a LJL single individual, though it may in cases key out correctly. f. Rather it is meant for the identification of samples of at least prognathous isognathous isognathous retrognathous three individuals, so obtuse acute that some idea about variation can be Figure 7. Morphological characters useful for identification of rock- achieved, and means dwelling haplochromines (parts d, e, f after Barel et al. 1977, altered). for morphometric val- a) Outline of a typical haplochromine cichlid with body parts relevant ues can be calculated. to species identification. Nevertheless, because b) Standard length (SL), body depth (BD), and head length (HL). of the extreme indi- c) Eye length (EyL) and lower jaw length (LJL) vidual variability of d) Ventral view of the anterior head to show lower jaw width (LJW) many Lake Victoria e) Dorsal view of the anterior head to show interorbital width (IOW) haplochromines, even f) Lateral snout outline. if this is kept in mind, the key

Figure 8. Tooth shape and tooth implantation (after Barel et al. 1977, unicuspid, unicuspid, unicuspid, weakly unequally subequally unicuspid, tricuspid stout, and slender, and blunt bicuspid bicuspid bicuspid obliquely altered). acutely acutely truncated a) The most common pointed pointed (scalpel- shaped) shapes teeth in the outer tooth row. b) Procumbent versus outer teeth implanted in outer teeth implanted an upright position procumbently upright im-plantation of the teeth in the outer tooth row.

59 cies. Thus the key is neither suitable to identify subadults, nor adult females. Many haplochromine species undergo considerable changes in body shape during their life. Including subadult individuals in a key would mean to increase the intragroup variation that much, that a key would become entirely unworkable. Many haplochromine species, and rock cichlids in particular, exhibit strong sexual dimorphism. Fe- males usually are smaller, often differ in head shape and dentition. Thus, considering females and males in one key would have an effect, similar to that of Chest squamation: generalized type (centre: H. parvidens-like) and including subadults. It rock-cichlid type (right: H. cryptodon; left: H. melanopterus). is possible to compile a separate key for fe- should not be used as a means of assign- males but I have not yet measured ing a sample of individuals to a taxonomic enough to do so. To begin with, I con- group, without reference to text and ta- sider a key for males most useful be- bles 3 and 4. Rather, it is meant as a guide cause male coloration will in many cases to tentative identification of the taxo- help to confirm identifications and allow nomic belonging of a species, which to identify a sample to species level af- should then be checked against the diag- ter it has keyed out to one species com- noses of species complexes, given in the plex. text. Once the identity of a species has Morphometric measurements were been traced down to the species com- taken in agreement with the definitions plex, the species diagnoses in text and given by Barel et al. 1977, and are shown tables, in combination with the photo- in text figure 8. The same holds for the graphs hopefully allows identification to dental morphology (text figure 9). Abbre- species level. viations used in text, tables and identifi- Furthermore it is important to note the cation key, are given in the text figures. following. The key is based on qualitative Readers with a deeper interest in haplo- and quantitative measurements taken chromine alpha-taxonomy are strongly from about 800 male individuals of 125 advised to read the work of Barel et al. species. Data from other publications (e.g. (1977). With a few exceptions, only adults Greenwood) have not been considered. in a size range between 70 and 100 % of With a few exceptions, all measured known maximum size of a species, and fishes were within a size range of 70-100 in most species only males, were meas- % of the known maximum size of a spe- ured for the data given in table 3. In the

60 exceptional cases, males of the respec- unicuspid or weakly bicuspid; inner teeth tive size had not been available. For meas- small and arranged in broad bands uring I used a digital calliper with glued- anteriorly in both jaws or only in the upper on divider needles, but normal divider and jaw (one exception); IOW 24.0-27.8 % of ruler do as well. Measurements are taken HL (rarely only 23 %) ...... Ptyochromis directly as the distance comprised between the points of the dividers placed DHP not decurved and/or dentition on the measuring points (Greenwood different and IOW < 24 % of HL ...... 7 1956a, Barel et al. 1977). The number of 7. DHP shallow and straight; inner teeth tall, vertical bars is generally counted be- slender, implanted so as to lie almost tween the pectoral fin, and the caudal horizontally, and deeply embedded in oral end of the dorsal fin, and includes the bar mucosa; head narrow, with an IOW from that often starts dorsally ahead of the pec- 21.1-23.8 % of HL (one case of 27.1 %); toral fin and, running slightly obliquely, lower jaw usually narrow and attenuated passes across the pectoral fin insertion...... Psammochromis Key to (sub)genera and species inner teeth not long and slender, not lying complexes of rock-dwelling almost horizontally, and not deeply haplochromines embedded in oral mucosa ...... 8 8. HL less than, or equal to 33 % of SL (when 1. anal fin with 4 or more spines 33 %, than EL > 24 % of HL) ...... 9 ...... Astatoreochromis alluaudi HL more than, or equal to 33 % of SL anal fin with 3 spines ...... 2 (when 33 %, than EL < 24 % of HL) ...... 10 2. 28-30 (maximum 32) lateral line pore 9. LJ shorter than, or equal to 34 % of HL scales ...... Astatotilapia nubila (one exception); jaws isognathous or LJ 33 or more lateral line pore scales (32 in shorter; mouth in most species lying rare individuals) ...... 3 horizontally; LJ and dental arcade narrow 3. Body depth < 27.5 % of SL; LJL <43 % of V- or narrow based U-shaped; DHP steep; HL; lower jaw length/width ratio > 2.5; chest scales small and deeply embedded ... distinct mid lateral and dorsal lateral .... Paralabidochromis “rockpicker” complex stripe present ...... LJ longer than, or equal to 34 % of HL; ...... H (?) “double stripe” complex jaws isognathous or lower jaw slightly Body depth > 27.5 %, or otherwise longer; mouth lying obliquely; LJ and different ...... 4 dental arcade not V-shaped, not narrow; 4. LJL > 45 % of HL, if between 43 and 45 %, DHP shallow to steep; chest scales neither then LJL/W ratio > 1.7 and IOW < 22.1 % small, nor deeply embedded (“short snout of HL ...... Harpagochromis scraper” group) or small and deeply LJL > 43 % of HL, if between 43 and 44 %, embedded and females bright yellow with then LJL/W ratio < 1.7 or IOW > 22.1 % of black chessboard pattern (“rockkribensis”) .. HL...... 5 Paralabidochromis “rockkribensis“ complex 5. outer teeth in the LJ implanted 10. Lips strongly swollen, often produced into procumbently, tooth band anteriorly in LJ lobes; prominent chessboard melanin describing a ventral inflection ...... 6 pattern ...... outer teeth in the LJ not procumbently ...... Paralabidochromis chilotes complex implanted, tooth band not describing a Lips not swollen to swollen but not ventral inflection ...... 11 produced into lobes; melanin pattern not 6. DHP decurved, outer teeth moderately to prominent ...... strongly (rarely only slightly) recurved, and .Paralabidochromis chromogynos complex

61 11. Large, protrusile mouth; either: outer teeth not moveably implanted, weakly bicuspid weakly bicuspid, short, in LJ inclined or unicuspid and crown recurved; inter- rostrad in some, deeply embedded in oral spaces between the teeth wider than the mucosa and often visible only after lips width of the tooth neck; DHP straight or are pushed up, inner teeth so deeply incurved; EL < 27 % of HL ...... 18 embedded in oral mucosa that they are outer teeth, also in adult males, moveably hardly visible; or: outer teeth longer, inner implanted; interspaces between the teeth and outer teeth not very deeply embedded, narrower than the width of the tooth neck; but outer teeth in LJ inclined rostrad (1 tooth shape, DHP, and EL variable ...... 19 species) ...... Lipochromis outer teeth strong, even in small fish not teeth neither deeply embedded in oral moveably implanted, unequally bicuspid mucosa, nor inclined rostrad ...... 12 and recurved; interspaces between the 12. Lower pharyngeal bone enlarged and teeth narrower than the tooth neck ...... stout, with enlarged, molariform ...... “Astatotilapia” II (barbarae, pharyngeal teeth; oral teeth stout and “black long snout”) unicuspid or weakly bicuspid 18. IOW usually above 24 % of HL, if between ...... Labrochromis 23 and 24 %, then both sexes with a Lower pharyngeal jaw not enlarged and prominent melanin pattern of 5-8 vertical not with molariform teeth ...... 13 bars on the flanks between the pectoral 13. Outer teeth bicuspid with a compressed fin and the caudal end of the dorsal fin; in major cusp that is laterally protracted and females generally 5-8 vertical bars ...... has a broad flange, or unicuspid and ...... H. (?) nyererei complex scalpel-shaped, strongly protracted, IOW below 24 % of HL; melanin pattern appearing obliquely truncated, moveably not prominent, consisting of 5-6 vertical implanted...... Haplochromis bars on the flanks between the pectoral teeth different ...... 14 fin and the caudal end of the dorsal fin, if 14. DHP decurved; LJ short and broad, length/ visible ...... H. (?) “deepwater” complex width ratio < 1.2; HL < 32.3 % of SL; teeth 19. EL > 25 % of HL; IOW < 27 % of HL; HL equally, subequally, and rarely unequally <34 % of SL; mostly < 3 rows of inner bicuspid, very closely set (usually teeth ...... H. (?) “pseudonigricans” complex contiguous); no or just a narrow gap EL > 25 % of HL; IOW < 27 % of HL; HL between inner and outer teeth in the >34 % of SL ...... “Astatotilapia” III lower jaw ...... Neochromis I (“incurved dorsal head profile”) DHP decurved or straight; length/width EL < 25 % of HL, if > 25%, then IOW >27% ratio of LJ > 1.2, if near 1.2, then teeth not of HL and HL > 34% of SL, mostly > 3 rows contiguously set; inner teeth separated by of inner teeth, arranged in broad bands.. 20 a distinct gap from the outer teeth ...... 15 20. outer teeth even in adult males unequally 15. BD < 32 % of SL ...... Neochromis II bicuspid ...... Xystichromis I (“unicuspid scraper”, “Bihiru scraper”) outer teeth in adult males unicuspid, BD > 32 % of SL ...... 16 coarse, and only slightly moveable; 16. HL < 32 % of SL, mid lateral stripe present upright, crowns not recurved ...... that can be restricted to the area above ...... Xystichromis II (“carp” complex) the anal fin ...... “Astatotilapia” I (brownae) HL > 32 % of SL, or no mid lateral stripe ...... 17 17. outer teeth at least in adult males coarse,

62 Introduction to the species discussions

On the following pages all cichlid species haplochromines do not constitute formal are discussed, that are currently known species descriptions. Undescribed spe- from rocky shores in south-eastern Lake cies are given descriptive names, pre- Victoria. Nevertheless, it is likely that sented in quotation marks. These names more, as yet unknown, rare or geographi- have no taxonomic validity and do not con- cally restricted species live even within form to the rules of zoological nomencla- the part of the lake that we have sur- ture. Formal descriptions of a number of veyed. Additionally some species are dis- the here discussed species are under cussed that are known from rocky places way. Descriptive names, when possible, in other parts of the lake. were designed so as to provide two types I have sorted the species into major of information about the “new” species: groups and discuss them group by group. most apparent morphological characters These groups will be referred to as (sub)- (e.g. coloration) and information either genera (see the chapter on taxonomy) if about ecology or taxonomical position. In they were described as such by Green- this way names like “yellow anal scraper” wood (1980), or species complexes, if and “red head nyererei” denote a spe- they are here described for the first time. cies that is an epilithic scraper with a yel- The species account within each major low anal fin, and a H. nyererei-like species group is preceded by a characterization of with a red head respectively. However, in the group. Ranges of morphometric val- cases where other descriptive names had ues that I give in these group characteri- already been used in previous publica- zations are generally ranges among the tions, I keep to those, to avoid confusion. mean values obtained from the different Each species account includes informa- species that comprise a group. The result- tion about coloration or other morphologi- ing implications for identification pur- cal characters that may help in identifica- poses are discussed in the chapter on tion. In this respect most species ac- identification. To characterize groups that counts are complemented by morphologi- were defined by Greenwood (1980), I re- cal information in two tables at the end fer to his original definitions, and in cases, of the book. In these tables mean values add information that I found to be of di- are given for a number of morphological agnostic value among the species I am characters that I found most helpful for discussing. To keep Greenwood's and my identification purposes. In table 3 results definitions apart, information cited or con- of my measurements of most of the new densed from his descriptions is printed and some described species are given. In in italics. For the purpose of this book, I table 4 the respective data are given, use from Greenwood's group diagnoses taken from previous publications, for a usually only characters that can be meas- number of described species that occur ured, counted, and observed on living at rocky shores, or are relevant to the fish. identification of rock-dwelling species. Many of the species discussed on the More detailed information on these and following pages are undescribed and other measurements will be given in here discussed and introduced to a forthcoming publications. broader public for the first time. It should Each species account is complemented therefore be stressed that the data given by a map, showing the geographical dis- in this book for various undescribed tribution of the species as currently

63 known within south-eastern Lake Victoria. These maps are presented at the end of the book. Furthermore, each species account includes information on the ecology of the species, such as habitat demands, food, and behaviour if particu- lars are known. Morphologi- cally and ecologically similar and/or closely related species, living sympatrically with the species under consideration, are mentioned in each species account. Further information is given optionally where it was considered relevant. The widely distributed matrix species (see the chapter on evolution) of the Mbipi species complexes are treated somewhat more extensively. Additional to the distribution maps more detailed maps are given on pages 17 and 65, showing Lake Vic- toria and the area surveyed for this book, with the most important places mentioned in the species accounts.

Angling for cichlids in rock holes.

Photographing Haplochromis in a cuvette at Zue Island.

64 Gana Island

Ukara Island

Ukerewe Island

Kunene Islands Bwiru Island Nansio Kiregi Is. Bay Namatembi Is. Nafuba Island Miandere Islands Mabibi Is.

Ikuru Island Ruti Is. Sozihe Vesi Island Archipelago Kome Island Bihiru Makobe Is. Ndurwa Point Island Mafwinki Is. Igombe Is. Senga Point Juma Mponze Point Island Chamagati Is. Bwiru Point Gabalema Is. Hippo Is. Capri Point Wakahangara Anchor Is. Kissenda Bay Point Nyegezi Rocks Ngoma Point Nyegezi Bay Kilimo Is. Nyamatala Is. Python Islands Mashoro Bay Shadi Rocks Luanso Bay Muranda Bay Luanso Is.

Marumbi Island

Figure10: Lake Victoria and the area surveyed for this book Buyago Rocks with localities mentioned frequently in the text. 0 5 10km

65 Vertical bar Mbipi

The haplochromine species joined in this dentition, but has a longer head, straight group share a melanin pattern on the flanks dorsal head profile and less broad jaws that consists of four to eight dark vertical (Xystichromis). Two lineages tend to have bars, that can in a few species occasionally shallow dorsal head profiles, longer jaws, be crossed by a faint mid lateral stripe. They and get coarse unicuspid outer teeth when are either narrow, numerous, regular in out- adult (H. nyererei complex, H. "deepwater" line, and densely spaced, or broad, some- complex). The last lineage consists of ana- times less regular in outline, and less many. tomically quite unspecialized fishes with All species have dark or bright male colora- large eyes (H. "pseudonigricans" complex). tion. A melanin pattern like that of the vertical Though adult individuals of the different bar Mbipi occurs also in a few species that species complexes within the vertical bar do not share other characters with this Mbipi can have very different tooth shapes, group, and do not live predominantly over all have in common a rounded, not mark- rocks, e.g. species of the Haplochromis lin- edly narrowed dental arcade, and a more eage, and H.(?) "thick skin". or less upright implantation of the outer teeth. Also those teeth anteriorly in the Blue, black, and orange — the lower jaw are upright, rather than pro- Neochromis complex of algae scrapers cumbent (figure 3 on page 30). The inner teeth are arranged in 2-7 (rarely 9) rows. In By the time of Greenwood's last revision subadult individuals, the teeth in the outer of the Lake Victoria haplochromines (Green- row are usually moveably implanted. This wood 1980), only one species of epilithic condition prevails in lineages with bicuspid algae scrapers was known from this lake, teeth in adults, whereas teeth become Haplochromis nigricans. Two others were firmly attached in adults of lineages with known from Lake Edward, H. serridens and coarse unicuspid teeth. All vertical bar H. fuscus. Greenwood placed them to- Mbipi that we found in southern Lake Vic- gether into the resurrected genus Neo- toria have small to very small, deeply to very chromis Regan, 1920. He defined this ge- deeply embedded chest scales. Some spe- nus as small haplochromines, with a very cies from Ugandan waters, however, have strongly decurved dorsal head profile (slop- large, and not deeply embedded chest ing at 70°-80° to the horizontal), a long, scales (see page 87ff). None of the known much coiled intestine (ca 3-4 times stand- species has thickened lips. ard length of the fish), and broad bands of Within the vertical bar Mbipi, five major inner teeth anteriorly and anterolaterally in lineages can be defined, that have differ- both jaws, not separated from the outer row. ent gross morphology. They overlap largely The outer teeth are equally or subequally in the measurements of individual morpho- bicuspid, their minor cusp is only a little metric characters, but fall apart on combi- smaller than the major cusp, their crown is nations of different characters. The first of compressed and broader than the body. these lineages is characterized by a real They are very closely set (contiguous), rock scraper anatomy, reminiscent of moveably implanted, tall, slender but ro- Pseudotropheus in Lake Malawi, with a bust, showing only a slight antero-posterior short head, usually strongly decurved dor- decline in their height and size. The dentary sal head profile and broad jaws (Neo- (tooth carrying lower jaw bone) is markedly chromis). A second one has similar scraper foreshortened, deep and stout, its anterior

66 margin strongly curved medially so that the wuchs, the biocover of rocks, and in particu- anterior outline of the lower jaw is almost lar the algae component (Bouton et al. sub- rectangular. The lower jaw is with 30-38% mitted). To obtain this food, a fish presses of the head length short and broad, with a its slightly opened mouth against the rock length/width ratio of 1.0 to 1.4. and closes it, scraping with the teeth over Greenwood suspected that there might the rock surface. Once filamentous Auf- be a second species of epilithic scrapers in wuchs is held between the densely set Lake Victoria and this was confirmed in the teeth, the fish performs a jerking move- late seventies, when two more species ment to tear it off. This so called "pull- were discovered in the Mwanza Gulf area scraping" is the main feeding technique of (van Oijen et al. 1981). Several further spe- Neochromis species (Seehausen et al. in cies were discovered only much later: in press [a]). However, other feeding tech- 1989 by myself in Uganda and in 1990 by niques are applied and unicellular algae and N. Bouton in Tanzania. During our survey of various animal components of the Auf- rocky shores in the southeastern (Tanzanian) wuchs are eaten as well. Even plankton is part of the lake, we discovered between frequently consumed (Bouton et al. submit- 1991 and 1996 about 15 more species. ted). Though their anatomical range exceeds in H. (Neochromis) nigricans (Boulenger some characters (mostly in tooth shape and 1906) is the only described Neochromis dorsal head profile inclination) the limits, set species in Lake Victoria. According to litera- by Greenwood's definition of Neochromis, ture it is widely distributed in the lake and I am rather confident that these species, or has been recorded from several places most of them, constitute a monophyletic along the north coast, from an island in the lineage to which I refer in the following as central region as well as from Mwanza in the Neochromis complex or Neochromis lin- the south (Greenwood 1956a, van Oijen et eage. al. 1981). However, the taxonomic status of I characterize the Neochromis complex by the different populations has to be investi- the combination of the following charac- gated. My data suggest a disjunct distribu- ters: Subequally to unequally bicuspid outer tion. In large parts of the area that we in- teeth (rarely weakly bicuspid outer teeth); vestigated, H. (N.) nigricans seems to be 3 (rarely 2) to 7 (rarely 9) rows of inner teeth, absent (see map). The measurements that arranged in broad bands anteriorly in the I took from southern populations deviate jaw, except in two species separated from slightly from Greenwood's data (Table 3 ver- the outer teeth by only a narrow gap or not sus table 4), however, this may partly be separated at all; a lower jaw that tends to due to personal differences in measuring. be square shaped and is on species aver- The species is characterized by a rather age (with the exceptions of one morph of deep body, steep dorsal head profile, a very one species) not longer than 38% and not broad lower jaw, many rows of inner teeth, shorter than 33% of the head length and the outermost of which are hardly shorter 0.9 to 1.2 (rarely 1.3) times as broad as long; than the outer teeth and by a male colora- a relatively short head (29.3-32.5% of SL); tion that is rather uniform among popula- and 5 to 8 dark vertical bars on the flanks. tions: light blue with five to seven, in width Most species have a rather steep, decurved and outline regular, darker grey to black ver- dorsal head profile, and all species from the tical bars between the pectoral fins and the southeastern lake have small, deeply em- posterior end of the dorsal fin, a pale grey bedded chest scales. to reddish grey anal fin with very small egg The species of the Neochromis complex dummies and a caudal fin that is usually en- are ecological equivalents to the species of tirely crimson. At some islands yellowish the genera Pseudotropheus and Tropheus and very dark males occur sympatric with of Lakes Malawi and Tanganyika respec- the normal coloured fishes. Females are tively. Their main food is epilithic Auf- usually yellowish brown with five to seven

67 Haplochromis (Neochromis) nigricans; male from Saa Nane Island.

H. nigricans; a territorial male in its natural H. nigricans; a male from an aquarium environment. population.

A female H. nigricans. A male H. nigricans from Igombe Island.

68 dark vertical bars but in some populations exhibit a warm deep yellow that is particularly bright on the chin. The chest scales are small and deeply embedded. H. (N.) nigricans inhabits the shallow waters of most rocky areas within its geographical ranges but can be absent from large steeply sloping rocks and islands H. "black nigricans" from Chamagati. dominated by such rocks. Its highest densities are found between 1 and 2m water depth, depending much on the water transparency and thus the abundance of epilithic algae. In these depths it can be the

H. "black nigricans", a freshly caught specimen.

A female H. "black nigricans"

Chamagati Island

A female H. "yellow-blue nigricans"

69 An OB female H. "black nigricans" dominant algae scraper. Males defend ter- morph, very similar to that of H. (N.) "velvet ritories of usually more than a square me- black" and H. (N.) "blue scraper", and with a tre surface on horizontal or gently sloping piebald morph resembling that of H. (N.) bottom. I frequently saw them chasing "blue scraper" from Makobe and Gana Is- away also males of other Neochromis from lands. The latter, however, is extraordinar- their territories. Detailed stomach content ily rare. Along the mainland shores be- analyses have shown that, though epilithic tween the islands west of Juma and the algae clearly dominate the food, its compo- Mwanza Gulf, black Neochromis popula- sition can vary considerably between sea- tions occur that are morphologically inter- sons and geographical locations (Bouton et mediate between H. (N.) "black nigricans" al. submitted). In periods of high abun- and H. (N.) "velvet black". I assume these dance of other food organisms, e.g. plank- species to be phylogenetically very closely ton, insect larvae or prawns, H. (N.) nigricans related geographical vicariants. This not- can switch to these in spite of its special- withstanding, H. (N.) "black nigricans" is mor- ized algae scraper morphology. In clear wa- phologically resembling H. (N.) nigricans ter areas filamentous green algae are eaten much closer than H. (N.) "velvet black" (ta- much more frequently than unicellular al- ble 3). Ecologically it resembles H. (N.) gae or filamentous blue-greens. However, nigricans and H. (N.) "blue scraper", living at the latter can form the major part of the diet 0.5 to at least 4 m water depth, and pre- in areas with murky waters like the central dominantly over horizontal to gently slop- Mwanza Gulf. H. (N.) nigricans is absent ing surfaces. Of three individuals from from areas with very murky waters, e.g. the Chamagati Island studied, two fed pre- southern Mwanza Gulf. dominantly on algae (filamentous green al- H. (Neochromis) "black nigricans" is a sec- gae and diatoms), one on caddis fly larvae, ond species of "Nigricans-shape", and fits zooplankton and diatoms. The two that ate Greenwood's definition of Neochromis. It predominantly algae contained consider- has a decurved dorsal head profile like H. able amounts of mineral grains in the stom- (N.) nigricans and a similarly broad lower ach, implying that they scrape the rocks jaw. Its chest scales are small and deeply with a powerful bite. We observed forag- embedded. In contrast to the aforemen- ing fishes frequently moving in small tioned species, H. (N.) "black nigricans" has shoals. entirely black males, that may have a nar- H. (Neochromis) "velvet black"black", discov- row red edge on the caudal fin. In the lat- ered by E. Witte-Maas & F. Witte in 1979, ter case they closely resemble H. (N.) "vel- is at many places among the most abun- vet black" in coloration, and so do the fe- dant Mbipi, particularly in the northern males which are dark brownish. This spe- Mwanza Gulf. It is widely distributed and cies is known from very gently sloping is- morphologically extremely variable. I con- lands with small and very small rock boul- sider all populations as belonging to one ders in the Sengerema region west of species because changes are gradual along Juma Island and from the steeper shores geographical transects. H. (N.) "velvet black", with bigger rock boulders of Juma Island or the "velvet black"/"black nigricans"/"blue itself. At Mafwinki Island it occurs sympat- scraper" superspecies, is known from al- rically with a population of H. (N.) nigricans, most each sampled rocky station and rep- at Juma Island with H. (N.) "Juma scraper", resents the matrix superspecies of the and at Chamagati Island it is the sole repre- Neochromis lineage. A taxonomical descrip- sentative of the Neochromis algae scrapers tion is under way, using populations from but, like at the other places as well, coex- the northern Mwanza Gulf (Witte-Maas in ists with the algae scraper H. (Xystichromis) prep.) which I therefore treat as the typical "copper black" that occurs at most sampled "velvet black". These populations are char- locations outside the Mwanza Gulf. H. (N.) acterized by a dorsal head profile that is less "black nigricans" is polymorphic, with an OB- strongly decurved than in H. (N.) nigricans,

70 by subequally bicuspid to unequally and lake and western Speke Gulf, H. (N.) "vel- weakly bicuspid teeth in the outer rows vet black" is in that region only known from (weakly bicuspid to unicuspid in large indi- the mainland shore. Populations from the viduals), by maximum five and at least two shores of the central Speke Gulf resemble inner rows, by male breeding coloration of those from the northern Mwanza Gulf. uniform black with a narrow red caudal fin Populations south of the Nyegezi area in edge, by usually not more than five vertical the Mwanza Gulf lack the OB morph, gradu- stripes on the flanks between the pectoral ally change from the algae scraper morphol- fins and the posterior end of the dorsal fin, ogy towards a more generalized morphol- and by the occurrence of an OB morph, very ogy (with less decurved dorsal head profile, similar to those of H. (N.) "black nigricans" longer lower jaw, smaller minor cusp on the and H. (N.) "blue scraper". Compared to H. outer teeth and fewer inner tooth rows), (N.) nigricans, H. (N.) "velvet black" has a and the deep black, typical of northern higher length/width ratio of the lower jaw, males, makes way for a coloration varying the inner tooth rows are (in most popu- between grey-black, grey-blue and yellow- lations) separated from the outer row by a green. In the southern Mwanza Gulf green- small gap and tricuspid outer teeth are a ish males with red chest are common. common feature laterally in the jaws but Along the same cline, the extension of red can even be found anterolaterally. The chest colour in caudal and anal fins is increasing. scales are small and deeply embedded. These southern populations are intermedi- The distribution of such characterized ate in dentition and coloration between the populations extends from Hippo and Gaba- typical H. (N.) "velvet black" and the typical lema Islands in the North, via the Mwanza H. (Xystichromis) "copper black" which is ab- town area, Capri Point, Saa Nane, Ascari and sent from the Mwanza Gulf south of Python Anchor Islands, to the Butimba and Nyegezi Island. The yellow-green morph resembles areas in the South. We consider this spe- the southern most known population of H. cies to be much wider distributed, the (X.) "copper black" very closely in coloration. populations, however, becoming increas- I do as yet not understand these patterns ingly different with increasing distance to but it appears a possibility to me that H. (N.) the northern Mwanza Gulf. Along a transect "velvet black" and H. (X.) "copper black", from the northern Mwanza Gulf via the Gulf while in the northern Mwanza Gulf and entrance to the exposed open mainland Speke Gulf representing two coexisting shores of the southwestern Speke Gulf, species, merge into a single polymorphic head shape and dentition gradually ap- species in the southern Mwanza Gulf (see proach those of the more specialized algae also the species H. (?) nyererei and H. (?) scrapers H. (N.) nigricans and H. (N.) "blue "zebra nyererei"). For a discussion of the ap- scraper". The population at Igombe Island parent contradiction arising from the fact finally has even more and more narrowly that the two species are assigned to two spaced tooth rows (5-7 inner rows) than the different subgenera, the reader is referred sympatric population of H. (N.) nigricans (5- to pages 94. This is the only Neochromis 6 inner rows). All these populations share algae scraper in the Mwanza Gulf south of the above described coloration with the Shadi area (H. (N.) nigricans has its south- "Velvet black" from the northern Mwanza ern most record at Shadi rocks). Gulf. They are usually found sympatric with Wherever H. (N.) "velvet black" coexists H. (N.) nigricans, H. (X.) "copper black", and with H. (N.) nigricans, it has its maximum frequently with H. (N.) "giant scraper" and population density in slightly greater water H. (N.) "unicuspid scraper". In contrast they depth than has H. (N.) nigricans and domi- have only exceptionally been found sym- nates over the latter around larger and more patric with H. (N.) "blue scraper" (see distri- steeply sloping rock boulders. Ecologically bution maps). While the latter is known al- it is thus similar to H. (N.) "blue scraper" that most exclusively from islands in the open replaces H. (N.) "velvet black" at the offshore

71 A male Haplochromis "velvet black".

A female H. "velvet black". The rocks at Hippo Island.

An OB male of Haplochromis "velvet black".

72 A territorial male H. "velvet black" from islands in the open lake and western Speke Gulf. The diet of H. (N.) "velvet black" is very variable not only between populations but also between seasons. In the northern Mwanza Gulf this species is rather omnivorous, preying upon epilithic algae (predominantly diatoms but also filamentous algae), various epilithic fauna (Bryozoa, insect larvae, prawns), as well as on plankton and juvenile fish (Bouton et al. sub-

Below: an OB female H. "blue scraper". An OB female of H. "velvet black".

73 mitted). At some islands moss animals Also female coloration differs between (Bryozoa) are the dominant food item of H. populations. In some they are light yellow- "velvet black". The species lives in the north- ish with a bright yellow throat, in others they ern Mwanza Gulf more cryptic (= hidden) are more brownish with a grey throat. In- than H. (N.) nigricans, males and brooding terestingly we found rare brown female in- females defending rock holes and crevices dividuals also in populations with yellow fe- as territories. However, in the southern males (e.g. Makobe Island), indicating that Speke Gulf (Igombe Island) we observed some of the interpopulation variation may adult males and females foraging in shoals, have arisen via polymorphism. mixed with other species, at 2 to 3m water All populations share an OB-morph very depth. similar to those of H. (N.) "black nigricans" H. (Neochromis) "blue scraper" is the and H. (N.) "velvet black", the frequency of third of three, over a large area complemen- which is much higher at some islands than tally distributed species, that represent one at others. Furthermore most, if not all popu- superspecies (H. (N.) "velvet black", H. (N.) lations have yellow males. Beyond that two "black nigricans", H. (N.) "blue scraper"). It is other morphs are known from the Makobe known from all sampled offshore islands in Island population: piebald (shared with H. the open lake and western Speke Gulf and (N.) "black nigricans) and brassy and black, from two less offshore localities (Ndurwa making it the most polymorphic of all Point, Nansio/Ukerewe Island) that are in- known Lake Victoria cichlid populations. The habited by two, in coloration somewhat ab- black morph, though not entirely black, is errant populations. The offshore "Blue rather similar to H. (N.) "velvet black" and H. scraper"-islands are separated from each (N.) "black nigricans" males, with a narrow other and from the mainland by 20 to 40m red edge on the caudal fin. The piebald deep water and populations differ fre- morph is also known from the population quently in male coloration and body shape. at Gana Island north of Ukerewe. Though At Makobe Island H. (N.) "blue scraper" is hard data about the factors driving the evo- shallow bodied, the majority of the males lution of the different male nuptial colour are bright sky blue with rather widely patterns in these isolated populations are spaced, weak vertical bars and red colour lacking, circumstantial evidence suggests, is completely absent from their fins. At Ruti that the coloration of other species that live Island H. (N.) "blue scraper" is very variable sympatric with H. (N.) "blue scraper" may in male coloration. The majority of the play a role. At Makobe Island, where H. (N.) males is also sky blue but somewhat darker "blue scraper" males lack dark vertical bars than at Makobe Island, have darker vertical and red in their fins, the species coexists bars on the flanks, vivid red in the pelvic with H. (N.) nigricans whose males have dark and anal fins and there are individuals with vertical bars and a bright red caudal fin. bright red caudal fins as well as individuals Where the latter species is absent (Ruti, with completely black pelvics. Some indi- Mabibi, Bwiru, Miandere Islands), H. (N.) viduals at Ruti Island are rather deep bod- "blue scraper" males have dark vertical bars ied. Males from the Mabibi Islands resem- and red in their fins. At Gana Island again ble the most common form of Ruti Island H. (N.) "blue scraper" lacks red in the fins, but the red in their pelvics is less vivid. and it coexists with two other blue algae Males of the deep bodied Bwiru Island popu- scrapers with extensive red in caudal and lation are dark in colours, with deep metal- anal fins (H. (N.) "red tail giant scraper" and lic blue flanks, and dark wine red anal and H. (Xystichromis) "red anal blue"). pelvic fins, those of Miandere Island are All populations of H. (N.) "blue scraper" quite similar in coloration, but not so deep have a specialized algae scraper dentition, bodied, and at Gana Island they lack any very similar to that of H. (N.) nigricans, with red, like at Makobe Island but have darker densely set outer teeth, many inner tooth and more narrowly spaced vertical bars. rows and no gap between outer and inner

74 tooth rows. H. (N.) "blue scraper" lives at body shape I consider H. (Neochromis) depths from 1 to 6m (Seehausen & Bou- "Vesi scraper" distinct from "blue scraper" ton 1996a), and inhabits steeply sloping until more is known. In the Vesi archipelago large boulder as much as gently sloping it coexists in sympatry and parapatry with small boulder shores. At the latter it over- four other Neochromis but is usually more laps largely with H. (N.) nigricans in depth abundant than the four other species to- but has its population density maximum in gether. Alongside with H. (N.) "giant slightly greater depth than that species. scraper", H. (N.) "short head nigricans", H. (N.) Thus its spatial distribution is very similar "eastern blue scraper", and H. (Xystichromis) to that of H. (N.) "velvet black". H. (N.) "blue "copper black" it inhabits gently, and mod- scraper" feeds primarily on epilithic filamen- erately steeply sloping shores, and is ab- tous green and blue-green algae where it sent from very steep islets that are inhab- inhabits gentle slopes and on blue-greens ited by H. (N.) "yellow anal scraper". Its depth and diatoms at steep slopes. Where it co- distribution ranges from the shallow water exists with H. (N.) nigricans it eats relatively to at least 5 m. less green algae and more animal food than H. (Neochromis) "Juma scraper" is an- the latter but it consumes some insects other poorly studied algae scraper that is everywhere (Bouton et al. submitted). Males similar, both, to H. (N.) nigricans and to H. at Makobe Island defend territories (N.) "blue scraper" (Seehausen 1994). It dif- throughout the depth range down to 6m. fers from the first by a lack of extensive red While in the shallow water territorial males coloration in the caudal fin, and by a longer frequently interact with those of H. (N.) lower jaw, from the latter by the absence nigricans and H. (X.) "copper black", they in- of OB-morphs, and by a deeper body, and teract with H. nyererei in deeper waters. from both by a reddish anal fin. Males are Females and nonterritorial males are less either light blue or yellowish with 6 to 7 gregarious than those of H. (N.) nigricans. darker vertical bars on the flanks. Fully A population that is similar to H. (N.) "blue grown they become dark metallic blue-grey scraper" and H. (N.) nigricans in anatomy but and exhibit a deep body with nuchal bump, does not agree with either of the two in col- while their dorsal head profile is not oration, lives at the Vesi archipelago in the rounded as in most H. (N.) nigricans but re- Speke Gulf. The metallic deep blue flank mains rather straight, giving the fish a char- coloration which is typical for H. (N.) "blue acteristic appearance. The red in the cau- scraper", is replaced by a blue-grey, the ven- dal fin is confined to a narrow rim. Females tral body outline is more convex than in that are brownish with yellow flush and a yel- species, and the caudal fin is often red like low throat. The species is known only from in H. (N.) nigricans. However, the vertical Juma Island, a big island with about 12 km bars are fewer (5-6) and less regular in out- of shoreline, in southern Lake Victoria. The line, the lower jaw is longer than in H. (N.) possibility that H. (N.) "Juma scraper" repre- nigricans and in contrast to the latter an OB sents a geographical form of H. (N.) "blue morph exists, similar to those of H. (N.) "blue scraper" cannot yet be ruled out. However, scraper" and H. (N.) "velvet black". Neither considering the polychromatic character of of these three more widely distributed spe- H. (N.) "blue scraper" (OB morphs), H. (N.) cies has been found in the Vesi archipelago. "Juma scraper" may well be specifically dis- Since each other sampled station in the tinct from it. It lives sympatrically with two Speke Gulf is inhabited by a population of other epilithic algae scrapers: H. (N.) "black either H. (N.) "blue scraper" or H. (N.) "velvet nigricans" and H. (X.) "copper black". From black", it is likely that the fishes from the both it is distinguished by male coloration Vesi archipelago belong to this lineage and (compare pictures) and by other morphol- should then probably be regarded a popu- ogy (table 3). It lives at depths down to 6m lation of H. (N.) "blue scraper". However, beat least and at shores composed of large cause of its peculiarities in coloration and rock boulders as well as at places with me-

75 Haplochromis "blue scraper", piebald male.

An OB male H. "blue scraper". A yellow male H. "blue scraper".

A black male H. "blue scraper"; below, a peppered female. A blue male H. "blue scraper"; below a female.

76 The north-eastern part of the Vesi Archipelago in the Speke Gulf

A male H. "blue scraper" at Makobe Island. A yellow male H. "blue scraper" at Makobe Island.

A male H. "Juma scraper". A male H. "Vesi scraper"

A female H. "Vesi scraper". An OB female H. "Vesi scraper"

77 dium sized boulders. More is not known the water and their ca. 7 vertical bars are about the ecology of this form. usually only faintly visible. Lips, snout and H. (Neochromis) "yellow anal scraper" is preorbital area are light blue. Females are a species of "blue scraper"-like appearance, usually brassy-yellowish. with an OB morph, whose specifically dis- We currently know H. (N.) "eastern blue tinct status is apparent from its coloration scraper" from three islands in the east of and its coexistence with H. (N.) "blue our survey area. It lives at gently sloping to scraper", H. (N.) "Vesi scraper", H. (N.) "east- moderately steep rock shores, with small ern blue scraper" and H. (N.) "velvet black" to medium sized boulders, and at water at different localities. In contrast to these depths between 1 and 6 m. At Sozihe Is- four species H. (N.) "yellow anal scraper" has land it is at relatively gentle slopes, at 1 to a yellow to orange-red anal fin in the male 3 m depth the most common haplo- sex, and a bright red caudal fin similar to chromine, numerically dominating over the that of H. (N.) nigricans. Male flank colora- sympatric algae scrapers H. (N.) "velvet tion is usually blue-green with many dark black" and H. (Xystichromis) "copper black". vertical bars, making a beautiful contrast to It is at Sozihe Island also the only Neo- the bright yellow to orange anal fin. From chromis species, that coexists with H. (N.) H. (N.) nigricans and H. (N.) "blue scraper" it "yellow anal scraper" in depths beyond 4 m. differs by a narrower tooth arcade, from H. At the very gently sloping small rock boul- (N.) nigricans, furthermore, in general ap- der and stone rubble shore of Zue Island pearance, anal fin coloration and a longer "eastern blue scraper" is numerically co- lower jaw. The species has a central/north- dominant with the algae scrapers H. (X.) ern Speke Gulf distribution but within its "copper black" and H. (X.) "short scraper". It range is largely absent from gently sloping seems to differ from the two others in shores. While very rare and rare at Ruti and depth distribution, being dominant at Mabibi Islands respectively, where it coex- depths between 2 and 4 m, while the oth- ists with H. (N.) "blue scraper", it is the only ers are more abundant at 0.4 to 2 m depth. abundant Neochromis at steep islets in the At Vesi Island "eastern blue scraper" is rare Vesi archipelago and is absent from more and coexists with the abundant H. (N.) "Vesi gentle slopes which are occupied by four scraper" that seems to occupy a similar other Neochromis species. At Sozihe Is- niche. H. (N.) "eastern blue scraper" is likely lands, where H. (N.) "yellow anal scraper" co- to be more widely distributed west of our exists with H. (N.) "eastern blue scraper", H. survey area. It lives sympatric with many (N.) "velvet black", H. (N.) "giant scraper" and other Neochromis species: H. (N.) "Vesi H. (X.) "copper black", it is restricted to wa- scraper", H. (N.) "yellow anal scraper", H. (N.) ter depths beyond 4 m, and was found "velvet black", H. (N.) "giant scraper" and H. down to at least 7 m. H. (N.) "velvet black" (N.) "short head nigricans". seems restricted to immediately inshore A species known only from the Vesi ar- shallow waters at moderately steep slopes, chipelago in the central Speke Gulf is H. and H. (N.) "eastern blue scraper" to shallow (Neochromis) "short head nigricans"nigricans". Males waters at gently sloping places. are light blue with 6-7 darker vertical bars, H. (Neochromis) "eastern blue scraper" a blue dorsal fin with red streaks and spots is another species of "blue scraper"-like ap- between the rays, a caudal fin that is dis- pearance that, however, lacks an OB tally red and proximally red streaked and a morph. From H. (N.) "blue scraper" it differs pale red anal fin with orange coloured egg furthermore by a steeper, dorsal head pro- dummies. Of all the epilithic algae scrapers file (the head appears more round than in studied by us, this one has on average the any other Neochromis), a somewhat shorter most extreme expression of equally bicus- head, and by male nuptial coloration. Males pid teeth in the outer row. Thus from a are on the flanks dark blue, frequently ap- morphological point of view it is one of the pearing purplish blue when taken out of most specialized epilithic algae scrapers.

78 "Short head nigricans" bears resemblance to represent a rare colour morph of H. (N.) H. (N.) nigricans and H. (N.) "giant scraper". "long black". From H. (N.) nigricans it is rather easily told H. (Neochromis) "Bihiru scraper" presents apart on the ground of coloration and a a taxonomical and ecological puzzle. Known shorter head, relative to the standard only from Bihiru Island at the western limit length. From sympatric H. (N.) "giant of the area surveyed by us, it may be more scraper" it differs most conspicuously by a widely distributed west of that area. H. (N.) steeper dorsal head profile, shorter lower "Bihiru scraper" seems to be a trophically jaw and more vivid red in the fins, particu- polymorphic species with an extraordinar- larly in the caudal fin. "Short head nigricans" ily wide range of body shapes. One morph seems not very abundant at Vesi Island. We has a more typical algae scraper morphol- found it only at one place, a moderately ogy with a shorter lower jaw (33.0-41% of steeply sloping shore with medium sized head length), more inner tooth rows (table rock boulders where it lives at about 4-5 m 3) and subequally bicuspid outer teeth that depth sympatric with H. (N.) "Vesi scraper", become unequally bicuspid only in fishes H. (N.) "giant scraper", H. (N.) "eastern blue of more than 100 mm standard length, and scraper" and H. (X.) "copper black". are still weakly bicuspid in full adults of 135 H. (Neochromis) "long black" is one of mm standard length. Particularly smaller in- several very slender and elongated Neo- dividuals of this morph have a relatively chromis algae scrapers. Due to the elon- steep, de curved dorsal head profile. In- gated body shape, the dorsal head profile dividuals of the second morph have a of these species is not as steep as in most rather straight dorsal head profile, irrespec- other Neochromis, though it is decurved as tive of their size. Their lower jaw is longer well. In coloration H. (N.) "long black" resem- (38.3-43.7% of head length), they have bles H. (N.) "velvet black", males being en- fewer inner tooth rows (table 3) and their tirely black with red dorsal fin lappets. Dif- outer teeth are weakly bicuspid already in ferent from H. (N.) "velvet black", however, fishes of 80 mm standard length, becom- the distal two thirds of the caudal fin are ing unicuspid in individuals above 90mm red and the anal is pinkish. We know this standard length. There are two reasons species from several localities around why I consider these two forms to belong Ukerewe Island. All have relatively steeply to a single species. (1) No consistent dif- sloping shores with medium sized to big ference in male coloration exists between rock boulders and most of them are surf ex- them. (2) The range of morphological varia- posed high energy shores. H. (N.) "long tion exhibited by females is less wide and black" lives sympatric with the following no two groups could be made. other Neochromis: H. (N.) "velvet black", H. Based on the same criteria, we consider (N.) "blue scraper", H. (N.) "yellow anal H. (N.) "Bihiru scraper" specifically distinct scraper", H. (N.) "giant scraper", H. (N.) "red from the sympatric population of H. (N.) "vel- tail giant scraper", H. (N.) "black tail giant vet black": Males of the first have usually a scraper" and with H. (Xystichromis) "copper light red anal fin and a caudal fin that is red black". It inhabits immediately inshore wa- on at least the distal two thirds while those ters of less than 2m and slightly more off- of the latter have a dark grey anal fin and a shore waters down to at least 7 m depth. black caudal fin with narrow red edge. Mor- We found it particularly abundant at Bwiru phological differences between the two Island. Here some spatial niche segregation species are among females as distinct as between "long black" and "blue scraper" can among males: H. (N.) "velvet black" (of the be observed with "blue scraper" numerically Bihiru population) have a steeper, more dominating in shallow waters, "long black" strongly decurved dorsal head profile, in depths beyond 5 m. Nothing more is clearly more densely implanted teeth and known yet about its ecology. At Bwiru Is- no gap between outer and inner tooth rows land we collected one red male that may (both morphs of "Bihiru scraper" exhibit

79 A male H. "short head nigricans".

A male H. "yellow anal scraper".

An OB female H. "yellow anal scraper". A male H. "eastern blue scraper".

A female H. "eastern blue scraper".

A male H. "long black" from Bwiru Island.

A red coloured male of Haplochromis "long A female H. "long black". black" from Bwiru Island.

80 The eastern shore of Bwiru Island, the natural habitat of H. "long black". gaps). Furthermore does "velvet black" have the OB morph which is characteristic for many of its populations. How- ever, more research, particularly on the ecology and on reproduction will have to show whether my current view is correct. A male H. "Bihiru scraper", a bicuspid morph. H. (N.) "Bihiru scraper" lives sympatric with two other Neochromis species, apart from "velvet black": H. (N.) "giant scraper" and H. (N.) "Labeo scraper", and also with H. (Xystichromis) "copper black". Big males of the bicuspid morph can be confused with "giant scraper" but differ by having the anal light red instead of dark grey with pink flush and by having fewer inner tooth rows (2 to 5 versus 4 to 8 in "giant scraper"). Big males of the unicuspid morph can be confused with A male H. "Bihiru scraper", a unicuspid morph. unicuspid individuals of "copper black" (see below) which, however, have the red in the caudal fin restricted to a narrow edge. Males of H. (N.) "Bihiru scraper" undergo considerable changes in general appearance during ontogeny. Individuals of all sizes and shapes between 80 and 135mm standard length are found in breeding dress. Those of the bicuspid morph when below 100 mm standard length resemble H. (N.) "long black" but differ in coloration and A female H. "Bihiru scraper".

81 have a higher lower jaw length/width ratio lineage). However, of the three individuals and fewer tooth rows, suggesting a lower whose stomachs we analysed, none had level of specialization for algae scraping. eaten snails although snails were abundant The shore of Bihiru Island slopes steeply in the environment (Igombe Island). Instead to very steeply and has big to huge rock two ate predominantly filamentous blue- boulders. Within the available range, how- green algae and smaller amounts of fila- ever, H. (N.) "Bihiru scraper" inhabits all mentous green algae, and one had an microhabitats and depth ranges from less empty stomach. All available evidence sug- than 2 to at least 9 m depth. It is every- gests that H. (N.) "unicuspid scraper" is an where the most abundant algae scraper, epilithic Aufwuchs eater. and in part of its depth range, beyond 4 m, In spite of its peculiar dentition H. (N.) "uni- it is the only one. cuspid scraper" is in general appearance It is not yet known whether an elongated and coloration very similar to H. (N.) "velvet Neochromis with a peculiar mouth, living black" and H. (N.) "blue scraper". Information sympatric with H. (N.) "Bihiru scraper" in shal- about reproduction behaviour is needed low water at Bihiru Island, represents an before its specific status can finally be con- extreme expression of the scraper morph firmed. This is particularly so because this of the latter or a separate species. This species, at all its known localities, coexists form has been named H. (Neochromis) in sympatry with either H. (N.) "velvet black" "Labeo scraper" because of a strongly (Igombe Island, Bwiru Point) or H. (N.) "blue retrognathous lower jaw, resembling the scraper" (Ndurwa Point, Makobe Island). At mouth of the cyprinids of the genus Labeo Igombe Island it is polymorphic for male and the Lake Malawi rock cichlids of the nuptial coloration with black males, resem- genus Labeotropheus. Its body is elongated bling H. (N.) "velvet black" and light blue like in "Bihiru scrapers" of similar size but males resembling H. (N.) "blue scraper". At the only obtained male is not black with a the other places it is in coloration more red anal fin but rather greenish-brown with grey-blue. Like "velvet black" and "blue a yellowish anal fin. It furthermore has big- scraper", it has also an OB morph. I think ger eyes than "Bihiru scrapers" of respec- that H. (N.) "unicuspid scraper", H. (N.) "vel- tive size. More individuals need to be col- vet black", and H. (N.) "blue scraper" form a lected before it will be possible to say complex of sibling species or that "unicus- whether it is specifically distinct from H. (N.) pid scraper" represents a trophic morph of "Bihiru scraper". one of them or even of both. The situation H. (N.) "unicuspid scraper" bears its name is very similar to that of the trophic morphs due to its dentition which, among the of H. (N.) "Bihiru scraper". However, the known epilithic algae scrapers, it shares forms in the "velvet black"/"blue scraper"/"uni- only with the unicuspid morph of H. (N.) cuspid scraper"-complex are further differ- "Bihiru scraper". Small fishes still have typi- entiated since the differences in head mor- cal algae scraper dentition with subequally phology and dentition are equally distinct bicuspid outer teeth but the outer teeth in males and females, and are accompanied soon become weakly bicuspid and are in by some differences in microdistribution full adults usually unicuspid, widely spaced and male nuptial coloration. and moderately to strongly recurved. The H. (N.) "unicuspid scraper" lives at gentle inner teeth are arranged in 3 or 4 rows and to steep slopes and at places with medium are frequently also unicuspid and recurved. sized to big rock boulders. Our data suggest They are always separated from the outer that it differs in microdistribution from H. teeth by a gap. Such dentition type was pre- (N.) "velvet black": At Igombe Island the lat- viously only known from snail eating ter species inhabits depths from 0.5 to cichlids that, with their teeth, pull the snail about 4 m, while "unicuspid scraper" lives body out from the shell (Witte & van Oijen between 2 and 6 m depth and can be ob- 1990; see the chapter on the Ptyochromis served at 6 m at the sand-rock interface.

82 Nevertheless, at intermediate depths we dorsal head profile. The species is rather observed adult nonterritorial males of both variable in respect of the number of inner forms in the same shoal. Other epilithic al- tooth rows and has, compared to H. (N.) gae scraper species that live sympatric with nigricans and H. (N.) "blue scraper" more "unicuspid scraper" are H. (N.) nigricans, H. unequally than subequally bicuspid outer (N.) "giant scraper" and H. (X.) "copper black". teeth. H. (Neochromis) "giant scraper" is the H. (N.) "giant scraper" lives sympatrically biggest member of the Neochromis com- with most other Neochromis algae scrap- plex (up to 150 mm standard length). It is ers: with H. (N.) nigricans and H. (N.) "velvet rather widely distributed in the Speke Gulf, black" at many places, furthermore with H. currently being known from the southern (N.) "blue scraper", H. (N.) "eastern blue mainland shore between the entrance to scraper", H. (N.) "yellow anal scraper", H. (N.) the Mwanza Gulf and Senga Point, from "unicuspid scraper", H. (N.) "short head most offshore islands in the western and nigricans", H. (N.) "Vesi scraper", H. (N.) "long central Speke Gulf and possibly from Uke- black", H. (N.) "large eye nigricans", H. (N.) rewe Island (one individual collected). Popu- "pseudoblack" and with the Xystichromis al- lations differ considerably in coloration with gae scrapers H. (X.) "copper black" and H. blueblack and black males at the mainland (X.) "short scraper". The specific status of H. shores of Bwiru and Ndurwa and the in- (N.) "giant scraper" is therefore certain. Nev- shore Igombe and Bihiru Islands, bright sky ertheless, we have found individuals, inter- blue males at the offshore islands Makobe mediate between this species and H. (N.) and Zue and more yellowish to grey-blue "velvet black", at one place at the southern males at the other localities. The coloration Speke Gulf mainland shore. of the population at Mabibi Islands is not From the Mwanza Gulf a Neochromis yet known. The bright blue males of form was reported by Witte et al. (1992) un- Makobe and Zue Islands resemble the most der the name H. "kruising", that is said to common morph of H. (N.) "blue scraper" at have been intermediate between H. (N.) many islands (but not Makobe and Zue!) nigricans and H. (N.) "velvet black" ("kruising" with red of variable extension and intensity = Dutch word for hybrid) but was charac- in caudal and anal fins. The anal fin is pale terized by a retrognathous lower jaw like "gi- to rather intensively red in blue males and ant scraper". It went almost extinct after the is usually dark grey with a wine red to pink- Nile perch upsurge and, in spite of exten- ish flush in dark males. The egg dummies sive search, only three individuals have are small and usually arranged in a single been seen at three different places over the line. The caudal fin is proximally blue-grey past five years: 1990 (Python Islands, N. with a faint red flush, distally reddish but Bouton pers. comm.), 1990 (Anchor Island, never as intensely red as in H. (N.) nigricans N. Bouton pers. comm.), 1995 (Nyegezi or H. (N.) "yellow anal scraper". Females rocks). It is possible that H. (Neochromis) have a brassy sheen and yellowish fins that "kruising" is a geographical variant of H. (N.) are particularly bright in the Zue Island "giant scraper" that seems to remain smaller population. than the Speke Gulf populations of the lat- H. (N.) "giant scraper" has a characteristic ter. head shape. The dorsal head profile is steep H. (N.) "giant scraper" is at gently sloping but rather straight, and the lower jaw shores more abundant than at steeply slop- slightly retrognathous and narrower than ing shores. The population from the Vesi ar- that of H. (N.) nigricans and H. (N.) "blue chipelago for instance inhabits only the scraper". The mouth has a slight beaklike moderately steep slopes of the main island appearance. Even subadult individuals are and is absent from some of the steeply readily recognized by their head shape. sloping rock islets off the main island, the Intraspecific variation between populations habitat of H. (N.) "yellow anal scraper". In the exists in the steepness and curvature of the rather well studied communities of Makobe

83 A blue male H. "unicuspid scraper". The only specimen known of H. "Labeo scraper" comes from Bihiru Island.

and Zue Islands, H. (N.) "giant scraper" is, among the Neochromis species, the one that has its maximum population density in the shallowest water. Adult males defend territories of several square metres, over rather horizontal bottom, at about 1 m water depth. Brooding females guard their fry in often not more than 0.5 m depth. However, the species is found down to 4 m depth. Nonbrooding fishes A black male H. "unicuspid scraper". often move in shoals. H. (N.) "giant scraper" seems, among the Neochromis species studied, the most strict vegetar- ian. Of three individuals from Makobe Is- land, that we checked, one had eaten predominantly filamentous green algae, another one filamentous blue-green algae, and a third one higher plant material and diatoms. All three had eaten smaller amounts also of the other algae types while animalous food was hardly present in their stomachs (sometimes some chironomid larvae). It is almost im- A female H. "unicuspid scraper". possible to catch "giant scrapers" with worm-baited hooks while most other Neochromis species readily take such baits. In the clear waters of Makobe Is- land we used to observe big males of this species paying no interest to, or even avoiding worm baited hooks, while other Neochromis (with similarly "specialized" algae scraper dentition) were quar- relling over them. Only at Gana Island, the northern most of our sampling stations, we collected two other large Neochromis that differ An OB female H. "unicuspid scraper". from H. (N.) "giant scraper" in coloration

84 The north-western corner of Gana Island. and mouth shape. It is likely that both, H. (N.) "red tail giant scraper" and H. (N.) "black tail giant scraper" have their major distribution outside of our sampling area. Males of H. (Neochromis) "red tail giant scraper" are blue with about six vertical bars, a caudal fin that is bright red in the distal half, a dark grey dorsal fin with red rim and red dots A male H. "black giant scraper".

A male H. "giant scraper" caught in the net.

A male H. "giant scraper" from Makobe Island. Filling up at Ruti Island.

85 between the rays and a silvery whitish anal species is H. (Neochromis) "large eye fin with pink flush. They differ from those nigricans" from Makobe Island. In spite of of the sympatric H. (N.) "blue scraper" popu- extensive collecting and ecological relation in coloration and body depth as well search at that island this species was as by having a shorter and relatively broader caught only occasionally (3 individuals in lower jaw. From H. (N.) "black tail giant 1992, 4 in 1993, none after that) and only scraper" they differ in coloration and jaw in the female sex. This is the more pecu- shape (see below), from H. (N.) "giant liar, as in most other Neochromis species scraper", that we did not find at Gana Island, males are caught much more frequently by shorter but isognathous jaws, and male than females. H. (N.) "large eye nigricans" coloration. Closer resemblance exists be- resembles on the first glance H. (N.) tween this species and H. (N.) "short head nigricans but has a much narrower snout nigricans", currently thought to be endemic and lower jaw, a different dentition (the to the Vesi archipelago in the Speke Gulf. outer teeth are much longer and bigger Too little is known about both populations, than the inner teeth) and exceptionally big to interpret these resemblances. eyes (table 3). Either the species is ex- H. (Neochromis) "black tail giant traordinarily rare or we have not yet de- scraper" is blueblack with about six verti- tected its major habitat. Three individuals cal bars. All fins are black, the dorsal fin in 1993 have been caught at only 1 to 1.5 with some metallic blue in the hard part m depth, in a rock garden within a reef at and red lappets, the caudal fins with a the SW coast of Makobe Island, and one narrow red rim. Lips, snout and preorbital at less than 1 m depth at the edge of the area are light blue. "Black tail giant reef. At the same spot H. (N.) nigricans, scraper" is deeper bodied than the H. (N.) "blue scraper" and H. (X.) "copper sympatric "blue scraper" population. A pe- black" are living. Of the three individuals culiar feature of this species is the shape caught in 1992 is only known that they of its lower jaw. It is V-shaped and at the were at less than 3 m water depth, to- tip slightly curved downwards, together gether with H. (N.) "giant scraper" and the with the slightly thickened upper lip cre- three above mentioned species. Recently ating a beak like appearance of the some very similar individuals (again fe- mouth. males) were collected at Nansio Island/ The three morphologically similar Neo- Ukerewe. chromis species that live sympatric at H. (Neochromis ?) "pseudoblack" Gana Island, have distinctly different male seems to stand rather isolated within the anal fin coloration. "Red tail giant scraper" Neochromis complex if it really belongs has, on a silvery whitish-pinkish fin, deep here. It has a rather narrow head of a dis- yellow egg dummies that are surrounded tinct shape and big eyes, resembling the by a clear white inner ring and a transpar- species of the "Pseudonigricans" complex ent outer ring. "Blue scraper" has, on a but with the strongly decurved and grey anal with pale white edge, pale yel- steeply sloping dorsal head profile of low egg dummies with only a transpar- many Neochromis. The dentition of 3 in- ent ring. "Black tail giant scraper" has dull dividuals, collected in 1991 is typical for yellow egg dummies on a black anal fin. the latter group (subequally bicuspid These coloration differences may help outer teeth, 5-6 inner tooth rows). Four females to recognize their males. H. (N.) fishes, collected in 1993 and 1995 had "red tail giant scraper" and H. (N.) "black only 3 inner rows and those two collected tail giant scraper" were caught, in the very in 1996 only 2-3, while all had subequally clear water of Gana Island, at 4-5 m depth bicuspid outer teeth. It is currently not among big rock boulders. possible to interpret these observations. One of the least known, and at the In fishes below 85 mm standard length same time most peculiar Neochromis the lower jaw length/width ratio is higher

86 than in typical Neochromis species but Boulenger (1911) and redescribed by bigger individuals have a typical Witte & van Oijen (1990)) chest scales are Neochromis-ratio of below 1.1. Males are small and deeply embedded. Several dark brown-black with five to six vertical hundred specimens which I saw from the bands and an indication of orange sheen southern lake exhibit this feature that behind the pectoral fins. All unpaired fins they share with all other southern are black with a narrow red margin. Fe- Neochromis. In contrast, the fishes from males are light brownish and also carry the Napoleon Gulf in Uganda, show a five to six vertical bars on the flanks. H. generalized chest-squamation without (N.) "pseudoblack" partly resembles in col- scale size transition. As a consequence oration and general appearance H. they have only four to six (mean 4.8) "scraper pseudonigricans", a Pseudo- scales between the pelvic and the pecto- nigricans-group species with scraper ral fin insertions, compared to 6-9 scales morphology, though H. (N.) "pseudoblack" in Greenwood' s H. (N.) nigricans, and in is morphologically a step further towards the H. (N.) nigricans from the Mwanza a specialized epilithic scraper. Neverthe- area (mean 6.9). (2) The length/width ra- less, phylogenetic affinities to the tio of the lower jaw is significantly higher Pseudonigricans-complex cannot be ex- in "large scale nigricans" than in H. (N.) cluded. nigricans from Mwanza (U test p < 0.05). H. (N.) "pseudoblack" might be endemic (3) Oral jaw dentition: While H. (N.) to Ruti and Mabibi Islands in the centre nigricans as described by Boulenger, and of the western Speke Gulf and seems re- our fishes from Mwanza have 5-7 rows of stricted to shallow (1-3 m depth) inshore inner teeth in both jaws, "large scale places where it occurs in low densities nigricans" from Jinja has only 4-5. The dif- among rock boulders, together with H. ference between major and minor cusp (Xystichromis) "copper black". We have of the outer jaw teeth is greater in "large not caught it in the slightly offshore ar- scale nigricans", and teeth are more un- eas beyond 2 m depth, the major habitat equally than subequally bicuspid. Thus, in of H. (N.) "blue scraper". dentition H. (N.) "large scale nigricans" re- The little data I have about Neochromis- sembles some populations of H. (N.) "vel- like algae scrapers from the northern parts vet black", rather than H. (N.) nigricans of Lake Victoria are mostly based on a from the southern lake. short sampling trip to Jinja, done in 1989. I cannot rule out the possibility that H. The few they are, they do indicate two (N.) "large scale nigricans" will turn out to things: First, that the northern lake shores be conspecific with H. (N.) nigricans once have a species-rich epilithic algae scraper more information will become available, community as well. Second, that there particularly about the populations living are some interesting differences be- along the shores between Jinja and the tween southern and northern Neochro- southeast. The dentition differences be- mis. I found five epilithic algae scrapers tween H. (N.) "large scale nigricans" and, occurring sympatrically, three of which what I consider the typical H. (N.) certainly belong to the Neochromis-line- nigricans, seem extreme for conspecific age but we did not find fishes that match populations of rock-dwelling haplo- with Boulenger's (1906) description, and chromines. Nevertheless, the differences Greenwood's (1956a) redescription of H. that we found among populations of H. (N.) nigricans. One species, H. (Neo- (N.) "velvet black" within the Mwanza/ chromis) "large scale nigricans"nigricans", resem- Speke Gulf area are of similar magnitude. bles the latter in coloration but differs Though the differences between H. (N.) from it in other morphological characters "large scale nigricans" and H. (N.) nigricans (Seehausen 1994): (1) Squamation of the are particularly striking because the chest. In H. (N.) nigricans (illustrated by intraspecific variability among southern

87 The rocky shore at Vesi Island.

A male H. "red tail giant scraper". A female H. "large eye nigricans".

A male H. "black tail giant scraper". A portrait of a male H. "pseudoblack".

88 populations of the latter is very little, den- populations from northern, eastern and tal morphology alone would therefore western shores of the lake should be not be sufficient evidence for the specific sampled, and the different populations status of H. (N.) "large scale nigricans". be compared to the type material of H. Stronger evidence may be the squama- nigricans and H. simotes. tion differences. H. (Neochromis) "slender scraper" is a Chest squamation has been considered second Neochromis from Jinja (See- such phylogenetically relevant a trait that hausen 1994). In its elongated body Greenwood (1979) defined the haplo- shape and little decurved dorsal head pro- chromine genus Thoracochromis (which file it resembles H. (N.) "unicuspid scraper" does not occur in Lake Victoria) based on and similar forms from the southern lake. the single synapomorphy (shared derived From the sympatric H. (N.) "large scale character) of small and deeply embedded nigricans" it differs also in dentition and chest scales. The value of this trait for squamation, having longer outer teeth high level phylogenetic work has been and very small and deeply embedded questioned after it was found that most scales in the chest and nuchal regions. In (southern) rock-dwelling haplochromines contrast to H. (N.) "unicuspid scraper" it of Lake Victoria have Thoracochromis-like bears in the outer row typical subequally chest squamation (Seehausen 1994). bicuspid epilithic scraper teeth. The inner However, no case is known to us in teeth are arranged in 6 to 8 broad bands. which conspecific populations of H. (N.) "slender scraper" lives sympatrically haplochromine cichlids would exhibit ba- with two other Neochromis: H. (N.) "large sically different patterns of chest scale nigricans", and H. (N.) "Jinja big squamation. Moreover, Boulenger (1911) depicts, as Haplochromis simotes, a typical small scaled H. (N.) nigricans from the Victoria Nile at Jinja. This implies that the small scaled species also occurs at Jinja. H. simotes was synonymised with H. nigricans by Regan (1922) because he considered the apparent difference in head shape between the holotype of H. nigricans and H. simotes an artefact of preservation in H. nigricans. To finally an- swer the question about the status of H. (N.) "large scale nigricans", nigricans-like H. "slender scraper".

A male H. "pseudoblack". A female H. "pseudoblack".

89 scraper", and with two other epilithic male. Young adults have a less strongly scrapers: H. (X.) cf. "copper black" and H. decurved dorsal head profile than both, (X.) "Jinja blue scraper". H. (N.) nigricans and H. (N.) "large scale As H. (Neochromis) "Jinja big scraper" nigricans". denoted is a bigger species from Jinja/Na- Small and deeply embedded chest and poleon Gulf (Seehausen 1994) that re- nape scales have been interpreted as a sembles the sympatric H. (N.) "large scale derived character of phylogenetic value nigricans" in its red caudal fin coloration, by Greenwood (1979) who described the but has grey to blackish flanks, a less genus Thoracochromis based on this strongly decurved dorsal head profile, and character. Greenwood did not find species a narrower lower jaw. Like those of H. (N.) (or "possibly one") with deeply embedded "large scale nigricans", the outer teeth of chest and nape scales in Lake Victoria. In H. (N.) "Jinja big scraper" are unequally to contrast van Oijen et al. (1981) had ob- subequally bicuspid, and much longer served that many rock dwelling haplo- than those of the inner rows. The only 4 chromines in the Mwanza Gulf of Lake to 5 scales between pectoral and pelvic Victoria exhibit this feature, and argued fin are large and not deeply embedded. that it might be an adaptation to life be- The species occurs sympatrically also tween rock boulders: large, loosely em- with H. (N.) "slender scraper", H. (X.) cf. bedded scales are easily lost when fast "copper black", and H. (X) "Uganda blue escape swimming trough narrow gaps is scraper". required. Greenwood worked mainly at From Entebbe/Uganda another species the northern lake coast, and of the four that is likely to belong to the Neochromis Neochromis species that I could examine lineage was exported in the late 70s to from Uganda, only one had small, deeply Sweden by B. Selbrink as H. (N.) nigricans embedded chest scales. In contrast, all of (Selbrink 1986). It turned out to be nei- the species that I examined from the ther of the two by then known Neochro- southeastern lake, between the Mwanza mis species nigricans and "velvet black" Gulf and Gana Island, had small, deeply and was given the name H. (Neochromis) embedded chest scales. Since many "cross dresser" (L. Kaufman, pers. comm.). southern species also have broader lower This is a somewhat elongate blueblack jaws, more subequally bicuspid teeth, and fish with red caudal fin. Very similar fishes more inner tooth rows (all characters as- have been exported from Kenya to Ger- sociated with epilithic scraping) than many in the early 1990s by an aquarium northern species, it might be argued that fish exporter, and may still be in the they are adapted to different environmen- aquarium fish trade. It is possible that tal conditions. Tank bred individuals of several similar species are referred to un- four populations of H. (N.) nigricans and der the name "cross dresser". several populations of other species from Yet another epilithic algae scraper spe- Mwanza and Speke Gulf, retained their cies that resembles H. (N.) nigricans and squamation pattern over several genera- likely came from Uganda or Kenya ap- tions. Thus, the type of chest squamation peared in the 1980s in the aquarium is certainly not merely a phenotypic re- trade under the name Astatotilapia nubila. sponse to different ecological conditions. I have suggested the working name H. (Neochromis) "red anal nigricans" Xystichromis-like (Seehausen 1994). Like several other epilithic algae scrapers northern Neochromis it has large, and not deeply embedded scales between pecto- Several of the above described species ral and pelvic fin (I found 6 to 7 in very do not fall into Greenwood's narrow defi- big fishes), but it differs from the others nition of the genus Neochromis, but with by having an entirely red anal fin in the two exceptions share the among Lake

90 Victoria cichlids unique combination of a broader than the neck and body. The head decurved dorsal head profile, a broad and length of Greenwood's Xystichromis spe- short lower jaw, subequally bicuspid cies ranges from 31.8 to 33.4% of stand- outer teeth and more than three inner ard length, the lower jaw length from tooth rows. One species (H. (N.) "unicus- 36.7 to 37.6% of the head length, and the pid scraper") and one morph of a second length/width ratio of the lower jaw from species (H. (N.) "Bihiru scraper") have a 1.4 to 1.6. shallow, and straight dorsal head profile, Three species are currently described a longer and less broad lower jaw and in Xystichromis Greenwood 1980. They unicuspid outer teeth. Nevertheless, I were frequently considered epiphythic consider them Neochromis because in algae scrapers and plant eaters (Fryer & general appearance and relative head Iles 1972, Witte & van Oijen 1990), how- length, they closely resemble other spe- ever, only for H. (X.) phytophagus Green- cies of that group, and at least one of wood 1966, is the evidence unambigu- them shares a colour polymorphism ous. Studied individuals had not only plant (black, blue, OB) with other Neochromis material in the stomachs but were in fact species. Several other species, though caught over sand bottom near rooted epilithic scrapers as well, deviate from vegetation (Greenwood 1966). Respec- Neochromis in relative head length, head tive habitat notes are not available for H. shape, jaw and tooth shape and lack the (X.) bayoni (Blgr.) 1911 that was collected OB-morph. With their straight dorsal from the Victoria Nile, connecting Lakes head profile in spite of a deep body, with Victoria and Kyoga (Greenwood 1960). H. a longer head (species means >33% of (X.) nuchisquamulatus (Hilgend.) 1988 head length with the exception of one was collected from exposed littoral zones species), a longer and less broad lower in Lake Victoria and even from the far off- jaw (36.8-40.9% of head length, length/ shore Godziba Island in the centre of the width ratio 1.30 to 1.49) and with usually lake (Greenwood 1956a). Stomach con- more unequally than subequally bicuspid tents, analysed by Greenwood (1960), outer teeth, they match better Green- were composed of fragments of plant tis- wood' s Xystichromis-, rather than his sue and numerous filamentous algae that Neochromis-lineage. Greenwood calls "epiphytic" without giv- Greenwood defined Xystichromis ing evidence for their actual growing on Greenwood 1980 as small haplo- plants rather than on rocks. He mentions chromines with a long, much coiled intes- Oedogonium which we frequently find in tine (ca. 3-4 times standard length), and the stomachs of epilithic algae scrapers broad bands (4-6 rows deep) of inner and Greenwood's habitat notes for H. (X.) teeth anteriorly and anterolaterally in both nuchisquamulatus, particularly its occur- jaws, narrowly , if at all separated from rence at the isolated rocky Godziba Island, the outer row. The dorsal head profile suggest to me, that this species might be slopes gently, and is not strongly an epilithic scraper. decurved. The dentary (tooth carrying H. (Xystichromis) "copper black" is the lower jaw bone) neither as deep and fore- most widely distributed of all known epi- shortened as in Neochromis, nor as slen- lithic algae scraper species, and the ma- der as in Astatotilapia. The teeth in the trix species of the Xystichromis-lineage. outer row of each jaw are very closely set, We found it at almost every rock shore moveably implanted, tall, and slender but and rocky island except in the southern strong, showing only a slight antero-pos- Mwanza Gulf, south of Python Island and terior decline in their height and size. at the very steep slopes of some islands They are unequally bicuspid, the minor in the northeastern Mwanza Gulf. A popu- cusp prominent but clearly smaller than lation that might be H. (X.) "copper black" the major one, the crown is not distinctly has also been found near Jinja at the

91 A male H. "red anal nigricans".

A territorial male H. "copper black" at Chamagati Island

A territorial male H. "copper black" at Makobe Island

92 northernmost end of the lake (A. Meyer pers. comm., pers. obs.). This is also ecologically probably the most ver- satile of all known epilithic algae scrapers, if not of all Mbipi. It inhabits various stony and rocky habitats, including stone-sand mixed substrates. It is, however, absent from some very steeply sloping rocky islands and mainland shores and is generally more abundant at gently sloping shores, with A male H. "copper black" from Makobe Island. small and medium sized boulders. At many such places it is the numerically dominant cichlid species, living between 0 and 6 m water depth with the maximum density between 0 and 2 m depth (Seehausen & Bouton in press). The well studied population from Makobe Island feeds on a variety of food items such as diatoms, filamentous epilithic algae, detritus, insect larvae and moss animals (Bryozoa) and is more appropriately described as an epilithic omnivore than as an algae A female H. "copper black" from Ruti island. scraper (Bouton et al. submitted). In this respect it is similar to the H. (N.) "velvet black" populations of the Mwanza Gulf. Like in H. (X.) nuchisquamulatus, fragments of rooted plants are usually found in the stomachs of H. (X.) "copper black". From Scuba observations at Makobe Island we know that they are usually taken in as detritus, sedimenting among the rocks and are usually not browsed off living vegetation. Similar feeding be- A female H. "copper black".

Amranda Point; small boulders are typically found in the habitat of H. "copper black".

93 haviour could account also for the pres- ference between breeding males of the two ence of plant material in H. (X.) nuchi- species remains distinct even after preser- squamulatus stomachs. vation. A further striking difference between Unfortunately live coloration of H. (X.) the two species is that H. (X.) "copper black" nuchisquamulatus is not known. How- lacks the OB morph which is so character- ever, H. (X.) "copper black" differs from the istic of all, except the southern Mwanza Gulf latter and from the other described Xysti- populations of H. (N.) "velvet black" and of chromis species in other morphological several other Neochromis: black blotches features, i.e. has a smaller lower jaw on orange to pink ground. Instead H. (X.) length/width ratio and a distinctly smaller "copper black" has an extremely rare OB-like eye diameter (compare table 3 with table morph that carries some orange blotches 4). In terms of coloration many popula- on a black or brown body. Among the sev- tions of H. (X.) "copper black" are similar eral thousand individuals that we have seen to H. (Neochromis) "velvet black", from of H. (X.) "copper black" were only two indi- which they differ mainly in having a wine- viduals of this morph that may be produced red anal fin rather than a black one with a by recurrent mutation. Both individuals red rim. The black males of "copper black" were males. Where H. (X.) "copper black" usually exhibit a coppery sheen on the and H. (N.) "velvet black" coexist, the first cheeks and sometimes on the flanks often tends to dominate in shallower and which we have never seen in H. (N.) "vel- more inshore reaches, the latter in deeper vet black". In some populations of H. (X.) and more offshore water or at steeper "copper black" males are yellow-green slopes. H. (X.) "copper black" occurs rather than black, in others they have blue sympatrically with all Neochromis and both lips. below described Xystichromis algae scrap- The relationship between H. (Xystichro- ers. mis) "copper black" and H. (Neochromis) "vel- My grouping of "copper black" and "velvet vet black" is one of the many taxonomical black" into two different subgenera of algae puzzles among the rock haplochromines. scraping haplochromines seems to contra- Prior to our extensive rock shore survey H. dict the close similarity observed between (X.) "copper black" and H. (N.) "velvet black" certain populations of them. This is due to were regarded as one species (see for in- unusually extreme intraspecific variation stance Seehausen 1994 who used the term among populations of "velvet black". While H. "velvet black Speke Gulf" for H. (X.) "cop- "copper black" matches better Greenwood's per black"). However, morphological and Xystichromis than his Neochromis lineage ecological data collected in the meantime (Greenwood 1980), and many populations show that these are two species that coex- of "velvet black" match Greenwood's ist in sympatry in large parts of their geo- Neochromis lineage, others deviate from graphical ranges. H. (X.) "copper black" dif- the latter into the direction of Xystichromis, fers from H. (N.) "velvet black" not only in while maintaining a typical "velvet black" ap- male coloration but also in head shape and pearance. Greenwood already arrived at the dentition. The first one has a less decurved, conclusion that Xystichromis and rather straight dorsal head profile and the Neochromis are closely related sister outer teeth are less closely spaced and of- groups within one superlineage (Green- ten more unequally, rather than subequally wood 1981). It is therefore not surprising bicuspid than in most (not all) populations to find closely resembling species in both of H. (N.) "velvet black". Where the two spe- groups or bridging between the groups. If cies coexist, males of H. (X.) "copper black" their resemblance indeed reflects particu- are often more yellow-green than black larly close phylogenetic relationship, the (e.g. Python Island), while they are black at separation of the lineages Neochromis and places from which H. (N.) "velvet black" is Xystichromis may have to be revised. absent (e.g. Makobe Island). The colour dif- A second deep bodied epilithic algae

94 scraper with a straight dorsal head profile orange-red in its lower half, and grey with is the small H. (Xystichromis) "short red streaks and a red rim in the upper half. scraper" that is known only from Zue Island, At Gana Island this is the only Xystichromis- near the northern shore of the central Speke like rock scraper that we found but it coex- Gulf, but may be distributed more widely ists with at least four Neochromis and two east of the surveyed region. This fish, Paralabidochromis rock scrapers. Its habitat though on first glance easily confused with consists of large to huge rock boulders that H. (X.) "copper black", is of a characteristic form extensive rock gardens with steep appearance. Relative to its large head, its walls as well as sun exposed vertical sur- body is short, giving it a heavy headed ap- faces with algae cover. We found H. (X.) "red pearance. The upper lip of males is dis- anal blue" at water depths from 3 to at least tinctly swollen and the premaxilla rather 5 m, where it was one of three abundant bullate. Males occur in two colour morphs: rock scrapers, H. (Neochromis) "blue one green-blue, the other one yellowish. In scraper" and H. (Paralabidochromis) "yellow contrast to H. (X.) "copper black", this spe- anal picker" being the other two. Possibly cies does not exhibit any red in its fins. Fe- H. (X.) "red anal blue" is north of our survey males are often very dark, almost blackish area more widely distributed. brown. Little is known about the ecology Similar to H. (X.) "red anal blue" is a spe- of H. (X.) "short scraper". It inhabits the very cies that is known only from the northern, gently sloping small boulder shore of Zue Ugandan lake shore. It had been imported Island from the surf zone to at least 4.5 m to Sweden by B. Selbrink under the name depth. In some very shallow places it can Psammochromis riponianus (Selbrink 1985). be the numerically dominant cichlid spe- Probably the same species has more re- cies, occurring in very high population den- cently been collected again at Jinja/Uganda sity. At other places H. (X.) "copper black is by A. Meyer (pers. comm., pers. obs.). I had more abundant and again at others H. (Neo- suggested the descriptive name "Uganda chromis) "eastern blue scraper". A third blue scraper" for it (Seehausen 1994). Males epilithic scraper, living sympatric with "short of H. (Xystichromis) "Uganda blue scraper" scraper", is H. (N.) "giant scraper". The dark, are bright blue with red in all unpaired fins brown-black colour of "short scraper" fe- and 6-7 vertical bars on the flanks. Both males suggests that also females are fre- sexes have quite long subequally bicuspid quently territorial. outer teeth and 3-4 inner rows. Nice pho- H. (Xystichromis) "red anal blue" is a lit- tographs were published in the Swedish tle known species that we recently discov- journal "Akvariet" (Selbrink 1985). The spe- ered at Gana Island northwest of Ukerewe, cies lives sympatrically with H. (X.) "copper the northern most point in our survey. In black" and (N.) "large scale nigricans" at Jinja. body shape it is very similar to H. (X.) "cop- H. (Xystichromis) "large eye black" is per black" though its head is somewhat known from only a few individuals, caught shorter relative to its standard length and at Igombe Island in the southern Speke its lower jaw shorter relative to its head Gulf. This species resembles H. (X.) "copper length (table 3). In these measurements it black" and H. (N.) "velvet black" in coloration is close to Neochromis but it has the straight but differs from both in the anatomy of the dorsal head profile typical for Xystichromis jaws (longer and less broad lower jaw), in and a Xystichromis-like dentition with un- dentition (less closely set outer teeth, equally rather than subequally bicuspid fewer inner rows) and most strikingly by its teeth in the outer row. This is a brightly col- larger eyes. The eye size of "large eye oured species with two male colour black", unlike eye sizes of the four above morphs. One is light blue, the other one discussed species, matches that of the yellowish. Both have a bright orange-red three described Xystichromis species. anal fin with orange coloured egg dum- Males of H. (X.) "large eye black" are entirely mies, and a caudal fin that is similarly bright black with an anal fin that is bright red in

95 A male H. "short scraper" from Zue Island. A female H. "short scraper".

A male H. "large eye black" from Igombe Island. A female H. "large eye black".

A male H. "red anal blue" from Gana Island. A female H. "red anal blue". its distal half, and with a red caudal fin larly large eyes is H. (N.) "large eye edge. The dorsal head profile resembles nigricans" from Makobe Island. Like that that of H. (X.) "copper black" but is slightly one, also H. (X.) "large eye black" has been more decurved and not concave. At Igombe collected in shallow immediately inshore Island H. (X.) "large eye black" coexists with waters (1-2 m depth) among rock boulders H. (Xystichromis) "copper black" and with and in rockpools but not in deeper parts of several Neochromis algae scrapers: H. (N.) the habitat, as opposed to what could have nigricans, H. (N.) "velvet black", H. (N.) "giant been expected from the large eyes. The scraper", H. (N.) "unicuspid scraper". The only two species, however, differ considerably epilithic algae scraper population with simi- in head shape and dentition. Nothing is

96 known about the feeding habits of H. (X.) standard length) and broad head, a lower "large eye black". jaw that is longer than in Neochromis complex species (38.1-40.7% of head length), The "carp" complex of big a dorsal head profile that is straight or Xystichromis-like algae scrapers slightly concave, and a dentition that differs from all dentition types previously known The species of this species complex from Lake Victoria cichlids: unequally bicus- share the following morphological fea- pid to blunt, sometimes slightly recurved, tures: large to very large size with a rela- unicuspid (big fishes) teeth in the outer tively long (species means 32.0-32.7% of tooth row and many inner rows that form

Fishermen at Makobe Island

A territorial male H. "carp" in the aquarium.

97 broad tooth bands, like in the other algae Makobe Island is separated by 5 km of scrapers, but are separated from the outer more than 20m deep water, from Ndurwa row by a well developed gap. Small indi- Point, from Bwiru Island off the southwest- viduals have, like other epilithic algae scrap- ern corner of Ukerewe Island and from ers, subequally bicuspid teeth in the outer Chamagati and Matwinki Islands in the row. None of the species of the H. "carp" Sengerema region. It seems thus a rela- complex is described and thus this group tively widely distributed, though often rare was not considered in Greenwood's revi- species. Male coloration varies little among sion of 1980. My morphometric measure- and within populations, but variation in ments fall into the range of described other morphological characters such as Xystichromis species, with which "carp" dentition is sometimes that pronounced complex species also share the straight dor- within populations, that one could think of sal head profile and the broad bands of in- different species. ner teeth. In spite of their deviation from H. (X.) "carp" inhabits exclusively the shal- described Xystichromis in terms of outer low reaches (0 to 2 m depth) of gently slop- tooth shape, I assign the species of this ing small rock boulder shores, in particular complex to Greenwood's Xystichromis-lin- somewhat surf protected microhabitats like eage until a revision of the algae scraping little bays. We observed big territorial males Lake Victoria haplochromines clarifies their in only 0.5 m water depth under floating position. roots of reed mats. H. "carp" is a solitary and Most members of the "carp" complex live rather cryptically living species that occurs at gently sloping to moderately steep in low population density. Of three individu- shores. Our stomach contents analyses of als of which we studied stomach contents, several species suggest that they are real two had eaten almost exclusively filamentous algae scrapers, in spite of the filamentous green algae, the other one fila- outer tooth shape and rather wide setting mentous blue-green algae. All had also in- which appear unsuitable for the loosening gested small amounts of plant (phane- of firmly attached filaments. Laboratory ex- rogam) tissue and two had eaten a few in- periments with Haplochromis "carp" had sect larvae. shown earlier that this species feeds by Close to H. (X.) "carp" is Haplochromis scraping and snapping (Seehausen 1994). (Xystichromis) "orange carp" which is All species seem to have a geographically known only from Zue Island, near the rather narrowly restricted distribution and, northern shore of the central Speke Gulf. with one exception, all are allopatric. From H. "carp", this species differs in mela- Haplochromis (Xystichromis) "carp" was nin pattern and male coloration. It has the first species discovered in this species more (7 to 9) and more closely spaced complex. N. Bouton first collected some in- vertical bars on the flanks. Males exhibit dividuals in 1991 at Makobe Island in the a warm orange coloration on the flanks Speke Gulf (Seehausen 1994). H. "carp" is and a reddish anal fin instead of the blue one of the few bright yellow Mbipi. On the anal fin with red edge of H. "carp". Like H lemon yellow flanks males carry 5 to 6 black (X.) "carp" at other localities, does also H. vertical bars. The caudal fin is reddish, the "orange carp" inhabit at Zue Island the very anal fin bright sky blue with a broad red shallow water. Between 0.4 and 1 m edge, the dorsal fin in the spinous part me- depth it is more abundant than H. (X.) tallic sky blue, in the soft part more hyaline "carp" is at other islands. However, "or- with numerous bright red streaks and dots. ange carp" seems even more narrowly re- There are usually two or three broad black stricted in its depth distribution. Among blotches on the dorsal fin basis. Females many hundreds of cichlids that we col- are light yellowish brown. Discovered at lected in nets set at depths between 1 Makobe Island, we know the species mean- and 5 m at Zue Island was only one "or- while also from Mponze Point, from which ange carp", caught in 1.2 m depth.

98 Haplochromis (Xystichromis) "red carp" haplochromine species after H. is one of three known bright red epilithic (Neochromis) cf. "velvet black". Like the algae scrapers (the others are H. (Para- latter it inhabits the pure rock, as well as labidochromis) "rock macula" and H. (P.) the rock/sand mixed substrates. Of three "red short snout scraper"). Young individu- studied individuals, one had an empty als are morphologically rather similar to stomach. One had been feeding pre- H. (X.) "carp" and H. (X.) "orange carp". dominantly on filamentous blue-green al- Adults have a shorter lower jaw and very gae, as well as on some filamentous different coloration. Males have opercu- green algae. The third one had been feed- lum and the lower half of the flanks ing to about equal parts on moss animals bright red to pinkish red. The upper half (Bryozoa), may fly (Ephemeroptera) larvae of the flanks is grey with a reddish sheen and filamentous blue-greens. but occasionally the entire flanks are red. Kissenda Island consists of medium The head is greyish, lips and cheek usu- sized rock boulders and slopes rather gen- ally light blue. The anal fin is red with or- tly to moderately steep. Algae scrapers ange coloured egg dummies, the caudal are not very abundant at this island. H. fin is reddish with a deep red edge, the (Neochromis) cf. "velvet black", H. (N.) dorsal blueish grey with red lappets and nigricans and an unidentified species of red dots in the soft part. The pelvic fins the H. (Paralabidochromis) "short snout are half black and half red. H. (X.) "red scraper" complex, coexist with H. (X.) "red carp" is known only from two localities, carp". The latter is numerically the domi- about 30 km apart from each other, along nant one and lives between 1 and at least the western shore of the Mwanza Gulf. It 4, possibly 6 m depth. Though morpho- may be more widely distributed along logically a scraper, the only individual of that shore, although we did not find it at which we checked the stomach had three collecting localities between the eaten almost exclusively chironomid lar- two places. The two populations differ in vae. It is noteworthy that this species co- coloration (the kind of red, see photos) exists at Kissenda Island with a second and at least in one morphometric value, very big and entirely red Haplochromis: H. suggesting that they are geographically (Ptyochromis) "red giant sheller" (see page isolated from each other. The population 200) from Buyago rocks (southwestern At Bwiru Island off the southwest cor- Mwanza Gulf: Smith Sound, H. "pink giant" ner of Ukerewe Island we collected a in Seehausen 1994) has a significantly male of a "carp"-group species that in col- smaller lower jaw length/width ratio. Fur- oration resembles H. (X.) "red carp" but dif- thermore it has a steeper dorsal head pro- fers in dentition. At a size (104.9 mm SL), file than the population from Kissenda Is- at which the latter has weakly bicuspid land (northwestern Mwanza Gulf). teeth in the outer row, it has still sub- to Buyago rocks are of a rather rare kind unequally bicuspid teeth. Until more in- of rock habitat: large, almost unbroken dividuals become available we treat the slabs slope gently down from Euphorbia Bwiru Island population as H. (Xysti- forest covered hills to the shallow bottom chromis) cf. "red carp"carp". Interestingly it lives of the Smith Sound. Below the water sur- sympatric with H. (X.) "carp" which seems face the slabs are eroded into rock boul- to have the normal "carp"-group tooth ders of 30 to 60 cm diameter. These boul- shape but has unusually few (3) inner ders form pure rocky substrate only down tooth rows compared to 5 in cf. "red carp". to a water depth of 1-1.2 m, followed by rock/sand mixed substrate or rocks, covered with a layer of sand. H. (X.) "red carp" lives here at 0.5 to at least 2 m water depth and is the most abundant

99 In all the colours of the rainbow — species of this complex was discovered the Haplochromis nyererei complex and described just after Greenwood's (1980) generic revision of the Lake Victoria The Haplochromis nyererei complex is a haplochromines. Until today Haplochromis rather close-knit species complex that very nyererei Witte-Maas and Witte 1985, re- probably represents largely a monophyletic mained the only described species in the group, and may in a revision of the rock- complex. However, meanwhile we know dwelling cichlids of Lake Victoria be given about 20 undescribed species, delimitation (sub)generic status. A (sub)generic name is among which is not always simple (See- currently not available, because the first hausen 1996), and which I will refer to as

A female H. "carp" in the aquarium. Rocky slab shore at Buyago Rocks

A male H. "orange carp" from Zue Island. A male H. "red carp" from Buyago Rocks".

A male H. "red carp" from Kissenda Bay. A male H. cf. "red carp" from Bwiru Island.

100 Nyererei species. Among the rock-dwelling lineages of algae scrapers and may repre- Lake Victoria cichlids the complex can be sent a sister taxon to them. The species characterized by the combination of the fol- complex is rather diverse in terms of feed- lowing characters (revised after Seehausen ing ecology, however, most species prey 1996): Straight to moderately incurved dor- predominantly on Aufwuchs-associated in- sal head profile; unequally bicuspid to uni- vertebrates and/or plankton. A few species cuspid teeth in the outer tooth row (fe- scrape predominantly algae. The most males of one species have subequally bi- commonly used feeding techniques are cuspid teeth), that are in adult males coarse, picking, snapping and some scraping tech- with recurved crowns, and set so that the niques (Seehausen et al. in press [a]). Most interspaces among the teeth are broader species exhibit pronounced sexual dimor- than the neck of the teeth; a rounded den- phism in size (females being smaller) and tal arcade; 2 to 5 rows of inner teeth; a gap anatomical characteristics associated with between the outer and the first inner tooth feeding. row; a lower jaw length of 38-43.5 % of Intragroup taxonomy among the H. head length; a lower jaw length/width ratio nyererei like species and populations is of 1.25-1.78 (1.89 according to Witte-Maas complicated because the range of variation & Witte 1985); a relatively broad head with in anatomical characters in the species an interorbital width of 23.6-29.1 % of head complex is rather narrow, and because

Haplochromis "zebra nyererei" lives among such shoreline boulders. length; small deeply embedded chest some species do across their populations scales; and a conspicuous and regular ver- display almost the entire range of some tical bar pattern (zebra-like appearance) at characters. Complicating are furthermore least in females, with 5 to 8 bars between the small geographical ranges of several pectoral fin and caudal end of the dorsal fin. species and the recurrent appearance of I think, the H. nyererei complex is closely similar male coloration in anatomically dif- related to the Neochromis and Xystichromis ferent forms at different geographical ends

101 of the surveyed area. The latter observation sexes are usually broader and less numer- indicates that similar male colour patterns ous than those in H. nyererei. The anal may in cases be due to parallel evolution, fin of males is red and the caudal fin fre- rather than to shared ancestry. As a first quently features red stripes and blotches rough approach to group species according as does the dorsal fin. The teeth in the to similarity, and independently of male col- outer tooth row are in adult fishes usu- oration, I did a cluster analysis (UPGMA ally unicuspid, pointed and widely method) by means of which species were spaced. The slightly longer head and ordered according to similarity in morpho- lower jaw of H. "zebra nyererei" gives it a logical characters, excluding male colora- slightly more predatory facies than has tion. This produced several species groups the red H. nyererei. We know it from to which I will refer in the following account most surveyed places except the most of the species as "morphological groups". gently sloping small boulder habitats in The reader should bear in mind that these the Sengerema region and at Zue Island are groups, composed of species with simi- (see below). Considerable regional differ- lar shape, and that they are not necessarily ences exist in coloration. For instance particularly closely related phylogenetically, populations from the Mwanza Gulf and though some may well be. Thus these from Ukerewe do have uninterrupted ver- groups should not straight forward be un- tical bars while fishes from some places derstood as phylogenetic groups. in the Speke Gulf have their vertical bars rather dissolved into blotches which ap- Nyererei with long heads and long pear like an interrupted mid lateral and jaws dorsal lateral stripe. H. "zebra nyererei" is one of the most The first morphological group, I call it cryptically living rock cichlid species. It in- the "Zebra nyererei group", consists of habits predominantly crevices among the species that have a rather long head and, rocks in shallow water (Seehausen 1996) relative to the head length, long lower and is only very occasionally encountered jaws. One of them, Haplochromis "zebra outside. In many diving hours at Makobe nyererei"nyererei", is the geographically most Island in the Speke Gulf I met it only widely distributed species of the H. twice: Once an adult male crossing a nyererei complex and the only one that small stretch of sand between two heaps was found throughout the surveyed area. of rocks. The other time an adult male, Similar populations have occasionally standing in water of less than 1m depth, been imported to Europe by aquarium between the rocky substrate and a water fish traders from the northern lake shores. hyacinth (Eichhornia) layer. H. "zebra H. "zebra nyererei" may represent a basal nyererei" lives at moderately and steeply member of the H. nyererei complex. It is sloping shores with medium to large rock the species that has the highest record boulders but is absent from very gently of sympatric occurrences with other sloping shores with small boulders. Its members of the complex, and is what I diet is dominated by benthic insect larvae, consider the matrix species of the predominantly of may flies (Ephemer- Nyererei complex. It closely resembles optera), which it collects between the the much better known red H. nyererei rocks. Zooplankton, small fish, and moss anatomically but has a male coloration animals, scraped off from the rocks, are that is strikingly different from that of H. occasionally also eaten (Bouton et al. sub- nyererei: It entirely lacks any red on the mitted). body. Instead it carries its 4 to 6 vertical H. "zebra nyererei" lives sympatric with bars on a light grey to bright white back- most other species of the H. nyererei ground, creating a characteristic zebra-like complex. There is strong evidence that all appearance. The vertical bars in both these are specifically distinct from H. "ze-

102 bra nyererei" (but see under H. nyererei and A second population resembling "zebra H. "big blue"). There is little doubt also about nyererei" in male coloration deviates from the distinctiveness of the uniquely col- it in another way. Haplochromis "zebra oured H. "red head", though it is allopatric Miandere"Miandere", living at Miandere Islands in the with H. "zebra nyererei". The situation is west of our survey area, has a much shorter more complicated with a few other forms head, and relative to the head length some- that resemble H. "zebra nyererei" in male what shorter lower jaws. In fact this popu- coloration, and live at places at which "nor- lation has one of the shortest heads found mal" H. "zebra nyererei" does not occur. among species of the H. nyererei complex. Their mutually exclusive distribution with H. Its habitat is a steeply sloping big boulder "zebra nyererei", suggests that they are rep- shore, and like "zebra Senga", also "zebra resenting the "zebra nyererei" lineage. How- Miandere" is frequently seen outside of ever, considerable differences in morphol- rocky crevices. Though caught in crevices ogy between these and other populations as well, it is by far the most abundant of "zebra nyererei" suggest that they could haplochromine species outside of crevices, as well be derived from other species and down to at least 6 m depth, living to- within the complex. Since, moreover, they gether with H. (Neochromis) "blue scraper", differ also ecologically from all other H. "orange dorsal nyererei", H. "yellow anal populations of "zebra nyererei", I prefer not nyererei" and H. "yellow chin pseudo- to lump them with H. "zebra nyererei" until nigricans". It is interesting to note that all more is known. three species of the H. nyererei complex One of these populations, Haplochromis that live at Miandere Island, have a very "zebra Senga"Senga", lives at a geographically iso- short head (table 3). It may imply selective lated rocky peninsula in the central Speke advantages of a short head in the particular Gulf (Senga Point). It differs from other H. environment of this island. "zebra nyererei" mainly by having much Haplochromis "red chest" is a small form, shorter lower jaws, in spite of a rather long known only from Anchor Island in the north- head, and bicuspid instead of unicuspid eastern Mwanza Gulf, one of several very teeth in the outer tooth row. Thus it seems steeply sloping islands composed of big to adapted to other feeding habits. In contrast huge boulders, from which typical "zebra to typical H. "zebra nyererei", the fishes of nyererei" has not been recorded. H. "red the Senga Point population are frequently chest" differs from "zebra nyererei" mainly seen moving about openly over the rocky by a shorter lower jaw and in male colora- substrate and go down to at least 5 m tion, bearing a large orange-red area on the depth. In this respect, they resemble spe- operculum and behind the pelvic. Moreo- cies like H. "big blue" and H. "red head". Mor- ver, its vertical bars are less dark and are on phologically they are closest to some yellowish, rather than whitish flanks. The endemics of the southern and central question whether "orange chest" is an ab- Speke Gulf (H. "blue nyererei", H. "red anal errant population of H. "zebra nyererei" or a nyererei" and H. "Bwiru nyererei") and there distinct species can currently not be an- is rather strong variation in some charac- swered. Though H. "red chest" seems not ters. I assume that this either is a popula- to coexist with the latter in sympatry, the tion of H. "zebra nyererei" that is occupying locality of its occurrence lies within the geo- a niche different from that of most other graphical range of the latter. The population populations, and is therefore exposed to a density of "red chest" at Anchor Island is low, quite different selection regime, or that it and until today less then 10 individuals is a population that originated by have been collected. introgression between "zebra nyererei" and Morphologically close to H. "zebra one of the three above mentioned Speke nyererei", but a distinct species that lives Gulf endemics. All three, just like typical "ze- sympatric with the latter at some places, bra nyererei", are absent from Senga Point. is Haplochromis "big blue"blue". It exhibits a

103 Male H. "zebra nyererei" from Igombe Island (top) and Makobe Island (centre).

A male H. "zebra nyererei" from Python Is. A female H. "zebra nyererei" from Python Island.

104 particularly pronounced sexual dimorphism. Males reach a size of more than 120 mm SL and deviate from H. "zebra nyererei" by a steeper dorsal head profile, a shorter head, broader jaws, often bicuspid teeth in the outer tooth rows, and smaller eyes. Females do not grow bigger than 80 mm SL, have unequally bicuspid teeth in the outer rows, and A male H. "zebra nyererei" from Hippo Island. more inner tooth rows than female "zebra nyererei". They retain the H. nyererei- like general appearance though with a steeper dorsal head profile than most other members of the complex. Full adult males, in contrast, attain a peculiar shape with indications of a nuchal bump. H. "big blue" is known only from the Sengerema region where it inhabits islands and some places at the mainland shore. At Chamagati Island and at the mainland shores, males are bright sky blue with 6 to 8 dark vertical bars on the flanks, and much red in the fins: anal fin, A male H. "zebra nyererei" from Namatembi Is. soft part of the dorsal fin and often pelvic fins. With this coloration they resemble H. (Neochromis) "blue scraper" and they are similarly variable in the extension of the red colour in their fins. A second, more rare male morph of H. "big blue" at Chamagati Island is pinkish red on the flanks between the vertical bars. Females are yellowish brown with red- brown head surface and bright yellow chin, underside, and fins. The populations from Matwinki and Juma Islands differ in morphology (bigger eyes, bigger interorbital width) but in particular in A male H. "zebra Senga" from Senga Point. male coloration from the others. Males are not blue but pinkish red. Unlike the red morph males from Chamagati Island, the Matwinki and Juma males have their red not restricted to the flanks but it is the dorsum, dorsal fin, head surface and snout where the red is most intensive. Though the red populations have a bigger inter orbital width, and a slightly shorter head and lower jaw, and thus were in the cluster analysis of morphometric data placed closer to some other species of the H. nyererei complex, I A male H. "zebra Miandere" from Miandere Is. consider them conspecific with the

105 blue populations with which they share a the Sengerema region, between Kujunju general appearance of males that is unique Point and the Mwanza Gulf entrance, we in the species complex, as well as the small found populations that are intermediate yellow females. between H. "big blue" and H. "zebra nye- Most habitats of both, the blue and the rerei" in coloration and anatomy (abbrevi- red form of "big blue", consist of very gen- ated cf. Zny-Bbl in the distribution map). tly sloping shores with small rock boulders Some of these populations are either very and rubble substrate. H. "big blue" is found variable or represent an admixture of two from very shallow water down to at least 4 species: H. "big blue" and H. "zebra nye- m depth. The very big fully adult males are rerei", converging towards an intermediate restricted to the deeper parts of the spe- phenotype, or actually hybridizing. These cies´ depth range and can easily be taken observations may suggest that H. "big blue" for a separate species, particularly since fe- is closely related to H. "zebra nyererei", de- males are rarely caught together with them spite different feeding ecology and trophic at depths below 1.5 m. Females and morphology. Along the Sengerema main- smaller males live predominantly in the land shore, typical "big blue" are found at shallow water. We caught females almost very gently sloping small boulder shores, exclusively at less than 1.5 m depth. Males while the variable populations that are in- of 10 cm SL become already territorial in termediate between "big blue" and "zebra the shallows. Both sexes feed here pre- nyererei", live at shores that have bigger dominantly on epilithic Aufwuchs, and the boulders and slope slightly steeper. Typical most frequently used feeding techniques "zebra nyererei" inhabit steeper sloping rock are pullscraping and picking (Seehausen shores with boulders large enough to cre- unpubl. data). The algae scraping habit of ate a rich interstitial of crevices. More eco- H. "big blue" is unique among the known logical work on these populations will be H. nyererei complex species. Like many al- needed to clarify these interesting patterns gae scrapers of other species complexes, but it seems that the two forms, in one way it inhabits with high population densities or the other, replace each other ecologically. the shallow waters, and is among the nu- It is particularly interesting that at Juma Is- merically dominant species at several land, "zebra nyererei" and the red form of places, following in abundance after H. "big blue" live in sympatry, and that no in- (Xystichromis) "copper black" and H. (Neo- termediate forms are found at this island. chromis) "black nigricans". Most other H. If "big blue" is considered the representa- nyererei complex species live either in tive of the "zebra nyererei"-lineage at gen- deeper waters or in crevices between the tly sloping shores, than it is furthermore rocks. At Juma Island most of the shore important to note that it lives sympatrically consists of bigger, steeper sloping boul- with a second species that is similar to H. ders. However, "big blue" does occur there "zebra nyererei", though grouped in a sepa- at the most gentle places down to at least rate morphological group: Haplochromis "all 4.5 m depth. Its population density is there black nyererei"nyererei". This is a rare species, of much lower than at the very gently sloping which we caught 12 individuals at Cha- small boulder islands. magati Island and at Kajunju Point (Sen- H. "big blue" coexists with four other H. gerema mainland). Males are entirely black nyererei complex species: H. "all black with a reddish anal fin and often rather dull nyererei", H. "pink anal", H. "zebra nyererei" orange coloured egg dummies. Females and H. nyererei (with the latter two only at are blackish brown. It lives at depths be- Juma Island). Some variation between its tween less than a metre and at least 4 m, populations exists. So do the fishes from and has been caught only at very gently the mainland have more an insectivorous sloping shores. Nothing more is known yet dentition than the fishes from Chamagati about its ecology. H. "all black nyererei" Island. At several rocky mainland shores of shares with H. "zebra nyererei" a long head.

106 In fact this species has the longest head H. "red head" lives at Zue Island between found among H. nyererei complex fishes the surf line and 5 m water depth, openly but it has at the same time one of the short- on the rock substrate, moving sometimes est lower jaws found among them, and so in mixed shoals with algae scrapers of the H. "all black nyererei" differs in this respect Neochromis lineage. At Mabibi Islands we considerably from "zebra nyererei". Morpho- observed it between 1 and 4 m water logically "all black nyererei" is similar to H. depth among boulders as well as moving "zebra Senga". It lives sympatrically with at freely across the rock bottom. Within the least two other Nyererei complex species, available range of microhabitats at Mabibi, H. "big blue" and H. "pink anal". it seems to prefer the less steeply sloping Morphologically and ecologically most areas. H. "red head nyererei" is the only H. similar to H. "big blue" is Haplochromis "red nyererei complex species at Zue Island but head nyererei" that is entirely different in coexists with H. "blue nyererei" at Mabibi male coloration. It is another non-cryptically Islands. living shallow water dweller. Known only from two islands in the central-northern Plankton eating Nyererei with a broad Speke Gulf, it is very abundant at Zue Is- head land, much less so at Mabibi Islands. Zue Island has a very gently sloping shore with In a second morphological group, let it be small rocks and stones, while Mabibi Is- called the "red nyererei group", species are lands have medium steep slopes and very joined that share a rather broad head (big different rock substrate (see the chapter interorbital width) and a medium long lower about habitats). The two populations differ jaw. Many of them are shoaling inhabitants anatomically, those of Mabibi Islands being of either the water column above the rocky more elongated and having a shorter head substrate, or of wide and deep gaps among than those from Zue Island, but they are al- big rock boulders. However, also the red most identical in their brilliant male colora- form of H. "big blue" fell in the morphologi- tion. Snout, cheek, branchyostegal mem- cal analysis into this group, as did H. brane, gill cover, chest and anterior parts of "Miandere zebra", which I regard as close to flanks and belly are deep blood red, the H. "zebra nyererei". With the exception of posterior parts of the flanks and the caudal these, all species in this group share a dark peduncle are greenish, the dorsum yellow- body coloration, and bright yellow, orange ish. The dorsal fin is metallic blue with red or red fins. In the case of H. nyererei, the dots and streaks to entirely red. The caudal bright red includes the dorsal aspects of the fin is reddish, the anal fin blue around the body. I propose the following hypothesis egg dummies, and reddish on the spinous concerning the ecological significance of part. Females are yellowish, though not as this coloration pattern. All species with this yellow as those of H. "big blue". Both sexes type of coloration about which information have 6 to 7 (rarely 8) vertical bars on the from Scuba observations is available (H. flanks. Males differ from those of H. "all red nyererei, H. "black and orange nyererei", H. nyererei" and H. "red flank nyererei" by the "lemon fin nyererei", H. "orange dorsal different topographical arrangement of col- nyererei"), have ecological and behavioural ours on head and body, and by the blue and traits in common: all are predominantly red instead of orange coloured anal fin. plankton feeders, the females of which for- Moreover, the other two do not belong to age in shoals in the water column above the group of species with a long head and the rocks. Males have their spawning terri- long lower jaws. The body shape of H. "red tories on the rocky substrate, and have to head" is similar to that of H. "big blue" but approach and court their females from be- is usually less deep, and with a less steep neath. Females see territorial males mostly dorsal head profile. H. "red head" seems to from above. Conspicuously coloured body remain considerably smaller. outlines, in particular dorsal and caudal

107 Chamagati Island, habitat of Haplochromis "big blue".

A territorial male H. "big blue" from Chamagati Island.

A female H. "big blue" from Chamagati Is. A red male H. "big blue" from Matwinki Island.

108 A territorial male Haplochromis "red head nyererei" in the aquarium.

A male H. "all black nyererei" from Chamagati. A female H. "red head nyererei"

A male H. "red head nyererei" from Mabibi Is. A male H. "red head nyererei" from Zue Island. fins, but also the anal fin that is held for instance, flank coloration. I propose somewhat aside of the longitudinal axis the term "body outline coloration" in con- of the body during quiver display and lead trast to "body centre coloration". The lat- swimming, may in such situations be ter occurs in many species of the "zebra more "eye-catching" to the female than, nyererei" and the "blue nyererei groups"

109 and a good example is H. "red head body depth, lower jaw length and width, as nyererei". well as details of male coloration. Fishes Haplochromis nyererei Witte-Maas & from Ngoma Point, Python, Nyamatala and Witte 1985 is the only described species Kissenda islands in the Mwanza Gulf are in the H. nyererei complex, and at the same deeper bodied than those from Nyegezi time one of the most conspicuous of all rocks (Mushroom Island is part of the Lake Victoria cichlids. It has a somewhat dis- Nyegezi rock peninsula), Anchor Island, junct geographical distribution, living in the Ascari Island, Capri Point, Igombe, Makobe Mwanza Gulf, at a few places along the and Ruti Islands. Males of the deep bodied southern Speke Gulf shore, and at two off- form have sky blue underparts, those of the shore islands in the Speke Gulf (Makobe other one have black underparts. Females and Ruti Islands). It is absent from other of the first are lighter yellowish brown than places at the southern shores of the Speke those of the second. Similar to the deep Gulf, as well as from the more eastern and bodied populations from some places in northern Speke Gulf islands, the northern the Mwanza Gulf is the population from Speke Gulf shore, and from the region west Senga Point in the central Speke Gulf. It is of Juma Island. However, at Gana Island separated from that of Kissenda Island by northwest of Ukerewe we collected fishes more than 40 km shoreline which at sev- that resemble it, and may well be H. eral places is inhabited by populations of nyererei. Furthermore, it has been reported the shallow bodied, more elongated type. even from the Kenyan lake shore (Kaufman We do not yet know whether such similari- & Ochumba 1993) and a photograph of that ties and differences among H. nyererei population has recently been published populations reflect population phylogenetic (DeMason 1996a). Judging from this photo, patterns or repeatedly evolved adaptations the Kenyan fishes differ from those from to local environmental situations. the southern lake shores, but may indeed H. nyererei lives in sympatry with H. "ze- belong to the same species. In the Mwanza bra nyererei" throughout its geographical Gulf south of Python Islands, populations range, with H. "black and orange nyererei", are found that are intermediate between H. H. "Bwiru nyererei", H. "red rim anal", and H. nyererei and H. "zebra nyererei" (see below). "pink anal" at Igombe Island, with H. "pink H. nyererei is morphologically similar to anal" also at Makobe Island, with H. "blue H. "zebra nyererei", the most conspicuous nyererei" at Ruti Island, and with the red difference being male coloration. Males ex- form of H. "big blue" at Juma Island. H. cf. hibit a bright orange-red to crimson-red nyererei of Gana Island coexists with H. head, dorsal fin and dorsum, the red colour "lemon fin nyererei" and H. "all red nyererei". extending downwards on the flanks to be- H. nyererei and H. "zebra nyererei" occupy low the lateral line, a lemon yellow ante- different microhabitats, though at some rior flank region and black or blue (depend- places with considerable overlap. It has ing on the population) lower jaw, cheeks, been shown that they mate assortatively gill cover, and lower side. Females are light (females of the red form mate with red brownish. Both sexes have 6 (rarely 5) to 8 males, those of the "zebra" form with "ze- vertical bars on the flanks. The description bra" males) and behave as two biological of H. nyererei (Witte-Maas and Witte 1985) species over much of their common geo- was based on fishes from two geographi- graphical range. However, in the southern cally nearby and morphologically similar Mwanza Gulf, south of Python Islands, they populations, Mushroom and Kilimo Islands seem to behave as two interbreeding col- in the northeastern Mwanza Gulf. Now, that our morphs of a single species and individu- H. nyererei has been collected from many als of all intermediate colour types exist other places, we know that morphological (Seehausen 1996, abbreviated cf. Zny x variation between populations is consider- Nye in the distribution map). At Makobe Is- able in both sexes, particularly regarding land entirely black males are frequently

110 found among the normal red ones (see metres in an hour. The plankton photo). Because intermediate phenotypes (copepod) density was so high that a fish also occur which are black with some red would almost permanently, and with on the dorsal head surface, dorsum, or dor- hardly any locomotion, perform the cyclus sal fin, and because no ecological differ- of sucking-in, processing, and swallowing. ences between the morphs have been ob- During plankton grazing, individuals kept served yet, I consider the black and the red some distance to each other but closed males as colour morphs of one species. up, as soon as the shoal was disturbed However, behavioural experiments, like and moved on. Adult reproductively ac- those done with H. nyererei and H. "zebra tive males become territorial, and at nyererei", are desirable to confirm this. Makobe Island defend areas of about 1.5 H. nyererei populations vary also eco- m diameter. These territories are fre- logically. At most places H. nyererei is pre- quently centred around rocks that are dominantly zooplanktivore, however, at slightly larger than the average, and that Ruti Island it is predominantly phyto- are surrounded by also slightly deeper planktivore and at Anchor Island crevices. Brooding females do not stay benthivore, eating mostly insect larvae inshore in shallow waters like the brood- and prawns (Bouton et al. submitted). It ing females of many Neochromis and inhabits predominantly slightly deeper Xystichromis do, but keep to deeper parts waters below 2m depth and has been ob- of the rocky shores below 4 m depth served down to 14 m depth at Ruti Island where they defend small territories. in the Speke Gulf. At steeply sloping These consist usually just of a crevice be- shores it occurs also in water shallower tween some rocks. Such territorial fe- than 2 m immediately inshore, and even males are very dark brown. In spite of in crevices among the rocks at the surf their small size (often no more than 5 cm line. At moderately steep shores it can be SL) they are very aggressive and success- confined entirely to the deeper parts of ful defenders of their offspring. the rocky habitat, and at very gently slop- Anatomically very similar to the elongated ing shores it is absent, possibly because form of H. nyererei is Haplochromis "black at such places rock substrate makes way and orange nyererei"nyererei", which is very differ- for sand before a suitable depth is ent in coloration. Males are deep black or reached. At most localities, H. nyererei is brown-black, with anal, caudal, and often among the numerically dominant haplo- pelvic fins bright orange, and the dorsal fin chromines in its major habitat and seems black with orange lappets. They can show to dominate other planktivorous species a trace of orange on the neck. Females are where they coexist. Females, subadults, dark and more grey-brown than H. nyererei and nonbreeding males of open water females. This species may grow a little big- dwelling (plankton eating) populations for- ger than the elongated form of H. nyererei. age in shoals, sometimes directly above Many males have filament-like extended the rocks, sometimes several metres pelvic fin rays. The lower jaw width is very higher in the water column, probably fol- variable within populations. We know H. lowing the highest plankton densities. At "black and orange nyererei" from three Makobe Island I had the opportunity to places at the southern mainland shore of observe such foraging shoals in some de- the Speke Gulf. It resembles H. nyererei tail at 4 to 6 m depth. They consisted of in its ecology. We found it at moderately ten to over a hundred individuals that var- steep and steep shores with medium ied relatively little in size. I saw shoals sized to very big rock boulders. At steep containing only fishes of 3 to 5 cm SL and shores it inhabits already immediately in- others with only larger individuals. When shore waters where it lives in deep and grazing plankton, the shoals were station- wide gaps between big boulders, while ary and moved often no more than a few at the moderately steep shore of Igombe

111 A territorial male Haplochromis nyererei from Igombe Island.

A male H. nyererei from Python Island. A female H. nyererei from Ruti Island.

A male H. nyererei from Makobe Island An orange male H. nyererei from Igombe Is.

112 A male H. nyererei from Senga Point. A male H. nyererei from Nyegezi Rocks.

A black male H. nyererei from Makobe Island. A male H. nyererei from Anchor Island.

A male H. "black & orange nyererei" from Ndurwa Point.

113 Island it is confined to slightly offshore ar- we caught four males of a colour morph eas of beyond 2 m depth. Males occupy with wine red instead of orange fins that territories at 3-4 m water depth and otherwise resemble the orange fin males deeper, among big boulders, from where in morphology but are similar in colora- they court bypassing or foraging shoals tion to H. "red anal nyererei" of the Vesi of females. The stomachs of four males archipelago in the central Speke Gulf. (two of the orange fin morph, two of a red Populations that differ considerably in fin morph) from Igombe Island that we general appearance and morphology from examined, contained unexpected items. H. "black and orange nyererei", but whose Two individuals (one of each morph, see male coloration is similar to that of the lat- below) had fed almost exclusively on ter, occur far away from the range of H. filamentous green algae and some blue- "black and orange nyererei" at steeply slop- green algae. The other two had small ing big boulder shores in the Sengerema amounts of algae as well, but predomi- region (Bihiru and Miandere Islands: H. nantly detritus consisting of macerated "orange anal nyererei", H. "small mouth macrophyte tissue. No traces of zoo- nyererei") and at Gana Island north of plankton were found. From Scuba work I Ukerewe (H. "lemon fin nyererei"). These know that the plant material is of rotting similarities may well be due to parallel water hyacinths (Eichhornia crassipes) that evolution of male nuptial coloration under form many centimetre thick layers at the similar environmental light conditions. southern Speke Gulf rocky mainland This in particular since a very similar col- shores since about one to two years ago. our pattern also occurs in species of the Before 1995 the rocks at these shores Haplochromis "pseudonigricans" complex were rather free of detritus. (see below). Future research must show The very close anatomical and ecologi- whether this hypothesis holds true. cal resemblance between H. "black and The two populations of black and orange orange nyererei" and H. nyererei sug- coloured fishes from the Sengerema is- gests that the two are closely related. lands Bihiru and Miandere fell into one mor- They may have evolved as geographical phological group with H. nyererei and H. vicariants. However, their current distribu- "black and orange nyererei". Nevertheless, tion is not entirely exclusive. At Igombe they are in general appearance and lower Island, both species live in sympatry. The jaw length/width ratio not only that much red H. nyererei is there much less abun- different from each other, that I consider dant than "black and orange" but its abun- them two different species, but Haplochro- dance seems to either have increased mis "small mouth nyererei" of Miandere Is- very recently, or to undergo considerable lands differs from all other H. nyererei com- fluctuations over several years. Some of plex species in general appearance. It is the red H. nyererei males that we found very deep bodied and does with its small there seem to bear some characteristics snout resemble species of the H. "pseudo- typical for H. "black and orange nyererei" nigricans" complex. Another peculiarity of (more orange than red colour, fila- this species is the large number of egg mentously extended pelvic fins). Thus, dummies that occupy much space on the occasional hybridization between the two anal fin of the male. Males are black with a forms cannot be excluded. However, bright yellow anal fin, yellow dorsal fin since both forms clearly maintain their lappets, and a caudal fin that is orange col- phenotype in sympatry, I consider them oured in its distal most third. Females are as two species. H. "black and orange brown with about seven narrow vertical nyererei" furthermore lives sympatrically bars on the flanks. At Miandere Islands H. with H. "zebra nyererei", H. "Bwiru "small mouth nyererei" lives at water depths nyererei", H. "pink anal" and H. "red rim ranging from 3 to at least 6 m. It lives in anal". At Igombe Island and Ndurwa Point sympatry with H. "Miandere zebra" and H.

114 "orange dorsal nyererei" as well as with a Nansio Island, a rocky hill, separated from population of H. "yellow chin pseudo- the main Ukerewe Island by water of just nigricans" that has a male coloration similar about 20 m width, this species lives be- to that of H. "small mouth nyererei". tween 3 and at least 5 m depth over me- At Bihiru Island, separated from the dium sized and big boulders, while rocky Miandere Islands by about 10 km of maxi- crevices in shallow water are inhabited by mum 40 m deep water, lives Haplochromis H. "zebra nyererei". At Bwiru Island H. "orange anal nyererei"nyererei". In coloration it hardly "Ukerewe nyererei" lives at depths between differs from "small mouth nyererei". Males less than 1 and at least 7 m. Its population are a bit more blueish, the fin colour is density is particularly high at about 4 m more orange than yellow and the egg dum- depth, in a partly surf exposed area. It is fre- mies are fewer on the anal fin. The vertical quently encountered also in rocky crevices, bars on the flanks, particularly of females, together with H. "zebra nyererei". The lat- are broader than those of "small mouth ter species is at Bwiru Island much more nyererei". Occasionally males are encoun- rare than at Nansio. tered with an anal fin that is black and has only a narrow orange coloured rim. The Plankton eating Nyererei species is with up to 125 mm SL one of with a short and broad head the largest members of the complex. It lives in crevices among big rock boulders A third morphological group, "lemon fin at only one metre depth, as well as at the group", consists of only two species, H. steep slope outside of crevices down to at "lemon fin nyererei" and H. "orange dorsal least 9 m depth. It is the second most abun- nyererei", that have a very short and very dant haplochromine species at the steep broad head (large interorbital width) in com- shores of Bihiru Island (after H. mon. Judging from male coloration ("body (Neochromis) "Bihiru scraper"). H. "orange outline coloration") they are very close to anal nyererei" lives sympatrically with H. "ze- the group with H. nyererei and H. "black and bra nyererei" that, however, is restricted to orange nyererei". However, since they are crevices in shallow water, and with H. cf. also ecologically similar to these species, "orange dorsal nyererei" that is very rare at living in very similar microhabitats, the re- Bihiru and was collected only once in wa- semblance in male coloration may as well ter of about 7 to 9 m depth. be due to parallel evolution. Haplochromis The last species in this morphological "lemon fin nyererei" lives at the island ar- group is Haplochromis "Ukerewe nyererei"nyererei". chipelago northwest of Ukerewe, where This is a species of the southern Ukerewe we found it at Gana Island. Possibly it is area, known from Ukerewe Island itself more widely distributed along the western (Nansio) as well as from Bwiru Island, that shores of Ukerewe and/or further north. in ecology (i.e. microhabitat distribution) Reaching more than 130 mm SL in the and morphology closely resembles H. male sex, it is the largest species known in nyererei. In male nuptial coloration it is quite the H. nyererei complex. Adult males are different from the latter and more similar of a very distinct appearance. They are deep to H. "zebra nyererei" (with which it lives in bodied and have a relatively steep dorsal sympatry) and H. "black and orange head profile, comparable to that of big H. nyererei". Males are dark metallic blue with "big blue" males. With this, and with a rela- black vertical bars on the flanks. The anal tively broad lower jaw, they are morphologi- fin, outer third of the caudal fin and dorsal cally very different from H. "black and or- fin lappets are either crimson or orange col- ange nyererei", a species to which "lemon oured. Females are grey-brown with metal- fin" is close in coloration. Males are grey lic silvery iridescent spots on the flank with six dark vertical bars to entirely black. scales. Both sexes bear, like in H. nyererei, The anal fin, almost the entire caudal fin, 6 to 7 narrow vertical bars on the flanks. At and the lappets of the dorsal fin are

115 lemon yellow. Females are grey-brown Island. There it lives in large shoals of fe- with six to seven vertical bars on the males and subadult males in deep can- flanks and have lemon yellow dorsal fin yons, steep "wells" and several metre lappets as well. deep and wide caves among big to huge "Lemon fin" inhabits in rather high abun- rock boulders. As a diver, one usually sees dance the fantastic rock gardens of Gana these shoals foraging on plankton, for

A male H. "small mouth nyererei" (Miandre). A female H. "small mouth nyererei" (Miandre).

A male H. "orange anal nyererei" (Bh). A female H. "orange anal nyererei" (Bh).

A male H. " Ukerewe nyererei" (Bwiru Is.). A female H. "Ukerewe nyererei" (Bwiru Is.).

116 which they disperse over several metres depth in the water column of "wells" and canyons, sometimes occupying a whole "well". While plankton grazing, the individuals in these shoals keep several centimetres distance from each other, and remind the observer of the plankton grazing shoals of Neolamprologus brichardi in Lake Tanganyika. When disturbed by a diver the shoals A male H. "orange dorsal nyererei" from Miandere Islands. withdraw into the large caves that are characteristic for their habitat. Adult and reproductively active males live differently: they defend territories around rock holes, in front of which they display upwards into the water column to the shoals of females that forage above them. Thus their behaviour is very similar to that described above for H. nyererei. Like the males of the latter, do also "lemon fin" A male H. "lemon fin nyererei" from Gana Island. males "jump" up into the lower parts of female shoals, to perform a lateral display and a gether with it at Miandere, this species quiver, and return to their cave in lead has a short, and yet narrow lower jaw, swimming position. When observing this and thus a small mouth. The "body out- courtship dance from above, a diver sees line coloration" of this species is reduced not more than yellow flashes, flickering to a broad orange band in the dorsal fin, around the female shoals. H. "lemon fin orange caudal fin corners and a pale yel- nyererei" lives at water depths from 2 to lowish white anal fin edge. The orange at least 5 m. The territories of the males dorsal band is so vividly coloured that it are at depths of 4 m and beyond. The appears almost fluorescent. The body is species lives sympatrically with three not entirely black as in H. "black and or- other species of the H. nyererei complex: ange nyererei" and similar species, but H. cf. nyererei, H. cf. "zebra nyererei" and more grey-brown, with usually five broad H. "all red nyererei". black vertical bands on the flanks. At Haplochromis "orange dorsal nyererei" Bihiru and Miandere Islands "orange dor- lives at the Sengerema islands Miandere sal nyererei" lives at open steep slopes and possibly Bihiru (identification of the with big and huge rock boulders. At Bihiru fishes from Bihiru has still to be con- it was collected only once at about 9 m firmed). It may be more widely distributed depth. At Miandere we collected adult west of the surveyed area. Similar to H. males in nets at depths of circa 4 to 6 m, "small mouth nyererei", which lives to- and observed subadults and females in

117 mixed shoals together with H. "yellow cal bars on the flanks. Males from Ruti Is- chin pseudonigricans" which have a very land are generally darker blue-black col- similar coloration. When these shoals are oured but otherwise like those from the seen from above by a diver, it is often only other islands. The population from Sozihe the orange dorsal bands that are visible Island differs from the others mainly in male like fluorescent light against the very dark coloration. These fishes are rather dull dark coloured rocks. H. "orange dorsal grey to black, the lips often being the only nyererei" lives sympatrically with H. blue part of the body. The red colour of the "Miandere zebra" and H. "small snout anal fin is frequently more intensive. Some nyererei". big males from Sozihe Island have a "zebra nyererei"-like general appearance. Typical H. Nyererei with narrow heads "zebra nyererei" have not been found at and short jaws Sozihe Island, and I cannot exclude the possibility that they form here an introgressive The last morphological group within the population with H. "blue nyererei" (compare H. nyererei complex, let it be called the with the situation at Senga Point, page 103). "blue nyererei group", consists of species Although one of the smallest species in the that share short lower jaws and a rather nar- H. nyererei complex, males of "blue nye- row head (small interorbital width). Most rerei" have in the outer tooth row rather seem to be substrate oriented feeders. stout, widely spaced unicuspid teeth which Judging from general appearance and male are often recurved. Small males and fe- coloration this group again can be split into males sometimes have some weakly bicus- two. One has reddish males, the other one pid teeth. blueish to blackish males. Endemic to is- Little is known about the ecology of H. lands in the central Speke Gulf is "blue nyererei". It lives at moderately steep Haplochromis "blue nyererei"nyererei". Males from to steeply sloping large boulder shores, Vesi and Mabibi Islands are dark metallic where it inhabits crevices between the blue with a bright metallic blue dorsal fin, rocks in shallow water, as well as the open and a proximally blueish, but distally often slope down to greater depths. Its micro- pale reddish caudal fin. The anal fin is proxi- habitat distribution differs among islands, mally black, followed by a pale silvery iri- and circumstantial evidence suggests that descent zone and is distally faint red. The these differences in microhabitat distribu- egg dummies are few and pale yellow. The tion are caused by some sort of interfer- anal fin coloration helps to distinguish "blue ence with other members of the H. nye- nyererei" from the sympatrically and partly rerei complex. At Sozihe Island H. "blue syntopically living, rather dull coloured Vesi nyererei" is abundant in all microhabitats, in Islands population of H. "zebra nyererei". rock holes and on the slope from shallow The latter has a red anal with often many water to at least 5 m depth. It is the only orange coloured egg dummies. Further- species of the H. nyererei complex at more, its flanks are more brownish rather Sozihe. At Mabibi Islands females and than blue. Females of H. "blue nyererei" are subadult males are among the most abun- light brown with iridescent blue spots on dant fishes in the shallow littoral of 1 to 2 the flank scales and have a greyish-yellow m depth over slabs and among boulders, anal fin. They are more elongated than the but reach down to at least 12 m depth. females of H. "zebra nyererei" and have a Adult males live between 2 and at least 12 characteristic head shape. Their snout ap- m depth and are there by far the most abun- pears (compared to that of other Nyererei) dant cichlids. The only other species of the too short for the slenderness of body and complex at Mabibi islands is H. "red head head, and gives the impression that the fish nyererei", which is quite rare and numeri- has got a "blow on its nose". Females usu- cally dominated by H. "blue nyererei". In the ally bear five, males six rather broad verti- Vesi archipelago, "blue nyererei" is the nu-

118 merically dominant haplochromine at many sembled the Ruti Island population, im- small, steeply sloping islets. It lives there plying that the species occasionally in the crevices among big boulders, and at reaches mainland rock shores but does the slope of these boulders, between 1 not establish populations there. and at least 6 m depth. In contrast, it is rare Territorial males of H. "blue nyererei" at at the moderately steep slopes of the main Ruti Island court females in much the same island. At the steep islets it coexists with manner as described above for H. nyererei. H. "zebra nyererei" but this one is restricted Females of both species live in mixed to crevices, and is even there outnumbered shoals. While plankton grazing, these shoals by "blue nyererei". In contrast, the moder- stay usually about 2 to 3 m above the ately steep slope of the main island is substrate and are courted by males of both densely populated by H. "red anal nyererei". species. At 10 m depth the metallic blue At Ruti Island H. "blue nyererei" is restricted dorsal fin of H. "blue nyererei" appears black to depths beyond 8 m. The crevices in shal- as does the body of the fish, but the pale low water at Ruti are populated by H. "ze- anal fin and caudal rim appear light whitish bra nyererei", and the open slope by H. and are well visible. nyererei that occurs in great abundance In the Speke Gulf west of Ruti Island and from the shallows to about 12 m depth. H. in the Sengerema region H. "blue nyererei" nyererei and H. "blue nyererei" seem to have is replaced by a similar species, Haplochro- similar microhabitat demands for the estab- mis "pink anal"anal". Though it differs just slightly lishment of territories. Both species settle from "blue nyererei" in the morphometric at Ruti Island in great density on the only measurements that I took, large males of slightly sloping rock rubble ground that ex- this species are not readily recognized as ists at the foot of the steep cliff, from about members of the Nyererei complex. This is 10 m depth downwards. However, territo- probably due to the steep dorsal head pro- ries of H. nyererei are also found at lesser file and the small eyes. Due to the combi- depths among the boulders and this spe- nation of morphological analysis, analysis of cies numerically dominates on the rubble distribution patterns and behavioural obser- shelf down to 11 m depths, while below vations by Scuba, it became apparent that that the ratio turns in favour of H. "blue "pink anal" is a member of the Nyererei nyererei". Territorial males of the latter oc- complex and probably closely related to cur down to at least 14 m depth. Territorial "blue nyererei". The recent discovery of the density at Ruti Island is very high and terri- much smaller females confirmed this clas- tories are relatively small, those of H. sification. nyererei seem to be larger than those of Young males can be blueish but full H. "blue nyererei". At Mabibi Island territo- adults are deep black and have a very char- rial H. "blue nyererei" males are observed acteristic anal fin coloration that is unique already in 6.5 m depth. Their territorial den- among the known Mbipi species, but sity is little compared to that at Ruti Island similar to that of "blue nyererei". The anal and their territories are either not border- fin is bright pink between the hard ing each other or are larger than at Ruti Is- spines. Behind the spinous part of the fin land, all indications of lower competitive the pink gradually fades away making pressure. These observations indicate that place for white with a faint pink flush. The the restriction of H. "blue nyererei" to very small yellowish egg dummies are situ- deep waters at Ruti Island may be ex- ated along the interface between the plained by its competitive inferiority to H. white distal part of the fin and the dark nyererei. Competitive interference with grey fin basis. The caudal fin is blackish other species may also explain the ab- with a pink margin of irregular width, the sence of "blue nyererei" from mainland black dorsal fin has pink lappets. shores of the Speke Gulf. We once H. "pink anal" inhabits mostly gently caught one male at Igombe Island that re- sloping shores with medium sized to

119 small rock boulders. At islands and in main- tat of "pink anal", the pale pinkish-white anal land habitats in the Sengerema region we fin is well visible, just like the pale fin of caught most of our "pink anals" in shallow "blue nyererei" at Ruti Island described water of less than 2.5 m depth, at Igombe above. Island and nearby mainland in the Speke I observed females and subadult males Gulf at depths between 2.5 and 6 m. At of "pink anal" foraging on plankton in shoals, Makobe Island, where we could study the mixed with H. nyererei, several metres species in some more detail, it is also found above the bottom. Of seven males that we in low densities in shallow water but it be- checked, four had eaten almost exclusively comes suddenly abundant at depths be- zooplankton (copepods), one exclusively yond 4 m. Scuba observations revealed that may fly (Ephemeroptera) larvae, one almost "pink anal" lives at Makobe Island a life very exclusively filamentous green algae similar to that of "blue nyererei" at Ruti Is- (Cladophora) and the last one detritus, may land, about 10 km further east. Syntopically fly and chironomid larvae. Only the last two with H. nyererei, males defend spawning had sand grains in their stomach, reflecting territories on the weakly sloping rock bot- their benthic feeding. Thus the food de- tom. While H. nyererei numerically domi- mands of "pink anal" seem to be quite simi- nates down to 5.5 m depth, the ratio lar to those of H. nyererei and competition changes in favour of H. "pink anal" beyond between the two species appears possible. 5.5 m. In the relatively dark spawning habi- Superficially "pink anal" can easily be

A male H. "blue nyererei" from the Vesi Islands. A female H. "blue nyererei" (Vesi Islands).

A mixed school of H. nyererei and H. "blue A male H. " blue nyererei" from Mabibi Is. nyererei" at a depth of 14 metres at Ruti Island.

120 confused with H. (Xysti- chromis) "copper black", with whom it lives sympatrically throughout its known geographical range. However, it differs from the latter in coloration and dentition. The superficial resemblance with "copper black" is also the reason for the late discovery of Haplochromis "red rim anal"anal", a form that is anatomically much more similar to "pink anal" and "blue nyererei", and inhabits the rocky mainland shores of the southern Speke Gulf. It is the largest of the three species and the one that gets the deepest body and steepest dorsal head profile. Large males are deep grey to blue-black with about seven black vertical bars on the flanks. All fins are black, the dorsal has red lappets and the caudal a narrow red edge that is more regular than in the other two species. Most distinct is the anal fin that is grey to black with a very sharply demarcated deep red rim along its distal mar- The southern islet at Mabibi Islands, the habitat of H. "blue nyererei" and H. "red head nyererei".

A male H. "pink anal" from Makobe Island. A male H. "pink anal" in the aqaurium.

121 gin. With this coloration the anal fin can al- played a role in the evolution of the sibling most certainly not function as the "light spot" trio and that the ecological impact has partly in deep waters, that the pale silvery metal- been masked by subsequent expansion of lic or whitish anal fins of the two similar geographical and ecological ranges. species are. Indeed does "red rim anal" live Though the three forms rarely occur in quite different habitats. It is found at gen- sympatrically, all three have been recorded tly to steeply sloping places with small to from Igombe Island. big boulders and lives at water depths be- Morphologically very close to H. "blue tween 2 and 6 m. At Ndurwa Point we nyererei" is also Haplochromis "red anal found a particularly high population density nyererei"nyererei". It is known only from the main of this species at about 3-4 m depth among island and some islets in the Vesi archi- large rock boulders, were it co-dominates pelago where it lives sympatrically with H. the haplochromine community, alongside "blue nyererei" but inhabits a different with H. (X.) "copper black" and H. "black and microhabitat (see above). It has, like the lat- orange nyererei". Scuba observations in that ter, a rather elongated body shape but adult habitat did not reveal evidence for shoaling males are black with an entirely bright red behaviour as displayed by "blue nyererei" anal fin, a red caudal fin rim and red dorsal and "pink anal". fin lappets. Distribution and brightness of H. "red rim anal" and the sympatric H. (X.) the red colour is reminiscent of the "body "copper black" can be told apart by male col- outline coloration" in species of other mor- oration (H. "copper black" males have the phological groups, however, it is unlikely largest part of the anal fin red) as well as that in deeper water the relatively dark red by morphology (the lower jaw is broader in has a similar effect. Males in nonbreeding the latter, the teeth subequally rather than dress have about six vertical bars on the weakly bicuspid and unicuspid, and the in- flanks. Though males are quite common, ner rows are more; see table 3). females have not yet been found. H. "red It seems likely that the three species H. anal nyererei" inhabits predominantly mod- "blue nyererei", H. "pink anal" and H. "red rim erately steep slopes with medium sized anal" are derived from one common ances- rock boulders, and depths between 3 and tor, although between the first and the third 5 m. It is completely absent from some one the assumed close relationship be- steeply sloping large boulder islets. A comes evident just when the second one morph with red fins that is similar to H. "red is considered as an, in some characters, in- anal nyererei" in coloration, frequently oc- termediate form. The major argument for curs among H. "black and orange nyererei" considering the three, largely allopatric at Igombe Island and Ndurwa Point. It forms specifically distinct, is that the varia- agrees in morphometrics with "black and tion among them in morphology and col- orange" and not with "red anal", and may ei- oration is geographically discontinuous. The ther represent a morph of H. "black and or- forms of Vesi, Mabibi and Ruti islands are ange nyererei" or populations of H. "red anal as unambiguously assignable to "blue nyererei" that, living under the same condi- nyererei" as are those of Makobe Island and tions, have converged to the morphotype the Sengerema region to "pink anal" and of "black and orange". those of the southern Speke Gulf mainland The fifth species in this group is Haplo- shores to "red rim anal". The forms are not chromis "Bwiru nyererei"nyererei". With its elon- connected by a cline along a geographical gated body and short head it deviates from gradient, neither do they represent pheno- the typical H. nyererei shape. The head types along an ecological gradient. There shape of adult males tends to resemble is no consistent correlation between habi- that of some algae scrapers with a slightly tat characteristics and phenotype expres- incurved dorsal head profile. Males are sion. Nevertheless, it is possible that adap- rather dull blue-grey, with variable amounts tation to different ecological situations has of faint red in anal and caudal fins. Females

122 are more conspicuously coloured, being dens of Gana Island northwest of yellowish brown to brassy on body and fins, Ukerewe were we found it in open wa- and often have yellowish lips. Both sexes ters among big rock boulders at depths carry about six vertical bars on the flanks. between 3 and 7 m. A single individual From H. "blue nyererei" it differs in male and of a very similar form has been collected female coloration, general appearance and at Ndurwa Point in the southern Speke by having a bigger mouth (longer and Gulf. H. "red flank nyererei" is known only broader lower jaw relative to the head from Nansio Bay. At the small Nansio Is- length). H. "Bwiru nyererei" was named af- land it lives in medium boulder size habi- ter the place of its discovery, Bwiru Point, tats between 1.5 and at least 5 m water the peninsula between Speke Gulf and depth. Both forms are not very abundant Mwanza Gulf entrance. Meanwhile we in their habitats. "All read" lives know it also from several other places along sympatrically with H. cf. nyererei, H. cf. the southern Speke Gulf shoreline. It is ab- "zebra nyererei" and H. "lemon fin sent from the offshore Speke Gulf islands. nyererei". "Red flank" lives sympatrically It lives in shallow water, between the surf with H. "zebra nyererei" and H. "Ukerewe zone rock boulders, and in deep rock pools, nyererei". together with H. "zebra nyererei" and down Apart from the red H. cf. nyererei, a few to at least 2.5 m depth. H. "Bwiru nyererei" other H. nyererei group species are lives in sympatry also with H. "pink anal", H. known also from northern waters of Lake "red rim anal", H. nyererei and H. "black and Victoria. A very beautiful species has re- orange nyererei". cently been imported to the US and Eu- The remaining two species of this mor- rope under the misleading trade name phological group are anatomically and in Haplochromis "nyererei red", and has coloration rather similar to each other and been introduced to hobbyists under the differ mainly in the lower jaw length/ name Haplochromis "crimson tide" width ratio that is higher in Haplochromis (DeMason 1996). Though of typical H. "red flank nyererei"nyererei", and in the distribu- nyererei appearance it is very different tion of red colour on the body: In males from red H. nyererei as described by of H. "red flank nyererei" red colour is Witte-Maas and Witte (1985) and defined most intensive on gill cover, chest and above. A better name, because describ- anterior flank while in Haplochromis "all ing a unique feature of the species, red nyererei" it is most intensive on gill would be "red snout nyererei". Males have cover, chest, dorsal head surface and dor- a light whitish grey dorsum, a dark green- sum. In both forms the lower body half ish to black underside and caudal pedun- behind the red area, including the caudal cle, and 6-7 black vertical bars on the peduncle, is greenish. The pelvic fin is flanks. The gill cover and the flanks be- black in the spinous, and red in the soft hind the pectoral fins, between the light part. The dorsal fin is metallic blue with dorsum and the dark underside, are bright red lappets and streaks that are more yellow to orange-red, interrupted by the prominent in "all red" than in "red flank". vertical bars. As a peculiar feature the The anal fin is bright orange coloured in snout is bright orange-red, as if dipped "all red" and red in "red flank". H. "all red into dye. Judging from my experience nyererei" has weakly bicuspid teeth in the with it in aquaria, the species may remain outer tooth rows. H. "red flank nyererei" rather small. Nothing is known about its can have unequally bicuspid teeth that life in nature. Considering its "body cen- are, for a species of the H. nyererei com- tre coloration" pattern I would postulate plex unusually closely spaced. Both forms that it is a bottom oriented rather than a have to be studied further to understand predominantly pelagic forager. A light the relationship between them. H. "all red blue coloured species of the H. nyererei nyererei" is known only from the rock gar- complex with about eight narrow vertical

123 Ndurwa Point, the habitat of H. "red rim anal". A male H. "red rim anal" (Ndurwa).

H. "red anal nyererei" at Vesi Is. A male H. "all red nyererei" at Gana Island.

A male H. "Bwiru nyererei" from Bwiru. A female H. "Bwiru nyererei" from Igombe Is.

bars on the flanks and a shape that vaguely reminds of H. "big blue" has many years ago appeared in the aquaria of European hobbyists (photos in Bleher in press). Details about its origin are not available, but most likely it had been sent out by exporters from Uganda or Kenya. A big, blueish species with steep dorsal head profile, that in general appearance resembles H. "big blue", has also been collected by staff of the Kenyan Fisheries Research Institute and A male H. "red flank nyererei" from Nansio Is. was exhibited in the aquarium of the Ken-

124 The beautiful "rock garden" at Gana Island at noon time. Four different species from the H. nyererei complex are found here.

Large heads — the Haplochromis "deepwater" complex

This is a small complex of species, that inhabit the deeper parts of rocky littorals, or crevices and holes in shallower water. They A male H. "crimson tide" in the aquarium. feed on a variety of food types, ranging from zooplankton to yan Natural History Museum. A male of epilithic sponges. None of the species is what might be the same species, was col- described. Anatomically the "Deepwaters" lected near Kisumu around 1989 (L. are very close to the species of the H. Kaufman pers. comm.), and was sent in a nyererei-complex, in particular to those of mixed shipment of Victorian cichlids to the "zebra nyererei group". Like those, most Germany (Seegers 1990). This patchy in- "Deepwaters" have a relatively long head formation indicates that the H. nyererei and long lower jaws. The complex can be species complex has a lake wide distri- characterized by a combination of the fol- bution and that many more species are lowing characters: straight to incurved dor- to be expected. sal head profile; unequally bicuspid to

125 unicuspid, acutely pointed teeth in the of the Sengerema region with its very gen- outer tooth row, that are in large males tly sloping shores. rather coarse, with recurved crowns; 2 to 5 Haplochromis "deepwater" was the first rows of inner teeth; a gap between the species in this complex to be discovered outer and the first inner tooth row; a lower by N. Bouton and myself in 1991. It is found jaw length of 40-42.4% of head length; a at many moderately steep to steeply slop- lower jaw length/width ratio of 1.4-1.9; a ing shores in the central and northern relatively narrow head with an interorbital Mwanza Gulf and has a scattered distribu- width of 22.6-23.8% of head length; small tion in the Speke Gulf. Though its southern- deeply embedded chest scales, and a faint most record, the Python Islands, are near melanin pattern of 5 to 6 vertical bars be- to the northernmost records of H. "blue tween pectoral fin and caudal end of the deepwater" and H. "yellow deepwater", it dorsal fin. Most characteristics overlap con- has never been found sympatric with any siderably with those of the H. nyererei- of these. The exclusive distribution and the complex and it appears plausible that the correlation between geographical distribu- "Deepwaters" are indeed very close to the tion and coloration, that resembles a simi- latter. Nevertheless, "Deepwaters" share a lar correlation in the H. "pseudonigricans" distinct appearance. Probably the best cri- complex (see below), suggest that H. teria to tell them apart from "Nyererei", are "deepwater" and the two southern forms the melanin pattern, the large egg dum- ("blue deepwater" and "yellow deepwater") mies, and the combination of a big, "heavy" are eco-geographical vicariants of one line- head with a narrow interorbital width. Fur- age. thermore, in some "Deepwater" species the Apart from the anal fin which can be sensory pores of the lateral line system on wine red or black and carries large to very the head are enlarged. They probably serve large yellow egg dummies, males of H. to locate prey in the dim-light environment "deepwater" are entirely velvet black. A me- of the "Deepwaters". Notwithstanding the tallic iridescent silvery-blue band in the close anatomical resemblance between spiny part of the dorsal fin helps to tell H. "Deepwater" and "Nyererei" species, a con- "deepwater" apart from the sympatric H. siderable resemblance in male coloration (Neochromis) "velvet black" and H. (Xysti- between the species of the "Deepwater"- chromis) "copper black" (that have, of complex, H. (Neochromis) "velvet black" and course, very different dentition). Females H. (Xystichromis) "copper black" deserves are dark brown and resemble in coloration mentioning (see photos). closely those of H. (N.) "velvet black". Male Taxonomy within the Haplochromis H. "deepwater" have a straight to slightly "deepwater" complex is still problematic. incurved dorsal head profile, that of females Currently I consider one species as rather is frequently slightly decurved. Though widely distributed (the matrix species of the rather variable in this respect, H. "deep- complex), while three others are known water" have strong jaws and somewhat only from the Mwanza Gulf. The pattern of thickened lips. Adult males in many popu- geographical distribution along the Mwanza lations are easily recognized by their body Gulf is very similar to that observed among shape: body depth decreases rapidly from species of the H. "pseudonigricans"-com- slightly behind the head to the caudal pe- plex (page 133): There is a consistently duncle, making it that the head appears par- sympatric species pair in the southern ticularly heavy. Among all known Mbipi, Mwanza Gulf, made up of one blue and one this species has the most prominent sen- yellow-red species, which from the central/ sory pores on lachrymal, preoperculum and northern Mwanza Gulf into the Speke Gulf dentary bones. is replaced by a black species. Like the H. "deepwater" lives at moderately steep "Pseudonigricans"-complex, the "Deep- and steep rocky shores and at most places water"-complex is absent from large parts at water depths from 2 to at least 10 m. Like

126 many other deep water dwelling Mbipi, it there a niche that is occupied by H. nyererei stays offshore at moderately steep shores, in the northern Mwanza Gulf. Of two indi- but can be found among inshore rock boul- viduals we checked stomachs. One was ders and in deep rock pools at less than 2 empty, the other one contained only uniden- m depth, where the shore drops off steeply. tifiable, mashed material that may have The microhabitat choice, deep water ver- been zooplankton. At most of its record lo- sus inshore rocky crevices, may depend calities H. "yellow deepwater" coexists with also on the presence or absence of H. "ze- Haplochromis "blue deepwater". However, bra nyererei" in the crevices. Where the lat- at the southern most rock shore that we ter is abundant, H. "deepwater" does not sampled (Buyago rocks in the Smith Sound), live in inshore crevices. The well studied we found only the yellow species. population that inhabits the Python Islands Haplochromis "blue deepwater" is mor- feeds predominantly on insect larvae, phologically very similar to H. "yellow epilithic sponges, ostracods, plankton and deepwater", though adults are usually some- the detritus layer covering the rocks at what deeper bodied, and give a generally greater depths (Bouton et al. submitted, more robust impression. Their dentition is pers. obs.). At Hippo Island, we found H. more typical for a benthic feeder, with more "deepwater" feeding also on small snails. inner tooth rows (see table 3). Males are It is unclear, how far the species´ absence metallic blue all over and in contrast to the from many places in the Speke Gulf is due other species in this complex, often lack to ecological factors. It is there absent from any red pigment in the fins. Their egg dum- many gently sloping shores and islands but mies are smaller than those of H. "yellow also from several steeply sloping large boul- deepwater". Females of these two species der habitats, like Ndurwa Point, Ruti and can hardly be told apart. The distribution of Vesi Islands, and one of its records from the H. "blue deepwater" resembles that of H. northern Speke Gulf (Zue Island) is from an "yellow deepwater", though it is absent island with gentle slope and small boulders. from the Smith Sound in the south and ex- Haplochromis "yellow deepwater" is tends slightly further north to Mashoro Bay known only from the southern Mwanza and the northern Luanso Bay. In contrast to Gulf, where it has been recorded at all sam- its yellow sibling, H. "blue deepwater" lives pled stations of rocky habitat south of, and inshore, in crevices and rock pools in shal- including Luanso Island. Males are yellow- low water. It is hardly ever found at depths ish grey to yellow with a reddish head, dor- beyond 2 m or more than a few metres off sum and dorsal fin. Characteristic is a me- shore. It therefore occupies in the south- tallic iridescent area on each scale, giving ern Mwanza Gulf a niche very similar to that the fish a silvery "pelagic" appearance. The occupied by H. "zebra nyererei" in the cen- caudal fin is transparent with many red- tral and northern Mwanza Gulf and Speke brown dots and streaks. The faint reddish Gulf. At an islet near Python Islands in the anal fin carries several, often large, pale yel- central Mwanza Gulf, we did not find H. "ze- low egg dummies. Females are greyish. H. bra nyererei" in rocky crevices, as on Python "yellow deepwater" has usually a straight Islands proper, but a population of cichlids dorsal head profile, and a bigger mouth similar to H. "blue deepwater". If these (longer and broader lower jaw) than its fishes prove to be H. "blue deepwater", this sympatric sibling, H. "blue deepwater". It species would live at least parapatrically lives slightly offshore, beyond 2 m water with H. "deepwater", the latter inhabiting depth, and is absent from the inshore rock greater depth. Information about the diet of pools and crevices that are inhabited by H. H. "blue deepwater" is not yet available, "blue deepwater". H. "yellow deepwater" is however, the combination of morphology usually the most abundant haplochromine and microhabitat suggests that the spe- in the deeper parts of the rocky habitat in cies lives mainly on insect larvae and the southern Mwanza Gulf. It occupies animalous epilithic Aufwuchs.

127 A so-called high energy shore at the south-east corner of Bwiru Island.

A male H. "deepwater" from Python Island. A female H. "deepwater" from Python Island.

The least known species in the "Deep- dividuals, male breeding coloration is not water complex" is Haplochromis "slender yet known. Available males were dull deepwater"deepwater". To date it has been recorded dark brown with a faint reddish sheen on only from Hippo Island in the northern the back and a reddish anal fin with or- Mwanza Gulf and just a couple of indi- ange coloured egg dummies. viduals have been collected. All were caught at the south-west corner of the island at depths between 6 and 9 m. They The Haplochromis "pseudonigricans" live sympatrically and syntopically with H. complex "deepwater", from which they differ conspicuously by a shorter head, more elon- This is a complex of anatomically rather gate body and more pointed snout. Also generalized Mbipi species, that are ana- the eyes are bigger. Since no brightly col- tomically not far removed from unspecial- oured male was among the collected in- ized "Astatotilapia"-like cichlids. Despite

128 this, it is my experience that most of its 30.2% of head length). The lower jaw is members, can on first glance be recog- not as broad as in Neochromis species nized as such. It is much more difficult, (length/width ratio 1.4 to 1.9, in one ex- however, to give a definition of the spe- ception 1.2), and usually shorter than in cies complex. It could be described as Nyererei species (36.9 to 39.1% of head taking a position between the Neo- length, in two species about 41%). The chromis-complex of algae scrapers and inner teeth are arranged in fewer (2-4), generalized insectivorous haplochro- and less broad bands than in Neochro- mines (e.g. Astatotilapia nubila). In many mis. The outer teeth are not contiguously measurements the H. "pseudonigricans"- set, but the interspaces between them complex broadly overlaps also with the H. are usually relatively small (not as wide nyererei-complex, though the species of as the width of the tooth neck). The the two lineages are of quite different shape of the outer teeth is less scraper- general appearance. like than in most Neochromis (unequally Some "Pseudonigricans" species have a bicuspid to unicuspid, rarely subequally steep, decurved dorsal head profile like bicuspid). Nevertheless, they are usually Neochromis species, others have a moveably implanted even in adult males, straight dorsal head profile and in most and not as coarse as in most "Nyerereis" species it is intermediate. It is usually not and "Deepwaters". The scales on the as shallow as that of many H. nyererei- chest are small and deeply embedded. complex species, and only very rarely The melanin pattern consists of 4 to 6 incurved. The head is usually narrower vertical bars between pectoral fin and (21-26.4% of head length) than that of caudal end of the dorsal fin, which are most Nyererei species, and the eyes are broader, and often less distinct than those bigger than in most other Mbipi (25.1- in Nyererei species. They are frequently

A male H. "blue deepwater" (Luanso Bay). A female H. "blue deepwater".

A male H. "yellow deepwater" (Luanso Bay). A male H. "slender deepwater" from Hippo Is.

129 broader than the interspaces between Nyegezi rocks (Mwanza Gulf). This is an area them, a situation that is very rare in Nye- that harbours an exceptionally high diver- rerei species. Females of most "Pseudo- sity within the Haplochromis "pseudonigricans" species are considerably much nigricans"-complex. We know from there at smaller than males, and frequently have least five sympatric forms, while the high- more scraper-like dentition, with closer est number elsewhere is three, and over set bicuspid teeth. large areas only one or two forms can be Overlap with other species complexes found at any one place. While some of the of Mbipi is considerable in most of the in- forms at Nyegezi rocks belong to more dividual characters that I measured, but widely distributed species (H. "black and clustering various Mbipi species based on yellow pseudonigricans", H. "long snout the combination of these characters, pro- pseudonigricans", H. "pseudoblue"), two duced a tree, in which 11 of 13 have never been seen anywhere else. One "Pseudonigricans-species" clustered to- of them is H. "large eye pseudonigricans", gether. This confirms the subjective im- the other one is H. "black pseudonigricans". pression, that we are dealing with a The situation is particularly difficult, since group of species that have more in com- we found intermediate phenotypes like mon with each other, than with species large eyed "black and yellow" pseudo- of other groups. nigricans", that suggest that occasional Also ecologically the species of the H. gene flow occurs among the narrowly en- "pseudonigricans"-complex occupy on aver- demic forms and H. "black and yellow age a position between specialized algae pseudonigricans". Nevertheless, material scrapers, more generalized insectivores and field notes, collected as early as 1978 and planktivores. Most species are either by Els Witte-Maas and Frans Witte, indicate epilithic Aufwuchs-feeders that feed pre- that these three forms have been persist- dominantly on animal components of the ing at Nyegezi rocks over quite some time. Aufwuchs, and to lesser extends on plank- Conform with the species recognition cri- ton and epilithic algae, or plankton feeders teria outlined in the chapter about tax- that forage, similar to H. nyererei, in the cur- onomy, I therefore consider these forms rents that prevail at the edge of rocky reefs, different species. Research on mate choice but consume also some Aufwuchs. Their and heritability of anatomical and coloration feeding behaviour consists mainly of snap- differences should proof this right or wrong. ping, picking and scraping (described by Males of H. "black pseudonigricans" can Seehausen et al. in press [a]). While the ma- on first glance be confused with H. (Neo- jority of the Neochromis, and some chromis) "velvet black". They have metallic Nyererei species inhabit shallow waters blue-black to velvet black, flanks, and at over gentle slopes, all known "Pseudo- some islets some purplish sheen on the nigricans"-complex species live either some- dorsum. The anal fin is reddish to grey what offshore at greater depths, or inshore transparent and carries small yellow egg at very steep slopes. In this respect they dummies. The caudal fin is grey with a red resemble species of the "Deepwater" com- flush to red. The dorsal fin carries a red rim plex, but it is more uncommon to find them and frequently red streaks in the soft part, in rock holes. though the red is not as prominent as in H. Haplochromis "black pseudonigricans" "blue pseudonigricans" and H. "large eye was the first form of the species complex pseudonigricans". We consider a form that to be discovered (Witte et al. 1992), and is is entirely grey with a weak metallic blue taxonomically one of the most difficult ones sheen on the flanks and transparent fins as of all. We know it only from a geographi- a morph of H. "black pseudonigricans", be- cally very restricted region, the rocks at the cause it does not differ anatomically from southern Nyegezi bay entrance and near- the blue-black form. H. "black pseudo- bye islets, collectively referred to as the nigricans" has a peculiar dorsal head profile.

130 It is relatively steep and straight, but with a among rock-dwelling Lake Victoria cichlids, small concavity, above the eye. As a conse- and known only from one other deep wa- quence, the premaxillary pedicels appear ter dwelling species (H. (Ptyochromis) prominent, giving the snout a tapering ap- "striped sheller"). "Large eye pseudonigri- pearance. It is with difficulty that females cans" has never been caught by angling rod, of the five sympatric species from the and seems to live slightly deeper and/or Nyegezi rocks can be assigned to the more offshore than H. "black pseudo- males. However, also females exist in five nigricans". Judging from its morphology it different forms. The type of females that I probably feeds upon animal Aufwuchs from believe to belong to the H. "black pseudo- the rocks. I consider this species closely re- nigricans" males, has a similar dorsal head lated to H. "blue pseudonigricans". It might profile (straight or "tapering mouth") and ex- be derived from a population of that south- hibits a blueish sheen on the flanks. ern species, that met at Nyegezi rocks with The habitat of H. "black pseudonigricans" another, (northern) ecologically similar spe- is characterized by steeply sloping, almost cies of the complex (H. "black and yellow vertically dropping huge rock boulders. Here pseudonigricans"), and as a result of com- it lives immediately inshore between 1 and petitive interaction with this, was pushed at least 4 m water depth. We never caught into a different ecological niche. However, it in nets that were set a few metres fur- this speculative hypothesis may prove dif- ther away from the shore. ficult to test. Haplochromis "large eye pseudonigri- The Nyegezi rocks are a meeting point cans" differs from the foregoing species between the southern and the northern anatomically and in coloration. Also from all group of the "Pseudonigricans" complex. other members of the "Pseudonigricans" From their fauna one can easily move on complex it differs by its small lower jaw via H. "black and yellow pseudonigricans" to length/width ratio that resembles that the northern group, and via H. "large eye found in Neochromis species. With the lat- pseudonigricans" to the southern one. In- ter it shares furthermore the steep and terestingly the "Pseudonigricans" species of slightly decurved dorsal head profile. Its the Mwanza Gulf display a geographical dis- very large eyes are comparable in size only tribution pattern in relation to male colora- to those of H. (Neochromis) "large eye tion that is very similar to the one described nigricans" and H. (Xystichromis) "large eye for the "Deepwater" complex (see page black" (table 3), and its large egg dummies 126). The southern and central Mwanza to those of "Deepwater" complex species. Gulf is inhabited by two sibling species, one Males have grey-blue to bright metallic blue red (H. "red pseudonigricans"), the other one flanks, and on the dorsum frequently a red- blue (H. "blue pseudonigricans"), while black dish flush. The dorsal fin is blue in most of male coloration dominates at the steeply the spinous and bright red in most of the sloping shores further north. soft part. The colour border is often sharp Most widely distributed in the south is and runs obliquely across the fin (see Haplochromis "blue pseudonigricans"pseudonigricans". It is photo). This dorsal fin coloration is basically currently known from each sampled station the same as that found in many individuals in the southern Mwanza Gulf between of H. "blue pseudonigricans". Like in the lat- Marumbi Island and Python Islands. We did ter, caudal and anal fin of "large eye not find it, however, in the Smith Sound, the pseudonigricans" are bright red. Different south-western extension of the Mwanza are the egg dummies. While orange col- Gulf. North of Python Island it is restricted oured in H. "blue pseudonigricans", they are to the western shore of the Mwanza Gulf, yolk yellow in H. "large eye pseudonigri- where it is known from Ngoma Peninsula, cans". Moreover they are bigger and com- Kissenda Bay and Hippo Island. This is a monly two or more flow into each other. conspicuously coloured species. Males are Confluent egg dummies are a rare feature bright metallic iridescent sky blue all over,

131 Nyegezi Rocks, habitat of several Pseudonigricans complex species.

A male H. "black pseudonigricans" (Ngoma). H. "large eye pseudonigricans" (Ngoma Point).

A male H. "blue pseudonigricans". A male H. "blue pseudonigricans" from Python Is.

132 with a bright blood red anal fin, caudal fin lower jaw than other populations, and fe- and soft dorsal fin. The red area in the dor- males are darker brownish. sal fin, often dissolved into streaks on the South of Python Islands, H. "blue pseudo- fin membranes, is usually sharply demar- nigricans" lives at moderately to steeply cated against the metallic blue anterior parts sloping shores. Its relative to the Mwanza of the fin. The egg dummies on the anal Gulf asymmetric geographical distribution fin are usually orange. At one place in the north of Python Islands correlates with dif- central Mwanza Gulf (Matumbi Island) we ferences in the rocky habitat between the found a rare colour morph with a red dor- steeply sloping eastern shore and the mod- sum, that was otherwise identical with the erately steep western shore of the north- blue males. Females are light yellowish ern Mwanza Gulf. Thus it seems in the brown with a silvery metallic sheen northern Mwanza Gulf to avoid steep ("pelagic" appearance). H. "blue slopes, which are inhabited there by H. pseudonigricans" is one of the deeper bod- "black and yellow pseudonigricans". At mod- ied species in the complex. Its dorsal head erately steep slopes adult H. "blue profile is straight, often slightly concave. pseudonigricans" usually stay several me- Most populations, particularly in the south, tres offshore, at steep slopes they can be have very regularly arranged, small and fine caught immediately inshore. The species teeth that are usually bicuspid. In the south- lives at depths from 1 (rarely) to at least 8 ern parts of its range, populations of H. m, and seems to frequently forage in the "blue pseudonigricans" from different is- open water column above, or next to the lands do not, or hardly differ from each rocks. other. Towards the north, however, more Of five individuals that we checked, two distinct morphological differences become had empty stomachs, one had eaten pre- apparent between populations. So does dominantly zooplankton, one caddis fly the population of Kissenda Island (northern larvae (Trichoptera) and one may fly larvae Mwanza Gulf) have a significantly longer (Ephemeroptera). None had any bottom

A territorial male Haplochromis "red pseudonigricans" in the aquarium.

133 material or epilithic algae in its stomach. yellowish to greenish. In less intensively Thus this species seems to forage pela- coloured males the red is restricted to the gically in the water column and by pick- lower chest. It is readily possible to tell this ing up insect larvae from the rocks with- species apart from the reddish morph of H. out scraping from the Aufwuchs. In feed- "blue pseudonigricans". The red in H. "red ing experiments in aquaria, it behaved re- pseudonigricans" is ventrally oriented and spectively and different from H. "black and does not reach the upper dorsum. In con- yellow pseudonigricans" that used to trast, the red in the red morph of its sibling scrape Aufwuchs (unpublished data). In species is dorsally oriented, and in fact re- parts of the geographical range of H. "blue stricted to the dorsum and the dorsal fin. If pseudonigricans", large zooplankton the red on the flanks of H. "red (Daphnia) can be very abundant. While it pseudonigricans" males entirely fades away, is south of Python Islands the most abun- they are still recognizable by the absence, dant haplochromine in deeper parts of the or but faint expression of red in the fins. If rocky habitats, H. "blue pseudonigricans" faintly present, this red in the fins is more is less abundant in the north-western wine-red, while it is blood-red in H. "blue Mwanza Gulf and very rare at Hippo Is- pseudonigricans". Moreover, H. "red land. pseudonigricans" has a red flush on the gill From Python Islands southwards, H. "blue cover, and its egg dummies are usually pseudonigricans" lives sympatrically with an smaller than in H. "blue pseudonigricans". anatomically extremely similar, but dramati- Some males have a part of the pelvic fins cally differently coloured species: H. "red red. Females of "red pseudonigricans" are pseudonigricans". In the northern part of its usually more brassy, and less silvery in col- range, it also lives in sympatry with H. our than those of "blue pseudonigricans". "pseudoblue", H. "long snout pseudo- Ecologically H. "red pseudonigricans" dif- nigricans" and H. "yellow chin pseudo- fers slightly from H. "blue pseudonigricans". nigricans". It lives more inshore, has rarely been caught Haplochromis "red pseudonigricans" and in nets set some metres away from the H. "blue pseudonigricans" appear to be sib- shore or in open water and never at depths ling species. Like the blue species, we beyond 4 m. As far as the limited data al- found also the red one at each sampled low to judge, the more littoral orientation place in the southern and central Mwanza is reflected in the feeding habits of "red Gulf. However, we found it also in the Smith pseudonigricans". Of three individuals that Sound (Buyago rocks), while it does not we checked, one had the stomach empty, reach as far north as the blue one does. This one had eaten only insect larvae (of may distribution pattern resembles that of the and caddis flies), and one exclusively small yellow and blue sibling species of the snails. The teeth of "red pseudonigricans" "Deepwater" complex in astonishing detail tend to develop a larger flange, thus a more (see page 126). scraper-like shape, suitable to take up food Between Marumbi Island and the Luanso items from a rock surface. Finally, the fe- Bay the two species are caught in approxi- male coloration does indicate that this spe- mately equal frequencies. At Python Islands cies lives more inshore and bottom ori- and at Ngoma rocks, the northernmost ented than its blue sibling. records of H. "red pseudonigricans", this Haplochromis "black and yellow species is much less abundant than the pseudonigricans" is known exclusively blue one. In fact it has not been caught from the steeply sloping rock shores of the there before 1995 in spite of extensive north-eastern Mwanza Gulf were it is rather sampling. Males of H. "red pseudonigri- abundant at the mainland as well as at is- cans" are bright blood red on the anterior lands. Its southernmost record is from the flanks, head and belly. The posterior third Nyegezi rocks, where it overlaps with a of the flanks, and the caudal peduncle are number of other species (see above). It is

134 morphologically (Table 3) very close to the detritus and fish (Bouton et al. submitted) large group of populations inhabiting islands and employs mainly picking, scraping and in the Speke Gulf and open lake, which I snapping as feeding techniques (See- consider populations of a single species, H. hausen et al. in press [a]). "yellow chin pseudonigricans". The major ar- More than any other known species of gument, to consider the north-eastern the "Pseudonigricans" complex, "black and Mwanza Gulf populations specifically dis- yellow pseudonigricans" is restricted to tinct, is a different female coloration (see habitats with large, very steeply sloping below). "Black and yellow pseudonigricans" boulders, where it lives immediately in- has nowhere been found sympatrically shore between 0.5 m and at least 5 m wa- with "yellow chins", but the Hippo Island ter depth and in crevices between the population of the latter is separated from shore rocks. It is usually found sympatrically the Gabalema Island population of "black with H. (Neochromis) "velvet black". These and yellow pseudonigricans" by less than two species are the numerically dominant 2 km of open water. Mbipi in most of the "steep-slope commu- Males are entirely deep black with dor- nities" of the north-eastern Mwanza Gulf. If sal fin lappets that are yellow in the spinous the hypothesis is correct that I formulated part and orange to red in the soft part, a rela- on page 110 to explain the function of body tively broad red caudal fin edge, and a outline coloration, "black and yellow bright yellow, sometimes orange anal fin pseudonigricans" might resemble some with small yellow egg dummies. This col- Nyererei species in a sexually dimorphic oration resembles in astonishing detail the ecology, with males defending territories body outline coloration of some species in on the rock bottom and females living so the H. nyererei complex (see page 113). Fe- to speak above the heads of the males. The males of H. "black and yellow pseudo- low water transparency in the Mwanza nigricans" are brownish with five broad ver- Gulf prevented underwater observations so tical bars on the flanks. They have a rather far. Females of "black and yellow steep, usually slightly decurved dorsal head pseudonigricans", however, in contrast to profile, and lack the bright yellow chin and those of Nyererei species with body out- throat that is characteristic for the females line coloration, do not appear to be shoal- of H. "yellow chin pseudonigricans". Fe- ing. Nevertheless, by Scuba I observed at males and subadult males of H. "black and Miandere Island in the open lake a popula- yellow pseudonigricans" can be told apart tion of H. "yellow chin pseudonigricans" from those of H. "blue pseudonigricans" by with body outline coloration, the behaviour the lack of the silvery iridescence on the of which is indeed conform to the hypoth- flanks (no "pelagic" appearance), and by the esis. dorsal head profile. From those of H. "large Under the name Haplochromis "yellow eye pseudonigricans" they differ in eye size. chin pseudonigricans" I unite a large With the decurved dorsal head profile and number of phenotypically similar, but geo- relatively broad lower jaw, H. "black and yel- graphically well isolated populations. low pseudonigricans" reminds of some They mostly inhabit offshore islands in epilithic algae scrapers of the Neochromis the open lake and in the Speke Gulf, but lineage. Also otherwise its morphology is, are largely absent from mainland shores. more than that of H. "blue pseudonigricans" Some share with H. "black and yellow and H. "red pseudonigricans", that of a fish pseudonigricans" the yellow anal fin of that removes firmly attached food items the males, in others it is red and several from a rock surface. So are the outer teeth populations are polymorphic in this re- stronger. The well studied population at spect. The small egg dummies are yellow Anchor Island feeds predominantly on in- to orange-red. The males from Miandere sect larvae, epilithic Aufwuchs (filamentous Island in the west of our survey area have blue-green algae, moss animals), prawns, a pale, yellowish white anal fin with or-

135 H. "red pseudonigricans" from Shadi Rocks.

Luanso Island, habitat of H. "blue pseudonigricans" H. "red pseudonigricans" from Ngoma Point. and H. "red pseudonigricans".

ange-red egg dummies. The body coloration of males is blue-green to green-black, at some islands almost black. Females are yellowish brown, similar to those of H. "black and yellow pseudonigricans" but are characterized by a bright yellow chin, brachyostegal membranes and throat. Like H. "black and yellow pseudonigricans" also these populations lack the metallic iridescence that gives species like H. "blue H. "black & yellow pseudonigricans" (Ngoma). pseudonigricans" and H. "pseudoblue" their "pelagic" appearance. The dorsal head profile of H. "yellow chin pseudonigricans" is less steep than that of H. "black and yellow pseudonigricans" and is often straight, rather than decurved. Its lower jaw is usually longer. These anatomical differences suggest that "yellow chin" is less A peppered H. "black & yellow pseudonigricans". substrate oriented in its feeding. Our

136 A male H. "yellow chin pseudonigricans" from Miandere Island.

H. " yellow chin pseudonigricans" (Miandere). H. "yellow chin pseudonigricans", a female.

A male H. "orange pseudonigricans" (Sozihe). A female H. "orange pseudonigricans" (Sozihe).

137 data on diet corroborate this: Of three males oration exists. The red anal fin morph domi- from Makobe Island, the stomachs of which nates, or is the exclusive one at places with we checked, one had eaten predominantly more gently sloping shores and smaller rock zooplankton, another one zooplankton, may boulders, the yellow one at places with fly larvae, and filamentous blue-green algae, steeper slopes and big boulders. The body and the third one filamentous blue-greens, coloration is darker, and changes from blue green algae, diatoms, and to a lesser extent to green-black, the steeper the rocky slopes zooplankton and insect larvae. At Ruti Island, in the habitat are. It resembles at the very I observed females moving in shoals, partly steep Ruti Island the coloration of H. "black mixed with H. nyererei, and foraging on and yellow pseudonigricans". Distinct from plankton. At Miandere Island, I observed all other known populations is the popula- similar mixed shoals with H. "orange dor- tion from Miandere Island with its aberrant sal nyererei". Usually H. "yellow chin yellowish-white anal fin and a small maxi- pseudonigricans" lives slightly offshore and mum size of about 8 cm SL. Its aberrant at depths beyond 3 m. At steeply sloping characters may have evolved in response islands, however, it can be observed more to coexistence with two ecologically simi- inshore, though rarely at depths of less than lar species of the Nyererei complex: H. "or- 2 m. At the gently sloping Makobe and Zue ange dorsal nyererei" and H. "small mouth Islands adults are restricted to the more off- nyererei". The latter species resembles in shore regions and depths between 4 and coloration closely the yellow anal morph of 8 m, though individuals are now and then H. "yellow chin pseudonigricans". seen in shallower water. Subadults have I consider H. "yellow chin pseudonigri- been caught at Zue Island at about 1 m cans" the matrix species of the "Pseudo- depth. At Hippo Island the species goes at nigricans" complex. The correlation be- least as deep as 10 m. At the gently slop- tween habitat characteristics and male col- ing islands, H. "yellow chin pseudonigri- oration among its populations, together with cans" occurs in low or moderate densities, the distribution pattern of H. "black and yel- while it can be rather abundant at some low pseudonigricans", being restricted to a steep islands (e.g. in the Vesi Archipelago). region with steep shores, strongly suggest The distribution pattern of this species is convergent or parallel evolution of male col- very peculiar. It is widely distributed, inhab- oration in response to environmental light iting every offshore island in the Speke Gulf conditions. The mutually exclusive distribu- and those that have been sampled in the tions of "yellow chins" and H. "black and yel- Sengerema region (apart from the very low pseudonigricans" in the northern gently sloping ones), but occurs only very Mwanza Gulf, suggest that the two forms, rarely at mainland rock shores. At the main- though now different in more than just land on both sides of the Speke Gulf we male coloration, have evolved from one an- found very different species of the "Pseudo- cestral stock as eco-geographically nigricans" complex (compare maps). vicariating species. More work is needed, Considering the extreme patchiness of in particular about the evolution of male the habitat of H. "yellow chin pseudo- coloration, before these interesting patterns nigricans", and its wide distribution, geo- can be disentangled. graphical variation is to be expected. Its Very interesting in this respect is also a populations are isolated from each other by species that is so far known only from the deep open water over large geographical eastern most island that we sampled, Hap- distances. They differ from each other lochromis "orange pseudonigricans" from mainly in male coloration and attained maxi- Sozihe Islands in the Speke Gulf. Anatomi- mum size. The differences in coloration are cally it is close to the populations of the "yel- largely confined to the darkness of the low chin"/"black and yellow pseudonigri- flanks and to the anal fin colour, and a cor- cans" lineage. However, it is very distinct in relation between habitat type and male col- male coloration, and females resemble

138 those of the H. "scraper pseudonigricans" species that share a relatively steep dor- lineage (see below). When we spotted sal head profile, bicuspid outer teeth, an these fishes for the first time in our nets, increased number of inner tooth rows we thought for a moment of H. nyererei. and a rather inconspicuous blueish-grey Males have the H. nyererei-colour pattern to black male coloration. For as yet not with a bright orange dorsum, dark vertical understood reasons, non of the islands in bars, and a blueish to greenish lower half the Speke Gulf is inhabited by such fishes of the body and caudal peduncle (compare and thus they rarely live in sympatry with with the photo of H. nyererei from Senga "yellow chins". Haplochromis "scraper Point). Lappets and soft part of the dorsal pseudonigricans" is known from all rocky fin, the caudal fin, and the anal fin are bright peninsulas along the southern Speke orange. However, the vertical bars on the Gulf shore, between the Bwiru peninsula flanks are fewer and broader than in H. at the corner to the Mwanza Gulf, and nyererei, which is in females particularly Senga Point. This is anatomically the most apparent. Females bear a high-contrast extreme scraper type within the H. melanin pattern with about six broad verti- "pseudonigricans"-complex. Its dentition cal bars. This is rare among "Pseudonigri- resembles that of the least specialized cans" species, and is found otherwise only Neochromis species with subequally bi- in females of the H. "scraper pseudonigri- cuspid outer teeth and up to 5 rows of cans"-group. Females of "orange pseudo- inner teeth. The outer teeth, however, nigricans" lack a yellow chin. From H. are rather widely spaced, and the gap be- nyererei, that does not occur at Sozihe Is- tween outer and inner teeth is wide. The lands, "orange pseudonigricans" differs ana- dorsal head profile is decurved and rather tomically. steep, the snout has often the tapering The habitat of H. "orange pseudonigri- appearance that I described above for H. cans" is variable. It lives at Sozihe Islands at "black pseudonigricans", the lower jaw is places with gentle, moderate and steep not as broad as in H. "large eye slopes, and over small to very big boulders. pseudonigricans". Males are light blue Females and subadult males are found in with red dorsal fin lappets at the places shallow waters of just 1 m depth as well with the clearest water. They become as at greater depths. Reproductively active black towards the interior of the Speke adult males in breeding dress are restricted Gulf. In contrast to most other "Pseudo- to depths beyond 2.5 m and are abundant nigricans" species, this one has a grey to in waters of 5 to at least 7 m depth. In the black anal fin without any red. The small greater depths they are, together with H. egg dummies are yellow. The caudal fin cf. "blue nyererei", the dominant haplochro- is grey to black with a small red upper mine species at Sozihe Islands. corner. At some places we found along- Males with coloration reminiscent of that side the blue-black type a second, yellow- of H. "orange pseudonigricans" occasionally ish colour morph with reddish-brown dor- occur within the geographical range of H. sum. Though we caught already about 25 "black and yellow pseudonigricans" males of this in 1995 discovered species, (Gabalema Islands, Anchor Island). It is likely we found only one female yet. It has a that these are rare phenotypes of the "black much narrower head than the males, and and yellow pseudonigricans" populations. shows a high-contrast melanin pattern, Their occurrence, however, may indicate a consisting of five broad vertical bars on certain disposition to orange male colora- brassy ground, similar to the pattern dis- tion and hint at possible phylogenetic rela- played by females of H. "orange pseudo- tionships. nigricans". The mainland shores of the Speke Gulf H. "scraper pseudonigricans" inhabits and of the open lake are home to a the steeply sloping large boulder shore number of "Pseudonigricans" complex at Bwiru Point, the moderately steep

139 Steep shores with large boulders are the habitat of Haplochromis "scraper pseudonigricans". slope of Igombe Island with smaller boul- of the Neochromis and Xystichromis lin- ders, and intermediate habitats at eages (H. (N.) nigricans, H. (N.) "giant Ndurwa and Senga Points. It seems to be scraper", H. (N.) "velvet black", H. (N.) "uni- absent from gently sloping, small boulder cuspid scraper", H. (X.) "copper black", H. habitats at Ndurwa Point. H. "scraper (X.) "large eye black"). At most places H. pseudonigricans" lives at water depths "scraper pseudonigricans" is much less between 1.5 and 6 m, and usually not di- abundant than some Neochromis and rectly inshore. Its diet is diverse. Of three Xystichromis. However, at Senga Point, a individuals that we checked, one had place with low haplochromine densities eaten almost exclusively filamentous in deeper waters, this species is numeri- green algae, the second one zooplankton cally dominant. and plant material and the third one A form that is very similar to H. "scraper aquatic mites (Sperchon) and caddis fly pseudonigricans" inhabits the northern larvae (Trichoptera). The change in male mainland (Ukerewe) shores of the Speke coloration that is observed among popu- Gulf. From the southern species it differs lations from west to east, is accompanied slightly in body shape, male coloration, by a change in dentition. Eastern popula- and quite considerably in eye size. I there- tions have a less typical scraper dentition, fore consider Haplochromis "Ukerewe with unequally bicuspid outer teeth and pseudonigricans" specifically distinct. It is fewer inner tooth rows, possibly reflect- currently known from the steep eastern ing different feeding habits in habitats in part of the Ukerewe shore between which algae growth is inhibited by less Namatembi Island and Nansio Bay, as transparent water. well as from Bwiru Island. Males are light H. "scraper pseudonigricans" lives sym- blue on the flanks. The dorsal fin is grey- patrically with H. "yellow chin pseudoblue with orange to red lappets. The nigricans" at Bwiru Point. However, the lat- edge of the caudal fin is red or orange on ter is very rare there. It furthermore lives its entire length and the anal fin is usu- in sympatry with several algae scrapers ally faint yellow with small egg dummies.

140 The dorsal head profile is often less steep pseudonigricans" from Gana Island north- than in "scraper pseudonigricans" and the west of Ukerewe. This is a very short- mouth opens to above rather than hori- headed species with a rather elongated zontally (slightly prognathous rather than body, and 5 broad dark vertical bars on isognathous snout outline). Finally, the flanks. The flanks are reddish on the up- eyes are considerably smaller. We found per half, yellow in the middle and yellow- "Ukerewe pseudonigricans" between 2.5 green ventrally. The dorsal fin is grey or and 5 m depth at moderate and steep grey with red streaks and has orange to slopes. Its dentition is less scraper-like red lappets. Caudal and anal fin are than that of H. "scraper pseudonigricans". largely orange to red. The steep dorsal In the species rich and densely populated head profile of adult males, combined communities at Nansio and Bwiru islands with the elongated body, can be reminis- it is very rare, while it is one of the com- cent of the condition in oral shelling snail moner species in the species poor low eaters. However, the dentition with density community at Namatembi Island mostly bicuspid teeth is that of an insect (compare with "scraper pseudonigricans" eater. At Gana Island we observed this at Senga Point). It lives in parts of its range species in water of about 4 to 6 m depth sympatrically with H. "black Ukerewe" as in a huge rock garden, composed of very another member of the "Pseudonigricans" big boulders. It lives in sympatry with H. complex. "black Ukerewe". A third species in the "scraper pseudo- Haplochromis "black Ukerewe" is a long nigricans" group is Haplochromis "Gana snouted black species with an elongated

H. "scraper pseudonigricans" (Bwiru Point). A female H. "scraper pseudonigricans" (Bwiru).

H. "Ukerewe pseudonigricans". A male H. "Gana pseudonigricans".

141 body that is known from several places pseudonigricans" is bright red between in the Ukerewe region. The different the soft rays. The outer teeth of adults are populations hardly differ in coloration usually bicuspid in both species but though they differ somewhat in other much more outworn in "long snouts" than characters, e.g. body depth and lower jaw in most "blue pseudonigricans". Finally, length. Males are dark grey with 4-5 broad the population of "blue pseudonigricans" vertical bars, to entirely black. Black is also does not have the long snouted appear- the dorsal fin that has red lappets, while ance. This is due to a dorso-ventrally less caudal and anal fin are in their proximal narrowed snout region and a straight, in- half greyish, in their distal half red. Fe- stead of an incurved dorsal head profile. males are brassy yellowish. H. "black We know little about the ecology of H. Ukerewe" lives at moderately steep "long snout pseudonigricans". It lives at shores as well as in the rock gardens of water depths of 1.5 to at least 6 m. At the Gana Island. We found it at water depths Nyegezi rocks it inhabits an almost verti- ranging from 2 to 6 m. Jaw anatomy and cally dropping slope with huge rock boul- dentition suggest that H. "black Ukerewe" ders, at Kissenda Island a moderately preys upon insect larvae and/or steep slope with boulders of medium zooplankton. size. It does not inhabit a gently sloping Two species have longer lower jaws area of Kissenda Island with smaller boul- than all other "Pseudonigricans" species, ders. H. "long snout pseudonigricans" and resemble in this respect the species lives sympatrically with several other spe- of the "Deepwater" complex. Neverthe- cies of this complex at Nyegezi rocks: H. less, their overall appearance is more that "black pseudonigricans", H. "black and yel- of "Pseudonigricans" species and I tenta- low pseudonigricans", H. "large eye tively include them in this species com- pseudonigricans" and H. "pseudoblue". At plex. Haplochromis "long snout Kissenda Island, where it coexists only pseudonigricans" is currently known from with H. "blue pseudonigricans", it appears two places on opposing shores of the that "long snouts" dominate more inshore, northern Mwanza Gulf: Nyegezi rocks at while "blues" are restricted to more off- the east shore and Kissenda Island at the shore waters. Judging from their num- west shore. The species is readily distin- bers in our nets, the densities of the two guished from other members of the spe- species are offshore approximately equal. cies complex by its long and acutely Anatomically similar to the aforemen- pointed snout and therefore shallow dor- tioned species is Haplochromis "pseudo sal head profile in both sexes. Its denti- blue"blue". However, it has a different dentition tion is of "Pseudonigricans" type. Males that is quite unusual for this species com- are dark blue to blue-black with dark red plex. Moreover, also its male coloration is in the anal and caudal fins, a few dark red unique within the "Pseudonigricans" spe- streaks in the soft dorsal fin, and dark red cies. We know "pseudoblue" only from a dorsal fin lappets. Females are brownish part of the Mwanza Gulf between Python with a metallic greenish flush on the Islands in the south and Hippo Island in flanks. H. "long snout pseudonigricans" the north. It is abundant only in the lives at Kissenda Island in sympatry with deeper waters of the Python Islands, a population of H. "blue pseudonigricans" while very rare at Kilimo Island, Nyegezi that has also exceptionally long jaws, but rocks and Hippo Island. Males at Python otherwise differs from H. "long snout Islands are dichromatic (Seehausen & pseudonigricans" anatomically and in Bouton 1996b). The coloration of the male coloration. Male flank coloration of blue morph ranges from iridescent metal- "blue pseudonigricans" is a lighter blue, fin lic blue flanks with a purple hue and a coloration a lighter red than in "long snout darker dorsum, to metallic purple-blue. pseudonigricans". The dorsal fin of "blue The blue is never as clear a sky blue as

142 in H. "blue pseudonigricans". All unpaired transparency at Python Islands prevents fins are purplish blue, the dorsal fin car- observations in its natural habitat. It lives ries a lighter silvery purple, metallic irides- at Python Islands at water depths ranging cent band in its spiny part. Instead of the from 2 to 8 m, but predominantly deeper bright red streaks that characterize H. than 4 m. It tends to occur slightly deeper "blue pseudonigricans", H. "pseudoblue" than H. "blue pseudonigricans". At the exhibits dull red-brown streaks on the other localities we caught it at similar membranes of the soft dorsal. Males of depths (4-7 m). Its clustered occurrence the red morph have the dorsal head sur- in our nets at Python Islands suggests face, dorsum, chest, and upper half of the that H. "pseudoblue" forages in shoals of flanks purplish red to red. The dorsal fin fishes of one size class. Our stomach con- is red, and the metallic iridescence on the tent analyses suggest that we are deal- flanks is more yellowish than blue. Large ing with a real zooplanktivore. The three adult males demonstrate up to eight big individuals checked had their stomachs orange-yellow to white-yellow egg dum- to 50%, 93% and 95% filled with mies. In smaller males the egg dummies zooplankton (Daphnia). The rest were a are much smaller in relation to the fin. few insect larvae, but one fish had eaten Females are greyish with a distinct silvery a larger amount of filamentous green al- metallic sheen on the flanks. Males from gae (Cladophora) that it must have in- Kissenda Island and Nyegezi rocks resem- gested in shallow water! The size class ble the blue morph from Python Islands distribution in our catches from different but have a reddish dorsum, more and seasons of the year suggests that H. brighter red in the unpaired fins and "pseudoblue" might be a seasonal smaller orange coloured egg dummies. spawner. In January and February we The only male from Hippo Island had lost caught almost exclusively semi-adults of its colours in the net. The coloration of between 8 and 10 cm SL, as well as the northern most population is therefore some very big adults. Until April the size not yet known. of the semi-adults increased, and be- When we first collected H. tween September and November only "pseudoblue" in 1991, at a time when very full grown adults were caught. The maxi- few rock-dwelling cichlids were known mum size attained is 110.2 mm SL. Like yet, N. Bouton and I were struck by the in many other Mbipi, females are much resemblance between this species and less frequently caught than males. the sympatric H. "blue pseudonigricans", The rarity of this species at any place, and therefore named it H. "pseudoblue". except Python Islands, may be a conse- However, "pseudoblue" differs distinctly quence of the Nile perch upsurge. Pisci- from H. "blue pseudonigricans" and H. "red vorous stages of the Nile perch are abun- pseudonigricans", both of which it coex- dant in the deeper waters at rocky shores ists with at the Python Islands. It has in the Mwanza Gulf. Several rock-dwell- longer lower jaws and a different tooth ing haplochromine species of deeper wa- shape. The teeth in the outer rows re- ters, that are elsewhere rare or entirely semble those of the other two absent, survived in a rocky trough be- "Pseudonigricans" species only in small tween the Python Islands that may offer individuals, while full adults have widely, more protection against Nile perch preda- but regularly spaced, strongly recurved tion than open slopes of rocky islands do unicuspid teeth, resembling those of H. (Seehausen et al. in press [b]). "deepwater". Large males can develop a In our extensive sampling of the peculiar lower jaw shape with a mental Mwanza Gulf we came about several in- prominence ("chin"). dividuals of "Pseudonigricans"-like fishes Not much is known about the ecology that do not fit into any of the known spe- of H. "pseudoblue". The very low water cies. It may well be that at least one more

143 A male H. "black Ukerewe" (Gana). A female H. "black Ukerewe" (Bihiru).

H. "longsnout pseudonigricans" (Kissenda). A female H. "longsnout pseudonigricans".

A male H. "pseudoblue" (Python Island). A female H. "pseudoblue" (Python Island). species of this complex occurs in the re- head shape and coloration. These fishes gion. N. Bouton and me collected in 1991 had a more decurved dorsal head profile, several large males at Python Islands were brassy-yellow rather than blue, and which we named Haplochromis "yellow had red fins similar to those of H. "blue blue"blue". On the same day large males of H. pseudonigricans". They were caught at "blue pseudonigricans" and H. depths beyond 4 m. In spite of intensive "pseudoblue" were collected, from which sampling at Python Islands, fishes of this H. "yellow blue" was clearly distinct in phenotype were not caught again after-

144 wards.

A small islet of the Vesi Archipelago.

145 Chessboard Mbipi

The rock-dwelling haplochromine species lower jaw is not always longer than broad. joined in this group share a melanin pattern Nevertheless, I think that with moderate re- in which the dark vertical bars are crossed vision of its definition, the taxon Para- by two dark longitudinal stripes (mid lateral labidochromis would be suitable to carry all and dorsal lateral stripe), reminding of a rock-dwelling cichlids with a chessboard chessboard. The so by melanin pattern de- melanin pattern. However, since Paralabid- fined group corresponds to Greenwood's ochromis was already by Greenwood de- Paralabidochromis lineage that was defined fined much wider than, for instance, on anatomical criteria. All known rock-dwell- Neochromis and Xystichromis, it may, alter- ing species with a chessboard melanin pat- natively, be split up into its distinct lineages. tern possess a dentition that is similar to Within the Chessboard Mbipi, two major the one described for Paralabidochromis groups can be identified, and within each (Greenwood 1980). This is: rather narrow of them again two species complexes, fall- dental arcades and the anterior teeth in the ing apart on gross morphology. The first lower jaw procumbently implanted, form- group is characterized by a long head (> ing with those in the upper jaw a forceps- 33% of SL). It contains the H. (Paralabido- like construction (compare text figure 5). In chromis) chilotes complex and the H. (P.) most species the teeth of the outer tooth chromogynos complex. The second group, row are more or less cylindrical in cross with a shorter head, falls apart into a rela- section, the crowns not flattened (some tively long jawed (lower jaw > 34% HL) species in the "Rockkribensis" complex group: the H. (P.) "rockkribensis" complex, make an exception to this). Most species and a short jawed group: the H. (P.) have slightly thickened lips, two have their "rockpicker" complex. At least one of the lips produced into lobes. The procumbent currently described Paralabidochromis spe- implantation of the outer teeth anteriorly in cies corresponds to each of these groups. the lower jaw is due to a slight to moder- Within Paralabidochromis a fifth group, the ate ventral inflection of the dentary (tooth H. (P.) crassilabris group, exists but has not carrying lower jaw bone), which the species been found in rocky habitats. of this group share with the species of the Psammochromis and Ptyochromis lineages, and with the non-rock-dwelling monotypic The short head Chessboard Mbipi: genera Macropleurodus, Platytaeniodus and Tiny mouths — the Haplochromis Hoplotilapia. In most species of the Chess- (Paralabidochromis) "rockpicker" board Mbipi the teeth in the outer rows are complex firmly, not moveably implanted. All but one lineage of Chessboard Mbipi have small, Within the chessboard species the spe- rather deeply embedded chest scales, cies of the "Rockpicker" complex are read- though the expression of this character is ily recognized by the combination of a in the "Rockpicker" group more distinct than short head (head length of less than 33% in others. In several characters the chess- of the standard length) with a short lower board group exceeds the limits set by the jaw (lower jaw length not exceeding 35% definition of Greenwood's Paralabido- of the head length). It is a rather close- chromis. So do several species exhibit a knit complex of species that share the fol- retrognathous lower jaw, rather than iso- lowing unique character combination: a gnathous jaws, and in the same species the peculiar small mouth; narrow, V- or U-

146 shaped dental arcades; strong unicuspid However, it does not seem to have a chess- or bicuspid and somewhat recurved outer board melanin pattern. Snoeks (1994) teeth; 2 to 5 rows of inner teeth; moder- when discussing the ecology of H. (P.) ately steep and slightly concave or paucidens, made the observation that, straight, to steep and decurved dorsal though evidently insectivorous, the species head profile; a lower jaw length of 28.5- does not correspond anatomically to the in- 34.3% of head length; a lower jaw sect eating haplochromines of Lake Victo- length/width ratio ranging from 1.07 to ria, as defined by Witte & van Oijen (1990). 1.31; an interorbital width of between However, as shown in the following, this 19.6 and 25.4% of head length; small apparent problem is explained by the cir- deeply embedded chest scales, and be- cumstance that few rock-dwelling cichlids tween 5 and 7 vertical bars on the flanks. were known by the time Witte & van Oijen I assume that the species of the "Rock- published their key to the trophic groups. picker" group form a monophyletic assem- Those few were meant by the authors, blage within the Paralabidochromis line- when they mentioned that a number of in- age. sect eating species have more specialized All known species of this complex are features. morphologically specialized epilithic Auf- Another species that corresponds wuchs feeders. In contrast to the other closely to some "Rockpickers" is the mys- large anatomically specialized epilithic Auf- terious H. (P.) victoriae Greenwood 1956. wuchs feeder complex, the Neochromis lin- It is, among others, this species that eage, the "Rockpicker" species forage pre- fuelled discussion about the phylogenetic dominantly on animalous components of relationship between Lake Victoria and the Aufwuchs, and on small insect larvae Lake Malawi (Nyasa) haplochromines (Chironomidae) in particular. These tiny in- (Greenwood 1956b, Greenwood 1983), sects live among the filamentous algae that since it closely resembles the Malawi ge- cover the rocks, and can sometimes at nus Labidochromis in some anatomical places occur in great densities. Different characters. Unfortunately described from from many other Mbipi, "Rockpicker" spe- a single individual, it is the type species cies hardly ever took our worm-baited of Greenwood's genus Paralabidochromis. hooks. This may reflect their specialized Different from all known "Rockpickers", H. feeding habits. At most places the species (P.) victoriae possesses large scales on the of this complex occur in low population chest. Judging from the morphometric densities. However, there are remarkable data available from the description of H. (P.) exceptions that show that "Rockpickers" can victoriae, it is otherwise closest to H. (P.) numerically dominate the rock cichlid com- "pseudorockpicker". munity. Like in all other species complexes The intra-complex taxonomy of "Rock- of Mbipi, one species has a wide geographi- pickers" is complicated by the circum- cal distribution (matrix species). Many oth- stance that most populations differ in col- ers have narrow geographical ranges, and oration and/or other morphological char- several have strongly declined recently, or acters from most other populations. This went even extinct. Circumstantial evidence indicates pronounced site fidelity among suggests that increasing water pollution these highly stenotopic species. Based could be a cause. on morphology (excluding coloration) Among the described species of the three groups can be identified. One has Paralabidochromis lineage, H. (P.) paucidens (within the above defined range) a rela- Regan 1921, an endemic of Lake Kivu, tively long, lower jaw that is isognathous shares all above mentioned anatomical char- with the upper jaw (compare text figure acters with the Lake Victoria "Rockpickers" 7f), and typical picker dentition that is and, based on these, should be considered similar to that of the Lake Malawi genus a member of the "Rockpicker" complex. Labidochromis (outer teeth usually tend-

147 ent sets of species. (2) In most, if not all cases of occurrence of more than one "Rockpicker" complex species at a locality, the sympatric species belong to two different morphological groups, one being morphologically more a picker, the other one more a scraper. In the following I discuss the species in the order of the three morphological groups. The first four species belong A reddish male H. "pseudorockpicker" from Kilimo Island. to the long jawed picker-type group.

Rockpickers with relatively long jaws

Haplochromis (Paralabidochromis) "pseudorockpicker" was first identified by Els Witte-Maas and Frans Witte who collected it at Mushroom Is- land/Nyegezi rocks in the northern Mwanza Gulf as early as 1978 (Witte et al. 1992). The name "pseudorockpicker" goes back to that A male H. "pseudorockpicker" from Kilimo Island. time. Witte-Maas and Witte had collected, and named H. ing to unicuspid, inner tooth rows fewer (P.) "rockpicker", what they first thought was than in the other groups). A second one only one species. Later they realized that has a very short, retrognathous, V-shaped their samples consisted of two different lower jaw that is as broad as it is long, a species and they called the second one peculiar dentition with a rather V-shaped "pseudorockpicker". It has a longer and nar- dental arcade, strong unequally bicuspid rower lower jaw than "rockpicker". Jaw and to unicuspid outer teeth that are inclined dental arcade are V-shaped rather than more towards the median, and a relatively U-shaped, the outer teeth tend towards uni- broad interorbital width. The third group cuspid and male coloration is different from has a moderately to very short, U-shaped that of "rockpicker". From "southern lower jaw, that is isognathous with the pseudorockpicker" it differs in male colora- upper jaw, or slightly retrognathous. It has tion and outer tooth shape, from "red a rather U-shaped dental arcade and pseudorockpicker" in male coloration and scraper dentition (bicuspid outer teeth, the number of inner tooth rows (table 3), many inner rows). In the distribution of from "chessboard picker" in male coloration "Rockpicker" species two patterns can be and eye size. All remaining species of the discerned: (1) The Mwanza Gulf and "Rockpicker" complex have shorter and Speke Gulf/open lake region have differ- broader lower jaws, and bicuspid outer

148 A male H. "southern rockpicker" (Sh). A male H. "red pseudorockpicker" (Anchor Is.). teeth. Males in breeding coloration have the other one filamentous blue-green algae light blue to light greenish flanks, rarely a and chironomid larvae. reddish flush on the dorsal head surface The populations of Kilimo and Hippo Is- and chest, a light blue dorsal fin with red lands differ in their morphology, particu- streaks in the soft part and red lappets, an larly in lower jaw length, and outer tooth entirely red caudal fin, and a yellow to light red anal fin (whitish in preserved specimens). The pelvic fins are usually black. Five to seven vertical bars are on the flanks. The population of Mush- room Island seems meanwhile extinct, but we found "pseudorockpicker" at three other places. At Python Islands we found, in spite of extensive sampling, only one individual. Another one has been caught in 1990 by Niels Bou- ton, but the species must be extremely rare at that place. Higher population densities exist at Kilimo and Hippo Is- lands. At Kilimo Island, H. (P.) "pseudorockpicker" can be found at water depths ranging from 1.5 to at least 4 m. At Hippo Island it seems to be confined to slightly deeper waters, between 4 and 8 m depth. At both places the slope is moderately steep and the boulders are of medium size. Of two males that we checked from Kilimo Island, one had eaten almost exclusively chironomid larvae, A south-west view from Python Island.

149 shape, probably reflecting limited gene flow a shorter lower jaw (compare tables 3 and between them. The species is not known 4). From the similarly coloured H. (P.) from outside the Mwanza Gulf. H. (P.) crassilabris, which lives in macrophyte "pseudorockpicker" used to live in sympatry rich sand bottom habitats in the Mwanza with H. (P.) "rockpicker" and H. (P.) "elongate Gulf, "red pseudorockpicker" differs by its rockpicker" in the past, and its Hippo Island narrower head and much narrower lower population lives geographically very near to jaw and dental arcades. The teeth in the a record locality of H. (P.) "yellow rockpicker" outer row are a mixture of unicuspids and (Kissenda Island), making at least parapatric bicuspids that are widely spaced and contact between the two species likely. A recurved. We know "red pseudo- superficially similar species that lives rockpicker" only from two individuals that sympatrically with "pseudorockpicker" is H. we found at Anchor Island in inshore shal- (P.) plagiodon. It is not a real rock-dweller but low water. The habitat consists of big, can sometimes be seen in rocky habitats. steeply sloping boulders. H. (P.) "rock- It differs from H. (P.) "pseudorockpicker" ana- picker" used to be found at Anchor Island tomically (table 3 versus table 4) and in col- as well in the past. I cannot exclude that oration. Male H. (P.) plagiodon have a red "red rockpicker", like other "rockpicker"- soft part of the pelvic fins. species of the Mwanza Gulf, was more Known only from the central and south- abundant and/or more widely distributed ern Mwanza Gulf south of Python Islands before the large scale ecological changes is a dull coloured species, Haplochromis in Lake Victoria began. (Paralabidochromis) "southern pseudo- Haplochromis (Paralabidochromis) rockpicker"rockpicker". It differs from "pseudo- "chessboard picker" was discovered just in rockpicker" in coloration and by having bi- 1996. Extreme rarity or geographically very cuspid, scraper-like teeth in the outer tooth restricted distribution are responsible for rows, although it has only two rows of in- this. We discovered "chessboard picker" in ner tooth. Both sexes are brassy yellowish waters of about 6 m depth at one of the with the metallic sheen and light underparts small islets in the Vesi Archipelago (central that is typical for many rock cichlids in the Speke Gulf). The islet consists of big to murky waters of the southern Mwanza Gulf. huge rocks that slope steeply. In this habi- We found it at a steeply sloping rock wall tat, it is the only "Rockpicker" at Vesi Islands, at about 6 m depth, as well as along a very while H. (P.) "yellow rockpicker" is very abun- gently sloping slab shore with sand/slab dant at less steep shores with smaller boul- mixed bottoms at 1.5 to 2.5 m depth. It ders. While in coloration this species re- was everywhere rare. minds of H. (P.) "rockkribensis", it is anatomi- Similarly rare and therefore poorly cally a member of the picker-type group in known is Haplochromis (Paralabido- the "rockpicker" complex, with unicuspid chromis) "red pseudorockpicker" which outer teeth. The only known male is yellow- we found only at Anchor Island in the ish with an orange flush on the chest and northern Mwanza Gulf. Its mouth shape two prominent longitudinal stripes. The is very similar to that of "pseudo- dorsal fin has red lappets and red streaks rockpicker" but it has a more elongate in its soft part, the proximally yellowish-grey body shape, a narrower head, and males caudal fin is distally red and the proximally are reddish with a red soft part of the pel- grey anal fin is distally orange-red with vic fins. Furthermore "red pseudo- small orange-yellow egg dummies. H. (P.) rockpicker" has more inner tooth rows. "yellow rockpicker" of Vesi Islands is much The number of vertical bars on the flanks more greenish, has a different body shape, is about six. From H. (P.) plagiodon, which less prominent horizontal but more promi- also occurs at Anchor Island, and has the nent vertical bars, and very different denti- same pelvic fin coloration, it differs by tion. Phylogenetic affinities of "chessboard means of the narrow head as well as by picker" are quite unclear.

150 Rockpickers with a short flanks (versus 5-6 in "blue rockpicker"), and V-shaped lower jaw less prominent longitudinal stripes. H. (P.) "yellow rockpicker" is known from Haplochromis (Paralabidochromis) "yel- almost all sampled places in the Speke low rockpicker" is a representative of the Gulf and from many places in the second morphological group of "Rock- Sengerema region. In the Mwanza Gulf, pickers". It is the geographically most in contrast, we found it only at one place widely distributed among the currently near the entrance of the Gulf (Kissenda known "Rockpicker" species, and the matrix Island) and it is very rare there. At Zue Is- species of the complex. It has one of the land in the north-central Speke Gulf it is shortest lower jaws that occur in "Rock- replaced by a similar but taxonomically pickers" and in Lake Victoria haplochro- difficult population (see below). mines in general. At many localities "yellow The habitats of "yellow rockpicker" are rockpicker" is characterized by yellow male quite diverse and include very gentle to body coloration, accompanied by red colora- moderately steep slopes. The species is tion in the unpaired fins: streaks in the dor- absent from habitats with very steep slopes sal fin, an entirely red caudal fin and an or- (several islets within the Vesi Archipelago) ange coloured to red anal fin. The pelvic fins or is very rare in such habitats (Ruti Island). are black. Females are light yellowish to It is very impressive in this respect, to ob- brownish. Both sexes carry 6 to 8 vertical serve the enormous abundance of "yellow bars and sometimes weak traces of two lon- rockpicker" at moderately steep slopes in gitudinal bands on the flanks. Males from the Vesi Archipelago (numerically a co-domi- Chamagati Island can exhibit an orange col- nant species, together with H. (Neo- oured chest. The intensity of the yellow chromis) "Vesi scraper"), and its complete flank coloration varies among populations. absence from steep slopes just a few hun- It is noteworthy that males are particularly dred metres further. At the moderately bright yellow in populations that live in steep slopes it lives at water depths rang- sympatry with other (always blue) "Rock- ing from 1.5 to at least 5 m and is most picker" species ("blue rockpicker" at Chama- abundant at about 3 to 5 m. At Mabibi Is- gati and Makobe Islands, "orange anal picker" lands, in a similar habitat, it is common, at Bwiru Island). In contrast, males in some though not dominant, at 2 to 5 m depths, populations at isolated offshore islands, and is absent from depths beyond 5 m. At that are inhabited by only one "Rockpicker" other localities it inhabits a similar depth (e.g. Ruti, Mabibi, Vesi Islands), are much range but is much more rare, so that actual less yellow, and often more greenish (see depth preferences are difficult to observe. photos). So do the flanks of the population While I frequently observed subadult and from Vesi Islands carry various hues from adult H. (P.) "blue rockpicker" when diving at yellow-green to deep blue, contrasting Makobe Island, I only once came across a beautifully with an orange to red anal fin. H. (P.) "yellow rockpicker". This was a full Apart from male coloration, also number of adult male that seemed to be having a ter- vertical bars, lower jaw shape, and dentition ritory at only 1.5 m water depth. At the help to tell "yellow rockpicker" apart from well studied island, "yellow rockpicker" "blue rockpicker". The outer teeth of "yellow seems to undergo considerable fluctua- rockpicker" are stronger, the rows of inner tions of population density. Of three teeth fewer, separated from the outer ones males from Makobe Island whose stom- by a broader gap (see table 3). Lower jaw achs we checked, two had eaten almost and dental arcade are slightly more V-shaped exclusively chironomid larvae, the third in H. (P.) "yellow rockpicker". Its dorsal head one filamentous green algae, unicellular profile is less rounded but more straight, algae and plant material, which it probably sometimes with a small concavity above ingested as detritus (see the discussion the eyes. It has 6-8 vertical bars on the under Xystichromis).

151 H. "yellow rockpicker" from Makobe Island.

A male H. "yellow rockpicker" from Bwiru Island. A male H. "chessboard picker" (Vesi Is.).

A male H. "yellow rockpicker" from Vesi Island. A female H. "yellow rockpicker" (Sozihe Is.).

H. (P.) "yellow rockpicker" occurs sym- (P.) "elongate rockpicker" at Ruti Island, patrically with H. (P.) "blue rockpicker" at with H. (P.) "orange anal picker" at Bwiru Makobe and Chamagati Islands, with H. Island and with members of all other spe-

152 A male H. "Zue rockpicker" (Zue Island). A female H. "Zue rockpicker" (Zue Island). cies complexes of the Chessboard Mbipi. am currently not able to say whether "Zue Anatomically very similar to H. (P.) "yel- rockpicker" is an aberrantly coloured popu- low rockpicker" is a "Rockpicker" popula- lation of "yellow rockpicker", an anatomi- tion from Zue Island in the central-north- cally aberrant "blue rockpicker", a result of ern Speke Gulf, which bears the melanin hybridization between the two species, pattern of H. (P.) "blue rockpicker" and a or something completely different. It lives male coloration that also resembles that between less than 1 and at least 4.5 m of the latter species. In contrast to all typi- water depth at a very gently sloping cal populations of "yellow rockpicker", this shore with very small boulders and stones one has only 5 to 6 vertical bars on the and can be rather abundant at about 3 m flanks which are crossed by well visible depth. longitudinal stripes. The anal fin is pale with slight pinkish flush and their is very little red also in the other unpaired fins. I

The north-eastern corner of Gana Island, habitat of H. "orange anal picker".

153 Rockpickers with scraper dentition "Rockpickers" do, about whose feeding habits we have information. H. (P.) "rockpicker" Haplochromis (Paralabidochromis) lived sympatrically with H. (P.) "pseud- "rockpicker"ockpicker", discovered by F. Witte and orockpicker" and H. (P.) "elongate rockpicker" E.L.M. Witte-Maas in the late 1970s in the at Nyegezi rocks, and with the latter and H. Mwanza Gulf (Witte et al. 1992), was the (P.) "red pseudorockpicker" at Anchor Island first known representative of the third mor- (Witte & Witte-Maas pers. comm.). Other phological group, the scraper type of sympatric Paralabidochromis were H. (P.) "Rockpickers". Witte and Witte-Maas used "rockmacula", H. (P.) "rockkribensis" and H. (P.) to collect it by worm-baited hook at rocks plagiodon. in the Nyegezi area and at Anchor Island Another species of the scraper type is (northern Mwanza Gulf). Between the early Haplochromis (Paralabidochromis) "orange 80s and the late 80s the species disap- anal picker"picker". It has an even shorter lower peared from its former localities. One male jaw than "yellow rockpicker" and there is was still collected by N. Bouton and me at hardly any overlap between the two spe- Mushroom Island in 1991. In spite of inten- cies with regard to this character. H. (P.) "or- sive sampling there and at many other ange anal picker" is known from only two places, no further individual was seen after localities, both in the western Ukerewe re- that. Considering the apparently very small gion, though quite far from each other. geographical range that this species had, it Since the extensive rocky shores between is possible that it is extinct. Male the two places have not been sampled, it "rockpicker" in breeding dress are, accord- is not known, whether the species has a ing to Witte and Witte-Maas (pers. comm.), closed distribution in western Ukerewe. grey-black to blue-grey, with 6-7 vertical bars However, the two populations differ in on the flanks. The dorsal fin is dark grey some morphometric measurements so with a blue sheen, red spots in the soft part considerably that I would probably consider and red lappets. Caudal and anal fin are them specifically distinct, would they not dark greyish near the basis and red distally. have a very similar male coloration. Males Females are dark grey-brown above, silver- are dark grey with a beautiful metallic deep white below, and bear a distinct chess- blue (Gana Island) or blue-green (Bwiru Is- board melanin pattern. The dorsal head pro- land) sheen on the whole body, and a me- file is decurved, the lower jaw slightly tallic blue or blue-green dorsal fin with red retrognathous. The lower jaw is U-, rather lappets. There are 7-8 (Gana Island) or 6-7 than V-shaped, the teeth in the outer row (Bwiru Island) vertical bars on the flanks. are bicuspid, and the number of inner tooth The caudal fin is bright red and the anal fin rows is increased (table 3). By these char- orange with small (Bwiru Island) to very acters H. (P.) "rockpicker" differs distinctly small (Gana Island) egg dummies. Females from the sympatric H. (P.) "pseudorockpicker" are known only from Bwiru Island yet. They and H. (P.) "red pseudorockpicker", but is are very similar to those of "yellow rock- close to H. (P.) "orange anal picker" and H. picker" but have an even shorter, more tiny (P.) "elongate rockpicker". From the first it dif- lower jaw. Small but significant differences fers in male and female coloration, from the between the two populations of "orange latter in dentition and body shape. anal picker" exist in maximum size (Gana > The rocky habitats that were inhabited by Bwiru), body depth (Gana < Bwiru), head H. (P.) "rockpicker" in the central and north- length (Gana < Bwiru), and eye length (Gana ern Mwanza Gulf, slope moderately steeply < Bwiru). Moreover, fishes from Gana Island to steeply and have medium sized to large have a more rounded dorsal head profile. rock boulders. The information available, H. (P.) "orange anal picker", particularly of suggests that the species was preying pre- Gana Island, are anatomically a peculiar ap- dominantly on chironomids (Witte & Witte- pearance. They have the shortest lower Maas pers. comm.), like all other jaw that I have measured among Lake Vic-

154 toria haplochromines (one male with a lower low rockpicker" and H. (P.) "red short snout jaw length of 27% of the head length; Witte scraper" that also live at Bwiru Island, have & Witte-Maas [pers. comm.] once had a their maximum abundance at 2-3 m depth. "rockpicker" with a lower jaw length of only Finally, "yellow anal picker" lives in sympatry 25.5% of the head length). This lower jaw with H. (P.) "sky blue picker" at Gana Island. is often retrognathous, however, since also F. Witte and E.L.M. Witte-Maas caught in the upper jaw is short, the jaws are some- the 1970s a species of the "Rockpicker"- times isognathous. With the elongated complex with a particularly shallow, elon- body and rounded dorsal head profile, the gated body shape. They named it Haplo- fishes strongly resemble algae scrapers of chromis (Paralabidochromis) "elongate the Neochromis lineage, and are almost cer- rockpicker"rockpicker". This species has disappeared tainly confused with such before one real- from its former localities in the northern izes the jaw shape. Mwanza Gulf (Nyegezi area, Anchor Island) H. (P.) "orange anal picker" is also ecologi- just like H. (P.) "rockpicker" has. We have, cally similar to Neochromis species, espe- however, recently come about a shallow cially to H. (N.) "blue scraper" with which it bodied "Rockpicker" at Ruti Island, that may lives in sympatry at both record localities. be identical with H. (P.) "elongate rockpicker". At Gana Island "orange anal picker" is ob- The male is rather dark blueish with an or- served at depths between 1.5 and 6 m. It ange-red anal, a dark dorsal fin with broad is at some spots the numerically dominant red lappets and a caudal fin that is dark at haplochromine at 2 to 4 m depth, espe- its base and red distally. This species has cially in the wide and deep crevices and one of the most extreme scraper dentitions gorges between big rock boulders. There within the "Rockpicker" complex. The outer seems to be a tendency for H. (Neo- tooth row consists of closely set subequally chromis) "blue scraper" to occur in more bicuspid teeth with slightly recurved shallow water, and I observed territorial crowns, and the inner teeth are arranged males only at less than 2 m depth. Males in about 5 rows (table 3). "Elongate rock- of "orange anal picker" defend territories picker" is morphologically very close to "or- among the big rock boulders at depths be- ange anal picker". It differs from the latter tween 1.5 and 3 or 4 m (beyond that depth by means of a longer lower jaw and nar- H. "lemon fin nyererei" are the dominant rower head. However, it is not impossible territory holders). Vertical and horizontal that the two forms turn out to be con- gaps in rock walls frequently form the cen- specific, once more is known about the tre of a territory. From such a hiding place very rare "elongate rockpicker". Nothing is males approach females, and I saw them known about the feeding ecology of this entering into the plankton grazing shoals species. It lives in sympatry with H. (P.) "yel- that are dominated by female H. "lemon fin low rockpicker" and is the more rare one of nyererei". Possibly female "orange anal the two. picker" join these shoals in the water col- Another typical representative of the umn of rocky gorges. A very similar picture "Rockpicker" species with scraper denti- of depth preferences is observed at Bwiru tion is Haplochromis (Paralabidochromis) Island: At water depths down to 2.5 m H. "blue rockpicker"rockpicker". It is characterized by (N.) "blue scraper" numerically dominates light blue male body coloration, and a and H. (P.) "orange anal picker" is very rare. prominent melanin pattern. The dorsal fin Just below 2.5 m depth the ratio of abun- is blue, sometimes with red streaks in dance turns sharply, and already at 3-4 m the soft part and with red lappets. Anal depth "orange anal picker" is often ten times and caudal fin are red. Under water the more abundant than "blue scraper", and can anal fin appears pale, while the caudal fin be the dominant haplochromine. At yet bright, resembling that in H. (Neo- greater depth (beyond 4.5 m) H. (Neo- chromis) nigricans. The pelvic fins are chromis) "long black" takes over. H. (P.) "yel- black. Females are silvery whitish, rather

155 than yellowish to brownish as those of the rockpicker" differs from that of all aforemen- sympatric H. (P.) "yellow rockpicker". Both tioned species. The outer teeth are rather sexes exhibit a distinct chessboard pattern fine, subequally to unequally bicuspid and of almost always six vertical bars and two very closely set. The inner teeth are very longitudinal bands on their flanks. The small tricuspids and arranged in up to 7 prominent appearance of the latter helps to rows in the upper, and up to 4 in the lower distinguish females of the blue species jaw. The gap separating them from the from those of the yellow species. The dor- outer teeth is less broad than in H. (P.) "yel- sal head profile of "blue rockpicker" is steep, low rockpicker", and can be almost absent. usually decurved, rounded and never con- The dentition of H. (P.) "blue rockpicker" thus cave. The dentition of H. (P.) "blue resembles that of the algae scrapers of the

A male H. "rockpicker" (Python Island). A male H. "elongate rockpicker" (Ruti Island).

A male H. "orange anal picker" (Bwiru Is.). A male H. "orange anal picker" from Gana Island.

A female H. "blue rockpicker" (Makobe). A male H. "blue rockpicker" from Nansio.

156 A male H. "blue rockpicker" from Makobe in the aquarium.

Neochromis-lineage. Like in some of the latter (e.g. H. (N.) nigricans), the inner teeth can be arranged in a zigzag pattern that partly ob- scures the rows. However, the outer teeth in the lower jaw are procumbent, and the dental arcades have the narrow shape of the "Rockpicker"-complex, though resembling more a U than a V. The jaws are isognathous. Similar to H. (P.) "blue rockpicker" are H. (P.) "elon- A female H. "blue rockpicker" with fry. gate rockpicker" and H. (P.) "rockpicker". The first one differs in body cies is absent from other islands with ap- shape, dentition and coloration, the second parently suitable habitat, e.g. Matwinki Is- one in dentition and coloration. land (near to Chamagati Island) and Igombe H. (P.) "blue rockpicker" has a relatively Island (near to Makobe Island). At Makobe wide geographical range but a very scat- Island adults inhabit slightly offshore water tered distribution. We know it from Nansio depths between 2 and 6 m, while we ob- (Ukerewe) at the northern shore of the served juveniles and subadults up to 5 cm Speke Gulf, from Makobe Island in the SL frequently in very shallow inshore wa- southern Speke Gulf entrance, and from ters. We measured maximum population Chamagati Island in the Sengerema region. densities of adults at 2 to 4 m depth (See- The habitat at two islands is a gentle hausen & Bouton in press). At Chamagati (Makobe) to very gentle (Chamagati) slope, Island adult "blue rockpicker" live between with small to very small rock boulders. At 1.5 and at least 3 m water depth. At both Nansio the slope is steeper, and the boul- places the species is more abundant than ders are bigger. It is unclear, why the spe- "yellow rockpicker". At Nansio it lives at

157 depths between 2 and at least 4.5 m. tensive red coloration in the dorsal fin in Scuba observations at Makobe Island re- form of dots and streaks. Furthermore, vealed that adults and subadults forage the chessboard pattern is less prominent. in small groups of two to four individuals The dentition with subequally bicuspid, or solitary. closely set outer teeth and 4 inner tooth Among 11 male "blue rockpicker" that rows in both jaws is, as already in "blue we checked (7 from Chamagati Island, 4 rockpicker", reminiscent of Neochromis from Makobe Island), four had empty dentition. In the case of "sky blue picker" stomachs, four had eaten predominantly even the lower jaw length resembles that chironomid larvae, one mayfly of many Neochromis, however, jaw and (Ephemeroptera) larvae, filamentous blue- dental arcade are narrower than in the lat- green and green algae, one diatoms and ter, and the teeth in the lower jaw are one unidentified material. All those that procumbent. The habitat of "sky blue had eaten insects had consumed consid- picker" is the already mentioned rock gar- erable amounts of various types of algae den of large, steeply sloping rock boul- with them. The one that had eaten pre- ders, where we found it at 3 to 4 m wa- dominantly diatoms had eaten also some ter depth sympatrically with H. (P.) "yel- chironomids. Thus the diet of this species low anal picker". is very similar to that of "yellow rockpicker", with whom it lives in sym- The Haplochromis (Paralabidochromis) patry (the ratio insect larvae versus algae "rockkribensis" complex is slightly lower in the blue species). Like most "rockpicker" complex species, also Among the chessboard species, the H. (P.) "blue rockpicker" almost never (one members of the "rockkribensis" complex exception) took our worm-baited hooks. are recognized by the combination of a In laboratory experiments in aquaria, "blue short head (head length less than 33% of rockpicker" foraged on epilithic Aufwuchs the standard length), with a relatively long predominantly by pullscraping and pick- lower jaw (more than 34% and up to 39% ing (Seehausen et al. in press [a]). Popu- of the head length). Their dental arcades, lation densities of "blue rockpicker" are though somewhat narrowed, are not as moderate at all three record localities, narrow as in the other species complexes meaning that the species is not difficult of the Paralabidochromis lineage. Their to find in its habitat but is not really abun- outer teeth are usually bicuspid, with a dant either. H. (P.) "blue rockpicker" lives flange on the major cusp, and the inner sympatrically with H. (P.) "yellow teeth are arranged in 2-4 rows. They have rockpicker" and three more Paralabido- a straight to slightly concave dorsal head chromis with scraper anatomy: H. (P.) profile that, compared to the dorsal head "short snout scraper", H. (P.) "red short profile of the "Rockpicker" complex spe- snout scraper" and H. (P.) "rockkribensis". cies, slopes gently. The lower jaw is usu- Haplochromis (Paralabidochromis) "sky ally slightly longer than the upper jaw, and blue picker" is yet another light blue always longer than broad, with a length/ "rockpicker" of the scraper group. We dis- width ratio between 1.16 and 1.6. The covered this species in early 1996 at interorbital width is, with 23.7-27.7% of Gana Island, at the northern limit of our the head length, larger than in the survey area. It may thus be more widely "Rockpicker" complex. The chest scales distributed outside the surveyed area. are in some species larger and less This species differs from the similarly col- deeply embedded than in most other oured "blue rockpicker" by a much longer rock restricted cichlids, or not deeply lower jaw (the longest one that occurs in embedded at all. On the flanks are be- the "rockpicker" complex), a dorsal head tween 5 and 7 vertical bars. profile that is straight to concave, and ex- Among the described species of the

158 Paralabidochromis lineage, H. (P.) plagiodon "rockkribensis" females is unique among fits anatomically into the definition of the Mbipi. It is very little variable among "rockkribensis" complex, and should be con- populations and seems therefore to be evo- sidered a member. However, it does not lutionary much more stable than the colora- exhibit a real chessboard melanin pattern. tion of males. Males of most populations In H. (P.) plagiodon, as well as in many spe- are characterized by a bright red chest, and cies of the "rockkribensis" complex, a ten- an otherwise yellow body, superimposed dency to an increased number of egg dum- by a faint chessboard pattern. Some mies can be observed, and these are fre- populations, however, differ considerably quently present also in females, though from this coloration and, following com- usually without the transparent outer ring. monly used species recognition criteria with H. (P.) plagiodon is sometimes caught in strict application of the SMRS concept (see rocky habitats, but is merely an occasional the chapter on taxonomy), might be re- intruder, coming from other habitats, rather garded different species. However, be- than a permanent rock-dweller (see the cause such different colour types represent chapter on frequent intruders). exclusively allopatric populations, because Intra-complex taxonomy seems relatively differences in coloration do not seem to be simple. There are two species that are dis- correlated with anatomical or ecological dif- tinct from all others: H. (P.) "rockkribensis" ferences, and because all populations share and (if considered a member of the com- the unique female coloration, I consider plex) H. (P.) plagiodon, and a complex of them as geographical variants of one spe- probably very closely related species, the cies. Among them are populations with en- H. (P.) "short snout scrapers". Some species tirely red males at islands in the northern of the Haplochromis (Paralabidochromis) Speke Gulf (picture) and a population with "rockkribensis" complex belong to the most very dark, almost black males from the ver- conspicuous Lake Victoria cichlids. Bright tically sloping Nyegezi rocks (picture). female coloration is a rare phenomenon Males from northern Lake Victoria are of- among Lake Victoria cichlids, and is charac- ten rather dark (Kaufman & Ochumba teristic for certain species of this complex. 1993), and a blueish form is known Such females exhibit a deep black chess- (Kaufman & Seehausen 1995). Unfortu- board pattern on yellow ground colour. nately little is known yet about the heritabil- Haplochromis (Paralabidochromis) ity of these colour traits in H. (P.) "rockkribensis" is together with H. "rockkribensis". At least in the case of the (Xystichromis) "copper black" probably the blackish form, pigmentation might partly be most widely and evenly distributed species a phenotypic response to the unusual light among the known Mbipi. It is, possibly to- conditions in the habitat. At the vertically gether with H. (P.) plagiodon, the matrix spe- sloping rocks upwelling light is almost en- cies of the "Rockkribensis" complex. H. (P.) tirely lacking. "rockkribensis" was discovered by F. Witte Also anatomically some variation among and E.L.M. Witte-Maas in 1978 (Witte et al. populations of "rockkribensis" exists, par- 1992). We found it at most rocky places that ticularly with respect to the feeding ap- we sampled, in big and small boulder habi- paratus. The outer tooth rows in both oral tats, and at steep and gentle slopes. It is jaws are composed of bicuspid teeth. known also from the northern shores of Most populations from the Mwanza Gulf Lake Victoria (Kaufman & Ochumba 1993). and Speke Gulf have a small flange on Maybe the most peculiar feature of H. (P.) the major cusp and three to four inner "rockkribensis" is its bright yellow female rows of teeth. However, some popu- coloration, with a dark black chessboard lations have a more epilithic algae scraper melanin pattern. Though a trend to similar like dentition with more inner tooth rows. female coloration exists in at least one other Other populations (e.g. that of Chamagati species (see below), the brightness of Island) have a more predatory appearance

159 Kissenda Island, habitat of the normal form of H. A male H. "sky blue picker" (Gana Is.). "rockkribensis".

A pull-scraping H. rockkribensis" at Chamagati.

A female H. "rockkribensis" (Python Island). H. "rockkribensis" (Python Is., normal form).

H. "rockkribensis" (Buyago, pale form). H. "rockkribensis" (Ngoma, black form).

160 Matwinki Island, habitat of the red form of H. "rockkribensis". A male of the red form of H. "rockkribensis".

A female and male H. "rockkribensis" (Chamagati, elongated form).

A male H. "rockkribensis" from the northern Mwanza Gulf.

161 with a slightly longer head and longer widely distributed but its representatives lower jaw. While populations in southern are almost everywhere very rare, and Lake Victoria possess rather fine teeth on were therefore overlooked for long. Ana- the pharyngeal jaws, some northern (Ken- tomically they are similar to "rock- yan) populations exhibit enlarged molari- kribensis" but have a shorter lower jaw form teeth on these bones, typical for and bigger eyes. Male coloration of some snail crushers (Kaufman & Ochumba species is very similar to that of the red 1993). The dorsal head profile of H. (P.) populations of "rockkribensis". Females "rockkribensis" does not vary much. It is are usually yellowish brown but those of straight and rather gently sloping. one species are yellow with a dark chess- H. (P.) "rockkribensis" is largely a shallow board pattern, though not as deep yellow water species, though individuals are as "rockkribensis" females. sometimes caught at depths beyond five The mouth of "Short snout scrapers" is metres. In the well studied community at smaller than that of H. (P.) "rockkribensis", Makobe Island, it has its highest densities but is not as small as that of "Rockpickers". between 2.5 and 4.5 m depth. The only Lower jaw and dental arcade are not as nar- three territorial males that I saw there in row as in the "Rockpickers". The outer teeth many diving hours, had their territories at are scraper teeth, mostly unequally bicus- 1.2, 1.5 and 6 m depth. At most places, pid with flange, similar to those of H. (P.) H. (P.) "rockkribensis" is found very regu- "rockkribensis", sometimes subequally bi- larly, but not in high abundance. However, cuspid and frequently, though not always, at a few islands it is abundant, and among closely spaced. The number of inner rows the dominant members of the commu- ranges from two to five. This dentition can nity. Such places have a gentle slope and be somewhat reminiscent of the condition small to medium sized rock boulders. H. in Neochromis. However, the "Short snout (P.) "rockkribensis" feeds in the southern scrapers" share with "rockkribensis" a char- range of its distribution predominantly on acteristic snout shape with a slightly prog- animalous Aufwuchs, including associ- nathous lower jaw, in contrast to the ated insect larvae, which it collects from isognathous jaws or retrognathous lower the rock surface. Of three males from jaw of "Rockpicker" complex and Neo- Makobe Island whose stomachs we chromis species. The dorsal head profile is checked, one had eaten predominantly less steep than in most Neochromis and filamentous blue-green algae and other "Rockpickers". In spite of the morphological epilithic algae, the second one caddis fly part-resemblance of some species with the larvae and sponges, and the third one al- Neochromis-complex, the "Short snout most exclusively chironomids. In labora- scrapers" are, with their lower jaw anatomy tory experiments H. (P.) "rockkribensis", and melanin pattern well embedded in the foraging on Aufwuchs-covered rocks, Paralabidochromis lineage. This is the only showed a preference for moss animals group of Mbipi in southern Lake Victoria (Bryozoa), which it scraped off from the that has predominantly rather large and not rocks. In its natural habitat, the species is deeply embedded chest scales (but see usually seen moving along the surface of page 90 for the northern lake). the rocks, and does not regularly inhabit We discovered most "Short snout the interstices between rocks, which are scraper" populations within the last two home to other epilithic invertebrate eat- years: So also a population at Makobe Is- ing Mbipi. H. (P.) "rockkribensis" lives at dif- land that had already by then the best ferent localities sympatrically with all sampled rock-cichlid community in Lake other species of this complex, and with Victoria. This just illustrates the low den- most other species of chessboard-Mbipi. sities in most "short snout scraper" A recently discovered group are the populations. With one exception, forms "short snout scrapers". The group is quite of the "Short snout scraper" group are

162 allopatric. This makes it in some cases slightly resembles H. (P.) "rockkribensis". difficult to decide, whether to consider Nevertheless, morphometrically it is different forms as different species or as close to the populations from Makobe populations of one. I currently distinguish and Mabibi Islands. At the gently sloping between five anatomically and in male Makobe Island H. (P.) "short snout scraper" and female coloration different species, lives in the more offshore waters beat least two of which live sympatrically. tween 1.5 m and 6 m depth. The major- I consider several geographically well ity of the fishes has been found in depths isolated, but anatomically and in colora- beyond 3.5 m. At Mabibi Island we found tion similar populations from offshore a big sexually active male at about 8 m Speke Gulf islands, as a single species, depth, at a vertically sloping rock wall. The to which I refer as Haplochromis (Para- fishes at Ruti Island live closer inshore labidochromis) "short snout scraper"scraper". Such and are most frequently found in populations are known from Sozihe, Vesi, rockpools. They, however, go down to at Mabibi, Ruti and Makobe Islands. In spite least 4 m depth as well. of much sampling we could not find any H. (P.) "short snout scraper" may be a dia- at the very gently sloping Zue Island. tom eater. Of three males whose stomachs From the two eastern most record locali- we checked, two had eaten almost exclu- ties, Sozihe and Vesi Islands, male colora- sively diatoms, the third one had eaten diation is not yet known because we found toms, filamentous blue-green algae and only females. Males at Mabibi and chironomid larvae to about equal parts. Un- Makobe Islands have grey-brown flanks fortunately its feeding behaviour has never with a reddish flush, a similarly coloured been observed. The species is so rare that dorsal fin with red lappets and streaks in I have never seen it under water, and nei- the soft part, a reddish caudal, and a red ther this one, nor any other member of the anal fin, with up to ten medium sized yel- group, has ever been kept in aquaria. At Ruti low-orange egg dummies. Females are Island, where "short snout scrapers" live in dark brown with yellow flush, particularly reach of angling lines, they readily take on the fins and a faint chessboard mela- worm-baited hooks, in contrast to species nin pattern on the flanks. Males of these of the "Rockpicker" complex. H. (P.) "short populations usually have a peculiar snout scraper" lives syntopically with H. (P.) rounded body shape (see photo) due to "rockkribensis", H. (P.) "yellow rockpicker", an unusually strongly convex lower body and H. (P.) "blue rockpicker", as well as with outline and a slightly decurved dorsal many scrapers of the Neochromis lineage. head profile. Females can have the same Haplochromis (Paralabidochromis) "red shape but can also be more elongate short snout scraper" is a large growing, with a much shallower dorsal head pro- brightly coloured species, known from file. Both sexes bear six to eight vertical Bwiru and Ukerewe islands in the north- bars on the flanks. ern Speke Gulf. Among the known "Short Fishes from Ruti Island differ some- snout scrapers" this one is anatomically, what from those of Makobe and Mabibi and in male coloration closest to H. (P.) Islands. Males have a greenish flush on "rockkribensis". It has a deeper body, a the flanks and frequently an orange col- broader head, a larger lower jaw length/ oured chest, somewhat resembling H. (P.) width ratio, and often a longer lower jaw "rockkribensis" that is absent from Ruti Is- than H. (P.) "short snout scraper" from the land. Fishes from Ruti Island are smaller, central Speke Gulf islands. Its outer teeth their head is narrower, the ventral body are often only weakly bicuspid. Moreover, outline usually less convex, and some its dorsal head profile is less steep, males have a less steep dorsal head pro- slightly concave and thus similar to that file, which contributes to the impression of "rockkribensis". Males are blue-grey to that "short snout scraper" from Ruti Island purplish grey, superimposed by a faint

163 The shore waters of this part of Hippo Island is home to H. "blue short snout scraper".

A male H. "red short snout scraper" from Bwiru Island.

164 A female H. "red short snout scraper" (Bwiru). A male H. "red short snout scraper" (Bwiru).

A female H. "blue short snout scraper". A male H. "blue short snout scraper" (Hippo Is.).

A female H. "short snout scraper" (Makobe). A male H. "short snout scraper" (Makobe Island).

A female H. "elongate short snout scraper". A male H. "elongate short snout scraper" (Hippo).

165 chessboard melanin pattern and bright land we found males of above 120 mm orange to red gill cover and chest. In standard length. many adult and sexually active individu- In its dentition this species differs slightly als the orange to red coloration extends from other "Short snout scrapers" as well. It across the entire lower two thirds of the has frequently one more inner tooth row. flanks. Such fishes are often extremely While anatomically otherwise similar to "red bright and closely resemble males of the short snout scraper" with its broad head and red populations of H. (P.) "rockkribensis". deep body, it differs from the latter by re- Less intensively coloured individuals can taining longer its bicuspid dentition. Even be quite similar to the "normal" (red chest) males above 100 mm standard length have type of the latter species. Females resem- real bicuspid outer teeth. These morphologi- ble in coloration those of H. (P.) "short cal differences correspond to differences in snout scraper" from the central Speke the ecology. At the, on average moderately Gulf islands, rather than those of H. (P.) steep slope of Hippo Island these fishes "rockkribensis". They are much less deep show a preference for the most gently slop- bodied than males. ing places. Females and subadult males H. (P.) "red short snout scraper" lives be- can be observed grazing Aufwuchs on im- tween 2 and at least 5.5 m water depth, mediately inshore, sun exposed, flat rock syntopically with H. (P.) "rockkribensis". We surfaces in the company of H. found its highest population densities (Neochromis) nigricans and H. (N.) "velvet (sexually active males) in relatively surf black". Females go down to depths of protected situations at 2-4 m depth. It about 3.5 m, large males live at depths be- lives sympatrically, furthermore, with H. tween 4 and maximum 7 m. Like the (P.) "yellow rockpicker", H. (P.) "blue sympatric Neochromis, and in contrast to rockpicker", H. (P.) "orange anal picker" and the sympatric H. (P.) "pseudorockpickers", with several Neochromis scrapers. this species readily takes worm-baited Haplochromis (Paralabidochromis) "blue hooks. short snout scraper" is another large grow- Hippo Island is the only place where we ing and deep bodied species. It is known found two species of the "Short snout from Hippo Island in the entrance to the scraper" lineage, living in sympatry. The Mwanza Gulf and from Matwinki Island in second species, "elongate short snout the Sengerema region. A similar population scraper" resembles in its reddish male from the nearby Kissenda Island may belong coloration the other species of the line- to this species as well. H. (P.) "blue short age. The evolution of the blue male col- snout scraper" is distinctly different from all oration of H. (P.) "blue short snout scraper" other members of the "Rockkribensis" com- may have been driven by this sympatric plex in its male coloration. Males are light situation, or, the other way around, may blue at Matwinki, dark blue with metallic have evolved in allopatry and conse- sheen on the flanks at Hippo Island, and can quently allowed the coexistence of two there on first glance be taken for males of closely related species that differ in male the sympatric H. (Neochromis) "velvet coloration. "Blue short snout scraper" fur- black". Different from the latter, they exhibit thermore lives sympatrically with H. (P.) an orange-red anal fin, an orange-red to red "rockkribensis", H. (P.) "pseudorockpicker", caudal fin, similarly coloured dorsal fin and H. (P.) "yellow rockpicker", and with lappets and streaks in the soft dorsal. Fe- the algae scrapers H. (Neochromis) males are yellowish-brown with greenish nigricans, H. (N.) "black nigricans", H. (N.) flush, and both sexes bear five to seven "velvet black" and H. (Xystichromis) "cop- dark vertical bars and two sometimes per black". prominent longitudinal stripes on the flanks. Haplochromis (Paralabidochromis) "elon- "Blue short snout scraper" grows larger than gate short snout scraper"scraper", the second spe- other species of the group. At Matwinki Is- cies from Hippo Island, has yellowish males

166 with a coppery to reddish coloured head, room Island, Nyegezi rocks). There it was and anterior half of the flanks. Females are probably already collected in 1978 by E.L.M. yellowish brown with a faint chessboard Witte-Maas and F. Witte (pers. comm.) who melanin pattern. Both sexes have about six took it for H. (H.) "macula", but already noted vertical bars on the flanks. From other spe- some peculiarities of the rock-dwelling cies of the "short snout scraper" lineage with "form". We observed and collected several reddish males, this form differs distinctly in more individuals in 1991. In 1993 we found body shape. It is much shallower than the just one male, in spite of much search. others, has the least steep dorsal head pro- Since then the species was not seen any file, and narrower jaws, resulting in a higher more. Considering its narrowly restricted length/width ratio of the lower jaw. From distribution and the disappearance of sev- the sympatric "blue short snout scraper", it eral other rock-dwelling haplochromines differs also by having usually fewer tooth from the northern Mwanza Gulf (e.g. H. rows and by greyish, rather than reddish (Neochromis) "kruising", H. (P.) "rockpicker", fins, including the anal fin. It is very rare at H. (P.) "elongate rockpicker"), H. (P.) "rock Hippo Island, and we found it only in greater macula" is to be considered critically endan- depths, between 6 and 7 m. gered, if it is not yet extinct. Haplochromis (Paralabidochromis) "rock Little is known about the ecology of this macula" resembles on the first glance H. (P.) species. Its former habitat is composed of "elongate short snout scraper", but differs huge, and mostly steeply sloping rock boul- from it in several characters. It has a greater ders. In this environment it was found at body depth, shorter, and relatively broader places where the huge rocks slope less jaws, and different female coloration. Its fe- steeply, which is an uncommon habitat situ- males are yellowish, with a dark chess- ation. All males that we found in 1991, were board pattern on the flanks, reminding of collected from crevices among rocks. H. (P.) "rockkribensis" females, though the colora- "rock macula" lived sympatrically with H. (P.) tion is less bright. This female coloration "rockkribensis", H. (P.) "rockpicker", H. (P.) sets the species apart from the other "short "pseudorockpicker", H. (P.) "elongate snout scrapers" with reddish males. Moreo- rockpicker" and some Neochromis algae ver, it differs from "red short snout scraper" scrapers. H. (P.) plagiodon is abundant on by shorter lower jaws, tooth shape (almost sandy bottoms adjacent to the huge rocks. subequally bicuspid outer teeth), and by a The study of unidentified fishes that narrower head, and from "short snout were collected by F. Witte and E.L.M. scraper" by a much less steep dorsal head Witte-Maas in the late 1970s in the profile, and a much less convex ventral body Nyegezi, area revealed one female indi- outline. Sexually active males are brightly vidual of a species of the "Short snout coloured: Gill covers, chest, belly and ante- scraper" lineage that does not seem to be rior flanks are deep blood red, the remain- "rock macula". Thus more species of this ing body yellow. A faint chessboard pattern complex may have existed in the central can be present. The dorsal fin is grey with Mwanza Gulf by that time. red spots and streaks, the caudal fin is grey with some red distally, and the anal fin is pale reddish. Females can have rather large, The long-head Chessboard Mbipi: and colourful egg dummies on their yellow Lobe-lipped cichlids — the anal fin. Haplochromis (Paralabidochromis) H. (P.) "rock macula" has its name because chilotes complex of its superficial resemblance with the epiphythic algae scraper H. (Haplochromis) This is a small complex (two species), "macula" (page 253). It is known only from that is probably closely related to the H. a small patch of huge rocks at the south- (P.) chromogynos complex. Among the ern entrance to the Nyegezi Bay (Mush- chessboard species its members are rec-

167 A male H. "rock macula" from Mushroom Island.

A female H. "rock macula" (Mushroom Island).

Habitat of members of the chromogynos Chamagati Island, H. chilotes and H. "short head complex: small rocks/sand interface at Bukoba. chilotes".

168 A piebald male H. chilotes.

A territorial male H. chilotes in the aquarium.

169 ognized by the combination of a long belonging to a single species, because head (head length 33 % of the standard never more than one colour type lives at length or more) with a relatively long any one place, and because differences in lower jaw ( 34.7-41 % of the head length), male breeding dress are little and not cor- small eyes (eye length between 21.2 and related with anatomical or ecological differ- 22.5 % of the head length, 25.4 % ac- ences. Geographical variation concerns pre- cording to Greenwood 1959a) and a dominantly the ground coloration, which is prominent melanin pattern. Their lower in most populations yellowish grey, in the jaws and dental arcades are slightly to fishes from Chamagati Island deep yellow strongly narrowed, their jaws iso- (in particularly lips and head underside), in gnathous, their outer teeth usually uni- those from Miandere Island almost orange, cuspid, and the inner rows arranged in 2- and in those from Senga Point pale grey 4 rows. The dorsal head profile is shallow, with pinkish lips. Males usually have light and sometimes slightly concave. The blueish flanks, tending to greenish on the lower jaw is 1.31 to 1.43 times as long caudal peduncle, a more or less orange col- as broad, the interorbital width with 21.3- oured chest, a blue dorsal fin with red 21.8 % (to 23.8 % according to Green- streaks in the soft part, a greyish red to red wood) of the head length very narrow. The caudal fin, and a pale reddish anal fin. The chest scales are small and deeply embed- intensity and extension of the orange chest ded and there are between 5 and 6 verti- coloration is again subject to geographical cal bars on the flanks. Both known spe- variation. The coloration of the flanks is su- cies develop strongly lobed lips. perimposed by a chessboard pattern, the Haplochromis (Paralabidochromis) longitudinal bands of which fade away dur- chilotes (Blgr.) 1911 is probably the most ing reproductive and territorial activity. unmistakable species among the rock- However, Greenwood (1959a) described dwelling cichlids of Lake Victoria. It is char- for H. (P.) chilotes from northern Lake Vic- acterized by a long, pointed snout, narrow toria an entirely different male coloration. tooth arcades with rather stout unicuspid The ground colour of his fishes was grey- outer teeth that are somewhat recurved, ish-black or black with a black dorsal fin that two to four inner tooth rows, lips that are had red lappets and red maculae between usually produced into small to very big the branched rays. The caudal fin was black lobes, and a conspicuous chessboard pat- basally, and light yellow or hyaline distally, tern. H. (P.) chilotes is one of the most the anal yellow or hyaline, with reddish-yel- widely distributed species among the rock- low egg dummies. If this coloration is in- dwelling Lake Victoria cichlids and the ma- deed typical for the northern populations, trix species of this complex. We know it then populations at the eastern and west- from 16 localities in the surveyed area. It is ern lake shores have to be investigated, be- furthermore known from several places in fore it can be said with certainty, whether northern Lake Victoria (Greenwood 1959a). the southern and northern forms belong to However, its distribution is less even than the same species. Recently H. (P.) chilotes that of other widely distributed Mbipi spe- from Uganda was imported to Europe by cies, such as H. (P.) "rockkribensis" or H . aquarium fish importers. These fishes re- (Xystichromis) "copper black", and H. (P.) semble in coloration the southern chilotes is absent from many sampled lo- populations (see photo in Kirner 1994). Un- calities. In spite of its unmistakable ana- fortunately neither the exact geographical tomical appearance, some doubts remain origin of these, nor of Greenwood's black about the identification of the fishes from males is known. southern Lake Victoria (see below). Females carry a prominent chessboard In spite of geographical variation in col- pattern on light silvery- to yellowish-grey, oration among populations, we consider all yellow, or grey-orange background. Fe- forms, known from the surveyed area, as males at Chamagati Island are almost as

170 bright yellow as H. (P.) "rockkribensis" fe- small rock boulders we observed H. (P.) males. A rare piebald colour morph exists chilotes moving in small groups at water in several populations of H. (P.) chilotes depths ranging from less than 1 to at least (Greenwood 1959a, 1974, Seehausen & 2.5 m. At other islands with medium sized Bouton 1996). We collected alive one indi- boulders, they lead a cryptic way of life in vidual from Zue Island (Speke Gulf), bear- crevices between rocks, from the shallow ing black blotches on a silvery whitish back- water down to greater depths. Greenwood ground (see photo). More recently I ob- (1959a) found H. (P.) chilotes not only over served two piebald individuals at Ruti Island, rocks, but also over sand in the northern a brooding female at 8 m water depth, and parts of the lake. F. Witte (pers. comm.) a subadult at 14 m depth. Under the light made similar observations in the Mwanza conditions at those depths, these fishes area of the southern lake. If these were not appeared bright white with prominent big just occasional invaders of sand bottom black blotches, dispersed over the entire habitats, then the species must have lost, body. after the Nile perch upsurge, a part of its Greenwood (1959a) described consider- former range of habitats. In seine nets able variation in the expression of the lip pulled over sand bottoms at many places, lobes, as well as in head length, and lower we only once had a H. (P.) chilotes, and that jaw length. In contrast to his observations, was directly adjacent to rocks. Greenwood we observed very little intra- and inter- seemed to observe that this species would populational variation in this respect. We be restricted to sheltered bays and gulfs. have never encountered adult individuals This is certainly not the case. It frequently without or with only weakly lobed lips. inhabits fully exposed places (e.g. the off- However, I suspect that Greenwood had inshore Miandere, Makobe and Ruti Islands), dividuals in his sample that belonged to an- probably even more often than sheltered other, as yet undescribed species, H. (P.) ones. "short head chilotes" (see below). Already The lobed lips in H. (P.) chilotes, and in small juveniles of H. (P.) chilotes are readily lobed lip cichlids from other lakes, are recognized by virtue of their narrow, pointed thought to play a role in feeding. Their ac- beaklike snout. In our aquarium raised tual function, however, is disputed (e.g. stocks, the lips start do swell at a size of Greenwood 1974, Fryer & Iles 1972). H. (P.) about 70 mm standard length. chilotes feeds primarily on insect larvae that H. (P.) chilotes inhabits predominantly are associated with epilithic Aufwuchs, gently sloping rock shores, and is abundant mainly on may fly (Ephemeroptera) larvae. only at places with very small rock boulders. Out of five males from Chamagati Island It occurs in low population densities also and two females and one male from at moderately steep shores with medium Makobe Island that we examined, three had sized rock boulders (e.g. in the Mwanza eaten predominantly may fly larvae, one Gulf), but is absent from steeply sloping may fly and chironomid larvae to equal shores with very big rock boulders. If, how- parts, two predominantly chironomids, and ever, at steeply sloping places the substrate two filamentous blue-green algae. The lat- at the foot of the steep rock cliffs is also ter two individuals, however, had very little rocky, H. (P.) chilotes can be found in those food in their stomachs, and also all others deep waters. Such a situation exists at Ruti had ingested some algae, indicating that al- Island, where the lobed lip cichlid can be gae are not the target items but ingested observed rather frequently at water depths incidentally when insects are removed from between 8 and 14 m. Sometimes, how- the rocks. Fishes that have only some al- ever, the population density or the com- gae in their almost empty stomachs, prob- plete absence of H. (P.) chilotes from a par- ably missed the target insect when pulling ticular place is not that readily explained by their lips over a rock. While pulling and habitat characteristics. At islands with very scraping its prey from the rocks, H. (P.)

171 H. "short head chilotes" and H. chilotes at Makobe island.

A male H. "short head chilotes" (Makobe). A female H. "short head chilotes".

H. cf. chromogynos from Zue Island.

172 Guano-covered rocks at Ruti Island.

173 chilotes presses its big lips against the chest and belly dull coppery red. The re- rocks, as aquarium observations have maining flanks are yellowish with a pink- learned (Seehausen 1993). It searches its ish flush, the area above the anal fin ba- prey predominantly on the underside of sis is blue-green. The orange-red chest rocks and in gaps between them. These area of H. (P.) chilotes is smaller then the are the preferred whereabouts of may fly coppery area of this male, and is in males larvae. To reach there, a fish turns from Makobe Island hardly present at all. sidewards before entering. In the Finally does H. (P.) "short head chilotes" aquarium H. (P.) chilotes is frequently ob- have six vertical bars on the flanks, ver- served digging for prey under slightly sus 4-5 in H. (P.) chilotes. overhanging rocks. Among the four individuals that we col- Females of H. (P.) chilotes are more fre- lected of "short head chilotes", two have quently encountered than males, while it fully lobed lips, one has an only slightly is the other way around in almost all other lobed lower lip, and a non-lobed but thick- species of Mbipi. This may hint at some, ened upper lip, and one has it the other as yet not known difference in ecology. way around. Greenwood (1959a) re- H. (P.) chilotes seems for instance, to be ported considerable variability of H. (P.) less territorial than most other Mbipi. It chilotes with respect to lip form, head usually lives sympatrically with other spe- length and lower jaw length. Though he cies of Chessboard Mbipi: with most spe- wrote that his sample as a whole shows cies of the H. (P.) "rockpicker" and the H. a complete intergradation of lip, jaw and (P.) "rockkribensis" complexes, with sev- dental characters (the reason for him to eral species of the H. (P.) chromogynos consider the fishes to represent a single complex and at several places with the species), he gives ranges and means of second lobed lip cichlid of Lake Victoria, measurements for two types, lobe-lipped H. (P.) "short head chilotes". and non-lobe-lipped fishes. Those of his Haplochromis (Paralabidochromis) non-lobe-lipped group are very similar to "short head chilotes" is maybe the most our measurements of "short head remarkable of the many new discoveries chilotes", while those of his lobe-lipped at rocky shores. The species, on first group exceed the range of our measure- glance, closely resembles H. (P.) chilotes, ments of chilotes (compare tables 3 and because it has very similarly lobed lips, 4). I suspect, that Greenwood had some and an equally prominent chessboard "short head chilotes" in his apparent melanin pattern on a silvery grey back- chilotes sample. Since he sorted the phe- ground. However, it differs considerably notypes according to lip size, and small from that species in a number of charac- or non-lobed lips are more common in ters. It has a much shorter head, and rela- "short head chilotes" than in chilotes, his tive to the head length, much shorter non-lobe-lipped group would have largely jaws (table 3). Hence the name "short consisted of H. (P.) "short head chilotes". head chilotes". Moreover, it differs from His lobe-lipped group would have con- H. (P.) chilotes in the coloration of both sisted of both species. If this is true, sexes. In the Sengerema region, and in "short head chilotes" would be much particular at Chamagati Island, H. (P.) more widely distributed, than is currently chilotes has a prominent yellow ground known. Apart from the lip shape, and the colour (see photo). In contrast, H. (P.) prominent melanin pattern, H. (P.) "short "short head chilotes" is at the same places head chilotes" is anatomically closer to the silvery whitish, without any trace of yel- species of the H. (P.) chromogynos com- low. The only known male in breeding plex, than to H. (P.) chilotes. dress, caught at Makobe Island, differs Basically nothing is known yet about from H. (P.) chilotes males from the same the ecology of "short head chilotes". All island and habitat, by having operculum, three record localities have moderate to

174 very gentle slopes, and medium sized to length). The chest scales are size reduced small rock boulders. At Chamagati Island and deeply embedded in three of the four we found H. (P.) "short head chilotes" at the known rock-dwelling populations and there very gently sloping shore at water depths are usually six vertical bars on the flanks. between 1.5 and 3 m, sympatrically with H. Only one species is described. We know (P.) chilotes, H. (P.) "blue rockpicker", H. (P.) from southeastern Lake Victoria currently "yellow rockpicker", and H. (P.) "rock- four populations of chromogynos-like kribensis". At the Sengerema mainland cichlids. All inhabit geographically narrowly shore we caught one female in even shal- restricted areas, and all are anatomically dif- lower water of less than a metres depth. ferent from each other. One may represent Finally, we caught a big male at Makobe Is- the real H. (P.) chromogynos but the others land at 5-6 m depth, again in sympatry with represent very probably undescribed spe- the above mentioned four other chess- cies. However, due to their rarity, they are board species, plus H. (P.) "short snout poorly studied. The scattered distribution of scraper". Like above described for H. (P.) this species complex almost certainly is the chilotes, we encountered also of "short result of recent extinction of populations. head chilotes" more females than males, Haplochromis (Paralabidochromis) contrasting with the ratio known from most chromogynos Greenwood 1959 is known other Mbipi in Lake Victoria. from several regions of Lake Victoria (Green- wood 1959). A form that may be this spe- The Haplochromis (Paralabidochromis) cies was frequently caught in the Mwanza chromogynos complex Gulf before the Nile perch upsurge (Witte et al. 1992). It inhabited rocky and sand bot- Within the Chessboard Mbipi the mem- toms in the Nyegezi and Butimba Bays but bers of this species complex are recognized has not been seen any more after 1982 by the combination of a long head (head (Witte et al. 1992). None of the forms that length more than 33% of the standard we found in the 1990s at rocky and pebble length), with a relatively long lower jaw* shores fits Greenwood's description of H. [footnote: *values of Greenwood (1959a) in (P.) chromogynos fully. One form is close to parentheses if deviating from mine] ( 35 it in body and head shape but differs in den- [32.5] to 38% of the head length), relatively tition. This H. (P.) cf. chromogynos is big eyes (eye length between 24 and known only from Zue Island in the north- 27.6[28.6] % of the head length) and a faint ern Speke Gulf. It has in both jaws five rows melanin pattern. Their lower jaws and den- of inner teeth, while H. (P.) chromogynos, tal arcades are narrow, though not as nar- according to Greenwood, should have row as those of H. (P.) chilotes. Most spe- three (rarely two or four). It is extremely rare cies have rather widely spaced, slender and at Zue Island and lives sympatrically with somewhat recurved unicuspid teeth in the the other Paralabidochromis species H. (P.) outer rows, and 3 to 5 inner rows of teeth chilotes, H. (P.) "Zue rockpicker" and H. (P.) that stand more or less upright. One spe- "rockkribensis" at about 4 m water depth cies has bicuspid outer teeth and only two over a gently sloping small rock and peb- inner rows. The dorsal head profile slopes ble substratum. I am not sure that there ex- gently to moderately steep and is straight ists actually a viable population. Greenwood or slightly concave. The jaws are usually (1959) described H. (P.) chromogynos as a isognathous (one species with a rare example of polychromatism, in which retrognathous lower jaw). The lower jaw is all females belong to the piebald morph always longer than broad, with a length/ but all males to the normal coloured width ratio between 1.26 and 1.53. The in- morph. The only individual that we caught terorbital width can be as narrow as in the at Zue Island is a female of the piebald H. (P.) chilotes complex, but also much morph, having black blotches on a clear wider (20.6-25.4[27.5] % of the head white ground. This piebald coloration is

175 A male H. "fleshy lips" from Mponze Point. very similar to that occurring in a morph of H. (Neochromis) "blue scraper", and differs from that observed in H. (P.) chilotes at Zue Island (black blotches on more silvery whitish ground). H. (P.) cf. chromogynos has slender, rather fine, widely but regularly spaced, acutely pointed, and recurved unicuspid teeth in the outer tooth rows, and unicuspid A female H. "fleshy lips" from Mponze Point.

H. "long teeth" from Matwinki Island.

176 teeth also in the inner rows, similar to what inhabited a very gently sloping shore at wa- Greenwood described for this species. Its ter depths from less than half a metre to at lips are thickened, particularly the upper least 1.5 m. We collected both sexes in ap- one and covered by a thick mucus layer. proximately equal numbers, a rare situation At Irondo Point, near the entrance to the among Mbipi. Males of "fleshy lips" are dark Mwanza Gulf, we found another representa- blue-grey on head and flanks, have a blue- tive of H. (P.) chromogynos-like cichlids. Also ish dorsal fin with red lappets and red macu- this one inhabits a rather gently sloping lae among the branched rays, a caudal fin shore, where it lives in an extensive rock that is greyish in the upper, red in the lower garden with interspersed patches of sandy half, and a red anal fin with often quite bottom. It has a longer head than H. (P.) large, dark yellow egg dummies. Males de- chromogynos, and differs in dentition from velop swollen, and sometimes very slightly the latter, and from all other forms that are lobed lips with a thick mucus cover all over, known in this complex: The outer tooth hence the name "fleshy lips". Females are rows are in both jaws composed of less golden yellow with red streaks and macu- slender bicuspid teeth, with a well devel- lae in dorsal and caudal fin and frequently oped minor cusp and a slightly recurved exhibit distinct orange yellow egg dum- major cusp, that is in most cases strongly mies with a white outer ring. Their lips are outworn. Moreover, there are only two rows less thickened than those of the males. of small inner teeth in each jaw. Finally, are Both sexes have about six vertical bars on lower jaw and dental arcade slightly the flanks. H. (P.) "fleshy lips" has slightly pointed, and not rounded, as in all other recurved unicuspid teeth in the outer rows, forms. The lips are hardly thickened, and are which are usually strongly outworn, and covered by just a thin mucus layer anteriorly less slender than in H. (P.) cf. chromogynos and anterolaterally. We caught only one fish and H. (P.) "long teeth". Its inner teeth are (a big male) yet, which was rather dull grey- long and unicuspid or weakly tricuspid as green in coloration. The considerable differ- well, and its chest scales are bigger, and not ences in jaw shape and dentition make it as deeply embedded as in the other most probable that this population repre- chromogynos-like forms and most other sents another species, for the time being Mbipi. called Haplochromis (Paralabidochromis) The second population with a narrow "pointed jaw chromogynoschromogynos". Its chest scales head, Haplochromis (Paralabidochromis) are small and deeply embedded. "long teeth" lives at Matwinki (Nyahihi) Is- Two further populations of H. (P.) land in the Sengerema region. It differs chromogynos-like cichlids differ from H. (P.) from H. (P.) "fleshy lips" by its finer, and chromogynos as described by Greenwood, acutely pointed outer teeth that are from our H. (P.) cf. chromogynos, and from strongly recurved in the upper jaw, and H. (P.) "pointed jaw chromogynos" by a nar- more widely spaced than in H. (P.) cf. rower head, and a longer and narrower chromogynos, by hardly thickened lips lower jaw. One of them Haplochromis that are free of mucus, by a more con- (Paralabidochromis) "fleshy lips" inhabits a cave dorsal head profile, and by a differ- small rock boulder and sand-rock mixed ent chest squamation with smaller and habitat at Mponze Point in the southwest- more deeply embedded scales. Its habi- ern Speke Gulf. Its population density was, tat is a very gently sloping small boulder when we discovered the species in 1993, and pebble shore. Its population density much higher than in the aforementioned must be extremely low. We caught two two populations. It was even the second individuals, one in 1993, the other one in most abundant species at Mponze Point, 1995, at exactly the same place among after H. (Xystichromis) "copper black". Unfor- small boulders and stones, at only 0.8- tunately, we could not find it there any 1.2 m water depth. They lived sym- more when we searched for it in 1996. It patrically with the Paralabidochromis spe-

177 cies H. (P.) "yellow rockpicker" and H. (P.) "rockkribensis", and with the anatomically similar H. (Psammochromis) cf. saxicola. From the latter species, "long teeth" can be distinguished by a narrower snout, the incurved dorsal head profile, and by dentition. Psammochromis of the size of "long teeth" still have many bicuspid teeth in the outer tooth rows, and deeper embedded inner tooth rows. One of the two so far collected individuals of H. (P.) "long teeth" is a black and white piebald female.

178 Rock-dwelling cichlid species of other lineages

Shallow heads and narrow jaws — scales on the chest are in the hard bottom the Psammochromis lineage dwellers (see below) distinctly smaller and more deeply embedded than those on Members of this group are rather large flanks and belly. Several species usually bear growing insect and mollusc eating species, outworn outer teeth with abraded crowns, that are anatomically not much derived from giving the teeth a bluntly incisiform appear- the generalized Astatotilapia like haplochro- ance. mine. Greenwood defines the genus At the rocky shores in southeastern Lake Psammochromis Greenwood 1980 as rela- Victoria we found at least eight haplochro- tively slender to moderately deep bodied mine species that fit Greenwood's defini- haplochromines with a characteristic physi- tion of Psammochromis. Three of them are ognomy, created by the combination of described, the others are new, undescribed thickened lips, a straight, moderately steep species. Apart from above mentioned char- dorsal head profile and a horizontal or but acters, they share a relatively long head slightly oblique mouth. The outer jaw teeth (32.2-35.9% of the standard length), rela- are tall and slender, unequally bicuspid and tively long lower jaws (37.3-43.9% of the unicuspid, their crowns recurved and either head length), a medium to large lower jaw compressed or finely acuminate in cross length/width ratio (1.3-1.7), and a relatively section. The tall and slender inner teeth are narrow head. All have size reduced chest usually implanted so as to lie almost hori- scales that are deeper embedded than zontally, are unicuspid in larger fish, and those on the body. Species of this complex deeply embedded in oral mucosa, so that can be confused with species of the H. they often appear small on first glance. The (Paralabidochromis) chromogynos complex, inner teeth being arranged in 2-4 (rarely 5) as well as with small piscivores. From the rows, there is a tendency to an increased first group, they differ by a usually longer number of tooth rows compared to the lower jaw, a higher jaw length/width ratio, generalized Astatotilapia like condition. As and by a stronger ventral inflection of the a consequence of a peculiar anatomy of the outer teeth in the lower jaw. Furthermore dentary (lower jaw bone), the outer tooth are the inner teeth in Paralabidochromis row in the lower jaw describes anteriorly a usually not long and slender. From fish eat- ventral inflection, so that the tips of the ing species, members of the Psammo- outer teeth are on one level with the much chromis lineage differ again by the inflec- shorter inner teeth. This Greenwood (1980) tion of the lower jaw teeth, by shorter lower considers "the most trenchant and diagnos- jaws and usually less acutely pointed teeth tic synapomorphy (derived character in the outer tooth row. shared by the members of a lineage) link- Similar to the Paralabidochromis line- ing members of the Psammochromis lineage, and different from the lineages of age", though it is shared with the species "Vertical bar Mbipi", the Psammochromis of the Ptyochromis and Macropleurodus lin- lineage is not confined to rock bottoms. eages, and is, less prominently, also in the H. (P.) riponianus, H. (P.) saxicola and H. (P.) Paralabidochromis lineage present. The aelocephalus have been described from mouth lies horizontally or but slightly ob- rock, shingle and sand bottoms (Green- lique, the dentary (teeth carrying lower jaw wood 1959a, 1960, Witte et al. 1992), H. bone) is very shallow, creating a slender and (P.) acidens from macrophyte vegetation, attenuated impression of the jaws. The and H. (P.) cassius from mud and sand bot-

179 The rocky shore of Ascari Island.

A male H. riponianus from Senga Point. A female H. riponianus (Sozihe Island).

A freshly caught male H. riponianus at Igombe Island.

180 A male H. cf saxicola (Sozihe Island). toms (Greenwood 1967, Green- wood & Barel 1978, Witte et al. 1992). Apart from the three described hard bottom dwelling species, we found five new species that, with the currently available information, qualify as Psammo- chromis, but may be restricted to rocky substrates. The occurrence or absence of Psammochromis in rocky habitats is sometimes not A female H. cf saxicola from Chamagati Island. easily explained by habitat characteristics. For instance Zue Island in the Mbipi. It consists of, sometimes rather ir- Speke Gulf, that is in terms of habitat and regularly set vertical bars, superimposed ecological composition of its cichlid com- by mid lateral and dorsal lateral stripes munity very similar to the Sengerema is- that are usually dissolved into elongate lands Chamagati and Matwinki, lacks any blotches. The most prominent mid lateral insectivore of the Psammochromis-line- blotch extends from the gill cover to the age that are very typical for the latter is- middle of the flank, ending between the tip lands. The species of the Psammochro- of the pectoral fin and the height of the anal mis lineage feed predominantly on large fin insertion. A second, less prominent insect larvae (may and caddis fly larvae), blotch can stretch along the caudal pedun- ostracods and small bivalves (sphaeriids). cle. The dorsal lateral stripe can be dis- Some species eat also snails. solved into blotches of varying size and po- Rock-restricted and non-rock-restricted sition but is usually most prominent along Psammochromis have a melanin pattern the central two thirds of the dorsal fin. I call that is similar to that of the Chessboard this melanin pattern a "broken chessboard".

181 Greenwood (1980, 1981) considered on large insect larvae (may fly and caddis Psammochromis to be part of a large sup- fly larvae) but to lesser extends also on posedly monophyletic complex, the bivalves and snails (Greenwood 1960). It Psammochromis-Macropleurodus super- usually has somewhat enlarged pharyn- lineage which, according to him, includes, geal jaws, provided with strong teeth, that apart from Psammochromis, all described tend to be submolariform and/or molari- oral shelling molluscivores of Lake Victo- form, suitable for crushing cases of ria and Paralabidochromis. However, caddis fly larvae and mollusc shells. We Lippitsch (1993), who studied scale and found H. (P.) riponianus at several places squamation characters of various Lake Vic- over sand bottom (mostly in not fully toria haplochromines, suggests that wave exposed areas), as well as over Psammochromis is neither a mono- sand-rock mixed substrates and habitat phyletic lineage, nor belonging into the interfaces. Less frequently we encoun- close relationship of the oral shelling tered it between 0.5 and 2.5 m depth molluscivores. over pure rock substratum, where the lat- Haplochromis (Psammochromis) ter slopes gently and consists of small riponianus (Blgr.) 1911 is known from boulders. Steeply sloping rocky habitats many sand bottoms and vegetation in the Mwanza Gulf are inhabited by zones, and from some rocky habitats in riponianus-like fishes that differ from this southern Lake Victoria, as well as from species in coloration and anatomy, and the northern parts of the lake (Greenwood most likely represent a separate, 1960, Witte et al. 1992, Kaufman & undescribed species. Ochumba 1993). However, since popu- Very similar to H. (P.) riponianus is Haplo- lations differ in coloration and at least one chromis (Psammochromis) saxicola Green- similar undescribed species exists, a re- wood 1960. It differs from H. (P.) riponianus vision of H. (P.) riponianus is desirable. in male coloration and by virtue of more Males active in reproduction, can be light slender pharyngeal bones that, in most in- metallic blue on the flanks with a light un- dividuals, do not carry molariform teeth. Fur- derside. Similarly light blue is in such thermore the lower jaw is frequently longer fishes the dorsal fin that bears a dark grey in H. (P.) saxicola than in H. (P.) riponianus, margin and red lappets. The caudal fin is the oral teeth are arranged in fewer rows, red distally, and the anal fin is light red and those in the outer row become earlier with small egg dummies. In other popu- unicuspid. Reproductively active males are, lations males are darker blue-grey to according to Greenwood (1960) dark grey- green-grey with similar, but also darker fin green, some scales on the flanks having coloration (Van Oijen 1978). In many popu- golden centres, chest and branchyostegal lations the pelvic fins are black in the membrane are black, and there is a cop- spinous, red in the branched ray part. pery-red flush on the opercula and flanks. Sexually quiescent males are silvery grey The dorsal fin is dark, with red lappets and with some blue flush on the flanks. Fe- maculae, the caudal fin is blue-grey with males are more or less intensely golden red margin, and the anal fin is blue-grey yellowish, and carry, like the males, six or with pink flush and bright yellow egg dum- seven vertical bars on the flanks. Individu- mies. The pelvic fins are black. Quiescent als collected by seine net over sand are males, as well as females closely resem- usually lighter coloured than those col- ble those of H. (P.) riponianus. lected over rocks or among vegetation. Several populations of Psammochromis However, compared to other rock-dwell- in southeastern Lake Victoria agree very ing cichlids, H. (P.) riponianus is generally closely with Greenwood's description of H. light coloured, similar to many sand- (P.) saxicola but have shorter heads and dwelling species. lower jaws than Greenwood's fishes. H. (P.) riponianus feeds predominantly Greenwood used exclusively individuals

182 from the northern lake shores for his de- crushed shells of some rather big sphaeriid scription. Since also our southern H. (P.) bivalves. riponianus have shorter lower jaws than his At Chamagati Island, H. (P.) saxicola lives northern ones, the difference between sympatrically with two similar species, an northern and southern fishes may reflect an unidentified Psammochromis and H. across-species impact of an ecological fac- ("Astatotilapia") "black long snout". At tor, correlated with geography. Like in other Matwinki Island it lives sympatrically with cases, mentioned before, final clarity, H. (Paralabidochromis) "long teeth", that has whether the southern and northern popu- a similarly long and pointed snout. Finally, lations belong to the same species or not, at Sozihe Island, it lives in sympatry with H. can also for H. (P.) saxicola be achieved only (P.) riponianus. From the unidentified after the eastern and western shores of Psammochromis, it differs by having a Lake Victoria have been investigated. In longer head, a more slender lower jaw, and agreement with Greenwood's description, a more pointed snout. From H. ("A.") "black even the biggest southern H. (P.) saxicola long snout" it differs in size, male coloration have rather slender pharyngeal bones, bear- and dentition, and from H. (P.) "long teeth" ing only slightly enlarged, non-molariform in lower jaw shape and dentition (see table teeth, and have fewer oral tooth rows than 3). H. (P.) riponianus. In both species, the teeth An undescribed Psammochromis that is are frequently strongly outworn, the apparently confined to rock substrate, and crowns abrased. Southern H. (P.) saxicola is known only from moderately to steeply have five to six vertical bars on their flanks. sloping shores of the central and northern We found H. (P.) saxicola at three places. Mwanza Gulf is Haplochromis (Psammo- At two of them (Chamagati and Matwinki chromis) "rock riponianus"riponianus". From both above Islands), the habitat is a very gently sloping described species it differs by virtue of a shore with small rock boulders and rubble. steeper dorsal head profile and a less The species occurs here already immedi- pointed snout, from H. (P.) riponianus by ately inshore in very shallow water. fewer tooth rows. Males have grey-green Subadults are found in waters of less than flanks, very much like H. (P.) saxicola, some- 0.5 m depth, full adults at 1 to at least 2 m times with a distinct red flush all over. The depth. At some places at Chamagati Island, dorsal fin is dark with red lappets and macu- in waters of less than 1 m depth over rub- lae between the branched rays, the caudal ble, H. (P.) saxicola is quite abundant. We fin is grey with red margin, to almost en- found an approximately equal sex ratio in tirely red, the anal fin is red, with orange our catches, which is a rare feature among coloured egg dummies. Unlike male H. (P.) rock-dwelling Lake Victoria haplochromines saxicola, the males of this species have the (but see under H. (Paralabidochromis) soft part of the pelvic fins red, and only the chilotes complex). At Sozihe Island in the spinous part black. In this respect, they re- eastern Speke Gulf, H. (P.) cf. saxicola (I have semble H. (P.) riponianus. There are six or not yet measured the fishes to confirm the seven vertical bars on the flanks. Females identification) is restricted to rather gently are darker brownish than those of the two sloping areas with small boulders and lives previously discussed species. there between 0.5 and 2.5 m water depth. H. (P.) "rock riponianus" lives at moder- According to Greenwood (1960), this spe- ately steep and very steeply sloping rock cies eats predominantly insect larvae shores that consist of medium sized to (chironomids and may fly larvae) and very big rock boulders. At such places it ostracods, less commonly also snails. In the lives at water depths from 1 to 4 m. It is stomach and intestine of one individual nowhere abundant and, like in other rock from Matwinki Island we found may fly lar- cichlids and unlike in the two above dis- vae, caddis fly larvae, chironomid larvae and cussed Psammochromis, males are en- ostracods to about equal parts, as well as countered clearly more often than fe-

183 A male H. "rock riponianus" from Shadi Rocks.

A male H. aelocephalus from Bihiru Island. males. The habitat requirements of H. (P.) Mwanza Gulf, possibly at more places. On "rock-riponianus" seem to be similar to superficial glance "rock-riponianus" can be those of the anatomically and in male col- mistaken for H. (Xystichromis) "copper oration very different H. (P.) aelocephalus, black", with which it lives sympatrically at and the two species live sympatrically at some places. Males of the latter, however, least at Hippo Island in the northern have entirely black rather than half red

184 pelvic fins, and very different dentition. four inhabit purely rocky habitats with Haplochromis (Psammochromis) aelo- moderate to very steep slopes in the cephalus Greenwood 1959 is the third de- central Mwanza Gulf, where they live as scribed species of the Psammochromis- real crevice dwellers. I cannot rule out the lineage that occurs in rocky habitats. possibility that the fishes from the two Greenwood described it, like the other very different habitat types represent two two species, exclusively from fishes col- different species. The males from the lected along the northern, Ugandan lake rocky islands are dark pinkish-grey all over. shores. According to him, the species A smaller or bigger orange coloured area lives over sand and rock. We found H. (P.) is situated behind the pectoral fin, fol- aelocephalus at five localities in the lowed by a large area, in which each scale southeastern lake. One population inhab- has a blue-green iridescent centre, ex- its a sand-rock interface at the southern tending along the lower half of the flank shore of the Bwiru Peninsula, the other

A female H. aelocepahalus from Bihiru Island. A female H. "Ruti-Psammo" from Ruti Island.

A male H. "blue sharp snout" from Bwiru island.

185 to the caudal peduncle. This coloration is Psammochromis lineage, and in particularly superimposed by five rather broad dark ver- a strong ventral inflection of the outer tooth tical bars. The dorsal fin is metallic purplish, row anteriorly in the lower jaw. Nothing is with red lappets and streaks among the known yet about the ecology of this rare branched rays. The caudal fin is dark pur- fish. Both individuals were caught some- plish-grey with red streaks and maculae; the what offshore between 6 and 10 m water anal fin is dark purplish-grey with bright or- depth. ange egg dummies. Females are darker Just as little known as the latter species brown than those of the previously dis- is also Haplochromis (Psammochromis) cussed Psammochromis species, and ex- "blue sharp snout" from Bwiru Island, south- hibit a slight purplish flush on the flanks. I west of Ukerewe. It is a large growing spe- do not know the male coloration of the cies that differs from the other Psammo- population that inhabits a sand-rock inter- chromis by a greater body depth and by a face. broader head. This causes its physiognomy Among all Psammochromis of the rocky to deviate slightly from that of the other shores, H. (P.) aelocephalus has the most species, its snout appears less attenuated. peculiar head shape. The snout is even Its assignment to the Psammochromis lin- longer, and more acutely pointed than in the eage is provisional. We have collected only others. However, we never found individu- one male yet. Apart from the broader head als with strongly thickened or even lobed and deeper body, it exhibits all the charac- lips as described by Greenwood (1959). The ters, typical for Psammochromis. It is dark fishes that we collected at Bihiru Island metallic blue with similarly dark fins. The have a much smaller mean and maximum dorsal fin has orange-red lappets and streaks size than those from the Mwanza Gulf. in its soft part, the caudal fin has a broad The dark coloration identifies H. (P.) aelo- orange-red margin, and the anal fin has a cephalus as real rock-dwellers. Apart from pale whitish margin with orange flush and the above mentioned individual, collected orange coloured egg dummies. The or- in a sand-rock mixed habitat, we found the ange-red coloration of the fin margins con- species in crevices between rocks in imme- trasts sharply with the dark blue coloration diately inshore waters of 1 to 3 m depth. of the fin bases and the body. We caught At the less steeply sloping Hippo Island we the fish at an exposed shore with medium found a very big male at about 5.5 m depth. sized boulders and moderately steep slope H. (P.) aelocephalus occurs usually in low in 8-10 m deep waters. Thus, this species population densities but is the numerically and H. (P.) "Ruti-Psammo" seem to live dominant crevice-dweller at Bihiru Island. It deeper than most other rock-dwelling deviates in habitat choice from the other Psammochromis species. The whitish anal rock-dwelling Psammochromis, which in- fin margin may at those depths play the role habit gently sloping small boulder shores, in courtship that I described earlier for or steeper slopes outside of crevices. deepwater-dwelling species of the H. Another morphologically peculiar species nyererei complex (H. "blue nyererei", H. "pink is Haplochromis (Psammochromis) "Ruti- anal"). Psammo"Psammo", which we collected at Ruti Island The deeper parts of rocky shores in the in the central Speke Gulf. To date we found southwestern Speke Gulf are inhabited by only two females and could not find this a species with distinctly enlarged pharyn- distinct species anywhere else. The fish geal teeth, Haplochromis (Psammochro- has a characteristic physiognomy, with a mis) "striped crusher"crusher". It seems to occupy Psammochromis head shape, but a some- in that region an ecological position similar what foreshortened snout, setting it ana- to that of H. (Labrochromis) "stone" in the tomically apart from the other Psammo- Mwanza Gulf. Apart from Astatoreochromis chromis species. Otherwise it shares all alluaudi, no anatomically specialized pharyn- above mentioned characters of the geal crushing snail eater is known from the

186 rocky shores in the southern and western outer teeth may serve this purpose, similar Speke Gulf. Thus, this representative of the to those of Ptyochromis species (see page predominantly insect eating Psammo- 190ff). H. (P.) "striped crusher" is one of the chromis lineage may have entered into a few rock-dwelling haplochromines, in vacant niche. H. (P.) "striped crusher" is in which females are more commonly en- every respect a typical Psammochromis. Its countered than males. long and slender outer teeth are in the H. "striped crusher" lives in sympatry with lower jaw strongly procumbent, and the H. (P.) riponianus at Makobe and Igombe Is- outer tooth rows exhibit a distinct ventral lands, but occupies a different microhabitat. inflection. The pharyngeal dentition of adult It differs from that species most obviously individuals is composed of submolariform in coloration. From the anatomically special- teeth. The scales on the chest are size re- ized snail crusher H. (Labrochromis) "stone", duced and somewhat deeply embedded, H. (P.) "striped crusher" differs in body though less so than in many other rock shape, oral and pharyngeal jaw dentition, cichlids. This is a rather inconspicuously col- and coloration. It is less deep bodied, has a oured species with rather little difference narrower head, a less steep dorsal head between the sexes. Both are greyish- profile, and a narrower, more attenuated brown with yellowish caudal, anal and pel- snout. The head does not have the heavy vic fins. The dorsal fin has red lappets and, appearance, that is typical for H. (L.) "stone" in the male, red streaks or maculae. Both and the Labrochromis lineage in general. A sexes have some metallic blueish to green- trenchant difference lies in the shape of the ish flush on the flanks that is in males usu- outer teeth, which are in H. (L.) "stone" ally somewhat more prominent. Both ex- much shorter, stouter, and implanted in an hibit rather prominent broken longitudinal upright position. The vertical bars on the stripes, while the vertical bars of the "bro- flanks of "striped crusher" are not as broad ken chessboard pattern" are much less dis- as those of "stone", and the latter does not tinct. exhibit the broken longitudinal stripes. H. (P.) "striped crusher" inhabits the rather Sympatry of the two species seems to be gentle slope of Makobe Island as well as rare and is known only from Hippo Island the steeply sloping Ruti Island. It is at all in the northern Mwanza Gulf, where "stone" places very rare. Like H. (Labrochromis) is frequently encountered in crevices "stone" (see below) it usually lives in sym- among inshore rocks, while we caught patry with Astatoreochromis alluaudi, but "striped crusher" only once, about 10 m off the two snail eaters are ecologically segre- shore in deeper water. H. (P.) "striped gated by depth. At Makobe Island, "striped crusher" has never been recorded further crusher" lives offshore at depths beyond 4 south in the Mwanza Gulf and is replaced m, while A. alluaudi lives predominantly in by other mollusc crushers in the northern shallower waters. At Ruti Island, "striped Speke Gulf. crusher" inhabits the only gently sloping Was "striped crusher", within the ana- small boulder habitat at the foot of the tomical range of Psammochromis, the huge, steep rocks, at 10 m depth and be- one that is most close to the anatomically yond that, while A. alluaudi again lives at specialized pharyngeal crushers, so is lesser depths. A male from Igombe Island, Haplochromis (Psammochromis) "red whose stomach we checked, had eaten zebra" anatomically close to the special- predominantly snails and ostracods, and ized oral shellers of the Ptyochromis line- smaller amounts of caddis fly and age. This very colourful species is known chironomid larvae. Surprisingly we did not from the Mwanza Gulf entrance and the find any shell fragments among the heaps mainland coast west of it, as well as from of snail soft tissue. This suggests that the Ukerewe (Nansio Bay). Quite some ana- fish shelled, rather than crushed its prey. tomical variation exists among individuals The strongly procumbent, long and slender from different places that can, however,

187 A male H. "red zebra" from the Mwanza Gulf entrance.

H. "striped crusher" from Ruti Island. not yet be evaluated because only one or long and attenuated, though not as long two individuals are known from each lo- as in the described Psammochromis spe- cality. H. (P.) "red zebra" has rather typical cies. It possesses all characters, used by Psammochromis shape with a straight Greenwood to define Psammochromis. dorsal head profile and a snout, that is However, slightly divergent from the typi-

188 cal Psammochromis condition, the inner tooth rows in the upper jaw of fishes from Ukerewe Island are arranged so as to form a broad band anteriorly in the jaw, reminiscent of the condition in the Ptyochromis lineage. Again similar to the condition in some species of the latter lineage, the lower jaw in some individuals of "red zebra" is retrognathous. Only males are known yet of H. (P.) "red zebra". They A male H. "yellow giant" from Mponze Point. carry usually seven narrow dark vertical bars on yellowish red to The pectoral fins are either entirely black, bright red flanks. The dorsal fin is or half black (spinous part) and half red anteriorly blue, posteriorly blood red or (soft part). The outer teeth in the lower with such coloured streaks and maculae. jaw are long, slender and strongly The caudal fin is bright red, the anal fin procumbent. The scales on the chest are pale reddish with yellow egg dummies. small and deeply embedded like in the

The rocky shore of Bihiru Island.

189 majority of rock restricted cichlids. that dissolve into maculae in the distal half At all four record localities, the species is of the fin. The blueish caudal fin has a simi- extremely rare. Its habitats are gentle, and lar blood red pattern of streaks and macu- moderately steeply sloping shores, with lae, and the anal fin is pale blue-grey with medium sized rock boulders. It lives slightly reddish flush distally and yellow egg dum- offshore at depths beyond 3 m. One male mies. There are six to seven vertical bars from Ukerewe, that we examined, had the on the flanks and broken horizontal lines. first part of its intestine packed with dark H. "yellow giant" lives at very gently slop- grey snail flesh, the last part with may fly ing to moderately steep shores with small larvae. Snail shell fragments were not to medium sized boulders, and over sand- present, implying that the fish shelled its rock mixed substrates. We found it also in snail prey. At several places, H. (P.) "red ze- a seine net that was pulled over sand bot- bra" lives sympatrically with H. (P.) "rock- tom. Thus, the species seems not to be re- riponianus", and with the anatomically more stricted to rocky substrates. It is that rare specialized snail sheller H. (Ptyochromis) everywhere that not much can be said xenognathus. At Hippo Island it lives in about its depth distribution. At Vesi Islands sympatry furthermore with H. (P.) "striped we caught two males at 3 and 4 m water crusher" and H. (P.) aelocephalus. From all depth. At Mponze Point we observed big four it differs in anatomy and coloration. In males in water, less than 1 m deep. This is coloration "red zebra" superficially resem- rather unusual for big insect eating haplo- bles H. nyererei, with whom it lives in chromines. The feeding habits of "yellow sympatry at Hippo Island. It differs from the giant" might be in line with those of some latter anatomically, and by having the red typical Psammochromis. A big male with extending over the entire flanks, while red enlarged pharyngeal teeth from Vesi Islands in H. nyererei is restricted to the dorsum had its intestine filled to 50% with and upper flanks. ostracods, 45% with crushed shells of Haplochromis (?) "yellow giant" is a rare, sphaeriid bivalves, and with a few larvae of large growing species that I cannot unam- chironomids and may flies. The species' biguously assign to any of the lineages geographical distribution is similar to that Greenwood described. However, it shares of H. (P.) "striped crusher". The two differ a number of characters with the species of distinctly in dentition and male coloration, the Psammochromis lineage, and is ana- and seem to be at least partly segregated tomically probably closer to them than to by habitat. any other lineage. Its dorsal head profile is steeper than in typical Psammochromis. Its Hook-teeth — the Ptyochromis lineage outer teeth are less slender than those of of snail shellers typical Psammochromis; the outer tooth row is anteriorly in the lower jaw but slightly Snail eating cichlids are in Lake Victoria ventrally inflected, and thus, the teeth are more species rich than in the other African but weakly procumbent. The teeth in the Great Lakes. Two basically different feed- inner rows stand upright, rather than being ing strategies have evolved. One group, more or less horizontally implanted; and with their specialized oral dentition, pulls the they are not deeply embedded in oral mu- soft body of the snail out from the shell, or cosa. Some individuals have the teeth on crushes the shell between the oral teeth. the lower pharyngeal bone in the median These are the so called oral shellers which rows enlarged, in others they are just are described in this chapter. The other slightly coarser than the other pharyngeal group crushes the shells of snails between teeth. Males have bright yellow flanks and enlarged pharyngeal jaws, provided with a more blue-grey dorsum. The dorsal fin is broad molariform teeth, the "pharyngeal blueish with, among the branched rays, a mill". These are the so called pharyngeal characteristic pattern of blood red streaks, crushers, described in the next chapter.

190 Both groups are dealt with in some detail tall, by more strongly recurved teeth in the by Greenwood 1974. outer rows, a broad lower jaw and a steep About 23 species of snail eating cichlids and often decurved dorsal head profile. are known from Lake Victoria that shell or From Neochromis, a lineage with similar crush their prey with their oral jaws, rather head profile, Ptyochromis differs in denti- than between the pharyngeal jaws (Green- tion and in the anatomy of the lower jaw, wood 1956b, 1957, Witte et al. 1992, from Paralabidochromis in dentition. From Kaufman & Ochumba 1993, Seehausen piscivorous haplochromines with unicuspid 1993, Kaufman & Seehausen 1995). The teeth, Ptyochromis differs by virtue of a majority of these share a particular head broad and short lower jaw, as well as in anatomy, and oral dentition with recurved tooth shape. unicuspid (or weakly bicuspid) teeth in the All described Ptyochromis, most of the outer tooth row and broad rows of usually undescribed species, and several oral shell- unicuspid inner teeth. For these species ing snail eaters of other lineages, share a Greenwood (1980) described the genus dominant melanin pattern consisting either Ptyochromis. I treat them here as species only of a prominent mid lateral stripe that of the Ptyochromis lineage. Greenwood de- can be interrupted, or of a prominent mid fined the lineage as haplochromines with lateral stripe, superimposed by a less promi- a dorsal head profile ranging from straight nent "broken chessboard". and steeply sloping to strongly decurved, a The described species of the Ptyochro- small, more or less horizontal mouth with mis lineage live predominantly over sand thickened lips and a lower jaw that is usu- bottoms. However, two, H. (P.) sauvagei and ally shorter than the upper. The slender H. (P.) xenognathus, are frequently found in outer teeth are very strongly recurved; rocky habitats. Where these species are anteriorly and anterolaterally in the lower found at rocks, sand bottom is usually not jaw implanted procumbently; in the upper far away, and is sometimes present in the jaw nearly vertical or very slightly pro- form of small pockets among the rocks. cumbent. The teeth of the inner series are During our rock shore survey we discovered small, predominantly unicuspid in fish of several new species of oral shelling snail over 90 mm standard length and arranged eaters that seem to be restricted to rocky in a broad band across the anterior part of habitats. Most of them are clearly assign- each jaw, narrowing abruptly to a single row able to the Ptyochromis lineage according posterolaterally. There are 3-9 (rarely 2) rows to their anatomy. Thus, this largely sand in the upper jaw and 2-9 in the lower jaw dwelling lineage seems to have, at least at (modal ranges 4-5, 3-4). The outer margin some places, been successful in establish- of the dentigerous surface of the lower jaw, ing itself in the complex Mbipi communi- over its anterior half, dips downwards and ties. slightly outwards (like described above for Haplochromis (Ptyochromis) sauvagei Psammochromis) so that the insertions of (Pfeffer) 1896 is one of four described spe- the outer tooth row lie below those of the cies of the Ptyochromis lineage. It is char- inner series. The range of the lower jaw acterized by a steep dorsal head profile, and length is 22-38% of the head length, the a dentition consisting of strong, recurved, lower jaw is usually longer than broad. I add in the lower jaw procumbent outer teeth. to this definition that the lower jaw is usu- Greenwood (1957) described males as ally between 1.1 and 1.6 times longer than grey-green or blue-grey with a coppery broad. sheen on the flanks and ventral aspects of A number of these characters are shared the operculum. Populations that have the with the species of the Psammochromis coppery sheen are known only from lineage, from which Ptyochromis differs northern parts of Lake Victoria. Such mainly by virtue of broad bands of inner populations, from Rusinga Island and the teeth which are relatively small rather than Migori River mouth, have several times

191 been shipped from Kenya to Europe and the covered (at Igombe Island) the first known U.S. under the names H. (P.) "Rusinga population that has, apart from blue males, sheller" and H. (P.) "Migori sheller" (Kaufman also yellow males (Seehausen & Bouton & Seehausen 1995). The populations in our 1996b). Males of the blue morph in this survey area in the southern lake, however, population are particularly bright blue on the have entirely light blue, males that can have flanks and have an anal fin that is proximally but a little coppery flush on the chest. pale blue-grey and only distally pale reddish. Interpopulation differences in male colora- Males of the yellow morph, in contrast, tion within our survey area, are in H. (P.) have an entirely red anal fin. I cannot, cur- sauvagei less apparent than among rock-re- rently rule out the possibility that we are stricted cichlids. However, recently we dis- dealing with a sibling species pair. Males

Haplochromis sauvagei; a male (above) and a female (below) from Kisumu.

192 of both forms have a blueish dorsal fin, and also among males (black blotches on whit- a proximally green-grey to blue-grey, distally ish grey ground; Seehausen 1993). The red caudal fin. Females of all known melanin pattern of H. (P.) sauvagei consists populations are brassy yellowish, and have of a simple mid lateral. Only frightened a bright yellow underside of the head. In males can show six to seven vertical bars some populations, however, entirely grey- on the flanks. While the mid lateral stripe ish females occur. is continuous in the northern populations, Many populations in our survey area have it is in southern populations interrupted. furthermore a piebald morph that is quite Until the long shores between our survey frequently found among females (black area and Rusinga Island in Kenya have blotches on brassy-yellow ground) but rarely been sampled, it must remain an open

Haplochromis xenognathus; a male from Kissenda Island (above) and a female from Nansio Island.

193 question, whether the northern populations ner teeth are usually arranged in many (3- are conspecific with those from the South, 9) rows, forming a broad band anteriorly in and which of them is Pfeffer's H. (P.) each jaw. The lower jaw is anteriorly often sauvagei, if they are not conspecific. North- that much decurved that the tips of the in- ern populations have been treated as spe- ner teeth, in lateral view, lie higher than cifically distinct from H. (P.) sauvagei by those of the outer teeth. The lower jaw is Kaufman & Ochumba (1993). often strongly retrognathous. However, in H. (P.) sauvagei lives in habitats in which all characters concerning lower jaw shape patches of sandy substrate are interspersed and dentition there is considerable indi- with patches of shelter. The shelter can be vidual variation. It remains to be investi- rocks, vegetation or both. It lives at water gated, whether the different types repre- depths from less than 1 m to at least 5 m. sent more than one species. Male H. (P.) This species is largely absent from pure xenognathus are bright light blue to green- rock habitats without pockets of sand. It ish grey. Blue males are usually caught over feeds predominantly on snails. Small, soft- sand, while most males from rocky shelled snails can occasionally be crushed substrates are more greenish. Blue males between the pharyngeal jaws. However, usually have all fins blue, the soft dorsal fin, the usual feeding mode is oral shelling: The and the upper half of the caudal fin bear red fish carefully approaches a moving snail streaks and maculae. Greenish males have and, when just a few millimetres are left more red in their fins, particularly in the anal that separate it from the prey, turns its head fin, and they frequently have half red pel- slightly sideways, waiting for a suitable mo- vic fins. Females are greyish, without the ment to get with a sudden quick bite hold bright yellow colour of H. (P.) sauvagei fe- of the soft body of the snail. This is fol- males. From Greenwood's description, it is lowed by jerking, to tear off the body or a not quite evident which of the two male part of it from the shell. Unlike in most real coloration types is the typical one for the rock cichlids, in H. (P.) sauvagei both sexes species. It may well be that H. (P.) are encountered in approximately equal fre- xenognathus from northern populations dif- quencies. It frequently lives sym- and para- fer in coloration from the southern forms patrically with other oral shelling cichlids of (Kaufman & Seehausen 1995). the Ptyochromis lineage, mainly H. (P.) xeno- While H. (P.) xenognathus, which is in our gnathus and, at Igombe Island, also with H. survey area widely distributed as an abun- (P.) "red rock sheller". dant sand-dweller, has at most rocky locali- Haplochromis (Ptyochromis) xeno- ties not or only exceptionally been caught, gnathus Greenwood 1957 seems closely it is a frequent member of the rock cichlid related to H. (P.) sauvagei and represents, community at a few places. Such a place is anatomically speaking, the more derived Igombe Island (southern Speke Gulf), and specialized one of the two. Though this where the blue sand-dwelling type intrudes species is very variable in head shape, it has into the sand-rock habitat interface from ad- generally a less steep dorsal head profile jacent sand bottom. Two others are than H. (P.) sauvagei, a longer, more pointed Kissenda Island (northwestern Mwanza snout, and a more extreme version of the Gulf) and Nansio Island (Ukerewe). The in- typical Ptyochromis dentition: The outer dividuals caught there are bigger than teeth are so strongly procumbent in the those caught over sand. At Nansio Island lower jaw that their necks lie almost hori- our bottom sampler indicated soft sub- zontally and form an extension of the lower strates among the rocks and subsequent jaw of considerable length. At the same Scuba diving revealed a peculiar habitat situ- time they are that much recurved that the ation. All horizontal rock surfaces were cov- tip of each tooth points almost vertically up- ered with organic sediment, grazed upon wards. Greenwood (1957) gives some by uncountable snails. From 5 m water drawings of this peculiar condition. The in- depth downwards, all pockets among the

194 rocks were filled up by more than 10 cm ten a faint flush of red on the dorsum. The thick layers of snail shells, among which dorsal fin is blue-grey with red lappets, the sediment had accumulated. This unique caudal fin proximally dark grey, distally red, situation, which may account for the very the anal fin is red, and the pelvic fins are high population density of H. (P.) xeno- black. A peculiarity is that several of the big gnathus, is probably due to the thick ever- orange coloured egg dummies on the anal green rainforest with high trees hanging fin are confluent in about 75% of the males. over the water, that provide continuous lit- The only other rock-dwelling haplochro- ter supply. I observed large solitary males mine that we found with confluent egg that, like other rock-dwellers, were hiding dummies, is H. (?) "large eye pseudo- in gaps among the rocks. nigricans". The chest scales of H. (P.) "striped The passibility that rock-dwelling popula- rock sheller" are usually small and deeply tions of H. (P.) xenognathus differ from the embedded. typical sand-dwelling form remains to be H. (P.) "striped rock sheller" inhabits mod- investigated. A population that has longer erately steep and steeply sloping rock lower jaws occurs at Zue Island (northern shores, with boulders of moderate and big central Speke Gulf). It lives slightly offshore, size. Subadults can be found in very shal- over small boulder rock substrate, at depths low water, but adults live at depths ranging beyond 3.5 m. The haplochromine commu- from 2 to at least 5 m, but probably down nity there is a real rock-cichlid one. Like to 15 m. The species is very rare. A big those from the other rock-dwelling male from Vesi Islands had exclusively dark populations, males from Zue Island differ grey snail flesh in its stomach. Like in many in coloration from the typical sand-dwelling other rock-dwelling cichlids, males of H. (P.) xenognathus. Moreover, the outer "striped rock sheller" are more frequently teeth are not as procumbent, as in the typi- encountered than females. cal H. (P.) xenognathus. The inner teeth are Known from just one locality is Haplo- arranged in 4 to 5 rows and the pharyngeal chromis (Ptyochromis) "deep water rock teeth are somewhat enlarged. Like in H. (P.) sheller"sheller". It is the most deep bodied of the sauvagei, and unlike in many other rock- rock-dwelling Ptyochromis species, has a dwelling cichlids, females and males of H. decurved, steep dorsal head profile, a very (P.) xenognathus are encountered over broad lower jaw, and a typical Ptyochromis rocks in approximately equal frequencies. dentition. The unicuspid teeth in the outer The species lives sympatrically with most rows are rather stout, resembling the con- other rock-dwelling Ptyochromis. dition in H. (P.) sauvagei and H. (P.) "striped A typical Ptyochromis is the new Haplo- rock sheller". Those in the lower jaw are chromis (Ptyochromis) "striped rock procumbent, those in the upper jaw are sheller"sheller". This very conspicuous species strongly recurved and also somewhat seems to be a real rock-dweller and is procumbent. The inner teeth are arranged known only from geographically isolated in 4 to 5 rows in the upper, and 3 to 4 in rocky islands in the central Speke Gulf. the lower jaw. They are unicuspid and im- Males grow large (table 3) and attain a planted so that they lie almost horizontally. stocky shape with a broad head and a The medial pharyngeal teeth can be en- steeply sloping dorsal head profile. Among larged but are not molariform. The chest the described species, H. (P.) sauvagei is scales are somewhat size reduced and closest to "striped rock sheller". The latter deeper embedded, though not as much as differs from H. (P.) sauvagei by a broader in many other rock cichlids. H. (P.) "deep wa- lower jaw, more blunt outer teeth and by ter rock sheller" differs from H. (P.) sauvagei coloration. Males carry a broad, deep black, by a broader lower jaw, from that species and uninterrupted mid lateral stripe, traces and from H. (P.) "striped rock sheller" in gen- of a dorsal lateral stripe and of three or four eral appearance and coloration. vertical bars on greyish ground that has of- H. (P.) "deep water rock sheller" is a poly-

195 Haplochromis "striped rock sheller" from Ruti Island.

A male H. "deep water rock sheller" from Python Island. morphic species. It has a blue and a yel- trasting, blood red caudal and anal fins. low-red male colour morph just like the The dorsal fin is blue-grey with red population of the snail crusher H. (Labro- lappets, and the black pelvic fins can chromis) "stone" that occurs sympatrically have some red in their soft parts. The with it (Seehausen & Bouton 1996). Males other, less abundant morph has a red of the blue morph have entirely light blue head and anterior dorsum, and greenish- flanks and, with these beautifully con- yellow lower and posterior flanks. The

196 anal fin is much less intensely coloured, only in small numbers. It is not clear, why but the dorsal fin is red. Females of this species seems to be endemic to a "deepwater rock sheller" are more colour- place that is geographically poorly iso- ful than those of most other Lake Victo- lated. Python Islands are situated in the ria haplochromines. Their flanks are grey- southern part of the central Mwanza Gulf, ish with yellow flush, their dorsal fin lap- not more than 600 m from the mainland, pets and caudal fin margin light red, and and the surrounding waters are not their anal fin beautifully orange-red with deeper than 8 or 9 m. Very few species orange-yellow spots in the position of in that region are endemic to a single is- male egg dummies. The melanin pattern land. I assume that H. (P.) "deep water rock is usually not prominent and consists of sheller" was in the pre-Nile perch time a mid lateral stripe and about five broad more widely distributed on rocky sub- vertical bars. strates in deep water, and that the popu- H. (P.) "deep water rock sheller" seems lation in the Python Islands trough is a rel- to be a real rock cichlid. It is known only ict population (Seehausen et al. in press from an eight metre deep rocky trough [b]). The species may have gone extinct between the two larger Python Islands, in most deep water rock-bottom habitats and is sometimes found elsewhere at when the Nile perch invaded them. The these islands. Different from the afore- narrow trough between the Python Is- mentioned species, it is rather abundant lands may offer more protection against in its habitat, or was abundant at least un- predation, than open rocky slopes do (see til 1993. In 1995 and 1996 we found it also under H. (?) "pseudoblue"). H. (P.)

Python Island as seen from the easternmost islet.

197 "deep water rock sheller" coexists there dark, uninterrupted mid lateral stripe. Their with other deep water dwelling rock dorsal fin is dark grey with red lappets, the cichlids like H. (L.) "stone", H. upper half of the caudal fin grey with a red "pseudoblue" and H. "deepwater". margin, the lower half entirely red, the anal Of three individuals whose stomachs we fin is wine red with large orange-yellow egg examined, two had eaten predominantly dummies. The pectoral fins are black. Fe- snails, of which we found only the dark grey males have not yet been found. H. (P.) "Zue flesh. One of these two fishes had also a sheller" lives sympatrically with H. (P.) case of a Trichoptera larva in its stomach. xenognathus on the gently sloping island The third individual had more diverse stom- shelf with small rock boulders. While the ach contents, consisting to equal parts of latter species is found from shallow to snail flesh and zooplankton (!, Daphnia) and deeper waters, H. (P.) "Zue sheller" is re- to a small part of chironomid larvae. We stricted to depths beyond 3 m and is quite found neither shell fragments nor sand rare. Two individuals, the stomachs and in- grains in the stomachs, confirming that the testines of which we examined, had eaten species is foraging on rock bottom (sand exclusively soft parts of snails. grains are commonly found in the stomachs While the above discussed three new of sand-dwelling Ptyochromis), and is shell- species of rock-restricted Ptyochromis live ing its snail prey. Some small pieces of predominantly in deeper waters, two other wood were found in one stomach. Python new species are found in shallow waters. Islands have dense evergreen vegetation of Different from the other rock-dwelling high bushes and small trees, many of Ptyochromis, they have a red male colora- which grow directly at the water level. Thus tion. Haplochromis (Ptyochromis) "red giant accumulation of dead wood at the ground sheller" must be very rare and is known of the trough between the islands is possi- from only two male individuals. One was ble and may form a grazing substratum for collected at Kissenda Island in the north- snails (see also under H. (P.) xenognathus). western Mwanza Gulf, the other one about H. (P.) "deep water rock sheller" lives in wa- 8 km further north at Irondo Point, immedi- ters beyond 4 m, and down to at least 8 m ately west of the entrance to the gulf. Both depth. Like with many other rock cichlids were bright red on the flanks, most inten- we caught many males and very few fe- sively so behind the pectoral fins. All un- males. paired fins were bright red, only the From "striped rock sheller" and "deep wa- spinous parts of dorsal fin and pelvic fins ter rock sheller" quite different in head were dark. "Red giant sheller" may deviate shape is Haplochromis (Ptyochromis) "Zue from other species of the Ptyochromis lin- sheller"sheller". The species, which is known only eage in its melanin pattern. The two males from Zue Island, is shallow bodied with a showed 6 to 7 vertical bars on the flanks dorsal head profile that is usually less steep but no mid lateral stripe. However, more in- than that of the other two species. Among dividuals have to be seen to know whether the known rock-dwelling Ptyochromis, this this is consistent. one has the least broad lower jaw (table 3). H. (P.) "red giant sheller" is one of the larg- Although it is narrow, the lower jaw is in est oral shellers. Its dorsal head profile is other respects typically Ptyochromis: decurved and slopes steeply. Its outer teeth retrognathous, dipping anteriorly down- are less strong than in the before dis- wards, and its outer teeth are procumbent. cussed steep headed species, and are The dentition of "Zue sheller" somewhat moderately to strongly recurved. The inner deviates from the typical Ptyochromis con- teeth in the upper jaw form a broad band, dition. The outer teeth are weakly bicuspid as is characteristic for the lineage. However, to unicuspid, and only slightly recurved. The in the lower jaw, this band is in one of the inner teeth are arranged in only 2 to 3 rows. individuals less broad. Even a large male Males are blueish grey on the flanks with a has many weakly tricuspid teeth in the in-

198 ner tooth rows, while individuals of the be- decurved, are procumbent, those in the up- fore discussed species of comparable size per jaw slightly to moderately recurved. The have usually only unicuspid inner teeth. inner teeth are arranged in just two rows Thus there are some indications that H. (P.) in both jaws, are tricuspid and slightly "red giant sheller" is anatomically less spe- recurved. H. (?.) "red rock sheller" is a rather cialized a snail sheller than most other elongated species with a decurved but not Ptyochromis species. steep dorsal head profile. Its chest scales At both localities this species lives at are small and deeply embedded like in the moderately steep rock shores with medium majority of the rock-dwelling haplochro- sized boulders and must be very rare at mines. Clearly is this species anatomically least at Kissenda Island (Irondo Point has less specialized for snail shelling than all only once been sampled). We caught both aforementioned Ptyochromis species. It is males at water depths of between 1 and 2 the slenderness, recurvature, and pro- m. At both places they lived in sympatry cumbent implantation of the outer teeth, with H. (P.) xenognathus, at Kissenda Island together with the relatively steep dorsal furthermore with the snail crusher H. head profile that suggest at least an ana- (Labrochromis) "stone", and at Irondo Point tomical relation of this small species to with H. (Psammochromis) "red zebra" that Ptyochromis. However, it may phylo- may be another little specialized snail genetically as well be closer related to sheller. From the latter species which has some small sand-dwelling insectivores. similar male coloration, H. (P.) "red giant Males are blueish with bright red gill cov- sheller" differs in head profile, head and jaw ers, chest and anterior flank. Further cau- width and dentition. From the second red dad the flanks become yellowish. The un- Ptyochromis like rock-dwelling species, H. derside is dark grey-green, the dorsum grey. (P.?) "red rock sheller", it differs in size, head The dorsal fin is bright blue with red streaks shape and dentition (table 3). Also from in the soft part, caudal and anal fins are other entirely red rock cichlids, it is distin- proximally blue and distally red, the anal fin guished by its head shape and oral denti- with mostly two big orange coloured egg tion. H. (Xystichromis) "red carp", that lives dummies. Females are light whitish yellow. in sympatry with H. (P.) "red giant sheller" at The melanin pattern deviates from the typi- Kissenda Island, has scraper dentition. H. cal Ptyochromis pattern, because the mid "red pseudonigricans" has a shallower dor- lateral stripe is very faint while vertical bars sal head profile, much narrower jaws, much are frequently seen. The light coloration of finer, bicuspid and not recurved outer teeth, females is typical for sand-dwelling, rather and fewer inner tooth rows. With its colora- than for rock-dwelling cichlids. The male is tion and deep body, H. (P.) "red giant sheller" rather dark, but its egg dummies are much resembles some pharyngeal crushing snail bigger than in other rock cichlids from shal- eaters from sublittoral mud bottoms (H. low waters. However, we never found this (Labrochromis) teegelari, H. (Labrochromis) species over real sand bottoms. Rather it mylergates). These species have strongly seems to be replaced over sand by the simi- enlarged molariform pharyngeal teeth. In lar H. (?) "bright red sheller" (Seehausen contrast, H. (P.) "red giant sheller" has only 1993). fine and slender teeth on the pharyngeal We know H. (?) "red rock sheller" from bones. just three places, Igombe Island, Mponze The second shallow water dwelling spe- Point, and Bwiru Point at the southern cies with red males and an oral sheller-like Speke Gulf coast. The habitat at Igombe dentition deviates even stronger from Island is a moderately steeply sloping Ptyochromis as defined by Greenwood. rock shore, composed of small to me- The outer teeth of Haplochromis (?) "red dium sized boulders. Here it is by far the rock sheller" are slender bicuspids. Those most abundant species in shallow waters, in the lower jaw, which is hardly anteriorly and lives as a real rock cichlid, though

199 A male H. "Zue sheller" from Zue island.

A male H. "red giant sheller" from the northwestern part of the Mwanza Gulf.

A male H. "red rock sheller" from Igombe Island.

200 large plains of open sand bottom are sur- A mill in the throat — rounding the rocky reef. Males occupy ter- snail crushing haplochromines ritories among the rocks from the surface to about 3 m depth, far away from the Apart from some above described spe- sand-rock interface. These territorial cies of the Psammochromis lineage that males are brightly coloured and very ac- have enlarged teeth on the pharyngeal tive, courting females that are passing by, jaws, there are currently about 13 species in much the same manner as other rock of anatomically specialized pharyngeal cichlids do. They are quite aggressive, crushing snail eaters known from Lake and frequently chase other rock cichlids Victoria (Witte & van Oijen 1990, Witte et away from their territories. Females al. 1992, Kaufman & Ochumba 1993, See- move in shoals, usually mixed with other hausen et al. in press [b]), seven of which species such as H. (P.) xenognathus, H. (P.) are described (Greenwood 1980). Unlike sauvagei, and female H. (Xystichromis) some Psammochromis that crush pre- "copper black" and H. (Neochromis) "vel- dominantly bivalves and ostracods, most vet black". These shoals struggle in the of these species crush also hard snail currents of the strong surf in shallow wa- shells like those of Melanoides tuber- ter, and are often seen over big flat rocks, culata. One, Astatoreochromis alluaudi is from which they pick up food items. not a member of the Lake Victoria spe- Shoals of female haplochromines are cies flock, and not endemic to the lake also seen at greater depths, but already (see page 255ff). Four of the five other at 2.5 m depth, H. (?) "red rock sheller" is described, and several undescribed spe- lacking among them. At Mponze Point cies inhabited sublittoral soft bottoms, "red rock sheller" lives in a very gently and disappeared in the course of the Nile sloping rock-sand mixed habitat in very perch upsurge (Witte et al. 1992). One de- shallow water. Its habitat at Bwiru Point scribed species (H. (Labrochromis) is characterized by very big, steeply slop- pharyngomylus) was living over sand being boulders, but the species is rare there. fore the Nile perch came up, and three High abundances of oral shellers in shal- undescribed species in our survey area low waters are characteristic for sandy inhabit rocks. We never found more than substrates while we did not find them one species at any one locality. Interest- over purely rocky substrates, except in ingly their distribution and that of bivalve the peculiar situation at Nansio Island crushing Psammochromis and similar (see under H. (P.) xenognathus) and at species (H. (P.) "striped crusher", H. (P.) Igombe Island. The high abundance of H. riponianus, H. (?) "yellow giant") is also (?) "red rock sheller" in shallow water at largely exclusive. However, species of Igombe Island is not quite explained. It both groups usually live sympatrically with may have to do with the proximity of the snail crusher Astatoreochromis large sandy plains. The stomach of a alluaudi. In most such cases A. alluaudi male from Igombe contained unidentifi- lives in shallower waters than the other able smashed material and lots of sand, species. its intestine contained to about equal Greenwood (1980) considered the pha- parts ostracods and may fly larvae, as well ryngeal crushing snail eaters of Lake Vic- as a few chironomid larvae. The presence toria with a strongly hypertrophic pharyn- of much sand in the stomach indicates geal mill (except Astatoreochromis that the fish has been foraging over sand. alluaudi) a monophyletic group, and gave Possibly H. (?) "red rock sheller", though them generic status, resurrecting the ge- reproducing among the rocks, partly uti- nus Labrochromis Regan 1920. He de- lizes the sand bottom that borders the fined this lineage as haplochromines char- rocks at 4-7m depth, as foraging ground. acterized by a massive hypertrophy of the pharyngeal mill, pharyngeal dentition that

201 is composed almost entirely of stout mo- about 150 mm standard length. Its distribu- lariform teeth, and stout but generalized tion is a typical central/northern Mwanza oral jaw teeth. The outer jaw teeth are of Gulf one, similar to that of some species the basic bicuspid type in small individuals of the Vertical bar Mbipi. We know it from and even in fishes bigger than 100 mm almost all sampled stations between Py- standard length, an exclusively unicuspid thon Islands in the south, Bwiru Point in the dentition is uncommon. The teeth are northeast, and Amranda Point in the north- moderately stout to stout with a west. It is absent from the southern subcylindrical neck and the crown is not Mwanza Gulf south of Python Islands and markedly compressed. The teeth forming is replaced by other pharyngeal crushing the inner series are small, tricuspid, and are species (mostly of the Psammochromis lin- arranged in 1-3 (rarely 4) rows anteriorly and eage) in the Speke Gulf. anterolaterally in both jaws. The fishes have Male H. (L.) "stone" occur in two basically a typically generalized facies, but most spe- different colour morphs (Seehausen & cies have a "heavy headed" appearance. Bouton 1996). One is blue on the entire Their mouth is horizontal or very slightly body, including the dorsal fin, usually with oblique, the jaws are isognathous, the pre- many red maculae in the soft part of the maxilla (tooth carrying upper jaw bone) is latter. The caudal fin is blue-grey with red not produced into a beak or shelf, the lips streaks and maculae, or largely red, the anal are not thickened. The dentary (tooth carry- fin is proximally blue-grey, distally faint red ing lower jaw bone) is slender and shallow, or almost entirely red, with yellow egg the lower jaw length ranges from 34-44% dummies. The other morph is bright red on of the head length (modal range 37-40%). the dorsal head surface, gill cover, anterior The only derived character in this defini- flanks, and anterior dorsum, and yellowish tion, which serves to distinguish the line- on the remaining flanks. The fin coloration age from other lineages with a generalized resembles that of the blue morph, but the facies, is the degree of hypertrophy of the red colour in the dorsal fin can extend into pharyngeal mill. This may not be a very the spinous part. Females are yellowish strong argument for a monophyletic origin brown with 4 or 5 broad vertical bars on the of the lineage. However, as I discussed in flanks. Adults can easily be identified by the chapter on taxonomy, I do not think the their massive head, and the strongly en- arguments that have been brought forward larged pharyngeal teeth that can be seen against Greenwood's phylogenetic hypoth- even with the naked eye on the living fish, esis, provide unambiguous evidence. if the latter is large enough. One just needs The rock-dwelling pharyngeal crushers to hold such a fish head-up, and look into have strongly enlarged pharyngeal bones its throat. However, if the fish is not very with broad molariform pharyngeal teeth, the large, an otoscope will be needed for mag- so called pharyngeal mill. Their oral teeth nification. Juveniles and subadults some- are usually blunt unicuspids in the outer what resemble fishes of the H. "pseudo- row and the inner teeth are arranged in not nigricans" complex. However, the red spots more than 2-2.5 rows. However, apart from in the soft part of the dorsal fin, that are usu- this, they have little in common and they ally prominent also in small individuals, of- may well be derived from more than one ten allow to identify H. "stone" even before phylogenetic lineage. Only one species ex- the pharyngeal teeth are noticeably en- hibits all characters that Greenwood used larged. At Python and at Kissenda Islands in the definition of the genus Labrochromis. we collected a few big blue males whose Another one is in some characters closer pharyngeal bones were only slightly en- to Psammochromis, and the third one to larged, and whose pharyngeal teeth were Paralabidochromis. not molariform but just a bit coarser than Haplochromis (Labrochromis) "stone" is generalized teeth. These individuals either the largest known snail eater, reaching represent a separate species or are a sec-

202 ond, most likely insectivorous trophic recurved as in "purple miller". morph of H. (L.) "stone". They live sym- At Zue Island near Nafuba in the north- patrically with blue and red males with ern Speke Gulf lives a quite different pha- hypertrophied pharyngeal mill. ryngeal crusher. This species, Haplochromis At most places where we found H. (L.) (?) "Zue crusher"crusher", differs from H. (L.) "stone" "stone", moderately steep to steep slopes, in head shape and coloration and from H. with medium sized to very large rock boul- (Psammochromis) "striped crusher" in oral ders, dominate the habitat. Subadult indi- and pharyngeal dentition, and in male col- viduals are frequently caught in shallow in- oration. Its dorsal head profile is less steep shore waters between rock boulders. than that of H. (L.) "stone", the head is less Adults inhabit greater depths. At the mod- broad and appears less heavy. The teeth in erately steep Python Islands they live some- the outer tooth row on the oral jaws are what offshore at depths ranging from 3 to short, stout, and blunt unicuspids of the at least 8 m. At very steeply sloping places, Labrochromis type. They are very slightly e.g. Anchor Island and Nyegezi rocks they recurved and slightly procumbent in the are found inshore and from 1.5 m depth lower jaw, somewhat reminiscent of the downwards. At such places the habitat over- condition in Psammochromis. The teeth in lap with the shallow water dwelling snail the inner rows are unicuspid. Males are crusher Astatoreochromis alluaudi is larg- wine reddish on the gill covers and ante- est. H. (L.) "stone" is at some places rather rior flanks, greenish on the remaining flanks. abundant, for instance in the deep trough The greyish dorsal fin has many wine red between the Python Islands, where it lives maculae in its soft part. in sympatry with another big snail eater, the I can currently not assign H. (?) "Zue snail shelling H. (Ptyochromis) "deepwater crusher" to any of the described lineages. sheller". The diet of H. (L.) "stone" consists It might be a member of the Labrochromis predominantly of snails but insect larvae lineage but may alternatively be an extreme are eaten as well (Bouton et al. submitted). crusher type of the Psammochromis line- H. (L.) "stone" seems to be one of the most age to which it shows some resemblance mobile rock-dwelling cichlids (Seehausen as well. It lives at the gently sloping small et al. in press [b]), and is one of the few spe- boulder shore of Zue Island offshore at cies that can sometimes be encountered depths beyond 4 m and is very rare. Once over sand, several metres away from rocks. more individuals have been found, its rela- While H. (L.) "stone" lives in sympatry with tionships may become more clear. oral shelling molluscivores and the crusher Haplochromis (?) "Sozihe crusher" is the Astatoreochromis alluaudi, we found it only second rock-dwelling pharyngeal crusher at Hippo Island sympatrically with another that I cannot currently assign to any lineage. pharyngeal crushing molluscivore. At Hippo Known as yet only in one individual from Island, H. (L.) "stone" lives predominantly in Sozihe Islands, the eastern most of our crevices among inshore rocks, while H. (P.) sampling places, this species may be more "striped crusher" occurs slightly offshore but widely distributed east of the surveyed is very rare. Before the Nile perch upsurge area. From both foregoing species it differs H. (L.) "stone" must have been in parapatric in coloration and dentition. The male is on contact with several mud bottom and sand the flanks blue with purplish sheen, and bottom dwelling species of the carries a melanin pattern reminiscent of a Labrochromis lineage. One of these is H. chessboard pattern, though the vertical bars (L.) "purple miller" (Witte et al. 1992) which are but faintly visible. The grey-blue dorsal is in coloration very similar to the red fin has red lappets, the caudal fin is largely morph of H. (L.) "stone". The latter differs red and the anal fin orange-red anteriorly from the mud bottom dwelling "purple and distally and grey-blue near the basis, miller" by virtue of smaller egg dummies where some small dark yellowish egg and outer teeth that are upright, rather than dummies are positioned.

203 A male H. "stone" from Anchor Island.

A female H. "stone" from Kilimo Island. A preserved H. "Zue crusher" from Zue Island.

The rocky shore of Python Island.

204 The teeth in the outer row of the 103 Mgobegobe — fish eating mm (SL) long fish are blunt and stout haplochromines of rocky shores weakly bicuspids and unicuspids, that are procumbent in the lower jaw. The inner The ecological group of fish eating haplo- teeth are weakly tricuspid and stand up- chromines can be subdivided into preda- right. The shape and arrangement of the tors of free-swimming fishes (piscivores teeth is somewhat reminiscent of that in sensu stricto), species that prey upon eggs some species of the Paralabidochromis and buccal-stage larvae in the mouth of lineage (e.g. H. (P.) crassilabris). So is the brooding females (paedophages) and those slightly thickened upper lip and the mela- that steal the eggs of spawning pairs (egg- nin pattern. However, the pharyngeal teeth snatchers) (Greenwood, 1974, Witte & van are molariform, but Kaufman and Ochumba Oijen 1990). Fish eaters sensu stricto are (1993) reported molariform pharyngeal characterized by relatively large heads and teeth in a member of the Paralabidochro- long jaws (the lower jaw length is usually mis lineage (H. (P.) "rockkribensis", page greater than 45% of the head length), 162). More fishes of this interesting snail armed with slender, recurved and acutely crusher need to be collected before more pointed unicuspid teeth, suitable to grasp about its taxonomical position can be said. their mobile prey. They are known at the We caught "Sozihe crusher" at a moder- southern lake shore as "Mgobegobe", and ately steep rock shore with rather big form the most species rich ecological group boulders, together with Astatoreochro- among the known Lake Victoria haplochromis alluaudi, but no other pharyngeal mines. More than 130 species are known crushing haplochromines. The water (M. van Oijen pers. comm.), of which 49 are depth was about 6 m. The male had eaten described. Though with the discovery of the predominantly planorbid snails of about large group of Mbipi it has become clear 2 mm shell diameter. They made up 90% that the proportion of piscivores in the Vic- of the complete intestine contents. The torian flock was overestimated in the past remaining 10% were made up by (Greenwood 1974, Goldschmidt & Witte chironomid larvae. 1992), their extraordinary diversity is an

A male H. "Sozihe crusher" from Sozihe island. Drawing by Helmut Seehausen.

205 outstanding character of the Victorian lateral regions of the premaxilla are not pro- cichlid species flock (Seehausen 1996). duced to form a distinct beak or peak. The Ecological and taxonomical diversity of outer jaw teeth are strong and recurved; piscivorous Lake Victoria haplochromines they are unequally bicuspid and unicuspid are dealt with in some detail in a number in fishes of less than 90 mm standard of publications (Greenwood 1962, 1974, length but exclusively unicuspid in fishes 1980, Fryer & Iles 1972, van Oijen 1982, over 120 mm standard length. The inner 1991). teeth are arranged in 1 or 2 (rarely up to 5) Regrettably a very large part of this preda- rows in each jaw. tor diversity has been recently lost. Con- Prognathochromis is defined as mostly comitantly with the population boom of the large growing haplochromines (70-230 mm top predator Nile perch, populations of standard length) with a body depth be- piscivorous haplochromines crumbled, and tween 24 and 45% of the standard length most of the species are currently consid- (mostly 30-34%) and a long lower jaw (41- ered extinct (Witte et al. 1992). During the 62% of the head length on species aver- past five years we found less than ten spe- age). The premaxilla are anteriorly distinctly cies of piscivores in southeastern Lake Vic- beaked or peaked. The outer teeth are toria, most of them at rocky shores. Some strong and recurved; they are mostly uni- of these have always been stenotopic rock- cuspid in fishes of more than 90 mm stand- restricted species, others lived in a wider ard length, and unequally bicuspid and uni- range of habitats before the Nile perch cuspid in smaller fishes. In one complex of boom, but survived only in rocky habitats. species, tricuspid teeth are laterally and In such cases it is quite unclear how sus- anterolaterally interspersed amongst the tainable in the long run the refugial unicuspids. The inner teeth are arranged in populations are. Most piscivores naturally 2 or 3 (rarely 1 or up to 6) rows in each jaw. occur in low densities (van Oijen 1992), and Furthermore important are a number of many of their current rocky refugia are qualitative and quantitative characters of the rather small rocky reefs and islands, that skull (neurocranium) that, according to may accommodate only a relatively small Greenwood, serve to separate the two gen- number of individuals of big predators. If era (see for details Greenwood 1980, van gene flow between these island popula- Oijen 1991). Van Oijen (1991) found that tions is limited or entirely cut off by Nile they seem to form continuous perch predation, the long term viability of morphoclines in which any division would populations is hard to predict. Another ef- be arbitrary, and thus, considered a splitting fect would be that the now isolated rem- of the piscivores into two genera not desir- nants of originally panmictic populations able. As I discussed in the chapter on tax- (with random mate exchange over larger onomy, I agree with him that some species distances), will go their own way in their may have been assigned by Greenwood further evolution. (1980) to the wrong genus, but I think the Greenwood divided the majority of the data are not suitable to reject Greenwood's fish eating Lake Victoria haplochromines subdivision of the piscivorous Lake Victo- into two lineages, which he allotted generic ria haplochromines into two major line- rank in his last revision: Harpagochromis ages. Therefore I keep here to Greenwood's Greenwood 1980, and Prognathochromis classification. Greenwood 1980. Harpagochromis is de- Regrettably very little is known about fined as large growing haplochromines hunting behaviour of Lake Victoria pisci- (146-200 mm standard length) with a body vores. This is so because the relatively depth between 30 and 42% of the stand- murky waters of Lake Victoria make under- ard length (modal range 34-36%) and a long water observation of fast moving predatory lower jaw (43-61% of the head length on fishes almost impossible. Greenwood did species average). The anterior and antero- some feeding experiments with Haplochro-

206 mis (Prognathochromis) gowersi in an Haplochromis (Harpagochromis) guiarti aquarium. He wrote the following about it. (Pellegrin) 1904 was in the "pre-Nile perch "For some minutes after the prey was in- era" the piscivore most frequently en- troduced, the predator remained stationary countered at rocky shores (F. Witte & M. or slowly aligned itself with the prey. Then van Oijen pers. comm.) in the Mwanza it suddenly darted forward, inevitably catch- Gulf. It has from that area entirely disap- ing the prey fish by the caudal peduncle. peared in the course of the 1980s. Un- Never once did I see a frontal attack. The like some other rock-dwelling cichlids prey struggled for a short while but soon that went extinct in the Mwanza Gulf, we became motionless. The predator appeared could not find H. (H.) guiarti, or similar to make no further attempt to swallow its species anywhere else either, nor has it food, although slight, almost trembling jaw been seen in northern parts of the lake and opercular movements were detectable (Kaufman & Ochumba 1993). It used to be (apparently to macerate the prey; my restricted to sand, shingle and rock words). After four or five minutes the prey substrates but was geographically widely fish was released. The greater part of its distributed in the lake (Greenwood 1962). caudal musculature had been grated away... According to Greenwood, males are The predator then either took hold of its dorsally malachite green, shading to sil- prey from behind, and continued to rasp ver ventrally. All fins are colourless except away the caudal region, or it positioned it- the black pelvics and the dark caudal fin. self for a frontal attack. The latter course re- The three to four egg dummies are bright sulted in the head and forepart of the prey orange. Females are similarly coloured being taken into the mouth. This, of course but have light yellow pelvics and but distorted the predator's mouth and small orange spots in place of the egg branchial region and it gulped and 'chewed' dummies. H. (H.) guiarti is an, anatomi- vigorously for some eight to ten minutes. cally speaking, relatively little specialized At the end of that time only the prey's cau- piscivore, whose body and head shape is dal fin protruded from the jaws. In all, the close to that of large "Astatotilapia"-like in- process of capture and ingestion took from sect eaters (e.g. "A". brownae), and whose fifteen to twenty minutes" (Greenwood inner teeth usually retain the tricuspid 1962). Thus H. (P.) gowersi and probably condition found also in various insect eat- many other Lake Victoria piscivores cannot ers. It feeds predominantly upon juvenile swallow their fish prey as a whole, but cichlids but also upon large insect larvae rather have to macerate it piece by piece and plant tissue (Greenwood 1962). between their fine and sharp pharyngeal Haplochromis (Harpagochromis) teeth (see also van Oijen 1989). serranus (Pfeffer) 1896 was one of the We currently know five species of pisci- first Lake Victoria cichlids to be described. vorous haplochromines that are still found It is a large growing species, but in terms at rocky shores. More species used to be of head anatomy also relatively little spe- found over rocks but disappeared during cialized, compared to many other fish eat- the last decade. Two ecological groups can ers (Greenwood 1962). Greenwood re- be distinguished among them. The species corded it from several places in Uganda, of one group inhabit the rock surface, and two in Kenya and one in Tanzania. We are never found in crevices between the found at two rocky islands in the Speke rocks (at least four species of "rock surface Gulf large growing fish eaters that prob- hunters"). Those of the other group live in ably belong to this species. They agree caves, crevices and gaps among the rocks, with Greenwood's redescription of the and are never found outside of such (two types in general appearance, coloration, known species of "rock crevice hunters"). All melanin pattern and dentition. A charac- rock-dwelling species belong to Green- teristic feature of the latter is that the lat- wood's Harpagochromis lineage. eral ones of the unicuspid outer teeth are

207 A female H. cf serranus from Zue island.

A male H. guiarti from the Mwanza Gulf, collected and photgraphed in 1978. more strongly recurved than the anterior lappets, the caudal grey with red margin ones. However, our rock-dwelling fishes and also the anal fin is proximally grey and are slightly more slender than the fishes only distally reddish. It carries a few or- Greenwood described. ange or orange-yellow egg dummies. Van Reproductively active males have a Oijen (in Greenwood & Barel 1978) pub- brownish grey dorsum and head, a white lished a very similar description of male lower lip, and are metallic greenish to coloration of H. (H.) serranus. Females of blueish on flanks and caudal peduncle. the rock-dwellers are light yellowish The dorsal fin is grey-blue with red brown, tending to silvery on the under-

208 A male H. "big blue hunter" from Vesi Island.

A female H. cavifrons from Vesi Island. sides, and exhibit a distinct dark mid lat- is easily recognized by its melanin pat- eral stripe and a dorsal lateral stripe that tern. is usually also distinct. The fins are trans- H. (H.) serranus differs from the similar parent. On the yellowish anal fin are usu- H. (H.) "big blue hunter" in having a more ally some distinct yellow-orange spots in strongly prognathous lower jaw, more the position of male egg dummies. If qui- strongly thickened lips, more bullate max- escent, also males exhibit distinct mid lat- illae, that are only partly covered by the eral and dorsal lateral stripes. Particularly preorbital bone, and fewer tooth rows. As when seen under water H. (H.) serranus its major habitat Greenwood (1962) men-

209 tioned shallow waters with mud, though (H.) serranus from the two rocky islands he found the species also over sand and in our research area are somewhat size shingle. We know the species from two reduced and deeper embedded, com- localities in our survey area, the gently pared to the generalized condition found sloping small rock boulder habitat of Zue in most non-rock-dwelling cichlids. Island in the northern Speke Gulf, and the Haplochromis (Harpagochromis) "big blue gently sloping rock habitat with medium hunter" is another big piscivore that prob- sized boulders of Makobe Island in the ably belongs into the closer phylogenetic southwestern Speke Gulf. The current neighbourhood of H. (H.) serranus. It differs disjunct distribution of H. (H.) serranus in from the latter by a slightly convex rather our survey area is almost certainly a con- than straight dorsal head profile; a shorter, sequence of the Nile perch upsurge. At broader and not prognathous lower jaw; Makobe Island the population density of maxillae that are completely hidden under this predator is very low and it is ob- the preorbitalia when the mouth is closed; served only in somewhat deeper waters finer outer teeth, that are less recurved, and between 3 and 6 m depth. At Zue Island are laterally not more recurved than its density is higher, or was so until a few anteriorly; more inner tooth rows; and by years ago, and the species lives at depths male coloration. Males are light blue on the ranging from less than one metre to at flanks, shading to blue-grey dorsally and on least 5 m. A 20 cm (SL) long male from the head. The dorsal fin is blue-grey with Makobe Island whose stomach contents red lappets like in H. (H.) serranus, but the we inspected, had swallowed a juvenile red in the caudal fin is more extensive and haplochromine. The length of the prey the anal fin is entirely pale red with rather fish, recalculated from the length of its small orange-yellow egg dummies. Until pectoral fins, was about 22 mm standard now we collected only males in breeding length. At a time when many females of dress. These did not exhibit a mid lateral algae scrapers guarded their fry in shal- stripe. However, females and non breeding low water at Zue Island, I observed H. (H.) males may exhibit it. serranus roaming around in those shal- H. (H.) "big blue hunter" is known only lows within shoals of non-brooding algae from the Vesi Archipelago in the central scrapers. In this way it may approach its Speke Gulf. It inhabits a moderately steeply prey, juvenile and subadult cichlids, unrec- sloping shore, with medium sized rock boul- ognized. ders, which is part of a broad rocky shelf, H. (H.) serranus seems to be one of the lying between the shore and at least 10 m few larger piscivorous haplochromines, water depth. We found large sexually ac- that survived the Nile perch boom at a tive males at depths of around 4 to 5 m, number of places. I found it in 1989 also but it is likely that the species uses the en- at rocks near Jinja/Uganda, and it has tire rocky shelf as its hunting ground. I con- more recently (1992) again been recorded sider this species critically endangered. from there by A. Meyer (pers. obs.), and Haplochromis (Harpagochromis) cavi- elsewhere in northern Lake Victoria by frons (Hilgendorf) 1888 is, like H. (H.) Kaufman and Ochumba (1993). Unfortu- serranus, one of the first cichlid species that nately nothing is known about population were described from Lake Victoria. It must structure and gene flow between the have been widely distributed before the refugial populations. For successful con- Nile perch boom, and had been recorded servation it is desirable to know whether from various places (Greenwood 1962). Af- the currently surviving populations are ter the Nile perch boom it had not been isolated from each other or not, and seen for many years. Recently we caught whether the rock-dwelling populations one female on the very rocky shelf in the are distinct from those that live or lived Vesi Archipelago, on which also H. (H.) "big over soft bottoms. The chest scales of H. blue hunter" survived. H. (H.) cavifrons has

210 a unique melanin pattern, consisting of ir- short rows of two to five, sometimes regular dots and blotches, distributed over more scales, rising from the fin basis. the entire body and head, and giving the Sometimes a single scale can exist be- fish a freckled appearance that is not known tween the rays, isolated from the fin ba- from any other Lake Victoria cichlid. Accord- sis. The fin scales can be very difficult to ing to Greenwood (1962) males have an ol- find since they can be embedded under ive to yellow-brown ground colour, shading a pigmented epithelium. to silvery below, and a dark brown head. Males of H. (H.) howesi are very dark Dorsal and caudal fins are yellowish, some- brown-black, with a flush of dark orange times with traces of deep red. The anal fin over the pectoral fins. The dorsal fin is is proximally dark, reddish to pinkish distally, dark grey with blackish lappets, the cau- and with yellow egg dummies. The pelvic dal fin brown-black, sometimes with a fins are black and pink. Females have simi- faint pinkish flush. The anal fin is dark red lar coloration, except that pelvic and caudal in its anterior two thirds and brownish fins are yellow. Our female does not have posteriorly. The egg dummies are orange- dark blotches on these fins, as described yellow to dark orange. The pelvic fins are by Greenwood. Greenwood mentioned black, with frequently some reddish flush that his females from the Mwanza area in the soft part. Van Oijen described fe- were dark olive green instead of yellow- males as dorsally dark brown-grey, brown. Our individual from the Speke Gulf ventrally on the flanks greenish, and hav- is of the yellow-brown colour type. ing greenish yellow fins. The individuals In contrast to the two above discussed used for the description were all from species, H. (H.) cavifrons has the outer teeth places in the northern Mwanza Gulf. We only slightly recurved. The individual from caught two big females in the Speke Gulf Vesi Islands agrees in this respect with the that had dark grey fins. One of them, as fishes Greenwood used for his re- well as all unidentified subadults of the description of the species, but it has a howesi group (see below) that we col- shorter lower jaw and narrower interorbital lected in the northern Mwanza Gulf, ex- width than the latter. H. (H.) cavifrons seems hibited a dark mid lateral stripe on the to predominantly inhabit hard bottoms. flanks. From its rarity in beach seine catches, and H. (H.) howesi lives in holes, crevices regular presence in offshore set gill-nets, and gaps among rock boulders, and has Greenwood postulated that it might not be never been recorded from other habitats. a member of the inshore community. Alter- These stenotopic microhabitat demands natively, pure sand habitats, that fishermen restrict the species to rock shores com- usually select for beach seining, may not posed of rather big to very big boulders. be its preferred habitat. Gill-nets can also Only among such are hollow spaces large be set over rocky or mixed bottoms. We and plentiful enough to offer suitable caught our fish on the rocky shelf in 4-5 m conditions for large growing cave-dwell- deep water. ers. The population densities of this spe- Haplochromis (Harpagochromis) cies may never have been very high. howesi van Oijen 1992 is the only de- However, within the past ten years, it al- scribed species of rock crevice hunters. most or entirely disappeared from its type Strictly speaking it is, according to van localities in the northern Mwanza Gulf Oijen (1992), not a piscivore but a crab (van Oijen 1992). The last record from the eater (see below). From all but one (oth- Mwanza Gulf, given as 1992 in van Oijen erwise very different) described pisci- (1992), is based on a mis-identification. vores of Lake Victoria H. (H.) howesi dif- The fishes, van Oijen referred to (as Bou- fers by having small scales between ton & Fermon, pers. comm.), were some spines and rays of dorsal and anal subadults, collected by Y. Fermon and me, fin. These scales are usually arranged in that turned out to be a new species (see

211 A male H. howesi from Gana Island.

A territorial male H. "orange rock hunter" from Gabalema Islands.

212 View towards Ukerewe from the plant-covered rocks of Zue Island.

213 below). The last record of H. (H.) howesi mm standard length, in van Oijen's study, in the Mwanza Gulf is from 1989 at An- preyed upon juvenile haplochromines. The chor Island (N. Bouton, pers. comm.). stomach of a female (113 mm SL) from Vesi However, in early 1996 we found H. (H.) Islands that we examined, contained a mac- howesi, or very similar fishes, at two is- erated fish that must have had about 15 lands in the Speke Gulf and at Gana Is- mm standard length. The intestine con- land northwest of Ukerewe. In the Speke tained another fish that was darkly Gulf we caught only females. They agree pigmented, and a heap of unidentified in- anatomically and with regard to the pecu- vertebrate eggs. liar fin squamation with H. (H.) howesi At Hippo Island and Bwiru Point in the en- from the northern Mwanza Gulf. I consider trance to the Mwanza Gulf, we caught a them conspecific, though final security few subadults of H. (H.) howesi-like fishes can not be achieved without knowledge in rocky crevices. Considering the fact that of male coloration. At Gana Island we howesi is not the only species of its kind caught a big male and a female. Also in the area, as was previously assumed, we these agree with the description, though cannot be sure of the identity of these the scales on the fins are fewer than in fishes. Though with the new records H. (H.) Mwanza Gulf and Speke Gulf individuals, howesi seems geographically more widely and are usually not more than two in a distributed than previously known, I con- row. sider it critically endangered. At all its recent Van Oijen sampled between 0.5 and 2.5 record places it was very rare. m water depth and got many subadults but Haplochromis (Harpagochromis) "orange very few adults of H. (H.) howesi. He as- rock hunter" is the second known species sumed that the latter may stay deeper (van of rock crevice hunters and probably closely Oijen 1992). This may well be the case. At related to H. (H.) howesi. Some subadults Sozihe and Vesi Islands adult H. (H.) howesi were collected by Y. Fermon and me at live at depths down to at least 7 m. The Gabalema Islands ("Rocky Islands" in van habitats are moderately to steeply sloping Oijen 1992) in the Mwanza Gulf entrance. rock shores which are composed of big boul- After they had reached adulthood in aquaria, ders. At Gana Island the species lives in big it became evident, that they belonged to a caves among huge boulders, and we found new species, and were not H. (H.) howesi. adults already at about 2.5 to 3 m depth. Their general appearance is very similar to Van Oijen (1992) gave extensive informa- that of H. (H.) howesi and they share with tion about food and feeding habits of H. (H.) the latter the squamation of anal and dor- howesi: In the northern Mwanza Gulf adults sal fin. "Orange rock hunter" differs from of this species used to prey upon crabs howesi by a shorter lower jaw and bigger (Potamonautes niloticus) and fish (Haplo- eyes (table 3), but most distinctly by an en- chromis and the cyprinid Rastrineobola tirely different male coloration. argentea). The biggest haplochromines Reproductively active males are light grey swallowed by this predator measured over on the dorsal head surface and the dorsum, 50 mm standard length. Van Oijen assumes and bright orange-red on cheeks, gill cover that crabs are not eaten entirely, but that H. and flanks below the lateral line. The dor- (H.) howesi is rather a kind of a crab para- sal fin is metallic blue in the spinous part, site, tearing off limbs which can be grown transparent with red streaks in the soft part, new by the crab. Subadults used to prey and has black lappets. The caudal fin is predominantly on fish, furthermore on transparent with red streaks and maculae, prawns (Caridina nilotica), insect larvae the anal fin anteriorly pinkish red, (mostly dragon flies, Odonata) and crabs. posteriorly whitish with light yellow egg The haplochromines eaten, were usually dummies that are much bigger than those epilithic algae scrapers (Neochromis line- on the anal fin of male H. (H.) howesi. The age). Already H. (H.) howesi of little over 50 spinous part of the pelvic fins is black, the

214 soft part pink red. Females are olive-green- howesi group) exist in the lake. In view of ish, lighter coloured than those of H. (H.) their rapid decline (H. (H.) howesi) and nar- howesi, but otherwise very similar to them. row geographical range (H. (H.) "orange rock Both sexes bear a prominent melanin pat- hunter"), the two known species are to be tern, consisting of a mid lateral stripe, an considered critically endangered. interrupted dorsal lateral stripe, and indications of five or six vertical bars. Various Egg and fry robbers of head markings can be prominent in males. the Lipochromis lineage The bright coloration, together with the streamlined body shape, give H. (H.) "or- Bizarre modes of feeding have evolved ange rock hunter" a most attractive appear- in a group of piscivorous haplochromines ance. that, rather than hunting juvenile haplo- To date "orange rock hunter" is known chromines, persecute mouthbrooding fe- from just one locality, Gabalema Islands in males to forcefully deprive them of their the Mwanza Gulf entrance. Even there it is brood. Amazingly little is known about extremely rare. Its habitat are crevices the hunting techniques of these so called among big rocks at a very steeply sloping paedophages ("children eaters"). This is place. Despite extensive search we never certainly due to the fact that very few of found adult individuals. Like those of H. (H.) them have been maintained in aquaria howesi they may stay at greater depths. yet. Greenwood (1959b) hypothesized "Orange rock hunter" may occur at more that paedophages obtain their prey force- places in the Mwanza Gulf entrance. How- fully by engulfing the snout of brooding ever, the circumstance that usually only females. Fryer & Iles (1972) held against subadult individuals are caught, makes the this view the alternative hypothesis that identification difficult. We found at Hippo females voluntarily jettison their eggs or Island and Bwiru Point subadults that are young under stressful conditions, and that similar to those from Gabalema Islands. the young would be eaten then by Their identification is pending. I also do not paedophages. This, they assumed, might know whether the two species, howesi and be a mechanism enabling cichlid popu- "orange rock hunter" live or lived sympat- lations to regulate their densities. In the rically at the same islands. The circum- meantime a few behavioural observations stance that H. (H.) howesi is now known on paedophagous feeding behaviour in also from the Speke Gulf, and thus from haplochromines have been made, all sup- south and north of Gabalema Islands, and porting the view that paedophagous that howesi and "orange rock hunter" differ cichlids obtain their prey by forceful means. in eye size and egg dummy size, indicates As bizarre as paedophagy may appear to that the two species are not merely replac- us, its appearance is very common in the ing each other geographically. evolution of lacustrine radiations of haplo- I observed a complete sequence of pa- chromine cichlids. Paedophagous species ternal mouthbrooding (three weeks) and are known also from Lakes Edward, guarding of post-buccal stage juveniles (an- George, Kivu, and Malawi (Nyasa). At least other three weeks) in a male of H. (H.) "or- those of the latter certainly evolved inde- ange rock hunter". This male was the only pendently of those in Lake Victoria. Ana- wild caught male that has been kept in tomical similarities between the phylo- aquaria until now, and performed brood genetically unrelated paedophages of the care only once (Seehausen in prep.). different lake systems are amazing (com- Crevice-dwelling piscivores are a consist- pare for instance Greenwood 1959b with ent component of cichlid communities at Stauffer & McKaye 1986). steeply sloping rock shores in Lake Victo- Greenwood (1980), in his last revision ria. It seems quite possible that more, as of the Lake Victoria haplochromines, as- yet unknown species of this group (H. (H.) signed all anatomically derived

215 A male Haplochromis cf melanopterus from Makobe Island in the aquarium.

These cichlid larvae were taken from the stomach of H. cf melanopterus.

216 An OB H. cf melanopterus from Makobe Island.

A mouthbrooding female H. cf melanopterus from Makobe Island. paedophages to the resurrected genus (modal range 42-48%). The lower jaw is Lipochromis Regan, 1920. He defines either broadly rounded, almost square this lineage as follows. Haplochromines shaped or boat-shaped with an anteriorly with a thick-lipped, widely distensible and rapid narrowing, making its outline more protractile mouth, and small teeth that are acute. The posterior end of the maxilla is deeply embedded in the oral mucosa (of- bullate, well visible even when the mouth ten invisible without dissection). The is shut. There are two distinct forms of body depth ranges from 27-47% of the outer teeth. In one, the tip of the crown standard length. The lower jaw length lies is inclined anteriorly or laterally ( a unique between 38 and 56% of the head length feature among Victorian haplochromines;

217 my words) whilst in the other it is either a male of an undescribed member of the vertical or fairly strongly recurved. The in- H. (L.) obesus group in an aquarium, attack- ner teeth are arranged in one or two rows ing and chasing brooding females of other and are predominantly unicuspid in large haplochromines, that Wilhelm had intro- fish (above 100 mm SL). Compared with the duced into the tank. When stressed, most teeth in equal-sized fish from other line- females spat out part of their brood which ages, those of Lipochromis are shorter. was eaten up immediately by the paedo- Moreover, the number of teeth in the outer phage. As long as the females retained rows is reduced, the posterior third of the some eggs or larvae, the paedophage con- premaxilla (teeth carrying upper jaw bone) tinued harassing them. After this led to ex- is devoid of teeth or teeth are present but haustion of females, the paedophage en- widely spaced. gulfed their snout with his mouth and Based on jaw and tooth shape, Green- sucked out remaining eggs or fry. Wilhelm wood distinguished two subgenera. Lipo- suggests that the origin of snout-engulfing chromis (or obesus complex) contains spe- may lie in a modification of fighting behav- cies with a broadly rounded lower jaw (one iour of a piscivorous ancestor, hunting on exception) and outer teeth, the crowns of already released young that are still guarded which are inclined anteriorly. Cleptochromis by the female. The snout-engulfing, as ob- (or parvidens complex) contains species served by Wilhelm, appeared to be the cli- with a boat-shaped lower jaw and straight max of a performance which resembled a or recurved teeth. By that time, seven spe- fight, and usually happened after the female cies were known and described, four in the was so exhausted that the paedophage first subgenus (one from Lakes Edward and was able to manipulate it. George), three in the other one. In this more These are interesting observations that recent work Greenwood revised his prior show that snout engulfing, that was hypoth- assumption (Greenwood 1974) that the two esized on basis of the anatomy of paedo- groups of paedophages would represent phages, indeed exists. However, the obser- two unrelated lineages, that had evolved vations tell relatively little about the impor- parallel. Since Greenwood's revision tance of such behaviour in nature. A female twelve new species of paedophages have in nature would likely try to escape as soon been discovered in Lake Victoria, though as she sees the paedophage, or latest, af- none of them has been described yet. ter it starts to attack her. Thus, it is most Greenwood (1980) already observed that unlikely that the paedophage would actu- the gap in jaw form between the sub- ally get the possibility to exhaust the female genera Lipo- and Cleptochromis can almost until she cannot resist snout engulfing any entirely be bridged by some species. Two more. More likely the paedophage must ac- of the new species from rocky shores add tually catch the female by surprise, or must considerably to this. Consequently in the engulf her snout in a real fight, in which she definition of the Lipochromis lineage the is attacking him frontally. This latter passi- lower jaw shape should be described as bility seems real, considering the circum- broadly rounded, boat-shaped to almost stance that brooding females of many litto- square shaped. H. (L.) obesus and one new ral haplochromines (all Mbipi species of species (H. (L.) "matumbi hunter") differ dis- which I observed brooding females) are ter- tinctly in melanin pattern from the other ritorial and quite aggressive. I have observed Lipochromis (see below). paedophages in nature, roaming around A few more words about the peculiar among the territories of brooding females. feeding behaviour of Lipochromis: Green- Though I have not seen them being at- wood (1974) observed in an aquarium-kept tacked by such females, it is possible that H. (L.) parvidens "behavioural patterns sug- this happens once a paedophage tries to gesting it might engulf the snout of a enter the hideout of a female. brooding female". Wilhelm (1980) observed It seems to me therefore that snout en-

218 gulfing in nature can function only in a situ- shores had size reduced and deeply em- ation, in which the paedophage intrudes bedded chest scales, while a species of the into a territory and hideout of a brooding H. (L.) parvidens complex that we caught female, thus in littoral and benthic habitats. in a weed bed, had generalized chest squa- In situations were brooding females are not mation. Knowledge about functional popu- territorial, for instance in open water dwell- lation size is an important prerequisite for ing cichlid species, a hunting technique de- developing strategies to protect the highly scribed by McKaye and Kocher (1993) on endangered species of the Lipochromis lin- paedophages from Lake Malawi, is likely to eage. A first step, however, is to know and be the more important one. These authors recognize the species. I hope the informa- observed three species of paedophagous tion on the following pages will contribute haplochromines ramming brooding fe- a little to this. males in the open water, and in this way Haplochromis (Lipochromis) melano- forcing them to jettison some eggs or pterus Trewavas 1928 is one of the taxon- young involuntarily. Many more underwater omically least studied cichlid species of observations are needed to understand the Lake Victoria. It was described from a sin- ecological and evolutionary importance of gle specimen that was caught early this the different behaviour patterns in paedo- century in the Smith Sound, the south- phages. Observations in large aquaria, how- western extension of the Mwanza Gulf. ever, could certainly contribute a lot. A third It's, for a paedophage unusually short strategy of paedophagy, "egg snatching", lower jaw, and the combination of an evolved in another lineage of Lake Victoria obesus complex head shape with a lower haplochromines, and is discussed on page jaw shape that resembles more the one 243. found in the parvidens complex, caused Greenwood's (1959b) data suggest that some reservation about the specific sta- most paedophages in Lake Victoria are not tus of the fish (Greenwood 1959b). This very much restricted to particular bottom in particular, because the specimen had types. However, with the exception of H. apparently suffered some postmortem (L.) melanopterus (known from a single fish) distortion. Much later (Greenwood 1980) and H. (L.) parvidens, all species were found more fishes were found that seemed to predominantly over hard bottoms, thus fit the description of H. (L.) melanopterus. sand, shingle or rock. Most paedophages However, details about these have never have become very rare after the Nile perch been published. upsurge and many disappeared entirely. So We found at rocky shores and islands in have H. (L.) maxillaris, H. (L.) microdon, and the southern Speke Gulf and Sengerema at least two undescribed species after 1986 region a paedophage that, with some prob- entirely disappeared from former places of ability, is H. (L.) melanopterus. It has a short occurrence in the Mwanza area (Witte et al. lower jaw, though not as short as in the type 1992). Most of the currently surviving spe- specimen (compare tables 3 and 4), has a cies are found only at rocky shores and is- typical obesus complex head shape and a lands. Unfortunately very little is known lower jaw similar to that of parvidens com- about their ecology and life history. Thus it plex species. It narrows rapidly in its ante- is not clear, whether these are local rior half and is pointed. This shape of the populations that are isolated from each lower jaw, and its small length, serve to dis- other by stretches of non-rocky bottom and tinguish H. (L.) melanopterus from the oth- deep water, like many other rock cichlids, erwise similar H. (L.) maxillaris. Other dis- or whether more frequent migration exists tinguishing characters are a chest and between localities. There is some evidence nuchal squamation with size reduced scales that Lipochromis species inhabiting rocks in H. (L.) melanopterus versus generalized are indeed particularly adapted to that habi- scales in the other species and a higher tat type: All species that we found at rocky number of scale rows on the cheek (3-4

219 The island Sozihe in the Speke Gulf.

A male H. cryptodon from Mafwinki Island.

220 versus 2-3). In all these characters the venile Haplochromis and a few Haplochro- fishes from rocky habitats in the Speke mis eggs (see photo). Though they had be- Gulf agree with the type specimen. gun to be digested, in some of the juve- Reproductively active males are very niles a yolk sac was visible. Thus they must beautiful, but rarely seen. Apparently only have been eaten in a stage in which they large males develop the bright colours. were still in their mothers mouth (buccal Their head, dorsum and caudal peduncle stage). I frequently had the opportunity to are dark brown, the underside is grey- observe H. (L.) melanopterus under water brown. Framed by these dark colours, the in its natural habitat. In times in which many flanks are bright yellow, interrupted by about brooding Mbipi females are hiding among seven dark brown-black vertical bars. All ver- rocks and under overhanging shore vegeta- tical fins are dark, the dorsal fin with a me- tion at Makobe Island, also the paedo- tallic blue band in the spinous part, and the phages accumulate in the very shallow wa- anal with some blueish flush and small yel- ter, and are seen roaming slowly around low egg dummies. Females are light among the Mbipi. Aquarium observations brown-grey, with up to eight narrow verti- learned me that females of H. (L.) cal bars, that are very regular in outline and melanopterus attempt to obtain eggs and position. This melanin pattern is reminis- fry even from brooding females of their cent of the one in the Mbipi species of the own species, but are never successful. Haplochromis nyererei complex. H. (L.) They slowly approach a brooding female melanopterus has an orange-blotched (OB) from laterally behind and above. When morph, similar to those of Mbipi species of about 10 cm from her mouth, they turn the Neochromis complex. slightly to one side, as if ready to attack. Lat- The type specimen of H. (L.) melano- est in this moment the brooding female pterus was caught over mud bottom. We simply turns away. When the harasser is at- have never found the species over such tacked by the brooding female, attacks con- bottom, nor in any other habitat apart from sist of lateral display and tail beats only. rocky shores and islands. We found it liv- They miss the frontal display and frontal ating at gently to moderately steeply sloping tack component that is otherwise charac- shores with small to medium sized boul- teristic for haplochromine combat behav- ders. We observed adults between 0.5 and iour and is performed in other species also at least 6 m water depth, but with higher by brooding females. A frontal display and frequencies in shallow waters. This is the attack would offer the paedophage oppor- most widely distributed and most fre- tunities to get hold of the mouth of the quently encountered paedophage at rocky brooding female. Thus, the avoidance of shores in southeastern Lake Victoria. This this component of combat behaviour may notwithstanding, it occurs generally just in be an adaptation, protecting brooding fe- very low densities. Quite extensive aqua- males of H. (L.) melanopterus against loss rium observations showed that this species of offspring through the feeding mode of exhibits a far lower level of intraspecific ag- its own species. The defence behaviour of gression than any other rock cichlid that has brooding H. (L.) melanopterus females been kept in aquaria, except H. (L.) shows with what simple means a paedo- "Matumbi hunter". Dominant males are very phage can be disarmed. tolerant of other males and females in all H. (L.) melanopterus lives in sympatry stages of sexual activity. Even brooding fe- with H. (L.) cryptodon and possibly with males are always accepted in the males H. (L.) "velvet black cryptodon" but is eas- territory and are not harassed as in other ily told apart from both on basis of body haplochromines. and head shape, and coloration. Of three individuals whose stomachs we Haplochromis (Lipochromis) cryptodon examined, two had empty stomachs, and Greenwood 1959 is according to Green- one had the stomach entirely filled with ju- wood the anatomically least specialized

221 among the described Lipochromis spe- fins are black and the pectoral fins have a cies. Greenwood (1959b) observed a very red flush. We have not yet seen females patchy distribution of the species, which but Greenwood (1959b) describes them as he could not explain on grounds of habi- dark green-brown, shading to light gold tat choice. Though he does not explicitly ventrally. The combination of coloration and state so, the predominant habitat seems to slender, streamlined body shape gives H. have been sandy bottom. Witte et al. (L.) cryptodon a very conspicuous and beau- (1992) found H. (L.) cryptodon or a very simi- tiful appearance. At rocky islands it has been lar species (H. "black cryptodon") in the pre- recorded at water depths ranging from 1.6 Nile perch times over a mud bottom in the to 5 m, at places with medium sized to small Mwanza Gulf. In the 1990s we found H. (L.) rock boulders, but is extremely rare. cryptodon only at gently to very gently slop- Similar and probably closely related to H. ing rocky islands. (L.) cryptodon, but at the same time very The taxonomic identity of H. (L.) crypt- distinct, is Haplochromis (Lipochromis) odon is problematic because the coloration "velvet black cryptodon"cryptodon". We discovered it of living males was not known to Green- in early 1996 and know it from only one wood and is thus lacking in the species de- place. It may however, be more widely dis- scription. Because of this uncertainty the tributed east of our survey area. This spe- cryptodon-like fishes from the Mwanza Gulf cies is distinctly bigger than rock-dwelling were usually referred to as "black cryptodon" H. (L.) cryptodon, but as slender as the lat- (Witte et al. 1992). However, the colour of ter. It has a broader lower jaw than any other preserved males from rocky shores agrees species of the parvidens-group, and bridges quite well with that described by Green- in this respect the gap between that group wood for preserved males of H. (L.) and the obesus-group. H. (L.) "velvet black cryptodon. They are not black (unlike those cryptodon" differs in coloration from H. (L.) of H. (L.) "velvet black cryptodon") but dark cryptodon as described by Greenwood, grey-brown. Greenwood included in his de- and as described above from rocky islands: scription four fishes that differed from the Males are entirely velvet black, literally from others by being more slender (27.5-31 % head to tail, including all fins. Sometimes of the standard length). Since these fishes some red streaks exist in the caudal fin and were amongst the smallest individuals, he the pelvics can have a flush of red. The egg considered the slender body a "juvenile" dummies on the anal fin of the males are character. All our individuals belong to this more, much smaller and darker than in H. slender type, and are in size indeed at the (L.) cryptodon. They are dull orange-yellow lower end of the range of individuals used instead of light yolk-yellow. Preserved indi- by Greenwood for the description of H. (L.) viduals retain a deep velvet black coloration, cryptodon. Nevertheless, it can at present a rare condition among Lake Victoria not be excluded that the current definition cichlids. It is known from a few Mbipi spe- of H. (L.) cryptodon includes more than one cies, but even those hardly reach the deep species. saturation of the black displayed by this Males of H. (L.) cryptodon from rocky habi- paedophage. "Velvet black cryptodon" also tats are dark grey-black on the head, dor- shares a beautiful orange blotched (OB) sum and underside, while the flanks are morph with those Mbipi. The only female slightly lighter grey with greenish to blue- that is known yet, is an individual of this OB ish metallic flush. The overall impression, morph. H. (L.) "velvet black cryptodon" has nevertheless, is that of a grey-black bodied short, stout, slightly recurved outer teeth fish. The dorsal fin is black with red lappets that are so deeply embedded in the oral and some red maculae, the caudal fin is mucosa that they are frequently not visible. largely bright red, and so is the anal fin, This attractive fish lives closely inshore which carries usually two large yolk-yellow at water depths ranging from 3 to at least egg dummies on red ground. The pelvic 7 m. The rock boulders in its habitat are of

222 moderate to large size, and the slope is outer row of two types of teeth. A rather partly moderate, partly steep. It is the only slender type with orange crown is recurved, paedophage that we found at that place, while a stouter type with white crown is though H. (L.) melanopterus may occur in upright, and the crowns frequently broken the low densities that are usual for this spe- off. cies. H. (L.) "velvet black cryptodon" seems Live coloration of H. (L.) microdon was not to behave more aggressive than H. (L.) known to Greenwood when he described melanopterus when kept in aquaria. the species. However, it was described later In the northern Mwanza Gulf we found a by M. van Oijen in a paper of Greenwood single female of another undescribed & Barel (1978). According to that paper and paedophagous species that is anatomically unpublished photographs (Witte pers. close to H. (L.) cryptodon and H. (L.) comm.), males of H. (L.) microdon are grey melanopterus, and thus, bridges the gap and yellowish with an extensive coppery to between the two groups of described bright red area behind the pectoral fins, paedophagous species. It has a shorter and somewhat reminiscent of H. (Paralabido- broader lower jaw than have H. (L.) chromis) "rockkribensis". Such fishes were cryptodon and H. (L.) "velvet black not caught over rocks. Our male from a cryptodon", which, however, is anteriorly rocky shore was metallic light blue on the rapidly narrowing. Its dorsal head profile is entire flanks, with orange-yellow anal fin, of the obesus-group type but its dentition, caudal fin, and dorsal fin lappets, and with recurved unicuspid outer teeth in both jaws, orange-yellow egg dummies. While more of the parvidens-group type. Without than five mud, sand and rock-dwelling knowledge of male coloration it is difficult paedophages are known that anatomically to judge whether this species is identical resemble H. (L.) cryptodon, "blue microdon" with one of several undescribed paedo- seems to be the first new species resem- phages that are known from sublittoral mud bling H. (L.) microdon. We found it in shal- bottoms, or whether it represents another low water at a gently sloping shore with rock-dwelling species. Its chest squamation small rock boulders in the northern Speke is of the type, found in all the other Gulf. Frans Witte (pers. comm.) had seen paedophages from the rocks and not of the similar fishes in the Speke Gulf before the type found in H. (L.) parvidens. It lives at a Nile perch boom. moderately steep slope, in a medium boul- Haplochromis (Lipochromis) "matumbi der size habitat and we caught it at about 2 hunter" is the most peculiar among the m water depth. new species of the Lipochromis lineage Discovered just in 1995, Haplochromis and one of the most peculiar of all the (Lipochromis) "blue microdon" is probably new rock-dwelling haplochromines. It is the most rare of the Lipochromis species the combination of an unusually slender currently inhabiting rocky shores in south- body, reminiscent of pelagic zooplankti- eastern Lake Victoria. With its incurved dor- vores, with an obesus-group "pug head" sal head profile, oblique mouth and rela- and an, among rock-dwelling cichlids very tively narrow lower jaw, "blue microdon" is rare melanin pattern, that creates this pe- among the rock-dwelling Lipochromis spe- culiar impression. However, the peculiari- cies most distinct. It resembles anatomi- ties are more and include dentition. The cally H. (L.) microdon (Blgr.) 1906 of the teeth of "matumbi hunter" are less deeply parvidens group, but has a slightly shorter embedded than in other species of the and narrower lower jaw (table 3 versus ta- lineage and those in the outer rows, par- ble 4). This deviation could be due to the ticularly in the lower jaw, can be rather relatively small size of the only individual long and slender. In spite of this, the outer that we collected and measured, but its col- teeth of the lower jaw are distinctly in- oration is distinctly different from that of H. clined rostrad like in other members of (L.) microdon. Its dentition consists in the the obesus-group. Those in the upper jaw

223 A territorial male H. cryptodon from Makobe Island.

A male H. "velvet black cryptodon" from Sozihe Island.

224 An OB female of H. "velvet black cryptodon" from Sozihe Island.

H. "Matumbi hunter" in the aquarium. A male H. "blue microdon" from Zue Island.

A male H. "Matumbi hunter" from Matumbi Island.

225 are upright or slightly recurved. The inner specific aggression. Females, while teeth are unicuspid even in fishes of be- guarding their fry, make an exception to low 100 mm SL. The lower jaw is nar- this. Among the species of Lake Victoria rower than in other species of the obesus- haplochromines that I have been breed- group but not anteriorly narrowing as rap- ing in aquaria, "matumbi hunter" behaved idly as in the species of the parvidens- exceptional in that it hardly preyed upon group. H. (L.) "matumbi hunter" shares fry of its own species. Fry grew up in the with H. (L.) obesus a melanin pattern of a midst of a group of adults. continuous and very broad mid lateral A real curiosity is a fish that is known stripe. I did not see such a stripe in any from only a single big male. On first glance other Lipochromis, nor in any other rock- it is a Haplochromis nyererei, closely re- dwelling cichlid. This may indicate a close sembling the males of the population with phylogenetic relationship between the which it lives sympatrically. Just when look- two species. The chest squamation of ing a second time we realized that the fish "matumbi hunter" is a typical rock cichlid has an obesus-like "pug head" with a snout squamation with small and deeply em- that is much broader than in any normal H. bedded scales. nyererei. Further inspection of the fish re- H. (L.) "matumbi hunter" is known from vealed that its anterior outer teeth are a single, topographically very narrowly re- stouter than in H. nyererei and that outer stricted place in the southern Mwanza and inner teeth are deeply embedded in Gulf. It is most likely that its occurrence oral mucosa. Moreover, the last part of its there is a remnant of a once more wide premaxillae is free of teeth, apart from one spread distribution, and I consider it to be tooth on the left side. Sympatric males of one of the most critically endangered H. nyererei have their teeth distributed in cichlid species at the rocky shores in regular distances over the entire length of southern Lake Victoria. Its habitat is shal- the premaxillae. Thus, anatomically the low water (1 to 2 m depth) over large, but pug-headed fish is a typical paedophage. I only moderately steep rock boulders. This would still consider it a morphologically ab- is a rare habitat situation, since large boul- errant individual of H. nyererei, were differ- ders mostly slope steeply. Peculiar habi- ences restricted to characters of jaws and tat demands may contribute to the rarity teeth. However, the egg dummies of the of this species, though one would see lit- "pug-head" are bigger than in any known tle reason for a paedophage to be that species of the H. nyererei complex and al- narrowly habitat specific. It may, however, most twice as big as in individuals of the be that this species depends more than sympatric population of H. nyererei (3.64 +/ other paedophages on a mixed diet. We - 0.39 mm versus 1.88 +/- 0.16 mm diam- studied the stomach contents of two in- eter, excluding the transparent outer ring). dividuals. One had eaten predominantly It appears not plausible why an individual fish eggs (probably of haplochromines), of H. nyererei that has an aberrant and to smaller amounts may fly (paedophage like) jaw structure, oral mucosa (Ephemeroptera) larvae and zooplankton and dentition, should also have an aberrant (Cladocera). The other one had eaten only egg dummy size. Rather it may represent may fly larvae (predominating) and a paedophage in H. nyererei dress. It could zooplankton. either be a morph or species of the H. We were lucky enough to build up a nyererei complex or a Lipochromis that has breeding stock of this unique species and adopted the coloration of H. nyererei, pos- one can hope that aquarium observations sibly mimicking the males of its prey spe- will contribute to an understanding of the cies. Until more is known I call it ecological significance of its peculiar mor- Haplochromis (?) "nyererei paedophage"paedophage". phology. Like H. (L.) melanopterus also this species exhibits very little intra-

226 Silver arrows of the are also black, though the caudal fin has a "Double stripe" complex red margin, and the anal fin is partly red. In contrast, males of H. tanaos become dark Anatomically very different from all other bluish dorsally, and black on the entire ven- rock-restricted cichlid species are the mem- tral half of the body. H. thereuterion has bers of a "new" species complex that has slightly larger teeth than H. tanaos, and the recently been described by Van Oijen & outer ones in the lower jaw are somewhat Witte (1996), and for which no (sub)generic procumbently implanted. Finally, H. name is available yet. These are tiny and thereuterion differs from H. tanaos by hav- very slender insect- and zooplankton eaters, ing the premaxilla (upper jaw bone) slightly probably closely related to Haplochromis laterally expanded (visible when the fish is (?) diplotaenia Regan & Trewavas 1928. H. viewed from dorsal, as a slightly convex, diplotaenia has long been a "forgotten" spe- rather than a straight snout outline) and cies until van Oijen & Witte redescribed it. slightly thickened lips. It is known from a single female that was H. thereuterion is known only from the caught in Uganda. Regrettably ecological northern Mwanza Gulf. It lived at rocky information is not available, neither about patches of mainland shore and at rocky its habitat, nor about its diet. H. diplotaenia islands, being recorded from the Nyegezi has a characteristic melanin pattern of two Rocks, Nyegezi Bay, Anchor, and Hippo Is- dark, thin longitudinal bars (mid lateral and lands. The one individual from Hippo Is- dorsal lateral stripe) on silvery whitish land in the north-western Mwanza Gulf ground. Similar fishes have been exported differed from those from the eastern from Uganda to Sweden under the name Mwanza Gulf localities in morphometrics Prognathochromis sp. aff. longirostris by B. and coloration, possibly suggesting re- Selbrink in the 1980s (Selbrink 1985b). Simi- stricted gene flow between populations lar fishes have also been observed in the across the Mwanza Gulf (van Oijen & southern Tanzanian part of the lake since Witte 1996). H. tanaos is known from 1978, and were referred to as H. "double several places along the entire southern stripe" (Witte et al. 1992, Seehausen 1995b, shore of Lake Victoria. Both "double Seehausen & Witte 1995). The recent stripe" species have declined dramatically taxonomical study demonstrates that the after the Nile perch upsurge. The two southern "double stripe" comprises two lived (past tense! See below) at some species, both of which are specifically dis- places parapatrically, e.g. in the Nyegezi tinct from H. diplotaenia. The two species Bay. The rock-dwelling H. thereuterion oc- differ most markedly in habitat and diet. curred more inshore than the sand- Before the Nile perch became abundant, dweller. Though being confined to rocky one of them, Haplochromis (?) tanaos van substrates, it is, in contrast to most other Oijen & Witte 1996, inhabited sand bot- rock cichlids, not a bottom oriented fish, toms. It is dealt with here only because of but seems to be more surface oriented. its great resemblance with the other one. It feeds predominantly on terrestrial in- Haplochromis (?) thereuterion van Oijen & sects which it apparently collects from Witte 1996, was observed exclusively over the surface. Furthermore it eats chiro- rocks. Though some morphometric differ- nomid larvae. The sand-dwelling H. ences between the two species exist (see tanaos used to feed predominantly on van Oijen & Witte 1996), females and qui- zooplankton and insect larvae in the pre- escent males look very similar. Quiescent Nile perch era (op. cit.). The great similar- males of both species are dorsally blue-grey ity in anatomy and male coloration, and with some red in dorsal and caudal fins, a the ecological differences among these red anal fin, and black pelvic fins. When species may indicate a recent speciation sexually active, those of H. thereuterion get event, in which diverging habitat a pitch black head and body, and most fins (substrate) affinities may have played a

227 A male H. "nyererei paedophage" from western Mwanza Gulf.

A female H. thereuterion from Kissenda Island. major role. earlier, we found a population of what For many years after the Nile perch up- seems to be a remnant stock of H. tanaos surge, neither of the two species was in the sublittoral waters of the Kissenda seen. However, in January 1996 we Bay over a mud bottom (Seehausen & caught one female of H. thereuterion at Witte 1995, van Oijen & Witte 1996, Kissenda Island in the north-western Seehausen et al. in press [b]). This popu- Mwanza Gulf. Hence there is some hope lation had not only changed its habitat but that this species is surviving, though in also its diet considerably, feeding now extremely low densities. About two years predominantly upon insect larvae. This

228 instance indicates how rapidly adaptation Teeth like a scalpel — algae scrapers to new ecological demands can occur. H. of the Haplochromis lineage thereuterion is by its melanin pattern easily told apart from pelagic zooplanktivores Dental specialization for scraping algae that occasionally enter into rocky habitats, has evolved among Lake Victoria cichlids and that have merely one, if any visible along at least two pathways with quite lateral stripe (see pages 256 and 257). different results. One type of algae scraper dentition is represented by the Neochromis lineage (page 66). The sec-

A male H. tanaos from the southern part of Lake Victoria.

A female H. tanaos collected at Juma Island.

229 ond type is characterized by bicuspid or uni- tooth crowns are usually strongly abrased. cuspid teeth with a compressed crown, and It may be interesting to note here that teeth a major cusp that is laterally strongly ex- of the Neochromis dentition type hardly panded (protracted) to form a broad scalpel- ever show traces of abrasion, and in this like scraping surface. In his last revision, way prove to be more durable in rock-scrap- Greenwood (1979) restricted the genus ing. Haplochromis Hilgendorf, 1888 to species Greenwood (1979) defined the genus with such tooth form. At least 17 are known Haplochromis (sensu stricto) as haplochro- from Lake Victoria (Witte et al. 1992, mines that have the crowns of their outer Kaufman & Ochumba 1993, Seehausen et jaw teeth compressed and noticeably ex- al. in press [b]), of which merely two are de- panded relative to their slender, cylindrical scribed. One more is described from Lake neck and body. The major cusp in bicuspid Nabugabo (H. (H.) annectidens), an eastern teeth is very much larger than the minor satellite of Lake Victoria, one from Lakes one. The compressed, anteriorly protracted Edward and George (H. (H.) limax) and one and dorsoventrally expanded major cusp from Lake Kivu (H. (H.) astatodon). I refer to gives the tooth, be it bi- or unicuspid, the them here as to the Haplochromis lineage appearance of having an obliquely trun- (sensu stricto). cated crown. The number of rows of inner While the Neochromis dentition type is teeth in both jaws lies between 4 and 6 (but restricted to fishes living in rocky habitats, between 2 and 6 according to Green- and is among them apparently very suc- wood's (1956b) species descriptions). The cessful, the Haplochromis dentition type is inner teeth are tricuspid, but in some spe- found predominantly in habitats with sub- cies the anterior and anterolateral teeth in merged vegetation, e.g. water-lily zones the outermost one of the inner rows may and floating reed mats. However, it occurs be identical with those of the outer row. All in a few species also in rocky habitats. jaw teeth are moveably implanted. Greenwood (1956a) noted that the two by Most of the new species with Haplochro- then known species with Haplochromis mis-like dentition from rocky habitats can dentition, H. (H.) obliquidens and H. (H.) unambiguously be assigned to the so de- lividus, inhabited vegetation rich areas as fined Haplochromis lineage. Apart from well as rocky piers. Though both are known tooth shape there is currently little evidence from south-eastern Lake Victoria, they are for the monophyly of the lineage. The spe- there largely restricted to vegetation, and cies regularly encountered at rocky shores therefore only briefly introduced in my spe- show either no distinct melanin pattern or cies account. However, at least three other one of vertical bars and a faint midlateral species are regularly or exclusively found stripe. Greenwood (1980) considers the at rocky shores. Haplochromis lineage to be closely related Many species of the Haplochromis line- to the Neochromis and Xystichromis lineage (sensu stricto) feed primarily on dia- ages. Support for this view has recently toms which they scrape off submerged come from a study of squamation charac- leaves and stems of plants, and off rocks. ters (Lippitsch 1993). In that case, it is likely To scrape off diatoms from leaves, the that Haplochromis is a close relative, maybe fishes take, according to Greenwood the sister group (?) of the whole complex (1981), a leaf in the mouth and, holding it of "vertical bar Mbipi". loosely, swim along its length, scraping off Haplochromis (Haplochromis) obliqui- the epiphytes as the leaf passes between dens Hilgendorf, 1888 was the first cichlid the teeth. Greenwood assumes that the to be described from Lake Victoria. It has peculiar tooth shape is an adaptation to this been recorded from various places along way of feeding. Nevertheless, the rock- the lake shore and even from the very iso- dwelling species engage heavily also in lated rocky Godziba Island in the centre of "rock-scraping". As a consequence, their the lake. However, since not of all these

230 populations male coloration is known, it re- and flanks. The dorsal fin is light blue with mains to be investigated whether they ac- orange coloured lappets and orange-red tually represent the same species. H. (H.) maculae in the soft part. The caudal fin is obliquidens is characterized by a unique bright orange to red, the anal fin yellowish- tooth shape that it shares, among the orange with relatively large yellow egg known species, only with H. (H.) "blue dummies. Females are silvery grey with obliquidens", and which represents the transparent fins. Both sexes exhibit be- most derived condition (= most far re- tween 6 and 8 vertical bars on the flanks. moved from the basic haplochromine tooth H. (H.) "blue obliquidens" is known only shape) within the Haplochromis lineage. from Makobe Island in the south-western The anterior teeth in the outer tooth rows Speke Gulf. It seems to be very specific in have on slender necks obliquely truncated its microhabitat choice, being found only crowns, on which the minor cusp is entirely inshore, at places that are slightly protected lacking so that the tooth is unicuspid, re- against the surf by offshore rocks, and usu- sembling a curved scalpel. Males are bright ally are near to some grassy vegetation, yellow-green, with a yellow-grey dorsal fin hanging into the water. We never saw it at with red lappets and maculae, an anal fin water depths beyond 2 m. It shares its habi- with some pinkish flush and yellow egg tat with several other algae scrapers of the dummies, and pelvic fins that are half black Neochromis, Xystichromis and Haplochro- and half faint pinkish. Females are silvery- mis (H. (H.) "purple yellow") lineages. The yellow, with yellow anal and pelvic fins. brightly coloured males have their territo- Though the combination of tooth shape ries on the rock surface. In aquaria they de- and male coloration should make H. (H.) fend their territories quite aggressively. I fre- obliquidens easy to identify, its name has quently observed "blue obliquidens" scrap- many times been used in aquarium litera- ing algae from the small and medium sized ture for yellowish haplochromines with dif- rock boulders in the shallow water. Its outer ferent dentition. In most cases these were teeth are usually strongly abrased. They H. (?) "thick skin"-like insect eaters. H. (H.) seem to be pulled over the rocks in such obliquidens is a shallow water dweller that an angle, that the abrasion shapes the origi- has been observed in the vicinity of emer- nally obliquely truncated teeth again ob- gent vegetation, in water-lily zones, at the liquely truncated. Of three individuals that margin of papyrus swamps, but also over we examined, one had eaten predomi- sand and rocky substrates (Greenwood nantly moss animals (Bryozoa) and in 1956b, Katunzi 1980, Witte & Witte-Maas smaller amounts diatoms and blue-green pers. comm.). At least in the Mwanza area algae. The other two had eaten almost ex- it has become very rare recently. I have per- clusively may fly larvae (Ephemeroptera) and sonally never seen it alive, and unfortu- some caddis fly larvae (Trichoptera). The nately cannot show it in a photo. H. (H.) stomachs of all three contained also some obliquidens feeds predominantly upon dia- fragments of plant tissue. H. (H.) "blue toms and detritus, but at least occasionally obliquidens" seems to be an opportunistic also on large amounts of insect larvae feeder just like H. (H.) obliquidens. Unlike (Greenwood 1956b, Katunzi 1981). in the Neochromis algae scrapers, that live Anatomically very similar to H. (H.) sympatrically with it, both sexes are found obliquidens is the rock-dwelling Haplochro- at about equal frequencies. mis (Haplochromis) "blue obliquidens"obliquidens". It has The second described Victorian species the typical obliquidens tooth shape and dif- of this lineage, Haplochromis (Haplochro- fers from that species predominantly in mis) lividus Greenwood 1956, has a less male coloration, has smaller eyes and small, extreme tooth shape. Even in the ante- deeply embedded chest scales, like the rior-most teeth, the minor cusp is still majority of rock-dwelling haplochromines. present and the major cusp is less pro- Males are bright light blue on head, dorsum tracted. Furthermore it differs from H (H.)

231 A male H. "blue obliquidens" from Makobe Island. obliquidens and H. (H.) "blue obliquidens" in male coloration. Males have light olive- green ground colour and a golden-red flush on the flanks, extending from the head to the caudal peduncle according to Greenwood (1956b). The fishes that we found in the Mwanza Gulf half bright orange-red flank coloration that extends

A mouthbrooding female H. "blue obliquidens".

A male Haplochromis lividus from Nyegezi Rocks.

232 A male H. "purple yellow" from Nyegezi Rocks.

A male H. "purple yellow" from Kissenda Island.

233 from the gill covers and dorsal aspects of the dorsum above the lateral line and the the head to near the caudal peduncle. dorsal fin are dull purple to bright pur- The dorsal fin is greyish with red streaks plish-red. The entire remaining parts of and lappets. The caudal fin is similarly head and body are bright yellow to green- coloured with a red margin. The anal fin ish-yellow. Rock-dwelling males can have is pale greyish with pink flush and rather a dark sooty-grey underside. The dorsal big yellow egg dummies, the pelvic fins fin carries in the upper half of the spinous are either entirely black, or half black and part a distinct metallic light blue band, half whitish transparent. Females are grey- sharply contrasting with the purplish red ish with yellow anal and pelvic fins. of the rest of the fin. The caudal fin is H. (H.) lividus is known from various greyish, with dull red streaks and macu- places along the lake shores, but again, I lae, and a blueish sheen in its lower half. consider the taxonomy of these popula- The blueish grey anal fin has a pinkish tions unresolved until male coloration is flush and carries quite many medium known of all. Greenwood (1956b) found sized yellow egg dummies. The pelvic it in shallow littoral zones in the vicinity fins are black. Females are light yellow- of emergent and submerged vegetation, ish grey, and both sexes carry between 6 in water-lily zones, at the margin of papy- and 8, usually very faint vertical bars on rus swamps, and commonly over rock for- the flanks. The chest scales are not as mations of piers. He found it sympatrically small as in H. (H.) "blue obliquidens" and with H. (H.) obliquidens and had indica- most other rock cichlids but are some- tions, that it lives predominantly in slightly what size reduced. They are not deeply deeper water than the latter. We found embedded. H. (H.) lividus a few times at rock-reed in- H. (H.) "purple yellow" is abundant in terfaces in the northern Mwanza Gulf but water-lily zones and at the edge of papy- I consider it there merely an occasional rus stands, but occurs also in purely rocky intruder in rocky habitats. Unlike the un- habitats, including islands that have no described rock-dwelling members of the submerged or floating vegetation at all. Haplochromis lineage, this species has In south-eastern Lake Victoria it is at rocky generalized chest squamation. The scales shores the most frequently encountered, are big and not deeply embedded. Its and most widely distributed species of major habitat probably are areas with sub- the Haplochromis lineage. However, it is merged vegetation. The species is in never very abundant at rocky shores, and south-eastern Lake Victoria not very com- always less abundant than the epilithic al- mon. gae scrapers of the Neochromis lineage. Much more common is Haplochromis It is entirely absent from most rocky (Haplochromis) "purple yellow"yellow", a species places in the Speke Gulf. At rocky shores with a lividus-type tooth shape that grows H. (H.) "purple yellow" lives in shallow wa- bigger than all other known species of ter at gently, moderately and steeply slop- this lineage. It has first been collected by ing shores with all boulder sizes. We F. Witte & E. Witte-Maas (Witte et al. never found it at depths beyond 2 m. It 1992). Its outer teeth are either bicuspid is most frequently encountered where with a broadly flanged major cusp, or reed grows among the rocks at the water- weakly bicuspid with a reduced minor line, and under floating carpets of the wa- cusp and an obliquely truncated major ter hyacinth (Eichhornia crassipes) (See- cusp. Its inner teeth are in the upper jaw, hausen, Kangwe & Samwel-Terry 1996). In more than in the other species, arranged the clear waters at Makobe Island I have in a broad band. Males of this species observed flocks of H. (H.) "purple yellow" have a very distinct coloration, that can be grazing epiphytes from the roots and stems confused only with that of H. (H.) "red of overhanging reed. A male from Nyegezi back scraper". Their upper head surface, whose stomach we examined, had eaten

234 predominantly plant tissue and diatoms. At lividus-type, bicuspid with strongly broad- Makobe I also observed them scraping al- ened major cusp. Its head is, relative to the gae from rocks. That this is a common feed- standard length, longer than in all other ing behaviour of rock-dwelling populations, species, except H. (H.) "red back scraper". is witnessed by frequent strong abrasion of Its straight dorsal head profile is less steep the outer teeth. The abrasion shapes the than that of the latter species. This is a very teeth less obliquely truncated than in H. (H.) inconspicuous fish. Males are light silvery "blue obliquidens" from the same place. greyish with a flush of orange behind and Haplochromis (Haplochromis) "red back above the pectoral fins. When sexually ac- scraper" is surprisingly similar to H. (H.) "pur- tive, the dorsum gets a beautiful violet ple yellow" in male coloration, but is ana- flush and the flanks caudad of the orange tomically quite different. It has a longer area become greenish. The silvery overall head than other species of the lineage, a appearance of the fish is due to the flank more decurved dorsal head profile than "pur- scales having a metallic iridescent margin. ple yellow" and fewer inner tooth rows in The fins are greyish transparent to sooty, the upper jaw (table 3). The shape of the the anal fin carries two or three orange-yel- outer teeth is of the lividus-type. Males low egg dummies. Females are greyish-sil- have the same colour bipartitioning of the very. Both sexes bear about 5 to 7 very faint body that was above described for H. (H.) vertical bars on the flanks and can have a "purple yellow". However, the upper head, trace of a mid lateral stripe. dorsum and dorsal fin are more orange-red "Orange chest silvery scraper" is known than purplish. The remaining body is yellow. from several rocky islands and mainland The dorsal fin has the same metallic light shores in the central Mwanza Gulf, but is blue band that the fin of "purple yellow" has, at all of these places rather rare. Only at and also the other fins are very similar in one place we found this species sympat- coloration to those of that species. The merically with H. (H.) "purple yellow". At three tallic blue in the lower half of the caudal is other record localities of "orange chest sil- brighter and the anal fin is entirely metallic very scraper" we could not find H. (H.) "pur- blue. Females usually exhibit a mid lateral ple yellow", which is otherwise widely dis- stripe and traces of a dorsal lateral stripe. tributed in the Mwanza Gulf. This may indi- Both sexes have 6 to 8 vertical bars on the cate some sort of ecological segregation flanks. The chest scales are rather small and between the two species. "Orange chest sil- usually deeply embedded. very scraper" lives only in very shallow wa- H. (H.) "red back scraper" is known only ter, between less than one and maximum from a rock-reed interface in the Nyegezi 2.5 m depth, and usually at places where Bay (northern Mwanza Gulf) where it lives reed grows among the littoral rock forma- in sympatry with H. (H.) "purple yellow". It tions. was collected by Y. Fermon and me in 1991 and has not been seen after that. An Anatomically unspecialized aquarium population exists, going back to haplochromines of the the fishes collected in 1991. Unfortunately Astatotilapia type not enough is known about the ecology of "red back scraper", to discuss the resem- When Greenwood revised the haplo- blance in male coloration between this spe- chromine species flock of Lake Victoria, cies and "purple yellow". a number of species could not be as- The last species of the Haplochromis lin- signed to any of the lineages that he al- eage that is regularly encountered at rock lotted generic rank, due to the lack of any shores, has a very long, unhandy, but nev- of the derived characters of those line- ertheless descriptive working name: Haplo- ages. He placed those of them, that were chromis (Haplochromis) "orange chest sil- anatomically unspecialized, and not much very scraper"scraper". Its tooth shape is of the different from the fluviatile haplo-

235 A territorial male H. "red beck scraper" from a rock-reed interface in the Nyegezi Bay

A female H. "red back scraper" from Nyegezi Bay.

236 A male H. "orange chest silvery scraper" from Nansio Island.

The floating water plant Eichhornia found in the rocky habitat at Marumbi.

237 chromines of eastern Africa, together the others as "Astatotilapia" in quotation with the fluviatile species into the genus marks until their phylogenetic relationships Astatotilapia Pellegrin, 1903 (Greenwood have been identified. To express the distinc- 1980). However, he was not satisfied tiveness of the riverine Astatotilapia (includ- with this solution, for two reasons: (1) ing its lacustrine representative) from the The members of the genus Astatotilapia endemic lacustrine lineages of Lake Victo- share no derived character which would ria, I do not include Astatotilapia among the set them apart as a group from the other lineages provisionally treated as subgenera lineages. In fact, all they share seems to of Haplochromis but as a distinct genus. be the absence of the derived characters With this alteration, the Astatotilapia lin- that characterize other lineages. (2) The eage is defined after Greenwood (1979, species endemic to Lake Victoria (or 1980), as small to medium sized haplochro- most of them) differ from the fluviatile mines with relatively deep body (35-40% species in a number of meristic charac- of standard length), and predominantly un- ters: They have on average more caudal equally bicuspid outer teeth in fishes of vertebrae (modal numbers 16 cf. 15) and less than 70 mm standard length, that are more scales along the lateral line (modal firmly (not moveably) attached to the bone numbers 32-33 cf. 28-30). Hence Green- and the crowns of which are neither much wood considered it uncertain whether his compressed nor clearly demarcated from Astatotilapia lineage was a monophyletic, the neck of the tooth. Some unicuspid and thus phylogenetically meaningful, teeth are present in fishes of all sizes. They group. dominate in larger individuals. The scales More recently it has been shown on the on the chest are not strongly size reduced, basis of squamation characters that the lin- and there are usually between 28 and 30, eage is indeed at least diphyletic. Species but never more than 33 lateral line pore from Lake Victoria turned out to be closer scales (= scales that carry a pore of the lat- related to other members of the Victorian eral line). Lake-dwelling (lacustrine) species flock, than to the river-dwelling Astatotila- of "Astatotilapia" fit into this definition, ex- pia (Lippitsch 1993). Greenwood's data on cept that they have a higher modal number vertebrae and scale counts had indicated of lateral line pore scales. this already. All other Lake Victoria haplo- Relatively few anatomically unspecialized chromines have vertebrae and scale counts haplochromines live at rocky shores. The similar to those of the lake-dwelling "Asta- most frequently encountered species is totilapia". Greenwood, however, had not Astatotilapia nubila (Blgr.) 1906. It is widely given these characters much value. He was distributed along the lake shores and is one in uncertainty about whether the differ- of the few haplochromine species of Lake ences that he found, were the result of Victoria that are not endemic to the lake. It shared ancestry, or merely of eco-pheno- lives also in Lakes Nabugabo, Kyoga and in typic responses to life in lakes versus life rivers in the Victoria basin. However, it in rivers. However, the correlation between needs more investigations to confirm that meristic values and habitat is incomplete. all the populations, that were identified as At least one species that meristically be- A. nubila, really belong to one species. A. longs to the river-dwelling group, lives in nubila is usually regarded the prototype of typical lake environments in Lake Victoria a generalized haplochromine and is (see below). This suggests that the meris- thought to be morphologically close to the tic differences between the two groups are founding ancestor of the species flock. This not the result of eco-phenotypic responses. is indeed, among the lacustrine species that I think Astatotilapia has to be restricted to I studied, the only one with less than 32 the predominantly riverine group which, (28-31) lateral line pore scales, and is, at least however, has a representative in Lake Vic- in this respect, a member of the riverine toria. I follow here Lippitsch (1993) to leave Astatotilapia. From other generalized rock-

238 dwelling haplochromines, it is distinguished cladistic gradualism. From an anatomical furthermore by its rather long head (tables point of view, an unspecialized insect eater, 3 and 4), by large and not deeply embed- similar to today's A. nubila, could have given ded chest scales, and by pelvic fins, the first rise to the anatomically specialized rock- rays of which are prolonged into a distinct dwelling Mbipi. Some species within the filament in males, sometimes also in fe- Mbipi, such as the epilithic scraper H. males. Males are dark green-grey to almost (Xystichromis) "copper black" and the epi- black, with a black dorsal fin that can have lithic picker H. (Paralabidochromis) chromo- red lappets, a caudal fin that is red in its gynos, are still fairly close to A. nubila, while distal third or half, a red anal fin with yel- others such as the epilithic algae scraper low to orange coloured egg dummies and H. (Neochromis) nigricans and the picker H. black pectoral fins. Females are rather light (P.) chilotes are much more specialized. The yellow-grey. In both sexes faint traces of puzzling question is, why do species evolve about six vertical bars can be visible on the various levels of specialization if others, liv- flanks. The outer teeth are upright bicus- ing alongside with them, perform well with- pids. Those in the upper jaw are sometimes out specializing. somewhat scraper-like with a small flange, Other species with black males, that live while those in the lower jaw have no flange. in sympatry with A. nubila are H. ("A.") "black Possibly because of the lack of derived cave", H. (?) "deepwater", H. (Xystichromis) or peculiar characteristics of A. nubila, many "copper black", H. (Neochromis) "velvet generalized haplochromines the coloration black" and H. (?) "black pseudonigricans". All of which roughly agrees with that given in differ from A. nubila in scale counts, denti- the description of A. nubila, have been iden- tion, head shape (table 3), and details of col- tified as that species. In fact quite a number oration, particularly the latter two differ in of different species have been regarded as anal fin coloration. A. nubila by aquarists. However, confusion Haplochromis ("Astatotilapia") "incurved with similarly coloured Lake Victoria species dorsal head profile" is once again a spe- can be avoided by a simple count of the lat- cies with an admittedly unhandy, yet in- eral line pore scales. formative, working name. Among all We found rock-dwelling populations of A. known rock cichlids, this one is anatomi- nubila only in the Mwanza Gulf and particu- cally closest to A. nubila. It has a some- larly often at steeply to very steeply slop- what broader lower jaw than has A. ing shores, composed of large rock boul- nubila, and on average more inner tooth ders. They live inshore at water depths rang- rows (table 3). Furthermore the red col- ing from 1.5 to at least 8 m. Population den- our in its fins is much less vivid, the pel- sities can locally be rather high, but usually vic fins form a less distinct filament, and this species is not among the abundant particularly females have a distinctly ones in rocky habitats. We have not ana- incurved dorsal head profile. Its chest lysed stomach contents of A. nubila, but it scales are large but somewhat deeply is likely that it is at rocky shores, like else- embedded. Thus, in several respects, H. where, an unspecialized opportunistic ("A.") "incurved dorsal head profile" appears feeder, that takes insect larvae as well as like a slightly derived form of A. nubila, plant debris and other items. better adapted to life among rocks. Amaz- The existence of rock-dwelling popula- ingly, also its number of lateral line pore tions of this anatomically unspecialized scales (33) is just slightly above that of A. haplochromine, and their coexistence in nubila, though already at the lower end stable communities with rock-dwelling of the range of lacustrine haplochro- cichlids of various grades of ecological and mines. anatomical specialization, is a nice example We know H. ("A.") "incurved dorsal head for the apparently paradoxical situation first profile" only from Igombe Island at the described by Greenwood (1980) as southern Speke Gulf shore. It lives only

239 A male Astatotilapia nubila from Anchor Island.

A male H. "black long snout" (Chamagati). A female H. "incurved dorsal head profile"

A male H. "incurved dorsal head profile" from Igombe Island.

240 in shallow immediately inshore waters in bedded chest scales (typical rock cichlid gaps between rock boulders and in squamation), and by male coloration. The rockpools, sympatrically with H. (X.) "cop- entirely black males lack the red in the anal per black" and H. (?) "zebra nyererei" and fin that is typical for A. nubila and have less is less abundant than either of these. red also in the caudal fin. The dorsal fin has Identification of these three sympatric red lappets. The biggest males have a species is in the field sometimes not standard length of just slightly above 60 easy. The coloration of H. ("A.") "incurved mm. The species inhabits the very gently dorsal head profile" barely differs from sloping shore of Chamagati Island in in- that of H. (X.) "copper black" but its snout shore waters of between 0.5 and at least is longer and its lower jaw narrower, 1.5 m depth, together with insectivorous though not quite as narrow as that of H. Mbipi, two insectivores of the Psammo- (?) "zebra nyererei". Furthermore it differs chromis lineage, and H. ("A.") cf. brownae. in dentition and the distinctly incurved It has a very low population density. In labo- dorsal head profile from both other spe- ratory experiments on feeding behaviour, cies. "black long snout" performed a large variety Another species that appears like a of feeding techniques. However, it foraged slightly modified A. nubila with a just predominantly by "picking", and frequently slightly increased number of lateral line by "pullscraping", "snapping" and "digging". pore scales (32-33) is Haplochromis Thus even this anatomically unspecialized ("Astatotilapia") "black long snout"snout". It is until species is able to perform, though probably today known only from Chamagati Island with low efficiency, pullscraping, the feed- and looks like a dwarf version of A. nubila. ing technique that became the most impor- However, it clearly differs from that species tant one among the rock restricted by having bigger and stronger outer teeth Aufwuchs-eaters of the Mbipi lineages. that are more strongly recurved and very It is a puzzling fact that the two above dis- firmly implanted, with a tendency to cussed nubila-like species do not only in procumbent implantation in the lower jaw their outer appearance resemble A. nubila, (somewhat reminiscent of the condition in but have also a lateral line scale count Paralabidochromis); by smaller, deeply em- that is very close to the condition found

A male H. "black cave" from Gana Island. Drawing by Helmut Seehausen.

241 in riverine Astatotilapia (32-33 versus 33- 1962. Greenwood found this species 37 in rock-dwelling cichlids of other line- over sand and shingle, where it fed pre- ages). The evolutionary significance of dominantly upon insect larvae (Diptera). this observation is at the moment difficult Some individuals had eaten many juve- to assess. However, if the specialized lin- nile fishes, and Greenwood sometimes eages of rock cichlids have evolved from considered this species to be related to a nubila-like ancestor, one could imagine certain anatomically little specialized fish that forms that looked like "incurved dor- eating haplochromines like H. (Harpago- sal head profile" and "black long snout", chromis) guiarti (Greenwood 1962). In rocky stood at the beginning of their radiation. habitats we found H. ("A.") brownae-like I do not think that these are actually sur- fishes at two places. A not yet further ex- viving ancestors of the anatomically amined population lives at Chamagati Is- more derived rock cichlids. It is merely a land, and seems to inhabit predominantly nice example of the situation that Green- areas with small stones and shingle at the wood (1994) described very aptly by com- very gently sloping small boulder shore. The paring the Lake Victoria cichlid species other population lives at the moderately flock with a car factory, in which proto- steep rocky shelf of the main island in the types of every model ever produced, and Vesi Archipelago, a purely rocky habitat with all successive versions of those models medium sized boulders. This population are still in production alongside the latest grows bigger than H. ("A.") brownae and dif- models ("cladistic gradualism", Green- fers from it by a shorter and relatively wood 1981, 1994). broader lower jaw (compare tables 3 and Another anatomically generalized spe- 4). cies is Haplochromis ("Astatotilapia") Males of this H. ("A.") "large brownae" are "black cave"cave". It is a small, shallow bodied light greenish-grey to blueish on the flanks species that differs slightly from the two and have light blue lips. The carrot-orange above discussed species in several meas- flush on cheek, gill cover, flanks and belly, urements, male coloration and in the which is characteristic for reproductively number of lateral line scale pores. With active males of H. ("A.") brownae, is but 36 such scales it is a typical lacustrine faintly indicated in some males of H. ("A.") species. Males are black all over with "large brownae". The dorsal fin is sooty with sometimes some golden metallic flush red lappets, caudal and anal fin are pale red, above the pectoral fins, and a dark red- the latter with pale yellow to orange-yellow dish-brown dorsal head surface and neck. egg dummies. The pelvics are black. Both The dorsal fin has red lappets, the caudal sexes bear six or seven rather broad and fin only the upper and lower corners red, irregular vertical bars on the flanks, and par- and the anal has a small dull red area in ticularly females have a short mid lateral its distal half. The egg dummies are small stripe, which is most distinct above the anal like in many other rock-dwelling cichlids, fin, as is described for H. ("A.") brownae. The and dull yellow. The chest scales are chest scales are somewhat size reduced somewhat size reduced but not deeply and are deeply embedded, like in the ma- embedded. We know this species only jority of rock-dwelling cichlids. H. ("A.") "large from deep, dark and cavelike rock pools brownae" lives in waters of between 3 and and crevices among big boulders at at least 5 m depth. Of two males that we steeply sloping islands in the northern examined, one had the stomach empty and Mwanza Gulf, and it seems to be very had insect fragments in its intestine. The rare also there. other one had only unidentifiable material A completely different species, ana- in stomach and intestine. It is likely that the tomically quite removed from the nubila population of the Vesi rock shelf is a rem- type by its short head, is Haplochromis nant of a once wider distributed species. ("Astatotilapia") brownae Greenwood Remnant populations of several other large

242 predatory species survived at the same time that the male circled, who several place. Witte et al. (1992) found H. ("A.") times chased her away. When the H. brownae-like fishes over sand in the nyererei female laid some eggs, H. ("A.") Mwanza Gulf. Their male coloration was barbarae rushed forward to snatch the eggs very different from that of the Vesi Island away. This often happened so quickly that fishes. the spawning female had not yet even Similar to H. ("A.") brownae is Haplochro- started to collect the eggs. If she was about mis ("Astatotilapia") barbarae Greenwood to do so, H. ("A.") barbarae pushed her away. 1967. It differs from H. ("A.") brownae in Witte-Maas coined the term "egg-snatching" some morphometric characters (narrower for this behaviour. She observed that the inter orbital width, smaller eye length), actions of the egg-snatcher were usually so fewer gill rakers (8 or 9, cf. 9-12), smaller fast that it managed to disappear from the chest scales, and stout bicuspid rather than nest before the male H. nyererei could at- weakly bicuspid and unicuspid outer teeth tack it. This strategy was so successful, that (Greenwood 1967). H. ("A.") barbarae is a the H. ("A.") barbarae obtained eggs in 10 polymorphic species, females of which fre- out of 11 attempts. Sometimes, however, quently (or even mostly) are of a piebald she attacked while the H. nyererei female morph. The latter exhibits on a silvery-yel- had not laid any eggs (Witte-Maas 1981). low to yellowish ground a variable number H. ("A.") barbarae declined dramatically af- of, in shape and dorso-ventral extent irregu- ter the Nile perch upsurge and disappeared lar blotch-like vertical bars. Less common entirely from its former record localities in is the "normal" morph, and rare an OB- the Mwanza Gulf (Witte et al. 1992). We morph (Greenwood 1967). Male live colora- have never seen this species anywhere else tion is unknown. The melanin pattern is in- either since 1989 (Seehausen et al. in press conspicuous in both sexes, and consists [b]), nor has it been seen recently at the according to Greenwood (1967) of a faint northern lake shores (Kaufman & Ochumba mid lateral stripe, a very faint dorsal lateral 1993). If it is not extinct, it is certainly criti- stripe (only in females?), and four even cally endangered. Several haplochromines fainter vertical bars. that have been named barbarae by H. ("A.") barbarae is known from the north- aquarists, belong to different species. Real ern, Ugandan waters, and from eastern and H. ("A.") barbarae have been sent in the early southern Tanzanian localities. Like H. ("A.") 80s from Uganda to Sweden by B. Selbrink. brownae, it lives over sand, shingle and rock I do not know what happened to that stock. (Greenwood 1967, Witte et al. 1992). In the Those fishes had apparently been caught Mwanza Gulf it was found predominantly together with H. (Paralabidochromis) over rocks (Witte et al. 1992). H. ("A.") "rockkribensis" at a rocky shore. barbarae, though anatomically rather One of the species erroneously identi- unspecialized, has a peculiar feeding habit. fied as barbarae, and which is rather com- It seems behaviourally specialized to steel mon in the aquarium fish trade, is eggs from spawning pairs. Witte-Maas Haplochromis (?) "red piebald" or "piebald (1981) described in detail how this species redfin". This is a small insect eating spe- feeds on eggs of other rock-dwelling cies which is caught in the Victoria Nile cichlids. In the observed case, a pair of H. near Jinja (L. Kaufman pers. comm.). It dif- (?) nyererei was spawning. A female fers in dentition (tending to Paralabido- barbarae observed the pair from below the chromis type, unicuspid outer teeth) and water surface at a distance of 60 cm, just coloration from H. ("A.") barbarae. The ma- outside the range within which the H. jority of its individuals are orange- nyererei male would chase away territorial blotched (both sexes). The "normal" col- intruders. Whenever the female H. nyererei oured males have deep blue flanks, a red circled in the nest, the H. ("A.") barbarae anal fin, red dorsal fin lappets, and some moved stealthily forward, to withdraw each red distally in the caudal fin. Details about

243 A male H. "large brownae" from Vesi Island.

A female H. barbarae, collected and photographed before the Nile perch upsurge. the ecology of this species are not Insect-eating rock cichlids known. that cannot be grouped

There are a few insectivorous rock- dwelling haplochromines that I cannot assign to any of the lineages and groups described until here. In some cases this

244 H. "red piebald" is one of the species erroneously identified as H. barbarae. It is a small insect- eating species which is caught in the Victoria Nile near Jinja in Uganda.

A male H. "brown narrow snout" from Gana Island.

245 may be due to their isolated taxonomic sized and big respectively. At Gana the position, however, in others it may sim- slope is steep and the boulders are very ply be due to lack of information. Some big. All three individuals from Hippo Island are now known only from one or two in- (two in 1993, one in 1996) were encoun- dividuals but are anatomically or in terms tered at the very same spot. This indi- of coloration so peculiar that I do not want cates that the species may have particu- to skip them in this account of rock-dwell- lar microhabitat requirements that are ing Lake Victoria cichlids. Some have not not yet understood. Unfortunately noth- yet been measured and are hence not ing more is known yet about the ecology represented in table 3. of the fascinating fish. Three species are anatomically similar At Gana Island we recently discovered and may form a small group on their own Haplochromis (?) "pale egg dummy"dummy". It but may on the other hand also be mem- may have a wider distribution north of bers of the Vertical bar Mbipi. They are ana- our survey area. We found "pale egg tomically very close to the "Pseudonigri- dummy" living sympatrically with three cans" complex of the latter, but differ similar species at about 4 m water depth somewhat in melanin pattern. Their about among big rock boulders. From "brown five broad vertical bars are superimposed narrow snout" it differs by having a less by a rather distinct mid lateral stripe that narrow snout, from that species as well is in one species dissolved into two elon- as from "orange belly" and "black gate blotches. The three species occur Ukerewe" ("Pseudonigricans" complex) it sympatrically. differs in male coloration: pale brown with A conspicuous appearance with its grey fins, the dorsal fin with pale yellow- long, pointed and narrow snout and brown ish lappets, the anal fin with big pale yel- male coloration is Haplochromis (?) low egg dummies. "brown narrow snout"snout". It is a rare species Also Haplochromis (?) "orange belly" is but apparently has a wide geographical known only from Gana Island, our north- range. We know it from only three locali- ern-most sampling point. It may be more ties which are so far apart that the range widely distributed at the Ukerewe west of the species seems to cover most of coast and north of our survey area. It is the region that we surveyed (see map). of a distinct appearance and differs from Male body and head coloration is brown, "brown narrow snout" and "pale egg sometimes tending to reddish, some- dummy" in its finer, acutely pointed and times with a metallic blue flush on the more strongly recurved outer teeth, lower body half. There is a dark mid lat- smaller eyes, and male coloration. The eral band. The dorsal fin is dark with or- male is on the upper parts of the flanks ange to red lappets, the dark caudal fin from the snout to the caudal peduncle has a narrow orange to red edge. The pale red. It is bright orange on the gill anal fin is pale reddish in the two north- cover and on the lower half of the ante- ern populations, but yellow in the south- rior flanks between the lateral line and ern (Hippo Island) population, with orange the anal fin insertion, and green on the coloured egg dummies. H. "brown narrow lower half of the posterior flanks and the snout" is at its record localities rare to caudal peduncle. The dorsal fin is metal- very rare. Until today we found three in- lic blue with red lappets, caudal and anal dividuals at Hippo Island, two at Mabibi fin are red, the pelvic fins black. The mid Islands, and five at Gana Island. All but one lateral stripe is dissolved into two broad from Hippo Island were males. The spe- and elongated blotches. "Orange belly" cies has everywhere been found at lives sympatrically with "brown narrow depths between 4 and 7 m. At Hippo and snout" and "pale eggdummy" in the huge Mabibi the shore slopes moderately rock gardens of Gana Island. We found it steep and the rock boulders are medium in 4-6 m deep water among big boulders.

246 Haplochromis (?) "duck snout" is a spe- dorsal head profile, a large mouth, and a cies with a peculiar snout shape that has square shaped lower jaw, this fish is been discovered in 1995 at the Bwiru Pe- unique within the Mbipi. A similar lower ninsula between the entrances of jaw shape was previously known only Mwanza and Speke Gulf, and has not yet from the oral shelling molluscivore H. been measured. It has large eyes and a (Hoplotilapia) retrodens which is other- rather long snout. The male is light wise very different from H. "shovel brownish dorsally with a pinkish flush and mouth". metallic greenish-yellow flanks, red lappets on the dorsal fin, a red margin to the caudal fin, and a reddish anal fin with orange egg dummies. The female is similarly coloured but lacks the red in the fins. Both have five rather broad vertical bars on the flanks. Nothing is known yet about the ecology of H. "duck snout". We collected it in a ca. 1.5 m deep, wide rock pool at a steeply sloping large boulder shore. Haplochromis (?) "blue insectivore" is probably a member of a Mbipi lineage. It is known only from Irondo Point (Sengerema shore) were it lives in very shallow water in strong surf over small rock boulders, sympatrically with H. (Neo- chromis) "velvet black" and H. (Xysti- chromis) "copper black". The male is beautifully metallic light blue all over, with only faint traces of red distally in the unpaired fins. Its coarse unicuspid outer teeth suggest an insect diet. The species has not yet been further investigated. Haplochromis (?) "purple rocker" is one of the eleven rock-restricted haplochromines that were discovered in 1978 by Frans Witte and Els Witte-Maas at the Nyegezi rocks (Witte et al. 1992). In spite of intensive sampling in that region we failed to find this fish. It cannot be excluded that it declined strongly or went even extinct. Several other species also disappeared from the rocky shores in the north-western Mwanza Gulf (e.g. H. (Neo- chromis) "kruising", H. (Paralabidochromis) "rockpicker", and H. (P.) "elongate rockpicker"). Haplochromis (?) "shovel mouth" is a peculiar fish of unique appearance of which we caught only one male yet at Igombe Island in a rock pool of 0.5 to 1 m depth. With a combination of a shallow

247 A male H. "pale eggdummy" from Gana Is. A male H. "duck snout" from Bwiru Island.

A male H. "orange belly" from Gana Island.

A male H. "blue insectivore" from Irondo Point. A male H. "shovel mouth" from Igombe Island.

248 A territorial male Haplochromis plagiodon from a sandy habitat.

A male H. (Macropleurodus) bicolor.

249 Frequent intruders from other habitats

The cichlid communities of rocky reefs Gulf in 1978/79, twelve had disappeared and shores are at most places to over by 1990 (Witte et al. 1992). Except two 90% composed of stenotopic species, (H. (?) "thick skin", H. (?) "kribensis"), none that live only at rocky shores. Most of the of them has been recorded again over remaining species have either rocks and rocks in the Mwanza Gulf between 1991 sand (e.g. molluscivores of the Ptyo- and 1996. One more (H. (Macropleur- chromis lineage), or rocks and submerged odus) bicolor) is still occasionally seen vegetation (e.g. algae scrapers of the over rocks at one place in the Speke Gulf. Haplochromis lineage) as their major habi- Sand-dwelling cichlids are most fre- tats. However, apart from these two quently intruding into rocky habitats, and groups of rock-dwellers, one does from particularly some oral shelling mollusci- time to time encounter haplochromine vores. We observed Haplochromis (Para- species in rocky habitats, that are bypass- labidochromis) plagiodon Regan & ing visitors from other habitats. These Trewavas, 1928 at a number of steeply "occasional intruders" (Witte et al. 1992) sloping rocky islands and rock shores in are species the habitats of which are ad- the Mwanza Gulf, as well as at the very jacent to rocks, thus littoral sand-dwell- gently sloping Chamagati Island in the ers, vegetation-dwellers and, at steeply Sengerema region. The species is cur- sloping rock shores, pelagic open water- rently one of the most widely distributed dwellers. We never found inhabitants of haplochromines in southern Lake Victoria, sublittoral mud bottoms, e.g. detritivores, inhabiting exposed sand beaches as well over rocks. Their general absence may be as sublittoral mud bottoms (Seehausen et explained by the circumstance that mud al. in press [b]). A detailed investigation bottoms are rarely directly adjacent to may reveal a complex of sibling species. rock bottom (at least in the littoral and Among the rock cichlids only members sublittoral regions of the lake). At the of the H. (P.) "rockkribensis" complex can same time soft bottom dwellers are usu- be confused with H. (P.) plagiodon, in par- ally more stenotopic with regard to sub- ticular H. (P.) "rock macula" and H. (P.) strate type than are sand and vegetation "short snout scraper". H. (P.) plagiodon dif- dwellers. Also among the sand bottom- fers from both by having a longer, and less and vegetation-dwellers some species broad lower jaw and a different male col- seem to be less stenotopic and/or more oration (see photos). From "rock macula" mobile than others, and more often than it furthermore differs in female coloration. others intrude into rocky habitats. I briefly Females of H. (P.) plagiodon are brown- introduce these species to the reader. grey with narrow and thin vertical bars. The species composition among occa- From "short snout scraper" it differs also sional intruders of rocky habitats has by its less decurved dorsal head profile. changed significantly during the past ten Haplochromis (Macropleurodus) bicolor years. This is probably a direct conse- (Blgr. 1906) disappeared as an occasional quence of the Nile perch upsurge. intruder of rocky habitats in the Mwanza Pelagic, but also sand-dwelling haplochro- Gulf, but is still found at one place in the mines were directly affected by the in- Speke Gulf. According to Greenwood creased predation pressure, and of the 13 (1956b) it is a widely distributed species, occasional intruders that were observed living over sand, rock and shingle but only over rocks in the north-western Mwanza rarely over mud. It is now almost extinct.

250 From northern lake shores it has not re- do reproductively active males of several cently been reported (Kaufman pers. populations exhibit an orange to orange- comm.) and the small population in the red flush on the gill cover and large parts Speke Gulf is the only one known to still of the dorsal and ventral aspects of the exist in southern Lake Victoria. The last flanks (see photo). The most diagnostic time that I observed some individuals characters, however, are the peculiar, lat- there by Scuba, was early 1996. They erally compressed, tapering mouth, and joined shoals dominated by H. (Ptyo- the long slender bicuspid outer teeth that chromis?) "red rock sheller" in strongly are much longer than the teeth of the first surf exposed shallow waters over rocks. inner tooth row. H. (?) "thick skin" shares Possibly the rocks offer them retreat from with some species of the Ptyochromis lin- predation pressure exerted by Nile perch. eage the habit of "diving" into the sand in In coloration the species is not very dis- case of danger. Riexinger (1996) gives tinct, though sexually active males in some more information about H. (?) "thick some populations become bright orange skin". or red on the anterior half of the flanks. Haplochromis (?) "small blue zebra" is Piebald (bicolor) morphs do occur. With its another small insect eater with conspicu- strongly decurved dorsal head profile, the ous male coloration. Judging from its unique shape of its upper jaw which light pigmentation it is likely to be a sand- curves downwards in the posterior third, dweller. However, it has only a few times and with its unique dentition of stout bi- been caught both over sand and over cuspid teeth in which the minor cusp is rocks in the southern Speke Gulf and cen- strongly reduced and lies anteriorly of the tral Mwanza Gulf. It was seen for the last elongate and buccally orientated major time in 1991. This is a tiny species that cusp, this species can hardly be confused can hardly be confused with any rock with any other Lake Victoria haplo- cichlid. It resembles in body shape some chromine. other small sand-dwelling insect eaters. A number of sand-dwelling insect eat- Males are light blue-grey with about six ers used to intrude into rocky habitats and regularly shaped vertical bars on the one still does so. The latter is Haplo- flanks and much red in the unpaired fins chromis (?) "thick skin" (Witte et al. 1992), (see photo). The coloration of the anal fin a beautiful small yellow fish that can be is distinct from that in otherwise similar quite abundant at some sand beaches. It sand-dwelling species. It is almost en- seems to be widely distributed since simi- tirely orange-red to red with two or three lar fishes have frequently been shipped rather big egg dummies, surrounded by out from Kenya and Uganda to Europe broad transparent rings. and America, and are better known A very little known insect eater is Hap- among aquarists than many other spe- lochromis (?) "blue and orange fins" that cies. Unfortunately they have in aquarium is sometimes found over rocks at the publications been dealt with under vari- southern Speke Gulf shore. We have not ous wrong names. Most often they were found its real habitat yet, but its large, not mis-identified as H. obliquidens and H. deeply embedded chest scales, com- brownae. Details of the coloration of bined with its being very irregularly "thick skin" can vary among populations, caught, suggest that it is not a real rock- but they all have bright yellow flanks with dwelling species. rather many vertical bars. The melanin After sand-dwellers the second most pattern often resembles that of species common intruders of rocky habitats are of the H. nyererei complex. The dorsal fin species that normally live among sub- carries either many red maculae, or is al- merged vegetation (e.g. in water-lily and most entirely red. Often it carries red Ceratophyllum beds), near emerging veg- maculae also in the females. Furthermore etation (reed, papyrus) or under floating

251 A male H. "thick skin" from Nyegezi Bay.

A male H. "small blue zebra" from Nyegezi Island.

252 Above: a freshly caught Haplo- chromis "blue and orange fins" from Ndurwa Bay.

Right: a female Haplochromis "macula" from Nansio Bay. The author frequently found it over the rock formations of piers.

A male H. "macula" from Nansio Bay.

253 vegetation (water hyacinths, reed, papy- broken mid lateral stripe. Hence the rus). Though submerged vegetation is in- name "macula". However, I have not seen habited by many species of insect eating this blotch in the fishes from the north- and algae scraping haplochromines, we ern Speke Gulf. With this blotch, the encountered of them only algae scrapers melanin pattern of H. (H.) "macula" differs in rocky areas. This may reflect the way somewhat from that of other species of how these species reach rocky areas. In the Haplochromis lineage (see page 230). many cases they are likely to arrive with Haplochromis (Haplochromis) "black islands of floating reed, papyrus, and wa- and red fins" is another epiphythic algae ter hyacinths. Insect eating species are scraper with a Haplochromis type denti- much less abundant than algae scrapers tion. It resembles in size and head shape in the vicinity of such vegetation. They H. (H.) "purple yellow" but has a different are more common in submerged vegeta- dentition, with more slender, often ob- tion. liquely truncated outer teeth, and fewer Haplochromis (Haplochromis) "macula" inner tooth rows. We found it in the vi- is a brilliantly coloured algae scraper with cinity of floating Eichhornia over rocks at a tooth shape that suggests it to be a Makobe Island. We know only males yet. member of the Haplochromis lineage. It They are black with red lappets to the dor- used to be abundant in littoral areas of the sal fin, a red caudal fin margin, and an en- northern Mwanza Gulf before the Nile tirely red anal fin that carries yellow egg perch upsurge and was found also slightly dummies. The latter are unusually small offshore over sand and mud bottoms for a species of the Haplochromis line- (Witte et al. 1992). After the Nile perch age. upsurge it disappeared from the offshore Witte et al. (1992) list five pelagic and parts of its habitat range but continued to half pelagic plankton eating haplochro- be found inshore in the vicinity of reed mines that used to intrude into rocky habi- and papyrus. We frequently found it over tats from the open waters. Four of them the rock formations of piers. Recently we have not been seen after 1987, neither discovered a population that inhabits shal- in their major habitats in the open water, low water of a rocky island in the north- nor at rocky shores. They are possibly ex- ern Speke Gulf. It lives at a place where tinct (Witte et al. 1992, Seehausen et al. litter from overhanging large trees accu- in press [b]). The phytoplankton eater Hap- mulates on the rock bottom, and possi- lochromis (?) "kribensis" (Witte et al. 1992) bly constitutes a food source for this spe- is very occasionally still seen at rocky is- cies. Its high abundance at that place sug- lands in the northern Mwanza Gulf. Early gests that it is there a real rock-dweller, in 1996 we discovered a remnant popu- rather than an occasional intruder of the lation at a rocky island in the northern rocky habitat. Males have a bright red Speke Gulf. It is the only known place head, chest, belly, and dorsal aspects of where H. (?) "kribensis" is not only occa- the flank. The remainder of flanks and sionally encountered but was, at least caudal peduncle are yellowish-greenish. during our visit, quite abundant. While The dorsal fin is red in the spinous part scuba diving I observed H. (?) "kribensis" and transparent with red streaks and in mixed shoals with H. (H.) "macula" in maculae in the soft part. The caudal fin is the upper 1 m of the rocky littoral. Possi- blue-grey with red maculae in the upper bly it should here not be considered an half, and the anal fin is pale whitish with occasional intruder but rather a rock- two or three big orange or orange-yellow dweller. This exceptional situation may be egg-dummies. Females and quiescent explained with the above mentioned unu- males frequently exhibit an elongate dark sual character of the rocky habitat at this blotch on the flanks above the anal fin place: litter from overhanging large trees which is part of an otherwise very faint, accumulates on the rock bottom, and

254 may form directly or indirectly an addi- Gulf. Four of them, H. (Prognathochro- tional food source for species that are not mis) macrognathus, H. (Prognathochro- normally rock-dwellers. mis) perrieri, H. (?) "big teeth" and H. (?) The coloration of male H. (?) "kribensis" "longurius" seem to be extinct in the re- is reminiscent of that in H. (Paralabi- gion, and have not been reported from dochromis) "rockkribensis". However, other parts of the lake either (Kaufman & "kribensis" males are less yellow and have Ochumba 1993, Kaufman & Seehausen a silvery iridescent area on each scale, 1995). Of the fifth species, H. ("Astatotila- giving them a silvery ("pelagic") appear- pia") "two stripe yellow green"green", we found ance. The light silvery females differ dis- in 1991 one individual in a beach seine, tinctly from the deep yellow "rock- pulled over sand bottom. After 1991 it has kribensis" females. The melanin pattern also not been seen any more. In 1989 we of "kribensis" consists of, in their width ir- caught a juvenile H. (Harpagochromis) regular vertical bars and broken mid and altigenis Regan, 1922 over rocks at An- dorsal lateral stripes, while "rockkribensis" chor Island in the northern Mwanza Gulf. has a typical chessboard pattern. H. (?) That was the last sighting also of this large "kribensis" is the only representative of predator. the ecological group of phytoplankton eating cichlids, that is still encountered in southern Lake Victoria. In the time before the Nile perch upsurge, it used to inhabit the surface layers of sublittoral waters (Witte et al. 1992). During the past five years it has much more often been seen in rocky littorals than in open water. The protection of its remnant populations should be given high conservation prior- ity. Haplochromis (?) "citrus" is another phytoplankton eater that used to be seen occasionally over rocks, but it seems to be extinct (Witte et al. 1992). In the pre-Nile perch era four pelagic or half pelagic zooplankton eaters had been recorded over rocks (Witte et al. 1992). For one of them (H. (?) thereuterion) rock bottom was the major habitat, and a recent study showed that it was mores insectivorous than planktivorous. I discussed this species in another chapter. Three others were occasional intruders. Neither the surface dwelling Haplo- chromis (?) "argens""argens", nor the more bottom orientated H. ("Astatotilapia") megalops Greenwood & Gee, 1969 and H. ("Astatotilapia") piceatus Greenwood & Gee, 1969 have been seen after 1987 anywhere in the lake. They are possibly extinct (Witte et al. 1992). Witte et al. (1992) finally listed five species of fish eaters as occasional intruders in rocky habitats in the northern Mwanza

255 A male H. "kribensis" from Kilimo Island.

A female H. "kribensis" from Nansio Bay. A female H. "argens".

A male H. "argens" in the aquarium.

256 A male H. piceatus in the aquarium.

H. "yellow insectivore", another 'golden' haplo- H. cf altigenis from Anchor Island. chromine from lake Victoria.

A courting male Astatoreochromis alluaudi in the aquarium.

257 Astatoreochromis, Oreochromis and Tilapia at rocky shores

Eight cichlid species that are not part of Lake Victoria, A. alluaudi inhabited also the Victorian species flock, live alongside sublittoral mud bottoms (Witte et al. 1992). with the several hundred species of the It is not any longer found in such habitats. haplochromine flock in Lake Victoria. At many rocky shores it constitutes a fre- Apart from Astatotilapia nubila that is quent member of the community. Even far treated in another chapter, these are the offshore islands, such as the purely rocky haplochromines Astatoreochromis Ruti and Mabibi Islands, and the Vesi Archi- alluaudi and Pseudocrenilabrus multicolor pelago in the Speke Gulf have their A. victoriae, four species of Oreochromis and alluaudi populations. We found unusual one or two species of Tilapia. anal fin coloration among males of these Pseudocrenilabrus inhabits marginal populations at the geographically isolated swamps and creeks and we never found it central Speke Gulf islands. Males at all at rocky shores. Astatoreochromis alluaudi places that I have visited along the main- Pellegrin 1903, though found in marginal land shores of Lake Victoria, including the waters as well, is a common inhabitant of northern, Ugandan shore, as well as at less truly lacustrine environments and is found far offshore islands, have a largely red anal at most rocky shores and islands. It is eas- fin. Males from Ruti Island frequently have ily identified by its golden yellow coloration entirely yellow anal fins, and at the Vesi Ar- that is particularly intense on the underside, chipelago as well as at Mabibi Islands a and by its high count of anal fin spines (4-6 form with a distinctly white anal fin lives cf. 3 in the species of the haplochromine alongside the normal form with a red fin. species flock and Astatotilapia nubila). Fur- These cases of aberrant coloration in popu- ther peculiarities are a rounded caudal fin lations of A. alluaudi at geographically iso- (cf. a truncate or subtruncate one in mem- lated rocky islands may indicate that migra- bers of the flock) a high number of egg- tion and therefore gene flow between such dummies arranged in up to four rows on populations and the inshore main popula- the anal fin of the male, and otherwise very tion is limited. muted differences in coloration of male and A. alluaudi in Lake Victoria feeds predomi- female. Based mainly on these characters, nantly on snails of the hard shelled species Greenwood (1959b, 1979) considered A. Melanoides tuberculata (Greenwood alluaudi distinct from the members of the 1959b) which it crushes between hypertro- species flock of Lake Victoria. This was phied pharyngeal jaws, equipped with mo- confirmed later by biochemical work (Sage lariform teeth. Over rock bottoms it fre- et al. 1984, Meyer et al. 1990). To avoid mis- quently coexists with other pharyngeal identifications, it should be said here that crushing haplochromines. Usually some meanwhile several Haplochromis species spatial segregation can be observed among are known from Lake Victoria that also have them, A. alluaudi inhabiting the shallower, many egg-dummies (see photos under H. the other species the deeper parts of the (Paralabidochromis) "rockkribensis" com- habitat. It is interesting that this non-en- plex) and at least one other species with demic, eurytopic, but trophically specialized golden male coloration, resembling that of haplochromine species is quite successful A. alluaudi. in species rich communities of stenotopic A. alluaudi lives now predominantly at rock cichlids. rocky shores and among or in the vicinity Of the tilapiine cichlids found in Lake Vic- of submerged and emerging vegetation. toria, only Oreochromis (Nyasalapia) Before the Nile perch became abundant in variabilis and O. (Oreochromis) esculentus

258 are native to the lake. The others have been observed reproductively active territorial introduced. Most of these introductions males in shallow water in the vicinity of have been done in the 1950s and early reed growing among littoral rocks. These 1960s (Oguto-Ohwayo 1990). Already be- territorial males are very beautifully deep fore that, the stocks of the two native spe- black with white spots on body and fins, cies had been declining considerably due and red eyes. They occupy territories of to overfishing. The introduced species are several square metres diameter, from now much more abundant, and much more which they keep other tilapiines, but also widely distributed in the lake than the na- larger haplochromines away. They are tives. In the following I give a short account quite aggressive and seem to dominate of the role of the tilapiine cichlid species at over the males even of the biggest rocky shores. For identification and more haplochromines that utilize the same detailed information, the reader is referred habitat (e.g. H. (Neochromis) nigricans, H. to Trewavas (1983) and Witte and Van (N.) "giant scraper"). Densen (1996). Tilapia (Coptodon) zillii (Gervais 1848), The Nile tilapia, Oreochromis (Oreochro- and probably T. (C.) rendalli (Blgr. 1896), mis) niloticus (Linnaeus. 1758), was prob- was or were introduced to convert macro- ably introduced to compensate the decline phytes into commercially exploitable food in stocks of endemic tilapiines. It became fish. These plant eaters are at rocky shores that much abundant in Lake Victoria that it less common than the two above dis- is now, together with the Nile perch and the cussed microphagous species, and are "lake sardine" (a cyprinid), dominating the probably merely occasional intruders commercial fish catch. Its juveniles and from habitats with submerged vegetation. subadults are common at rocky shores and In those habitats Tilapia species are likely may locally even dominate the fish com- to compete with plant eating haplochro- munities numerically. The larger subadults mines. However, again very little is of between 10 and 20 cm standard length known about any possible interaction. are the main target of many fishermen at Native to Lake Victoria is Oreochromis rocky shores. I have observed large shoals (Nyasalapia) variabilis (Blgr.) 1906. Known of juvenile O. niloticus grazing from rocks in parts of the lake region as Mbiru, it is in shallow waters. Particularly subadults mix the only representative of the tasselled with various species of rock-dwelling haplo- "Tilapias" (subgenus Nyasalapia) in the chromines. By Scuba diving I found spawn- Lake Victoria basin. Before man intro- ing arenas of this species at the foot of duced O. variabilis into a number of ponds rocky reefs, along the rock-sand border at in eastern Africa, its geographical distri- 6-8 m water depth. Males construct rather bution was restricted to Lake Victoria, the deep spawning pits of about one metre in Victoria Nile above the Murchison Falls, diameter and can turn the rock-sand inter- Lakes Kyoga, Kwania and Salisbury, and face into a crater landscape. Simply noth- the lower reaches of the Kagera and ing is yet known about the impact of the Orangi rivers, both draining into Lake Vic- Nile tilapia on haplochromine rock cichlids. toria (Trewavas 1983). In Lake Victoria this It is likely that, with its high population den- was originally the dominant tilapiine spe- sity, it exerts competitive pressure upon cies at rocky shores. Its young feed upon some haplochromines. the algae growth on rocks in shallow wa- Oreochromis (Oreochromis) leuco- ter. Adults feed predominantly on loose stictus (Trewavas 1933) was probably in- bottom algae, mostly diatoms, and to troduced accidentally together with O. lesser extents on phytoplankton (Fryer niloticus (Ogutu-Ohwayo 1990). It is less 1961, Fish 1955, Lowe-McConnell 1956). abundant and more restricted to inshore To spawn, adult fishes move into shallow areas with shallow water. At rocky shores areas with sandy bottoms and construct it can locally be quite numerous. I have nests. These are of a particular architecture,

259 A fry-guarding female Oreochromis variabilis.

A male Oreochromis variabilis at Makobe Island.

An OB morph of Oreochromis variabilis.

260 A male Oreochromis niloticus from Zue Island. Tilapia rendalli from Mashoro Bay.

Oreochromis esculentus from a satellite lake near the Mwanza Gulf (Ziwawa Wanawake).

Oreochromis leucostictus from Makobe Island.

261 a shallow saucer of 13-25 cm in diameter, Adult O. variabilis can be distinguished with an about 2 cm high raised rim, sur- from those of the other tilapiine cichlids in rounded by a circle of small pits, which is Lake Victoria by bright orange to red dorsal the centre of a larger pit of 30-90 cm diam- fin lappets. Juveniles lose their tilapia-mark eter (Trewavas 1983). (a dark blotch on the dorsal fin at the bor- The population of Oreochromis variabilis der between spinous and rayed part of the in Lake Victoria declined dramatically dur- fin) earlier than those of other species (at a ing the last decades. The species is now standard length of about 1.5 cm). Subadults extinct in large parts of its former range. In can be confused with those of O. the late 1970s it was still frequently en- leucostictus but have more (31-32 cf. 28- countered at rocky shores in the northern 31 according to Trewavas; 30-33 cf. 28-30 Mwanza Gulf (F. Witte pers. comm.). In spite according to Hest/Tafiri & FAO/Danida) lat- of intensive search, we could not trace any eral line pore scales. A peculiarity of O. individual during our rock shore survey. We variabilis is its polychromatism. Apart from found O. variabilis only at one offshore is- the "normal" blueish to greenish-grey land in the Speke Gulf (Makobe Island). At morph a very beautiful orange blotched that island we observed the species over (OB) morph, a rarer piebald, and a very rare several years and it appears that a small, but white morph are known. These brightly col- possibly viable remnant population exists. oured morphs, known as Maradadi, are The reasons for the decline of O. variabilis more frequent among females, but occur are not well understood. Competition with also among males (Trewavas 1983). Their introduced species of tilapiine cichlids coloration is similar to respective morphs seems to play a role in it. While the two among haplochromine cichlids. native species O. variabilis and O. The second native tilapiine cichlid of Lake esculentus were ecologically segregated Victoria, Oreochromis (Oreochromis) and did not compete much (Fryer 1961, esculentus (Graham 1928) has probably Ogutu-Ohwayo 1990), the habitat demands never been more than just a bypassing visi- of the introduced T. zillii/rendalli and O. tor at rocky shores. It used to live further leucostictus are similar to those of O. offshore than O. variabilis, and usually over variabilis (Ogutu-Ohwayo 1990, pers. obs.). soft muddy bottoms. It is closely related to Fryer (1961) described that young T. zillii O. niloticus, from which it was geographi- compete for food with young O. variabilis cally isolated before the latter species was and aggressively dominate over the latter. I introduced into Lake Victoria. Overfished observed at Makobe Island large habitat since the beginning of our century, O. overlap between O. variabilis and O. esculentus has been strongly declining in leucostictus of all size classes. In mixed Lake Victoria (Ogutu-Ohwayo 1990) and is groups of young fish O. variabilis were far now extinct in the largest part of the lake. outnumbered by O. leucostictus. However, The introduction of its larger growing rela- while adult males of the latter have their tive, O. niloticus, may have sped up this territories in shallow, immediately inshore process through competition and/or hybridi- waters, reproductively active males of O. zation. In the late 1980s we encountered variabilis are found somewhat further off- occasionally a juvenile of this species at shore. Competition with O. niloticus is also rocky shores in the northern Mwanza Gulf. probable. Subadults of the latter are abun- In the 1990s we never saw any. The pho- dant in shallow inshore waters, overlapping tograph shows an individual from a rem- with both other species. Adult males con- nant population surviving in a satellite lake struct their big spawning pits at 6-8 m near Mwanza. depth. Apart from competition with introduced species, increased predation pressure by man and Nile perch may play a role in the decline of O. variabilis.

262 Non-cichlid fishes from rocky shores

Though the fish communities of rocky bottom. Garra dembeensis is restricted shores, reefs and islands in Lake Victoria to this substrate. We found it only at fully are dominated by cichlids in species surf exposed shores and also there only numbers and individual numbers, a good in the uppermost littoral zone between number of other fish species occur in the surface and maximum one metre these habitats. Several of them occupy depth. It is most frequently seen over specific ecological niches, in which they horizontal and diagonal rock surfaces at dominate over cichlids. Examples are the depth of merely a few decimetres. Garra surf zone-dwelling Garra, the cave-dwell- dembeensis is a very mobile algae ing, nocturnal predatory catfish Bagrus scraper, mostly seen while busy rasping docmac, and the reef edge-dwelling, off algae from the rocks. Anatomically it predatory Nile perch. Many of these spe- is, with a very shallow, cylindrical body cies directly or indirectly interact with and a subterminal mouth, equipped with cichlids and therefore should be consid- a sucking disk, highly specialized for al- ered in any attempt to improve under- gae scraping in shallow, heavily moved standing of the ecology and evolution of water. It seems to utilize a niche that rock-dwelling cichlids. Nile perch and none of the algae scraping cichlids is uti- Bagrus are important cichlid predators. lizing. It is interesting that in Lake Malawi Small and juvenile Barbus and "lake sar- (Nyasa), where Garra does not occur, rock- dines" are food for fish eating cichlids, but dwelling cichlids (Labeotropheus) have may at the same time compete with adapted to fill the niche of the surf zone zooplankton eating cichlids for food. Garra algae scraper. Garra dembeensis has a and Labeo may compete with Aufwuchs wide geographical distribution in central eating cichlids for food, and catfishes of African rivers, reaching as far west as the genus Synodontis may compete for Cameroun. food with snail eating cichlids. A detailed Anatomically similar to Garra are the discussion of such interactions is beyond much bigger Labeo species (often com- the scope of this book. Nevertheless, I monly called African carps), of which one, would like to at least mention and show Labeo victorianus lives in Lake Victoria, in a few photographs the most frequently and is currently found almost exclusively encountered non-cichlids of rocky shores. over rocky substrate. This is a beautiful At least 44 species of non-cichlid fishes yellow species which is endemic to Lake have so far been recorded from Lake Vic- Victoria and probably closely related to toria (Hest/Tafiri & FAO/Danida). Of these the central African Labeo weeksii. Labeo we recorded 19 at rocky shores in our sur- victorianus, at the lake known as Ningu, vey area. Probably only one or two of is a migrator that ascends into rivers to them are restricted to rock habitats, but spawn during the rainy season in the life history of several others rocks (Cadwalladr 1965). Introduction of modern play an important role, and for some they gillnets, set at the river mouths during may be the most important places of re- migration, caused a strong decline of this treat from the high predation pressure in species since the 1950s, and the collapse more open habitats. Thus rocky habitats of a once commercially important fishery. may be important for the survival of However, the species is surviving at low some not habitat restricted species of population densities. Islands with exten- non-cichlids. sive rocky shelves that have medium Six species of cyprinids live over rock sized rock boulders in deeper waters (be-

263 Barbus tripleurospilus.

Labeo victorianus at Mashoro Bay. yond 5 m depth), are currently probably & Cordone 1974). In the rocky areas the most important habitat of half-grown Labeo probably feeds upon epilithic algae and adult Labeo in the lake. In former and debris. days the species used to live predomi- The only large growing barb of Lake Vic- nantly over muddy bottoms (Kudhongania toria, the endemic Barbus altianalis

264 Barbus kerstenii from Bukoba.

Bagrus docmac from Python Island. radcliffi, at the lake known as Kuyu, has than mud. We found it very occasionally dramatically declined as well, and is now at gently sloping rocky shores and islands considerably more rare than Labeo. Dif- in shallow water (not beyond 4 m depth). ferent from the latter, it used to prefer It feeds predominantly upon insect larvae, hard, sandy and stony bottoms rather crustaceans and molluscs (Balirwa 1984).

265 Like Labeo it migrates upstream into riv- chromines. A. pumilus stays a few deci- ers to spawn and has been strongly metres below the surface, A. loati is a overfished. Two small barbels, Barbus strict surface-dweller. jacksonii and Barbus kerstenii are encoun- Elephant-snout fishes (Mormyridae), at tered in sheltered inlets at rocky shores. the lake known as Domo-domo or Mbete The first species more frequently than (Ukerewe region), were probably once the latter. Both feed on zooplankton and more common at rocky shores than they small benthic invertebrates. Finally the are today. Seven species have been re- "lake sardine", in East Africa known as corded in Lake Victoria of which three are Dagaa or Omena (Rastrineobola argentea) endemic. All have strongly declined in re- is a regular visitor of rocky habitats. It is a cent years. The only species that we re- small "sardine like" shoaling cyprinid that peatedly observed in rocky habitats is feeds on zooplankton in the pelagic zone Mormyrus kannume, a large growing spe- of Lake Victoria. Shoals can frequently be cies that is widely distributed in the drain- observed at rocky shores, particularly age system of the Nile. When found over where steeply sloping rocky peninsulas rocks, it is usually over medium sized penetrate into the sublittoral pelagic zone. boulders, and slopes that are not too Contrary to most other species, this one steep. Just one time we caught a strongly increased concomitantly with the Marcusenius sp. over rocks. Two en- Nile perch upsurge and haplochromine demic species, M. rheni and M. victoriae decline (Oguto-Ohwayo 1990, Wanink live in the lake. Mormyrids feed in the 1991), and forms now an important rocky zone probably on insect larvae. pelagic fishery. Over rocky bottom, it is Three groups of piscivorous non-cichlid sometimes hunted by fish eating haplo- fishes play a role in the fish communities chromines (van Oijen 1992). over rocky bottom: Spiny eels, bagrid and Two other, largely pelagic, zooplankton clariid catfishes, and Nile perch. Spiny eating fishes that frequently live over eels (Mastacembelidae), which are quite rocks, are the alestides (formerly speciose at Lake Tanganyika rock shores, characides) Brycinus jacksoni and Brycinus are, in Lake Victoria, represented by only sadleri, locally known as Nsoga. They are a single species: Caecomastacembelus particularly frequently observed in shel- frenatus. This is a very beautiful one, dark tered places, where rocky habitats are brown with a net of light irregular lines surrounded by extensive shallow waters, on the head, a light dorsum and a light e.g. in the southern Mwanza Gulf. Such brown reticulated underside. Little is areas are relatively poor in planktivorous known about its ecology. We caught it at rock cichlids. The two species are fairly rocky islands in the Mwanza Gulf and I similar in habitus. However, B. jacksoni is met it a few times while diving at Makobe a representative of the large growing Af- Island in the Speke Gulf. It lives in crev- rican alestides and has large scales, remi- ices among small and medium sized niscent of the central African B. rocks and is usually seen only half, the macrolepidotus. B. sadleri remains much head and anterior body standing out while smaller and has smaller, less prominent the rest remains hidden among the scales. We found them frequently in rocks. sympatry. The cyprinodonts (killis) Aplo- At least five catfishes live over rocks, cheilichthys pumilus and Aplocheilichthys among them the two Synodontis species loati are two tiny and strictly littoral endemic to the Lake Victoria basin. Both zooplankton feeders, that can be ob- have strongly declined after the Nile served at somewhat sheltered inlets, perch upsurge (Goudswaard & Witte in where some reed grows among the press) and are generally rarely seen. We rocks. I saw them moving around rather observed the bigger S. victoriae, locally "respectless" above the heads of haplo- known as Gogogo, somewhat more often

266 over rocks than S. afrofisheri (Ngere) ern Mwanza Gulf, was 10 to 40 cm stand- which we found just two or three times. ard length (mean 16 cm). Stomach con- Both species feed on snails and insect lar- tent analysis of such rock-dwelling Nile vae. Six species of clariid catfishes, locally perch revealed that individuals of 15 cm known as Mumi or Kambale, are known standard length can swallow adult rock from Lake Victoria, two of them endemic. cichlids of 7 cm standard length! The We regularly encountered two species in adult haplochromines found in Nile perch rocky habitats. One is probably the small stomachs at Python Islands and Makobe Clarias liocephalus, the other one is still Island were Haplochromis nyererei of unidentified but is probably the juveniles both sexes. Apart from preying upon of C. gariepinus or another large growing haplochromines, rock-dwelling Nile perch species. Both live in crevices among prey on the larvae of dragon flies rocks. Possibly they prey upon small (Odonata) and on prawns (Caridina fishes. We never found adults of C. nilotica). gariepinus among rocks. The most com- At the offshore islands in the Speke mon catfish in rocky habitats is Bagrus Gulf that are surrounded by deep water, docmak, locally known as Hongwe. catch frequencies of Nile perch in shallow Hongwe and Nile perch are probably the inshore parts of the rocky habitat are low, most important fish eating predators in but I regularly observed larger Nile rocky habitats apart from birds and spot- perches, of between 40 and 60 cm stand- ted neck otters. We found Bagrus in all ard length, in deeper waters. They size classes, though full adults were rarely seemed to patrol the rocky reefs at be- caught by our gears. Bagrus hunt prob- tween 6 and 15 m water depth as if hunt- ably by night. During day light they hide ing haplochromines. However, in crevices among the rocks and as a diver haplochromines seemed not to pay very one can see hardly more than their long much attention to the predators. I met antennas peeping out from crevices. exceptionally many large and very deep Small individuals are usually found in shal- bodied Nile perches at 6 to 7 m depth at low waters, bigger ones stay deeper. the very steep slope of the offshore Last not least, also the top predator of Miandere Islands east of Kome Island/ the changed ecosystem of Lake Victoria, Sengerema region. Lying on the ground the Nile perch (Lates sp.), locally known or slowly moving along it, I had continu- as Sangara lives in rocky habitats. In the ously between two and four of them in past it was believed that Nile perch would my sight which was restricted by water not live among rocks (Witte et al. 1992). transparency to a radius of about 3 m. However, sampling of various rocky These Nile perches were rather homoge- shores and islands, and particularly scuba neous in size and were almost certainly diving at these, has in the meantime targeting at rock-dwelling haplochro- learned us that this generalization is not mines. The way they patrolled the steep (or not any more) correct. Nile perch is slope without being much scared by a quite a frequent member of the rock- diver, was reminiscent of sharks patrolling dwelling fish communities. However, its a coral reef. These scuba observations abundance differs between localities, and confirm reports of fishermen that they not at all places does it invade the shal- have discovered places with large Nile low waters. It appears that Nile perch in perches at rocky shores in the rocky habitats of the Mwanza Gulf are Sengerema region (D. Chitamwebwa smaller and invade shallower water than pers. comm.) those at offshore islands in the Speke Gulf and in the open lake. The size range of 31 Nile perch, caught in January 1996 over rocks at three islands in the south-

267 Mormyrus kanume from Makobe Island.

A subadult Lates sp. at Kissenda island.

Synodontis victoriae. Synodontis afrofisheri.

268 Evolution of rock-dwelling cichlids

Most discussions about the evolution of gest that the majority of these are derived the cichlid species flock of Lake Victoria had from only two different ancestral species been imprinted by the case of the endemic (diphyletic origin). At least the larger one of species of Ugandan Lake Nabugabo (e.g. the two major groups, the Vertical bar Fryer & Iles 1972, Greenwood 1980, 1994). Mbipi, is anatomically and in their feeding Lake Nabugabo is a small satellite lake of modes diverse, but is known exclusively Lake Victoria. It was a lagoon of the latter from rocky habitats. This implies that their until a drop in water level separated it from ancestor first evolved its specificity to this the main lake about 4000 years ago. habitat, before it underwent radiation into Populations of five haplochromine species different feeding niches within the rocky that got in Nabugabo isolated from the habitat. For the second group, the Chess- main lake populations, evolved into new en- board Mbipi, the evidence for their primary demic species (Greenwood 1965). The radiation in rocky habitats is less unambigu- short time that was sufficient for these ous, since several species have been found populations to speciate, suggested that the also in other habitats. mechanism of speciation in satellite lakes could be suitable to explain a major part of Evolution of habitat specificity and the explosive speciation of Victorian cichlids trophic primary radiation (Greenwood 1980). However, as discussed in the chapter on the history of the lake, When a lacustrine habitat began to de- that mechanism alone, appears unsuitable velop by back-ponding of rivers (see the to explain the radiation of the cichlid spe- chapter on the history of Lake Victoria), the cies flock. Distribution patterns among and riverine ancestor of the modern lake cichlids ecological characters of the lithophilic must have been confronted with a major cichlids provide evidence against the satel- new and vacant adaptive niche, the open lite lake hypothesis, and for other mecha- water body. A response to this situation, nisms of speciation (Seehausen 1996b, be- that has been documented in several low). This does not mean that the process groups of bony fish, is the splitting of the of speciation in satellite lakes does not take gene pool of the formerly riverine and now place. It certainly exists, and may have lacustrine species into a benthic (littoral- and played a role also in the explosive multipli- bottom-dwelling) and a limnetic (open wa- cation of species, but it is most unlikely that ter-dwelling) species (Bentzen & McPhail it was the only, or the most important proc- 1984, Bentzen et al. 1984, Skúlason et al. ess. Instead of discussing the shortcomings 1989, Taylor & Bentzen 1993, Taylor & of the satellite lake hypothesis, I want to McPhail 1994). A splitting into these two make an attempt to construct from several adaptive types may have been the first lines of circumstantial evidence an inte- speciation event in the cichlid population grated hypothetical picture of events and during the riverine-lacustrine transitional processes, that may have been of impor- period in the early days of Lake Victoria. As tance in the evolution of rock-dwelling the lake increased in size, habitat diversity cichlids in Lake Victoria. also increased, and the benthic species The distribution of melanin patterns and may have split up further into forms that in- patterns of chest squamation, as well as habited the major substrate types. By ad- similarities in the not yet decoded "general aptation to the different habitats in behav- appearance" of rock-dwelling cichlids, sug- iour and/or coloration, hence by changes

269 that incidentally affected the Specific Mate species that got adapted to the conditions Recognition System (SMRS, see the chap- in rocky habitats, made up one large popu- ter on taxonomy), reproductive isolation be- lation with frequent migration between tween these forms may have evolved. Simi- rocky habitat patches. lar events seem to still happen among the The dominance of Vertical bar Mbipi in modern cichlids of Lake Victoria (see under species and individual numbers, as well as H. (?) tanaos and H. (?) thereuterion). By that in ecological diversity in any rock cichlid time distances among patches of rocky community of modern Lake Victoria, sug- substrate must have been relatively small gests that it was their ancestor that first colo- and the shallow open water separating nized the rocky habitat in this early phase. them was no formidable barrier to migra- After it got behaviourally and anatomically tion. Similar situations exist at present in the adapted to the rocky habitat, and as habitat shallow southern parts of the Mwanza diversity continued to increase, this ances- Gulf, where each rock-dwelling species tral rock-dwelling species may have under- seems to form one outbreeding population, gone trophic radiation. Again, in analogy to and we found little, if any intraspecific geo- what is now known from other bony fish, graphical variation. Thus, it is likely that the the result of this trophic primary radiation

Figure10a: Examples of anatomical diversity of rock-dwelling haplochromines (X-ray photographs): (a) Major adaptive types within the Vertical bar Mbipi (left column) and Chess-board Mbipi (right column) as the result of trophic radiations within both lineages. Upper row: Species with a long head, moderate to relatively long lower jaws, and relatively small eyes feed predominantly on larger insect larvae (e.g. Ephemeroptera) but also on plankton. H. (?) “deepwater” (“Deepwater” complex, left). H. (Paralabidochromis) chilotes (Chilotes complex). Middle row: Species with a short head, short lower jaws, and large eyes feed predominantly on small insect larvae (e.g. Chironomidae). H. (?) “black and yellow pseudonigricans” (“Pseudonigricans” complex, left). H. (Paralabidochromis) “Zue rockpicker” (“Rockpicker” complex). Bottom row: Species with a steep dorsal head profile and short but broad and strong lower jaws predominantly scrape algae from the rocks. H. (Neochromis) nigricans (Neochromis complex, left), H. (Paralabidochromis) “short snout scraper” (“Rockkribensis” complex).

270 may have been more benthically and more modes and anatomy, the different modern pelagically feeding species. As compared lineages are on average clearly adapted to to the ancestral species, the benthic forag- different feeding modes. That these modes ers increased their biting power by devel- are complementary in the sense of trophic oping short jaws, supported by strong mus- niche partitioning is apparent from the cles to scrape unicellular and/or filamentous structure and composition of modern algae from the rocks, the others increased rock cichlid communities (Seehausen et their sucking efficiency by developing al. in press [a], Seehausen & Bouton longer jaws and a protrusile mouth to prey 1996a, see chapter on ecology). upon insect larvae and/or plankton (text figure 11a). Such could have been the origin A role for sympatric speciation? of the major species complexes of vertical bar Mbipi (Neochromis, Xystichromis, This trophic primary radiation is unlikely Nyererei complex, "Deepwater" complex, to have happened by allopatric speciation. "Pseudonigricans" complex). Though all of It does not appear plausible that the de- them exhibit some diversity in feeding scendants of an ancestral population, if

Figure 10b: adaptive types of other rock-dwelling lineages: Upper right: Snail shellers have a steep dorsal head profile, short but strong jaws, and very procumbently implanted teeth (H. (Ptyochromis) sauvagei). Upper left: Paedophages have a short head, relatively long and very strong lower jaws (H. (Lipochromis) “velvet black crypotodon”). Middle right: Fish and crab eaters have a long head and long lower jaws (H. (Harpagochromis) howesi). Lower left: Snail crushers are deep bodied and have strongly enlarged pharyngeal bones (H. (Labrochromis) “stone”). Lower right: Predators of large insect larvae and small molluscs have a longer head and slightly enlarged pharyngeal bones (H. (Psammochromis) saxicola).

271 X-ra An example of polychromatism: different morphs of H. "blue scraper" at Makobe Island.

An example of a sympatric sibling species pair: two series of males of H. "red pseudonigricans" and H. "blue pseudonigricans" caught in a single haul.

272 split over several lake basins, would in each tomically very similar species (sibling spe- basin evolve specificity to rock substrate, cies), that differed slightly in but evolve in each basin a different mode microdistribution and feeding, and distinctly of feeding, and that these feeding adapta- in male coloration. Their distribution pat- tions are complementary to each other. terns do not give any evidence for that they Moreover, that rock-cichlid communities in may have speciated in allopatry. These sib- Lake Victoria comprise the same species ling species pairs usually consist of one spe- complexes at all until now sampled places cies that feeds more benthically and one would not be explained. The possibility of that feeds more pelagically. Examples are speciation in the absence of geographical "blue" and "red" "pseudonigricans", "blue" and barriers (sympatric speciation) has always "yellow" "deepwater", and "zebra" and "red" been disputed. However, recent evidence Nyererei. from molecular genetical studies makes According to the basic model of sym- likely that sympatric speciation in fish, in- patric speciation (Maynard Smith 1966), cluding African cichlids, is happening (Taylor speciation in the absence of geographi- & Bentzen 1993, Schliewen et al. 1994, cal isolation is possible if (1) morphs of a Taylor & McPhail 1994). In all these cases, polymorphic species have different selec- the speciation events coincided with the tive advantages in different microhabitats, invasion of a riverine species into a vacant (2) display some differences in micro- lacustrine environment, and with a splitting habitat distribution, and (3) mate into more benthic and more pelagic feed- assortatively. It is most likely that the fre- ers. quent occurrence of polymorphism in Vic- Some circumstantial evidence from rock- torian rock cichlids is instrumental in dwelling cichlids suggests that Lake Victo- sympatric speciation. Polymorphisms are ria cichlids are able to speciate in the same most apparent if they concern male nup- way. We found several cases of almost iden- tial coloration (male nuptial poly- tical geographical distribution of two ana- chromatism). However, polymorphisms in

14,000 years BP 8,000 years BP

Distribution of Speke Gulf endemics Distribution of Mwanza Gulf endemics

Figure 11. Two major areas of endemism in relation to pre-historic lake level fluctuations. Thin line = boundaries of present days Lake Victoria. Bold line = boundaries of the pre-historic lake (after Stager et al. 1986). Further explanation in the text.

273 jaw anatomy and dentition exist as well gle basin had already formed before the among rock cichlids (see page 79). Par- secondary radiations began. ticularly differences in anatomy are likely to cause differential advantages in differ- Secondary radiations ent microhabitats with different food by allopatric speciation sources. When an anatomical polymorphism gets genetically linked with a male As the lake continued to grow, the wa- polychromatism, the conditions for sym- ters separating patches of rocky habitat be- patric speciation may be fulfilled. In that came deeper, distances between them case, morphs may evolve into reproduc- larger. This may have led to a restriction of tively isolated species. gene flow among the populations of rock- If sympatric speciation via polymor- dwelling cichlids. The dwindling chances of phism exists among the modern Lake an emigrant to reach another patch of rocky Victoria cichlids, it may have played an habitat would have reinforced habitat important role in the trophic primary ra- specificity. As a consequence of their in- diation of the rock cichlids. By the time creasing geographical isolation, populations of this assumed trophic primary radiation likely started to diverge from each other. of rock-dwelling cichlids the lake may still Today this situation can be observed among have been small and migration between populations of rock-dwelling cichlids in the rocky areas likely caused a spreading of northern Mwanza Gulf. Allopatric the new feeding types, each represented populations of several species that live at by one species, over the entire lake. patches of rocky habitat, separated by The more the entire Victorian species sublittoral areas of 10-15 m depth, differ in flock became diversified, the more differ- anatomy (N. Bouton, J. de Visser, C.D.N. ent lineages would have happened to in- Barel & O. Seehausen in manuscript), col- vade the rocky habitat. However, they met oration (see page 276) and genetics (Dorit with already well adapted Vertical bar 1990). Where gene flow was little enough, Mbipi and could establish themselves differences evolving among allopatric only here and there, where vacancies ex- populations likely affected the SMRS. Indi- isted in the Mbipi communities (text fig- viduals of such different populations when ure 11b). For instance the Ptyochromis they happened to meet again, would not lineage successfully occupied the niche have recognized each other any more, but of the oral shelling snail eater at rocky behaved as two different species. Distribu- shores. Apart from the Vertical bar Mbipi tion patterns suggestive of this kind of allo- only one lineage (the Chessboard Mbipi) patric speciation can be observed within underwent a real secondary radiation that complexes of sibling species in the Speke must at least partly (Paralabidochromis Gulf, where distances among populations "rockpicker" complex) have happened in are bigger, and the waters separating them the rocky habitat. This lineage may have deeper than in the Mwanza Gulf (e.g. "blue been pre-adapted to fill vacancies in the nyererei"/"pink anal"/"red rim anal", "blue communities of Vertical bar Mbipi, be- scraper"/"velvet black"). In this way rapid cause of their quite different jaw anatomy. multiplication of each adaptive type, origi- Alternatively, if nascent Lake Victoria still nally represented by one species, is likely consisted of more than one lake basin by to have happened (secondary radiation) as the time that habitat specificity had stasimorphic speciation (speciation without evolved, the two melanin pattern lineages major anatomical innovation) in among the rock cichlids may have had their intralacustrine allopatry. origins in two different basins. However, the There are several lines of evidence for distribution all around the lake of one or that intralacustrine speciation was common two species of most feeding types of both among rock-dwelling cichlids: melanin pattern groups, implies that a sin- (1) Regional endemism and lake basin

274 history. During a period of dry climate, Factors driving divergent evolution about 14,000 years BP, the water level of in allopatry Lake Victoria was about 70 m lower than it is now (Stager et al. 1986, see the chap- If gene flow between allopatric popula- ter about the history of the lake). If the tions is little, differences between popula- lake had not completely dried up by then, tions can arise merely by incidence. How- it was considerably smaller than it is now. ever, circumstantial evidence suggests that The part of the lake, the cichlid fauna of several factors may accelerate genetical which is described in this book, was en- changes in allopatric populations. Partial cor- tirely dry (see text figure 12). A large relations between male coloration and char- number of rock-dwelling cichlid species acters of the light regime in the specific en- is now endemic to parts of this relatively vironment of a cichlid population strongly recently re-flooded area. These species suggest that light regimes have an impact almost certainly have evolved after the on the evolution of coloration (see pages lake had come back, thus within the last 121ff and 138ff). Since coloration in turn is 13,000 years. Moreover, 8000 years ago likely an important component of the the lake level was much higher than it is SMRS, populations that have been living now (see text figure 12). Flat rocky is- under different light conditions (deep ver- lands, like Makobe Island in the Speke sus shallow waters, steep versus gentle Gulf must have been flooded that deeply, slopes, clear versus murky waters, among that stenotopic shallow water-dwelling versus above rock boulders etc.) sufficiently rock cichlids like Neochromis would not long enough to evolve genetical adapta- have found suitable conditions. The tions, are likely to be reproductively isolated Mwanza Gulf, as a peripheral arm of the upon secondary contact. Examples might lake, did not exist because the lake be the blue and the red coloured "Short shores were a good distance further snout scrapers" (page 163ff), and the blue south. It is likely that its endemic species, and black "Pseudonigricans" complex spe- and many endemics of other southern re- cies (pages 133 and 135). Such cases of sib- gions evolved within the past 8000 years. ling species differ from those suggestive Interesting in this respect may be a ma- of sympatric speciation in that the distribu- jor water level recession that has been tions of siblings are in large mutually exclu- dated back to about 3700 years BP (Tem- sive, and areas of overlap are marginal to ple 1967). their geographical ranges. Further evidence (2) Chain species. If deep water is a bar- for the impact of light regimes on male col- rier to gene flow among populations of oration is the recurrent appearance of col- rock-dwelling cichlids, gene flow should oration patterns in different groups living occur predominantly along the mainland under similar ecological conditions, and shores and along underwater ridges. If in therefore under similar light regimes. Ex- this way restricted gene flow leads to amples are black males with yellow fins in speciation within the lake, one would ex- the Nyererei and "Pseudonigricans" com- pect chain species to occur. Chain species plexes, and males with a red dorsum and consist of a series of geographically iso- blueish underside in the same two species lated populations, the first and the last of complexes. which behave as reproductively isolated Many species of rock-dwelling haplochro- species when they happen to meet, mines exhibit polymorphisms in male nup- though they are connected by intermedi- tial coloration (Seehausen & Bouton 1996, ate populations, among which gene flow further examples in this book). While pos- exists. An example of what looks much sibly playing a role in sympatric speciation, like a chain species is provided in text fig- these polychromatisms may also have an ure 13. accelerating impact on the process of allopatric speciation. They may often pro-

275 Figure 12: Distribution patterns among H. “zebra nyererei”, H. “big blue”, and intermediate populations, suggestive of a chain species complex. At Juma Island both species live sympatrically. vide the material upon which selection respective female recognition criterion. will act. A good example are the red and This may have happened in H. (?) "black yellow anal fin morphs of H. (?) "yellow and yellow pseudonigricans" and H. (?) chin pseudonigricans" (page 138). Selec- "blue pseudonigricans" that, compared to tive advantages of the one or the other the widely distributed H. (?) "yellow chin morph under different light conditions pseudonigricans", lost the red anal and the will, over sufficiently long time, lead to yellow anal fin morph respectively (pages the extermination of the second morph. 133 and 135). Probably the most important selective Interaction among species, may poten- advantage of a male colour morph over tially play a major role in the process of another one is a higher attraction for fe- allopatric speciation by sexual selection. males. In haplochromine cichlids females For instance interspecific competition for are usually the sex that chooses the part- territorial space or for food, may cause the ner (Dominey 1984, Seehausen 1996b). individuals of a population to diverge Thus, the more attractive male colour from their "usual" microhabitat, and to ex- morph leaves on average a higher pose themselves to aberrant light re- number of offspring. Hence its relative gimes (see the discussion under H. (?) abundance will increase on cost of the "blue nyererei"). The females' mate choice less attractive morph. Speciation by such would then do the rest. It is probably the sexual selection is completed, once the same polymorphisms that are instrumen- alternative colour morphs are eliminated tal in allopatric and sympatric speciation from two populations, together with the under different situations, and it is likely

276 that both mechanisms of speciation each feeding type, represented by one played a role in the secondary radiations species (the ancestor of each modern of rock-restricted cichlids, though the species complex) was distributed all number of sibling species pairs that ex- around the lake. As gene flow among hibit distribution patterns suggestive of populations decreased, some populations allopatric speciation is bigger. may have undergone genetic changes faster than others (due to small popula- Matrix species tion size and/or strong local selection regimes) and may have had changed their Possibly as a consequence of the above SMRS by the time that a new wave of im- sketched temporal sequence of events, migrants of the matrix species reached almost every rocky shore cichlid commu- their habitat patch. Such populations nity in modern Lake Victoria is composed would not crossbreed with the matrix of the same basic faunal elements: species, but would persist next to it as a Neochromis, Xystichromis, Nyererei com- second species. plex, "Pseudonigricans" complex and Due to the wide distribution of the ma- "Deepwater" complex (the latter two can trix species, the probability that it colo- be missing in particular habitat situations) nizes each rocky habitat patch repeatedly of the Vertical bar Mbipi, "Rockkribensis" is very high. The probability that it re-colo- complex, "Rockpicker" complex, and nizes a patch inhabited by a new species Chilotes complex of the Chessboard is bigger than the probability that a new Mbipi. Moreover, each of these com- species spreads to other patches. There- plexes comprises one species or fore, it is unlikely that a matrix species got superspecies that is distributed through- extinct from any place forever, as long as out the investigated parts of the lake, and migration was not entirely disrupted. At a much larger number (except in the the same time, it takes long for a local- Chilotes complex) of regional endemics. ized new species to spread. If in this way, I refer to the widely distributed species the large and widely dispersed ancestral as to the matrix species* [foot note: *my form of each species complex has been use of the term matrix species is slightly carried along until today in form of the different from the use by Kaufman (in matrix species, while butting off regional press)] of each lineage. I propose the fol- daughter species, the distribution pat- lowing explanation of this distribution patterns found among the modern Mbipi tern: By the time the lake started to grow species would largely be explained. At bigger, and rock-dwelling populations got present it is difficult to test the hypoth- geographically isolated from each other, eses proposed here because molecular genetics do not yet reveal enough variation among Lake Victoria haplochromines to re- solve the phylogenetic puzzle. However, with the further im- provement of techniques and the development of new ones, I am optimistic that this will become possible.

Lates sp. with and adult male H. nyererei in its stomach. In front a second H. nyererei of similar size for comparison.

277 Conservation of rock-dwelling cichlids

After the loss of large parts of the cichlid Despite the importance of rocky habitats fauna of Lake Victoria, following the boom as the most species rich cichlid refugium, of the Nile perch (Lates sp., Witte et al. species diversity is declining there as well. 1992), rocky habitats became a major Regrettably, in the pre-Nile perch era, rocky refugium for cichlid species diversity. Rocky shores have been studied only in a very habitats currently harbour the most species small area (northern Mwanza Gulf: Nyegezi rich, and the ecologically most complex fish Rocks, Kilimo Island, Anchor Island, Hippo communities (Seehausen et al. in press [b]). Island). From this area at least eight rock- This has been attributed to the absence of dwelling haplochromines (H. (Neochromis) Nile perch from such habitats. Recent in- "kruising", H. (?) "purple rocker", H. (Para- vestigations by use of Scuba, however, re- labidochromis) "rockpicker", H. (P.) "elongate vealed that Nile perch is quite abundant and rockpicker", H. (Harpagochromis) howesi, H. quite able to hunt in rocky habitats. It seems (H.) guiarti, H. ("Astatotilapia") barbarae, H. that the spatial complexity of the habitat al- (Lipochromis) melanopterus) and many oc- lows cichlid populations to withstand the casional intruders disappeared until 1996. predation pressure, and that the balance This is about one third of the original rock- between predator and prey is being main- dwelling fauna of the area. During the past tained better than in other habitats. five years we witnessed population de- Most of the cichlid species that are now clines of several other species. The abun- living at rocky shores are stenotopic rock- dance of H. (?) "zebra nyererei" at Python Is- restricted species like the Mbipi. Several lands decreased by the factor four between others, however, were more widely distrib- 1991 and 1995 (Seehausen et al. 1996). H. uted in the pre-Nile perch era, but survived (Neochromis) nigricans at the same islands in the Nile perch era only at rocks. Most, if declined even stronger (Seehausen 1996c). not all of these "rock refugees" have already H. (Paralabidochromis) "rock macula" may in the pre-Nile perch era occurred in rocky have gone extinct at Nyegezi Rocks be- habitats, while many similar species, that tween 1993 and 1996 (see page 167). have not occurred in such habitats, went Several factors seem to contribute to the extinct. Thus, the cichlid species assem- decline of rock-dwelling cichlids. Lake Vic- blage that survived the Nile perch boom in toria is undergoing eutrophication at an in- rocky refugia, is not a random sample, but creasing rate (see the chapter on limnology). rather a selection of pre-adapted species. As a consequence of eutrophication a con- Nevertheless, it is ecologically diverse, siderable decrease in water transparency is and includes species such as the fish eat- observed. This is likely to affect cichlid com- ers H. (Harpagochromis) cavifrons and H. munities via its impact on food resources (Harpagochromis) serranus, the paedo- and on the reproduction system. Most if phages H. (Lipochromis) cryptodon and H. not all haplochromine species of Lake Vic- (Lipochromis) melanopterus, the insect toria produce viable and fertile offspring if eater H. (Paralabidochromis) chromo- they happen to hybridize. As a consequence gynos, the pelagic phytoplankton eater H. of this reproductive compatibility, reproduc- (?) "kribensis", and the tilapiine Oreo- tive isolation among species is labile. If pre- chromis variabilis. The characters, that zygotic reproduction barriers break down, make these species pre-adapted to inhabit nothing will prevent hybridization, by which rocky substrates, may differ among spe- two species may become one. While many cies. haplochromines that inhabit murky waters

278 over sublittoral mud bottoms, are seasonal widely distributed species with higher dis- spawners and have distinct spawning areas persal abilities take over (Seehausen et al. (Goldschmidt & Witte 1990), many rock-re- in press [b]). Generally, the small population stricted species spawn throughout the year, sizes typical for many rock-dwelling cichlids, and their spawning sites are largely identi- make them particularly vulnerable to fish- cal with their foraging sites (Seehausen ing. 1996d and unpubl. data). Segregation of The water hyacinth (Eichhornia species by spawning area or spawning sea- crassipes) is a floating weed that origi- son can therefore not play a major role in nates from South America, but has maintaining reproductive isolation among spread over large parts of Africa. It is them, but specific mate recognition (see present in Lake Victoria since the 1980s the chapters on taxonomy and evolution) and underwent in southern Lake Victoria must be of particular importance. If male recently rapid expansion (Bwathondi & coloration plays an important role in mate Mahika 1996, Seehausen et al. 1996). It recognition, reproductive isolation can now covers large areas in sheltered situ- break down if decreasing transparency of ations like the Mwanza Gulf. Its impact on the water goes with changes in the aquatic the cichlids at rocky shores is consider- light regimes. Such changes are usually a able. Weed carpets deprive the rocky lit- narrowing of the light spectrum, and a rela- toral of light, cause water stagnation, tive loss of short wave lengths. As a con- rapid accumulation of rotting plant mate- sequence, colour signals can be set out of rial among the rocks, and, consequently, function which may lead to hybridization oxygen depletion. Fish densities under among otherwise similar species. Patterns permanent weed carpets are low, and suggestive of such or similar events have rock-restricted cichlids decline in favour been described (Seehausen 1996b) and are of more eurytopic species like H. corroborated by strong correlations be- (Haplochromis) "purple yellow". These tween species richness and aquatic light changes in species composition can be regimes (Seehausen, J.J.M. van Alphen and observed even under the relatively light F. Witte unpubl. data). water hyacinth carpets of unstable exten- Fishing pressure at rocky shores and at sion, that are found in exposed situations. many islands has increased considerably At places where the water transparency over the past decade. In some regions fish- allowed Scuba observations, I observed ermen target at rock-dwelling haplochro- that Mbipi species like the epilithic algae mines for human consumption. Almost scrapers of the Neochromis lineage, and everywhere they are collected in large pickers of the Paralabidochromis lineage, numbers as the most important bait in Nile are among the first species to disappear perch long line fishery. At many places, fi- under water hyacinths (Seehausen et al. nally, they are a common by-catch of tilapia 1996). Strongly affected are furthermore fishermen. The tilapia fishery with line and those species that live in rockpools and large hooks affects predominantly the larger rocky crevices in shallow water (e.g. H. (?) haplochromines. Particularly deleterious, "zebra nyererei"). Their habitat can be en- however, is the use of poison (insecticides) tirely destroyed when the interstices for fishing. This has become a rather com- among the rocks are getting filled up with mon practice of fishermen targeting at rotting plant material. Oreochromis niloticus at rocky shores. Eutrophication, overfishing, and spread- Cichlid communities at very small rocky ing of water hyacinth all have negative im- outcrops can be heavily affected by poison. pacts on the cichlid species diversity at It is likely that frequent poisoning leads to rocky shores. Nile perch predation may changes in species composition: stenotopic have its impact as well, as may have com- species with low dispersal abilities, thus petition with rock-dwelling stages of the most local endemics, disappear, while introduced tilapiine cichlids (see page

279 The load of sediments and nutrients flushed int the lake by rivers has increased tremendously because of deforestation.

258ff). Regrettably, little attention has of a once extraordinarily rich species di- been paid in the past to the fish commu- versity at rocky shores. nities of rocky shores in Lake Victoria. Some of the threats to the rocky shore Hence we are just beginning to under- fish communities are topographically lo- stand their importance in the ecosystem calized, and the establishment of fish and the factors threatening them. A con- parks would be useful to protect at least servation strategy to protect these com- some areas against them (e.g. munities is urgently needed. However, as overfishing). To choose appropriate sites many of the threats are problems that af- for park establishment, the taxonomical fect the ecosystem in its entirety, conser- and ecological inventory of species diver- vation strategies can only be successful, sity along rocky shores has to be contin- if incorporated in a larger management ued, and should encompass the shores plan for the Lake Victoria basin. all along the lake. Unfortunately it is at The most urgent need is to provide the present very difficult to secure the finan- industries and towns around the lake with cial means that are needed to implement efficient waste water treatment systems such tasks. Finally, the upcoming to slow down water pollution and aquarium fish trade with rock-dwelling eutrophication. More than 40 industrial Lake Victoria cichlids should consider the plants operate, and seven large towns are small population sizes of many of its tar- situated along the lake shores. According get species. If it is well organized and im- to Tanzanian news sources, over two mil- plemented, aquarium fish export may lion litres of untreated sewage and indus- help to create locally and globally aware- trial waste are discharged into the lake ness of the values that are at risk. How- every day. If this situation is not changed, ever, if not, it may as well pose one more we will very soon be left with mere re- threat to the survival of rock-dwelling membrance and fragmentary knowledge cichlids in Lake Victoria.

280 Distribution maps

281 282 283 284 285 286 287 288 289 290 Table 3 Taxonomic measurements of rock-dwelling haplochromines

Explanations The table contains a set of morphometric measurements and qualitative characteristics of the dentition that were found to be useful for species identification. For described and undescribed species the values obtained in this study are shown. They can deviate from values given by other authors (see table 4).The table is at places incomplete because not all measurements could be taken yet. For the definition of measurements see the chapter on identification, and Barel et al. 1977. SL = standard length, BD = body depth, HL = head length, JR = lower jaw length/width ratio, LJL = lower jaw length, EL = eye length, IOW = inter orbital width. Values are mean values, except for SL. SL is the maximum of the random sample of adult males used for these measurements. Thus, it is not necessarily the maximum attained SL of a species but merely an orientation help. OT, IR, and gap are dentition characters. Values in front of an oblique bar refer to the lower jaw, those behind it refer to the upper jaw. Only one value indicates equal conditions in both jaws. OT = shape of anterior teeth in the outer tooth row: 1 = equally bicuspid, 2 = subequally bicuspid, 3 = unequally bicuspid, 4 = eakly bicuspid, 5 = unicuspid, ° indicates obliquely truncated cusp shape (see page 59). IR = number of inner rows. A mean value between two integers indicates a range between the two integers, if ranges are larger, they are given instead of the mean. gap = the width of the gap that separates the outer tooth row from the first inner row: - = no gap, + = small, ++ = moderate, +++ = wide gap.

SL BD HL JR LJL EL IOW OT IR gap

Vertical bar Mbipi Neochromis complex H. (Neochromis) nigricans 92.6 37.4 32.3 1.0 34.5 2/2 5-6/5-7 -/-,+ H. (Neochromis) "short head nigricans" 114.7 34.6 29.3 1.0 33.2 27.1 1.8/1.8 4-6/5-6 -,+/+ H. (Neochromis) "large eye nigricans" H. (Neochromis) "giant scraper" 143.9 30.7 1.2 36 2/2 3-8/5-8 -,+/+ H. (Neochromis) "red tail giant scraper" 120.9 37 29.3 1.04 33.9 24.0 H. (Neochromis) "black tail giant scraper" 118.2 39.3 30.4 1.07 34.0 23.9 H. (Neochromis) "blue scraper" 105.2 35.7 31.1 1.0 35.9 2-3/2 4-6/5-7 -/-,+ H. (Neochromis) "eastern blue scraper" H. (Neochromis) "Juma scraper" 99.6 36.8 31.4 1.05 37.7 24.9 2/2 4-6/5-6 -/-,+ H. (Neochromis) "Vesi scraper" 105.7 38.5 31.0 1.0 36.8 2/2 4-8/6-9 -,+/+ H. (Neochromis) "yellow anal scraper" 102.0 36.0 30.6 1.05 36.0 2/2 4-5/5-6 +/+ H. (Neochromis) "long black" 101.0 30.4 29.8 1.1 36.7 2/2 4-5/4-5 -,+/+ H. (Neochromis) "Labeo scraper" H. (Neochromis) "black nigricans" 109.7 38.6 30.1 0.93 36.8 2/2 4-6 -,+ H. (Neochromis) "velvet black" 115.2 36.04 31.2 1.13 35.9 1-4/1-4 4-6/4-6 -,+/+,++ H. (Neochromis) "unicuspid scraper" 91.02 30.7 30.8 1.30 38.1 2-5/2-5 3-4/3-5 +,++,+++ H. (Neochromis) "Bihiru scraper" bicuspid morph 135.3 30.3 1.20 35.8 2-5/2-5 2-5/3-5 +,++,+++ H. (Neochromis) "Bihiru scraper" unicuspid morph 122.5 32.1 1.31 40.8 (4-)5/(4-)5 2-4/2-5 ++,+++ H. (Neochromis?) "pseudoblack" 83.6 33.8 31.6 1.45 36.7 2-3/2-3 2-5/2-6 ++/++

Xystichromis like algae scrapers H. (Xystichromis) "copper black" 38.4 33 1.30 38.5 24.7 28.4 2-3/2-3 3-5/4 +/++ H. (Xystichromis) "short scraper" 78.9 36.3 35.3 1.32 38.4 24.0 24.1 H. (Xystichromis) "red anal blue" 96.33 36.9 31.9 1.3 36.8 23.3 27.0 H. (Xystichromis) "large eye black" 98.6 37.9 34.5 1.49 40.9 30.5 28.3 2-3 3/2-3 H. (Xystichromis) "carp" 118.0 38.1 32.6 1.2 40.7 22.8 29.1 3.9/3.9 3-7/3-6 ++/++,+++ H. (Xystichromis) "orange carp" H. (Xystichromis) "red carp" 115.0 37.5 32.3 1.44 39.2 23.6 26.8 2-5/4-6 +,++,+++

291 SL BD HL JR LJL EL IOW OT IR gap

Nyererei complex H. (?) nyererei 103.7 38.4 33.2 1.53 42.6 25.2 26.1 4.9/4.9 2-3/2-4 ++,+++ H. (?) "black and orange nyererei" 91.8 32.3 32.4 1.68 41.2 24.7 27.3 4.7/4.5 H. (?) "orange anal nyererei" 125.2 37.7 32.0 1.25 41.8 23.7 26.4 5/5 2-3 H. (?) "small mouth nyererei" 109.8 40.6 31.3 1.64 40.7 24.8 27.2 5/5 H. (?) "Ukerewe nyererei" 96.7 38.7 33.4 1.50 41.4 25.3 26.3 5/5 H. (?) "lemon fin nyererei" 133.0 38.7 31.7 1.35 41.6 21.8 30.0 5/5 H. (?) "orange dorsal nyererei" 89.1 36.0 31.4 1.78 39.4 24.0 29.1 5/5 H. (?) "zebra nyererei" 110.3 38.7 35.1 1.55 43.4 23.1 25.2 5/5 2-3/2-4 ++ H. (?) "big blue" 113.1 39.5 33.7 1.43 43.5 21.8 23.8 4/3 2-3/3-4 +++ H. (?) "big blue" red 109.8 37.6 32.1 1.42 42.6 23.1 26.3 4.2/4.1 2-3/3-4 ++,+++ H. (?) "red head nyererei" 96.5 34.5 33.2 1.44 43.2 23.1 23.6 H. (?) "zebra Miandere" 110.6 37.6 31.6 1.44 41.1 23.8 28.3 5/5 +++ H. (?) "zebra Senga" 96.5 35.5 34.2 1.37 38.8 3.8/3.8 2-3 +++ H. (?) "all black nyererei" 77.0 36.2 1.33 39.7 5/5 2-3/2-5 +++ H. (?) "blue nyererei" 89.5 35.9 32.5 1.50 38.5 25.8 25.1 4.9/4.9 2-3/2-3 H. (?) "pink anal" 87.5 36.8 32.7 1.5 38.6 22.9 24.0 5/4.7 2.2/2.3 +++ H. (?) "red rim anal" 97.2 37.0 32.5 1.6 37.9 25.7 26.4 4.5/4.5 2/3 H. (?) "red anal nyererei" 97.0 34.8 31.6 1.49 38.0 26.1 24.5 5/4.2 2-3/2-5 H. (?) "Bwiru nyererei" 105.7 33.8 33.0 1.39 40.2 25.0 25.0 +++ H. (?) "red flank nyererei" 87.9 37.7 32.2 1.62 39.2 24.1 24.5 H. (?) "all red nyererei" 100.0 37.5 31.6 1.37 40.4 25.3 24.8

Haplochromis "pseudonigricans" complex H. (?) "yellow chin pseudonigricans" 103.9 35.5 31.9 1.41 38.0 26.9 22.3 3.9/3.6 2-3 +++ H. (?) "orange pseudonigricans" 96.0 38.3 32.3 1.5 36.8 27.4 23.8 4.8/4 H. (?) "black and yellow pseudonigricans" 85.1 36.5 31.7 1.42 35.2 27.9 22.1 4.6 2.56 H. (?) "large eye pseudonigricans" 91.1 36.8 31.8 1.2 37.5 30.2 23.8 3.6/3.9 H. (?) "blue pseudonigricans" 95.6 37.6 33.1 1.6 37.8 27.3 23.3 2.8/2.7 2/2-3 H. (?) "red pseudonigricans" 80.1 37.9 33.1 1.7 38.1 25.8 22.4 2.8/2.6 2-3/2-3 +++ H. (?) "scraper pseudonigricans" 101.4 35.4 32.2 1.4 36.9 28.3 24.4 2.4/2.2 3-4/3-5 +/++,+++ H. (?) "Ukerewe pseudonigricans" 98.8 36.6 31.3 1.5 37.4 25.1 26.4 2.8/2.8 H. (?) "Gana pseudonigricans" 95.8 34.7 30.0 1.4 37.4 25.7 24.9 3.5/3.5 H. (?) "black Ukerewe" 103.0 35.5 32.5 1.59 38.9 26.4 21.9 4.2/3.7 3/3 ++,+++/+++ H. (?) "long snout pseudonigricans" 92.6 37.4 33.4 1.6 41.2 25.5 21.0 4/3.6 2-3 +++ H. (?) "pseudoblue" 92.4 37.7 33.7 1.7 41.1 4.2/4.2 2/3 +++

Haplochromis "deepwater" complex H. (?) "deep water" 116.2 39.6 34.2 1.8 42.3 23.0 23.6 4.6/4.6 2/3 ++,+++ H. (?) "blue deep water" 101.0 38.8 34.0 1.52 39.7 22.4 23.8 4/4 3.6/4.3 +++ H. (?) "yellow deepwater" 99.0 36.3 34.2 1.4 41.0 4.3/4 3/3.6 +++ H. (?) "slender deep water" 105.4 36.5 32.7 1.9 42.4 26.5 22.6 4.8/4.8 2.5/3

Chessboard Mbipi Paralabidochromis "rockpicker" complex H. (Paralabidochromis) "pseudorockpicker" 88.7 37.4 31.7 1.28 32.8 30.3 23.0 4.8/4.6 2-3 ++,+++ H. (Paralabidochromis) "red pseudorockpicker" 83.7 34.5 29.9 1.27 33.2 24.4 19.6 3.5 H. (Paralabidochromis) "southern pseudorockp." 85.0 36.7 32.1 1.31 34.3 2.8/2.8 2 ++,+++ H. (Paralabidochromis) "chessboard picker" 77.0 34.8 31.2 1.39 31.2 26.2 20.1 5/5 2/2.5 H. (Paralabidochromis) "yellow pseudorockpicker" 99.1 36.8 30.1 1.07 29.0 25.5 25.4 3.3/2.6 2.5/2.8 (++),+++ H. (Paralabidochromis) "Zue rockpicker" 89.1 36.2 30.6 1.14 28.8 29.8 26.6 4.5/3.5 3.5/3 H. (Paralabidochromis) "rockpicker" 2-5 3-4(5) + H. (Paralabidochromis) "elongate rockpicker" 91.03 32.5 29.6 1.11 31.9 25.3 22.4 2/2 5/5 H. (Paralabidochromis) "orange anal picker" 110.2 33.5 28.0 1.11 28.5 26.0 24.9 3.3/2.9 3.6/4.1 H. (Paralabidochromis) "blue rockpicker" 114.5 36.6 30.1 1.16 30.7 26.4 24.6 2.8/2.2 3.7/4.3 +,++ H. (Paralabidochromis) "sky blue picker" 97.7 34.5 29.3 1.3 35.0 26.5 26.1 2 4/4 +/++

Paralabidochromis "rockkribensis" complex H. (Paralabidochromis) "rockkribensis" 102.0 37.3 32.3 1.32 38.9 23.7 2.5/2.2 3-4/3-5 +,++/++,+++ H. (Paralabidochromis) "short snout scraper" 111.2 37.8 31.9 1.16 34.5 27.2 24.6 2.8/2.6 2.2/2.7 ++,+++ H. (Paralabidochromis) "red short snout scraper" 105.7 41.1 32.0 1.3 36.1 26.2 27.7 3.8/3.3 2.8/3.5 +,++,+++ H. (Paralabidochromis) "elongate short snout scr." 90.4 33.4 31.6 1.6 35.6 26.0 23.8 2.8/2.8 2.5/3 +/+,++ H. (Paralabidochromis) "rock macula" 95.3 37.0 30.8 1.3 34.0 26.9 23.7 2.4/2.4 2.3/2.1 H. (Paralabidochromis) "blue short snout scraper" 125.6 38.1 31.6 1.2 35.1 25.0 27.4 2.6/2.6 3.5/3.8 +,++

292 SL BD HL JR LJL EL IOW OT IR gap

Paralabidochromis chilotes complex H. (Paralabidochromis) chilotes 121.5 37.6 36.8 1.43 41.0 21.2 21.8 5/5 2-3/3-4 H. (Paralabidochromis) "short head chilotes" 127.6 36.8 33.0 1.31 34.7 22.5 21.3 5/4.7 2/3-4 +++

Paralabidochromis chromogynos complex H. (Paralabidochromis) cf. chromogynos 95.8 35.5 34.1 1.27 36.0 26.0 25.4 5/5 5/5 +++ H. (Paralabidochromis) "pointed jaw chromogynos" 90.5 39.2 36.4 1.26 35.3 24.2 22.7 3/3 2/2 ++ H. (Paralabidochromis) "fleshy lips" 87.3 35.6 35.0 1.4 37.4 26.9 20.6 5/5 3/4 ++ H. (Paralabidochromis) "long teeth" 85.0 33.7 35.1 1.53 38.0 27.6 21.2 5/5 3/3.5 +++/++

Psammochromis H. (Psammochromis) riponianus 117.6 35.6 34.0 1.3 37.3 5/5 3.5/3.75 H. (Psammochromis) "rock riponianus" 100.0 37.1 33.1 1.5 37.7 23.9 23.8 4.5/4.5 2/2.8 ++,+++ H. (Psammochromis) cf. saxicola 118.8 36.9 34.1 1.7 39.0 25.9 21.1 4.3/4.3 2.3/2.6 +++ H. (Psammochromis) aelocephalus 124.6 37.2 35.9 1.7 43.9 25.0 22.8 4.3/4.3 2.2/2.2 ++,+++ H. (Psammochromis) "red zebra" 100.7 36.9 32.2 1.49 38.8 26.9 22.9 4.3/4 3/3.5 ++/+++ H. (Psammochromis) "Ruti-Psammo" 89.1 36.5 33.8 1.8 41.4 28.6 21.6 4/4 2.5/3 ++ H. (Psammochromis) "striped crusher" 92.7 34.9 34.3 1.55 37.8 24.5 21.5 4.2/3.7 2/2 ++/+++ H. (Psammochromis) "blue sharp snout" 115.1 43.3 34.3 1.47 40.1 23.3 27.1 5/5 2/3 +++ H. (?) "yellow giant crusher" 132.1 36.7 32.7 1.4 38.6 23.0 23.1 5/5 3.2/3.7 ++/+++

Astatotilapia-like insect eaters Astatotilapia nubila 79.1 39.8 35.7 1.74 41.7 28.5 26.5 3.1/3.3 2.3/3 ++,+++ H. ("Astatotilapia") "incurved dorsal head profile" 100.1 37.2 34.5 1.51 39.0 25.4 25.9 3.8/3.3 3.3/3 ++/+++ H. ("Astatotilapia") "black long snout" 61.3 36.0 35.7 1.95 38.4 27.6 21.2 3/3 2/3 +++ H. ("Astatotilapia") "black cave" 72.8 34.7 33.2 1.82 38.8 31.2 22.9 3.3/3 2/2 ++ H. ("Astatotilapia") "large brownae" 119.3 32.8 31.0 1.39 36.3 24.5 24.0 4.3/4.1 2.3/2.7 ++/+++

Ungrouped insect eaters H. (?) "brown narrow snout" 100.2 35.6 33.1 1.7 39.0 26.8 22.0 4/4 2/3 H. (?) "pale egg dummy" 89.8 36.5 32.2 1.39 39.6 25.1 21.7 H. (?) "orange belly" 90.5 37.5 32.2 1.51 41.2 27.2 24.7 4.8/4.5 3/3.5 ++,+++

Double stripe complex H. (?) thereuterion 69.8 26.0 34.1 2.69 39.6 29.0 19.8 3/3 2/2 +++

Snail crushers H. (Labrochromis) "stone", red & blue 144.9 38.3 33.7 1.39 39.9 24.6 26.7 5/5 2/2.25 ++,+++/+++ H. (?) "Zue crusher" 118.4 37.4 33.3 1.42 37.0 25.7 24.1 5/5 2/2 ++/+++ H. (?) "Ukerewe crusher" 103.1 39.4 32.8 1.28 36.3 24.9 25.5 4.5/4.5 2/2.5 +++

Snail shellers H. (Ptyochromis) xenognathus "rocks" 111.8 36.7 31.5 1.4 34.8 5/5 4.5/5.5 +++ H. (Ptyochromis) "striped rock sheller" 124.4 34.4 33.7 1.12 35.8 26.9 27.8 5/5 4.75/5.5 ++/++,+++ H. (Ptyochromis) "deep water rock sheller" 111.3 39.7 33.0 1.1 35.0 27.2 23.6 5/5 3-4/4-5 +++ H. (Ptyochromis) "red giant sheller" 118.6 38.5 32.2 1.24 38.5 25.2 25.0 5/5 3.5/4.5 +,++/++,+++ H. (Ptyochromis) "Zue sheller" 101.4 32.4 32.4 1.56 37.0 26.4 24.8 4.7/4.5 2.3/3 ++,+++/+++ H. (Ptyochromis?) "red rock sheller" 82.6 34.3 31.1 1.4 37.0 28.6 22.4 3/3.5 2/2 +/+++

Fish eaters H. (Harpagochromis) cavifrons 116.6 32.5 35.5 1.78 45.2 23.2 19.5 5/5 2/3 +++ H. (Harpagochromis) serranus "rocks" 156.0 32.0 36.8 2.06 48.1 19.7 18.9 5/5 1.8/2.8 +++ H. (Harpagochromis) "big blue hunter" 145.7 32.4 35.3 1.78 46.9 23.6 22.1 5/5 2.3/3.5 +++ H. (Harpagochromis) howesi 129.1 32.8 35.3 2.04 46.5 24.6 19.6 5/4.9 2/2 ++,+++/++,+++ H. (Harpagochromis) "orange rock hunter" 109.0 33.2 36.8 1.91 43.5 27.0 17.3 5/5 2/2 ++,+++/+++

Paedophages H. (Lipochromis) cryptodon 103.7 28.8 31.5 1.45 42.8 23.9 23.7 5/4.8 1/1 +++ H. (Lipochromis) "velvet black cryptodon" 125.6 28.4 32.1 1.38 42.1 23.7 24.8 5/5 2 +++ H. (Lipochromis) "blue microdon" 73.1 37.7 33.5 1.52 42.9 29.6 22.2 4.8/4.5 2/2 ++/+++ H. (Lipochromis) cf. melanopterus 96.8 35.3 32.9 1.13 40.5 29.6 26.9 5/4.9 1.5/1.5 ++/+++ H. (Lipochromis) "matumbi hunter" 85.1 28.3 31.0 1.3 41.4 25.6 26.5 5/5 2/2 ++ H. (?) "nyererei paedophage" 94.3 38.9 31.7 1.29 40.8 28.3 28.1 5/5 2/2 ++

293 SL BD HL JR LJL EL IOW OT IR gap

Algae scrapers of the Haplochromis lineage H. (Haplochromis) "blue obliquidens" 106.6 36.9 32.3 1.40 39.1 27.6 30.0 4°/3.9° 3.5/3.6 -,+,++/++,+++ H. (Haplochromis) "purple yellow" 36.4 33.4 1.50 36.5 28.2 29.7 2.8°/3° 3.2/4.3 +,++ H. (Haplochromis) "red back scraper" 36.0 36.5 1.48 40.0 30.7 27.1 3°/3° 2.8/3 ++/++,+++ H. (Haplochromis) "orange chest silvery scraper" 80.0 38.8 35.0 1.39 41.8 3/3 3/3 ++/+++

Table 4: Taxonomic measurements of described rock-dwelling haplochromines

Morphometric and dentition data of described haplochromine species that occur frequently or exclusively at rocky shores. All data are taken from Greenwood´s descriptions and re-descriptions (all references in the text), except those for H. (?) nyererei, H. (Harpagochromis) howesi, and H. (?) thereuterion, which are taken from Witte-Maas & Witte 1985, van Oijen 1992, and van Oijen & Witte 1996 respectively. If mean values were not available, either modal values (if available), or ranges are given. In none of the original descriptions, except those of H. (?) nyererei and H. (?) thereuterion, were sexes measured separately. For H. (?) nyererei and H. (?) thereuterion the values of males are given, to enhance comparison with the data in table X. For H. (H.) howesi only the values for fishes above 110mm SL are given, to enhance comparison with the data in table 3. For abbreviations and further explanations see table 3.

SL BD HL JR LJL EL IOW OT IR gap

Vertical bar Mbipi H. (Neochromis) nigricans 94 36.9 31.2 1.2 35.6 30.0 28.8 2,3 3-7 -,+ H. (Xystichromis) nuchisquamulatus 113 37.9 31.8 1.5 37.6 29.6 29.3 3 3-6/4-8 + H. (?) nyererei 86 39.9 33.4 1.45 37.5 29.2 5.2 3 2-4/3-5 ++

Chessboard Mbipi H. (Paralabidochromis) victoriae 76 33.0 31.6 - - 29.2 25 5/5 4/3 H. (P.) chilotes "lobed lips" 148 32.5-40.8 35.5 - 39.6 25.4 23.8 4-5 2-3(1,4) H. (P.) chilotes "non-lobed lips" 32.7 - 33.2 - - - - - H. (P.) chromogynos 110 35.0 33.2 1.3 32.5 28.6 27.5 4-5 3(2,4) -

Psammochromis H. (Psammochromis) riponianus 104 35.7 35.7 1.6-8 38.5 26.6 25.1 3-5 (1)2-4/(2)3-4 H. (Psammochromis) saxicola 123 37.8 37.8 1.8 43.0 24.4 26.9 (4-)5 2-3(4,5) H. (Psammochromis) aelocephalus 120 31.3-38.4 33.0-38.6 1.8-2.0 41.2-45.1 24.5-27.8 24.8 4-5 (2)3-5/3-6 Astatotilapia-like species Astatotilapia nubila 86 36.8 34.2 39.2 28.4 25.2 3 2-3/2-5 ++ H. ("Astatotilapia") brownae 104 35.1 31.6 2.0 40.3 28.6 29.8 3(4,5) 1-2/2-3(4) H. ("Astatotilapia") barbarae 106 34.0 33.5 1.5-7 38.0 25.5 25.8 3 2(3)/2-3

Double stripe complex H. (?) thereuterion 82 26.5 33.5 2.6 41.2 27.8 19.5 3,4,5 2/2 +++

Snail eaters H. (Ptyochromis) sauvagei 105 35.6 31.9 1.3 34.5 28.9 27.0 (4,)5 2-6/3-8 ++ H. (Ptyochromis) xenognathus 113 34.8 33.1 1.4 34.5 26.0 26.8 5 3-9 Astatoreochromis alluaudi 163 33.8-43.3 35.0 1.5-6 40.0 22.1 28.3 4-5 1-2

294 SL BD HL JR LJL EL IOW OT IR gap

Fish eaters H. (Harpagochromis) cavifrons 195 35.6 37.3 2.0 55.5 20.3 25.4 5 1-2(3)/(1)2-4 H. (Harpagochromis) serranus "rocks" 205 36.0 36.3 2.0 54.3 23.3 23.3 5(4) 2(1,3)/2-3(4) H. (Harpagochromis) howesi 163.3 34.8 36.4 2.4 49.7 22.5 21.2 5 2(3)/2(1,3)

Paedophages H. (Lipochromis) cryptodon 130 27.5-35.6 30.3-34.9 1.5-6 42.3 25.8 23.6 5(3,4) 1-2 H. (Lipochromis) melanopterus 127 35.5 33.5 1.3 37.6 28.2 27.0 5 2 H. (Lipochromis) maxillaris 160 32.0-42.8 30.0-34.8 1.5-8 50.0 27.3-33.2 26.5 5 1(2) + H. (Lipochromis) obesus 170 33.6-47.3 30.3-35.9 1.3 40.0-54.5 27.6 32.2 3,4,5 1(2) +

Algae scrapers of the Haplochromis lineage H. (Haplochromis) obliquidens 89 37.5 32.3 1.4 37.2 31.4 31.8 4°,5° 2-4 ++ H. (Haplochromis) lividus 90 36.5 32.7 1.6 37.2 31.4 29.7 3° 2-4/2-5 ++/+-

295 References

Akiyama, T., A.A. Kajumulo and S. Olsen. 1977. Sea- Mus. Genova 45: 64-78. sonal variations of plankton and physicochemical Bouton, N., O. Seehausen, J.J.M. van Alphen and condition in Mwanza Gulf, Lake Victoria. Bull. F. Witte. submitted. Trophic niche segregation Fresh. Fish. Res. Lab. 27: 49-60. among rock-dwelling haplochromines (Pisces: Baerends, G.P. and J.M. Baerends-van Roon. 1950. Cichlidae) from Lake Victoria. submitted to An introduction to the study of ethology of cichlid Oecologia. fishes. Behaviour (Supp.) 1: 1-243. Bwathondi, P.O.J. and G.C. Mahika. 1996. Distribu- Bailey, M. 1993. The undergravel filter. In: Konings, tion and infestation density of the weed Eichhornia A. (ed). Enjoying Cichlids. Cichlid Press, St. Leon crassipes on the Tanzanian shores of Lake Victo- Rot: 31-33 ria. Bull. Tanz. Fish. Res. Inst. in press. Balirwa, J.S. 1984. Ecological Separation in Barbus Cadwalladr, D.A. 1965. Notes on the breeding biol- species in Lake Victoria. M.Sc. thesis, University ogy and ecology of Labeo victorianus Boulenger of Dar es Salaam, 155 pages. (Pisces: Cyprinidae) of Lake Victoria. Rev. Zool. Barel, C.D.N., M.J.P. van Oijen, F. Witte and E.L.M. Bot. afr. 72: 109-134. Witte-Maas. 1977. An introduction to the taxonomy Cooke, H.B.S. 1958. Observations relating to and morphology of the haplochromine cichlidae quarternary environments in East and South Africa. from Lake Victoria. Neth. J. Zool. 27: 333-389. Trans. Proc. geol. Soc. S. Afr. 60, Annexure No. 5: Barel, C.D.N. 1983. Towards a constructional mor- 1-73 phology of cichlid fishes (teleostei, perciformes). DeMason, L. 1996a. What´s new across the world. Neth. J. Zool. 33: 357-424. Cichlid News 5 (1): 28. Barel, C.D.N., W. Ligtvoet, T. Goldschmidt, F. Witte DeMason, L. 1996b. Cover photo. Cichlid News 5 and P.C. Goudswaard. 1991. The haplochromine (2): 1. cichlids in Lake Victoria: an assessment of bio- Dominey, W.J. 1984. Effects of sexual selection and logical and fisheries iinterests. In: Keeleyside, life history on speciation: species flocks in Afri- M.H.A. (ed). Cichlid Fishes, Behaviour, Ecology, can cichlids and Hawaiian Drosophila. In: Echelle, and Evolution. Chapman and Hall: 258-279. A.A. and I. Kornfield (eds). Evolution of Fish Spe- Bentzen, P. and J.D. McPhail. 1984. Ecology and evo- cies Flocks. University of Maine at Orono Press, lution of sympatric sticklebacks (Gasterosteus): Orono, Maine: 231-249. specialization for alternative trophic niches in the Dorit, R.L. 1990. The Correlates of High Diversity in Enos Lake species pair. Can. J. Zool. 62: 2280-2286. Lake Victoria Haplochromine Cichlids: a Bentzen, P., M.S. Ridgway and J.D. McPhail. 1984. Neontological Perspective. In: Ross, R.M. & W.D. Ecology and evolution of sympatric sticklebacks Ollmon (eds). Causes of Evolution. A (Gasterosteus): spatial segregation and seasonal paleontological perspective. University of Cicago habitat shifts in the Enos Lake species pair. Can. Press, Chicago, London: 322-353. J. Zool. 62: 2436-2439. Eccles, D.H. and E. Trewavas. 1989. Malawian Bleher, H. in press. The Fresh- and Brackish Water cichlid fishes. The classification of some Haplo- Fishes of the World. chromine genera. Lake Fish Movies, Herten (W. Boehme, A. 1996. Erfahrungen bei der Pflege von Germany), 334 pages. Viktoriasee-Buntbarschen. DCG-Info (German Fish, G.R. 1955. The food of Tilapia in East Africa. Cichlid Association) 27: 8-12. Uganda Journal 19: 85-89. Boulenger, G.A. 1906. Descriptions of new fishes Fohrman, K. 1993a. Filtration. In: Konings, A. (ed). discovered by Mr. E. Degen in Lake Victoria. Ann. Enjoying Cichlids. Cichlid Press, St. Leon Rot: 25- Mag. nat. Hist. 17: 433-452. 30. Boulenger, G.A. 1911. On a third collection of fishes Fohrman, K. 1993b. Food and feeding. In: Konings, made by Dr. E. Bayon in Uganda, 1909-1910. Ann. A. (ed). Enjoying Cichlids. Cichlid Press, St. Leon

296 Rot: 34-37. cichlid fishes in Lake Victoria. Bull. Br. Mus. nat. Forsberg, L. 1993. The water. In: Konings, A. (ed). Hist. (Zool.) 3: 295-333. Enjoying Cichlids. Cichlid Press, St. Leon Rot: 19- Greenwood, P.H. 1957. A revision of the Lake Victo- 24. ria Haplochromis species (Pisces, Cichlidae). Part Fryer, G. 1961. Observations on the biology of the II. Bull. Br. Mus. nat. Hist. (Zool.) 5: 73-97. cichlid fish Tilapia variabilis Boulenger in north- Greenwood, P.H. 1959a. A revision of the Lake Vic- ern waters of Lake Victoria (East Africa). Revue toria Haplochromis species (Pisces, Cichlidae). Part Zool. Bot. afr. 64: 1-33. III. Bull. Br. Mus. nat. Hist. (Zool.) 5: 179-218. Fryer, G. and T.D. Iles. 1972. The cichlid fishes of Greenwood, P.H. 1959b. The monotypic genera of the Great Lakes of Africa: their biology and evolu- cichlid fishes in Lake Victoria. Part II. Bull. Br. tion. Oliver and Boyd, London, 641 pages. Mus. nat. Hist. (Zool.) 5: 163-177. Goldschmidt, T. and K. Goudswaard. 1989. Repro- Greenwood, P.H. 1960. A revision of the Lake Victo- ductive strategies in haplochromines (Pisces, ria Haplochromis species (Pisces, Cichlidae). Part Cichlidae) from Lake Victoria. In: Goldschmidt, T. IV. Bull. Br. Mus. nat. Hist. (Zool.) 6: 227-281. An ecological and morphological fieldstudy on the Greenwood, P.H. 1965. The cichlid fishes of Lake haplochromine cichlid fishes (Pisces, Cichlidae) of Nabugabo, Uganda. Bull. Br. Mus. nat. Hist. (Zool.) Lake Victoria. Ph.D. thesis, University of Leiden, 12: 315-357. Netherlands: 93-119. Greenwood, P.H. 1966. Two new species of Haplo- Goldschmidt, T. and F. Witte. 1990. Reproductive chromis (Pisces, Cichlidae) from Lake Victoria. strategies of zooplanktivorous haplochromine Ann. Mag. nat. Hist. 8: 303-318. cichlids (Pisces) from Lake Victoria before the Nile Greenwood, P.H. 1967. A revision of the Lake Victo- perch boom. Oikos 58: 356-368. ria Haplochromis species (Pisces, Cichlidae). Part Goldschmidt, T. and F. Witte. 1992. Explosive VI. Bull. Br. Mus. nat. Hist. (Zool.) 15: 29-119. speciation and adaptive radiation of haplochromine Greenwood, P.H. 1973. A revision of the Haplochro- cichlids from Lake Victoria: an illustration of the mis and related species (Pisces, Cichlidae) from scientific value of a lost species flock. Mitt. Lake George, Uganda. Bull. Br. Mus. nat. Hist. Internat. Verein. Limnol. 23: 101-107. (Zool.) 25: 139-242. Goldschmidt, T., F. Witte and J. de Visser. 1990. Greenwood, P.H. 1974. The cichlid fishes of Lake Ecological segregation in zooplanktivorous haplo- Victoria, East Africa: the biology and evolution of chromine species (Pisces, Cichlidae) from Lake a species flock. Bull. Br. Mus. nat. Hist. (Zool.) Victoria. Oikos 58: 343-355. Suppl 6: 1-134. Goldschmidt, T., F. Witte and J. Wanink. 1993. Cas- Greenwood, P.H. 1979. Towards a phyletic classifi- cading effects of the introduced Nile perch on the cation of the 'genus' Haplochromis (Pisces, Cichli- detritivorous/phytoplanktivorous species in the dae) and related taxa. Part I. Bull. Br. Mus. nat. sublittoral areas of Lake Victoria. Cons. Biol. 7: Hist. (Zool.) 35: 265-322. 686-700. Greenwood, P.H. 1980. Towards a phyletic classifi- Gophen, M., P.B.O. Ochumba and L.S. Kaufman. 1995. cation of the 'genus' Haplochromis (Pisces, Cichli- Some aspecte of perturbation in the structure and dae) and related taxa. Part II: the species from biodiversity of the ecosystem of Lake Victoria (East Lakes Victoria, Nabugabo, Edward, George and Africa). Aquat. Living Resour. 8: 27-41. Kivu. Bull. Br. Mus. nat. Hist. (Zool.) 39: 1-101. Goudswaard, K. and F. Witte. in press. The catfish Greenwood, P.H. 1981. Species flocks and explosive fauna of Lake Victoria after the Nile perch upsurge. evolution. In: Greenwood, P.H. and P.L. Forey (Eds). Env. Biol. Fish. Chance, Change and Challenge - The evolving bio- Graham, M. 1929. The Victoria Nyanza and its Fish- sphere. Cambridge University Press & British Mu- eries. A report on the fishing surveys of Lake Vic- seum (Nat. Hist.), London: 61-74 toria. London: Crown Agents for the Colonies, 256 Greenwood, P.H. 1983. On Macropleurodus, Chilotila- pages. pia (Teleostei, Cichlidae), and the interrelationships Greenwood, P.H. 1956a. A revision of the Lake Vic- of African cichlid species flocks. Bull. Br. Mus. toria Haplochromis species (Pisces, Cichlidae). Part nat. Hist. (Zool.) 45: 209-231. I. Bull. Br. Mus. nat. Hist. (Zool.) 4: 223-244. Greenwood, P.H. 1984. African cichlids and evolu- Greenwood, P.H. 1956b. The monotypic genera of tionary theories. In: Echelle, A.A. and I. Kornfield

297 (Eds). Evolution of Fish Species Flocks. Univer- ria region. S. Afr. J. Zool. sity of Maine at Orono Press, Orono, Maine: 141- Kaufman, L. and P. Ochumba. 1993. Evolutionary and 154. Conservation Biology of Cichlid Fishes as Revealed Greenwood, P.H. 1994. The species flock of cichlid by Faunal Remnants in Northern Lake Victoria. fishes in Lake Victoria - and those of other Afri- Cons. Biol. 7: 719-730. can Great Lakes. Arch. Hydrobiol. Beih. Ergebn. Kaufman, L. and O. Seehausen. 1995. Cichliden von Limnol. 44: 347-354. der ugandischen und der kenianischen Greenwood, P.H. and C.D.N. Barel. 1978. A revision Viktoriaseeküste. In: Stawikowski, R. (ed). of the Lake Victoria Haplochromis species (Pisces, Cichliden. Festschrift zum 25jährigen Jubiläum der Cichlidae). Part VIII. Bull. Br. Mus. nat. Hist. DCG: 147-151. (Zool.) 33: 141-192. Kirner, S. 1994. Neu importiert: Cichliden aus dem Harrison, K. 1991. The taxonomy of East African Nile Viktoriasee. D. Aqu. u. Terr. Z. 47: 347-348. perch, Lates spp. (Perciformes, Centropomidae). J. Konings, A. 1988. Tanganyika Cichlids. Lake Fish Fish. Biol. 38: 175-186. Movies & Verduijn Cichlids, Herten (Germany) & Hecky, R.E. 1993. The eutrophication of Lake Victo- Zevenhuizen (Netherlands), 272 pages. ria. Verh. Int. Ver. Limnol. 25: 39-48. Konings, A. 1995. Malawi Cichlids in their Natural Hecky, R., F.W.B. Bugenyi, P.B.O. Ochumba, J.F. Environment. Cichlid Press, St. Leon Rot (Germany), Talling, R. Mugidde, M. Gophen, L.S. Kaufman. 352 pages. 1994. Deoxygenation of the deep water of Lake Vic- Kudhongania, A.W. & A.J. Cordone. 1974a. Batho- toria, East Africa. Limnol. Oceanogr. 1476-1481. spatial distribution patterns and biomass estimates HEST-Haplochromis Ecology Survey Team. 1989. Fish of the major demersal fishes in Lake Victoria. Afr. Stocks and Fisheries in Lake Victoria. A Handbook J. Trop. Hydrobiol. Fish 3: 15-31. to the HEST/TAFIR,I & FAO/DANIDA, regional Lehman, J.T. and D.K. Branstrator. 1993. Effects of seminar, Mwanza, Jan.-Feb. 1989. Report of the nutrients and grazing on the phytoplankton of Lake Haplochromis Ecology Survey Team (HEST) and the Victoria. Verh. Int. Ver. Limnol. 25: 850-855. Tanzanian Fish. Res. Inst. (TAFIRI) 53, Leiden, Neth- Liem, K.F. 1980. Adaptive significance of intra- and erlands. interspecific differences in the feeding repertoires Hilgendorf, F.M. 1988. Fische aus dem Victoria- of cichlid fishes. Am. Zool. 20: 295-314. Nyanza (Ukerewe See) gesammelt von dem Lippitsch, E. 1993. A phyletic study on haplochro- verstorbenen Dr. G.A. Fischer. Sber. Ges. naturf. mine fishes (Perciformes, Cichlidae) of East Af- Freunde Berl.: 75-79. rica, based on scale and squamation characters. Holzberg, S. 1978. A field and laboratory study of J. Fish. Biol. 42: 903-946. the behaviour and ecology of Pseudotropheus ze- Lowe-McConnell 1956. Observations on the biology bra (Boulenger), an endemic cichlid of Lake Ma- of Tilapia (Pisces, Cichlidae) in Lake Victoria, East lawi (Pisces: Cichlidae). Z. Zool. Syst. Evol.-forsch. Africa. EAFFRO Suppl. Publ. 1:1-72. 16: 171-187. Marsh, A.C., A.J. Ribbink and B.A. Marsh. 1981. Hoogerhoud, R.J.C. 1984. A taxonomic reconsidera- Sibling species complexes in sympatric population of the haplochromine genera Gaurochromis tions of Petrotilapia Trewavas (Cichlidae, Lake Ma- Greenwood, 1980 and Labrochromis Regan, 1920 lawi). Zool J. Linn. Soc. 71: 253-264. (Pisces, Cichlidae). Neth. J. Zool. 34: 539-565. Maynard-Smith, J. 1966. Sympatric speciation. Hoogerhoud, R.J.C., F. Witte and C.D.N. Barel. 1983. American Naturalist 100: 637-650. The ecological differentiation of two closely resem- Mayr, E. 1942. Systematics and the Origin of Spe- bling Haplochromis species from Lake Victoria (H. cies. Columbia University Press, New York. iris and H. hiatus; Pisces, Cichlidae). Neth. J. Zool. McKaye, K.R. and W.M. Gray. 1984. Extrinsic barri- 33: 283-305. ers to gene flow in rock-dwelling cichlids of Lake Katunzi, E.F.B. 1980. Food preferences of four haplo- Malawi: macrohabitat heterogeneity and reef colo- chromine cichlids from Mwanza Gulf in Lake Vic- nization. In: Echelle, A.A. and I. Kornfield (Eds). toria. M.Sc. thesis, University of Dar es Salaam, Evolution of Fish Species Flocks. University of 132 pages. Maine at Orono Press, Orono, Maine: 169.183. Kaufman, L. in press. Asynchronous taxon cycles Meyer, A. 1993. Phylogenetic relationships and evo- in haplochromine fishes of the greater Lake Victo- lutionary processes in African cichlids. Trends Ecol.

298 Evol. 8: 279-284. An introduction to ecological and taxonomic inves- Meyer, A., T.D. Kocher, P. Basasibwaki & A.C. tigations on the haplochromine cichlids from the Wilson. 1990. Monophyletic origin of Lake Victo- Mwanza Gulf of Lake Victoria. Neth. J. Zool. 31: ria cichlid fishes suggested by mitochondrial DNA 149-174. sequences. Nature 347: 550-553. Oijen, M.J.P. and F. Witte. 1996. Taxonomical and Meyer, A., C. Montero and A. Spreinat. Evolution- ecological description of a species complex of ary history of the cichlid fish species flocks of zooplanktivorous and insectivorous cichlids from the East African great lakes inferred from molecu- lake Victoria. Zool. Verh. Leiden 302: 1-56. lar phylogenetic data. Arch. Hydrobiol. Beih. Owen, R.B., R. Crossley, T.C. Johnson, D. Tweddle, Ergebn. Limnol. 44: 407-423. I. Kornfield, S. Davison, D.H. Eccles and D.E. Mugidde, R. 1993. The increase in phytoplankton pri- Engstrom. 1990. Major low levels of Lake Malawi mary productivity and biomass in Lake Victoria and their implications for speciation rates in (Uganda). Verh. Int. Ver. Limnol. 25: 846-849. cichlid fishes. Proc. R. Soc. Lond. B 240: 519-553. Ochumba, P.B.O. 1987. Periodic massive fish kills Paterson, H.E.H. 1978. More evidence against in the Kenyan part of Lake Victoria. Wat. Qual. speciation by reinforcement. S. Afr. J. Sci. 74: 369- Bull. 12: 119-122. 371. Ochumba, P.B.O. and D.I. Kibaara. 1989. Observa- Piper, B.S., D.T. Plinston and J.V. Sutcliffe. 1986. tions on blue-green algal blooms in the open wa- The water balance of Lake Victoria. Hydrological ters of Lake Victoria, Kenya. Afr. J. Ecol. 27: 23- Sciences Journal 31: 25-38. 24. Raymond, M. 1995. Victorian Secrets. Trop. Fish. Oguto-Ohwayo, R. 1990. The decline of the native Hobb. 43 (8): 16-28. fishes of lakes Victoria and Kyoga (East Africa) Regan, C.T. 1922. The cichlid fishes of Lake Victo- and the impact of introduced species, especially ria. Proc. Zool. Soc. Lond. (1922) 157-191. the Nile perch, Lates niloticus, and the Nile Ribbink, A.J. 1991. Distribution and ecology of the tilapia, Oreochromis niloticus. Env. Biol. Fish. 27: cichlids of the African Great Lakes. In: Keeleyside, 81-96. M.H.A. (ed). Cichlid Fishes, Behaviour, Ecology, Oijen, M.J.P. van. 1978. The live coloration of cer- and Evolution. Chapman and Hall: 36-59 tain previously described Haplochromis species. Ribbink, A.J. and D.H. Eccles. 1988. Fish commu- Appendix to: Greenwood, P.H. and C.D.N. Barel. nities in the East African Great Lakes. In: Léveque, A revision of the Lake Victoria Haplochromis spe- C., M.N. Bruton and G.W. Sentongo (eds). Biology cies (Pisces, Cichlidae). Part VIII. Bull. Br. Mus. and ecology of African freshwater fishes. Orstom, nat. Hist. (Zool.) 33: 141-192. Paris: 277-300. Oijen, M.J.P. van. 1982. Ecological differentiation Ribbink, A.J., B.A. Marsh, A.C. Marsh, A.C. Ribbink among the haplochromine piscivorous species of and B.J. Sharp. 1983. A preliminary survey of the Lake Victoria. Neth. J. Zool. 32: 336-363. cichlid fishes of rocky habitats in Lake Malawi. Oijen, M.J.P. van. 1989. Notes on the relationship S. Afr. J. Zool. 18: 149-310. between prey size and pharyngeal jaw size in pis- Sage, R.D., P.V. Loiselle, P. Basasibwaki and A.C. civorous haplochromine cichlids from Lake Victo- Wilson. 1984. Molecular versus morphological ria. Ann. Mus. Roy. Afr. Centr. Sc. Zool. 257: 73- change among cichlid fishes of Lake Victoria. In: 76. Echelle, A.A. and I. Kornfield (eds). Evolution of Oijen, M.J.P. van. 1991. A systematic revision of Fish Species Flocks. University of Maine at Orono the piscivorous haplochromine Cichlidae (Pisces: Press, Orono, Maine: 185-201. Teleostei) of Lake Victoria (East Africa). Part I. Zool. Schliewen, U.K., D. Tautz and S. Paäbo. 1994. Sym- Verh. Leiden 272: 1-95. patric speciation suggested by monophyly of cra- Oijen, M.J.P. van. 1992. Haplochromis howesi spec. ter lake cichlids. Nature 368: 629-330. nov., a crab and fish eating cichlid from Lake Vic- Schraml, E. 1990. Haplochromisverwandte toria. Zool. Med. Leiden 66: 561-597. Buntbarsche aus Ostafrika. D. Aqu. u. Terr. Z. 43: Oijen, M.J.P. van. 1995. The generic classification 718-721. of the haplochromine cichlids of Lake Victoria, East Seehausen, O. 1992. Victorian Cichlids. Part I: In- Africa. Zool. Verh. Leiden 302: 57-110. troduction to taxonomy and ecology. The Cichlids Oijen, M.J.P., F. Witte and E.L.M. Witte-Maas. 1981. Yearbook 2: 54-59.

299 Seehausen, O. 1993a. Foraging and spatial behav- Seehausen, O., F. Witte, E.F. Katunzi, J. Smits and iour of rock dwelling cichlids in Lake Victoria in N. Bouton. in press [b]. Patterns of the remnant the light of competition, community structure, and cichlid fauna in southern Lake Victoria. Conserva- speciation. M.Sc. thesis, Universities of Hanover tion Biology. and Leiden, 234 pages. Seehausen, O., S. Kangwe and A. Samwel-Terry. Seehausen, O. 1993b. Victorian Cichlids. Part II: The 1996. Wasserhyazinthe und Fischerei bedrohen oral shelling/crushing molluscivores. The Cichlids Viktoriasee-Cichliden. D. Aqu. u. Terr. Z. 49: 112- Yearbook 3: 50-55. 116. Seehausen, O. 1994. Victorian Cichlids. Part III: The Selbrink, B. 1985a. Viktoriasjöns ciklider, del 3: epilithic algae scrapers. The Cichlids Yearbook 4: Psammochromis riponianus. Akvariet 59: 356-357. 42-51. Selbrink, B. 1985b. Viktoriasjöns ciklider, del 2: Seehausen, O. 1995a. The "Mbuna" of Lake Victo- Silverpilen - Prognathochromis spec. aff. ria. Trop. Fish. Hobb. 43 (8): 212-226. longirostris. Akvariet 59: 340-341. Seehausen, O. 1995b. Die Felscichliden des Selbrink, B. 1986. Viktoriasjöns ciklider, del 6: Viktoriasees. D. Aqu. u. Terr. Z. 48: 190-196. Neochromis nigricans - en "mbuna" fran Seehausen, O. 1996a. A reconsideration of the eco- Viktoriasjön. Akvariet 60: 44-45. logical composition of the cichlid species flock in Skúlason, S., D.L.G. Noakes and S.S. Snorrason. Lake Victoria before and after the Nile perch boom. 1989. Ontogeny of trophic morphology in four sym- In: van Densen, W. & R. Lowe-McConnell (eds). patric morphs of Arctic charr Salvelinus alpinus Lacustrine Fish Communities in SE-Asia and Af- in Thingvallavatn, Iceland. Biol. J. Linn. Soc. 38: rica. Samara Publishing Ltd., Cardigan, Dyfed, 281-301. Great Britain. Snoeks , J. 1988. Redescription d´Haplochromis Seehausen, O. 1996b. Distribution of, and reproduc- paucidens Regan, 1921 et description tive isolation among colour morphs of a rock-dwell- d´Haplochromis occultidens sp.n. (Pisces, Cichli- ing Lake Victoria cichlid, Haplochromis nyererei. dae) du lac Kivu en Afrique. Cybium 12: 203-218. Ecol. Freshw. Fish. 5: Snoeks, J. 1994. The Haplochromines (Teleostei, Seehausen, O. 1996c. A dynamic cichlid fauna as a Cichlidae) of Lake Kivu (East Africa). A taxonomic complex indicator of environmental degradation in revision with notes on their ecology. Ann. Mus. Lake Victoria. Bull. Tanz. Fish. Res. Inst. 2: in press. Roy. Afr. Centr., Sc. Zool. 270: 1-221. Seehausen, O. 1996d. Fortpflanzungsverhalten Stager, J.C., P.N. Reinthal and D.A. Livingstone. felsbewohnender Viktoriaseecichliden in ihrem 1986. A 25,000-year history for Lake Victoria, East natürlichen Lebensraum. D. Aqu. u. Terr. Z. 49: 666- Africa, and some comments on its significance for 669. the evolution of cichlid fishes. Freshwat. Biol. 16: Seehausen, O. and N. Bouton. 1996a. The commu- 15-19 nity of rock-dwelling cichlids in Lake Victoria. Stauffer, J.R. and K.R. McKaye. 1986. Description Bonner Zoologische Beitraege: in press. of a paedophagous deep-water cichlid (Teleostei: Seehausen, O. and N. Bouton. 1996b. Polychroma- Cichlidae) from Lake Malawi, Africa. Proc. Biol. tism in rock dwelling Lake Victoria cichlids: types, Soc. Wash. 99: 29-33. distribution, and observations on their genetics. The Sturmbauer, C. and A. Meyer. 1993. Mitochondrial Cichlids Yearbook 6: 36-46. phylogeny of the endemic mouthbrooding lineages Seehausen, O. and N. Bouton. in press. of cichlid fishes from Lake Tanganyika in eastern Microdistribution and temporal niche shifts in a Africa. Mol. Biol. Evol. 10: 751-768. rocky shore cichlid community in Lake Victoria. Talling, J.F. 1957. Some observations on the strati- Ecol. Freshw. Fish. fication of Lake Victoria. Limnol. Oceanogr. 2: 213- Seehausen, O. and F. Witte. 1995. Extinction of many, 221. survival of some. Trop. Fish. Hobb. 43 (7): 96-105. Talling , J.F. 1966. The annual cycle of stratification Seehausen, O., N. Bouton, J.J.M. van Alphen and and phytoplankton growth in Lake Victoria (East F. Witte. in press [a]. Feeding techniques, behav- Africa). Int. Revue ges. Hydrobiol. 51: 545-621. ioural versatility and quantitative specialization of Taylor, E.B. and P. Bentzen. 1993. Evidence for mul- rock-dwelling haplochromine cichlids from Lake tiple origins and sympatric divergence of trophic Victoria. Oikos. ecotypes of smelt (Osmerus) in northeastern North

300 America. Evolution 47: 813-832. ogy and fishery of haplochromine trophic groups. Taylor, E.B. and J.D. McPhail. 1994. Molecular tests Zool. Verh. 262: 1-47. of speciation scenarios in sympatric sticklebacks, Witte, F. and E.L.M. Witte-Maas 1987. Implications Gasterosteus. Second International Symposium on for taxonomy and functional morphology of intra- Stickleback Behaviour, Leiden/Netherlands, 28.08.- specific variation in haplochromine cichlids of Lake 01.09.1994. Abstract. Victoria. In: Witte, F. From Form to Fishery. Ph.D. Temple, P.H. 1967. Causes of intermittent decline in thesis, Leiden University, chapter II. the level of Lake Victoria during the Late Witte-Maas, E.L.M. 1981. Eggsnatching: an obser- Pleistocene and Holocene. In: Steel, R.W. and R. vation on the feeding behaviour of Haplochromis Lawton (eds). Liverpool essays in geography: 43- barbarae Greenwood, 1967 (Pisces, Cichlidae). 63. Neth. J. Zool. 31: 786-789. Trewavas, E. 1983. The tilapiine fishes of the gen- Witte-Maas, E.L.M. and F. Witte. 1985. Haplochro- era Sarotherodon, Oreochromis and Danakilia. Brit- mis nyererei, a new cichlid fish from Lake Victo- ish Museum (Natural History) London, 583 pages. ria named in honour of Mwalimu Julius Nyerere, Turner, G.F. and M.T. Burrows. 1995. A model of sym- president of Tanzania. Brill Publishers, Leiden, 13 patric speciation by sexual selection. Proc. R. Soc. pages. Lond. B. 260: 287-292. Wanink, J.H. 1991. Survival in a perturbated environment: the effects of Nile perch introduction on the zooplanktivorous fish community in Lake Vic- toria. In: Ravera, O. (ed). Terrestrial and aquatic ecosystems: perturbation and recovery. Ellis Horwood Ltd., Chichester. Wilhelm, W. 1980. The disputed feeding behaviour of a paedophagous haplochromine cichlid (Pisces) observed and discussed. Behaviour 74: 310-323. Wilson, E.O. 1994. The Diversity of Life. Penguin Books Ltd., 406 pages. Witte, F. 1984. Ecological differentiation in Lake Vic- toria haplochromines: comparison of cichlid species flocks in African lakes. In: Echelle, A.A. and I. Kornfield (eds). Evolution of Fish Species Flocks. University of Maine at Orono Press, Orono, Maine: 155-167. Witte, F. & W.L.T. van Densen (eds). Fish Stocks and Fisheries of Lake Victoria. A handbook for field observations. Samara Publishing Ltd., Cardigan, Dyfed, Great Britain. Witte, F., T. Goldschmidt, J. Wanink, M. van Oijen, K. Goudswaard, E. Witte-Maas and N. Bouton. 1992. The destruction of an endemic species flock: quantitative data on the decline of the haplochromine cichlids of Lake Victoria. Env. Biol. Fish. 34: 1-28. Witte, F., T. Goldschmidt and J. Wanink. 1995. Dy- namics of the haplochromine cichlid fauna and other ecological changes in the Mwanza Gulf of Lake Victoria. In: Pitcher, T.J. and P.J.B. Hart (eds). The Impact of Species Changes in African Lakes. Chapman and Hall, London: 83-110. Witte, F. and M.J.P. van Oijen. 1990. Taxonomy, ecol-

301 Species Index

The index lists all cichlid species that are dealt - "all red nyererei" 123, 124, 284 with in the text and that are shown in photo- - "big blue" (blue) 105, 108, 283 graphs. Vertical bar and Chessboard Mbipi are - "big blue" (red) 106, 108, 283 sorted alphabetically within the species com- - "black & orange nyererei" 114, 113, 284 plexes. All other species are sorted alpha- - "blue nyererei" 118,120, 284 betically, their (sub)generic assignment fol-lowing - "Bwiru nyererei" 123, 124, 284 the species name. Page numbers printed in - "crirnson tide" 123, 44, 125 boldface refer to illustrations. Page numbers in - "lemon fm nyererei" 116,117, 284 italics refer to distribution maps. - nyererei 110, cover, 28, 1l2, 1l3, 283 - "orange anal nyererei" 115, 1l6, 284 "Deepwater" complex - "orange dorsal nyererei" 118, 1l7, 283 - " blue deepwater" 127, 129, 284 - "pink anal" 119,121, 284 - "deepwater" 126, 128, 284 - "red anal nyererei" 122, 124, 284 - "slender deepwater" 129, 129, 284 - "red chest" 105, 283 - "yellow deepwater" 127, 129, 284 - "red flank nyererei" 123, 124, 284 - "red head nyererei" 107,109, 283 Neochromis complex - "red rim anal" 122, 124, 284 - "Bihiru scraper" 79, 81, 281 - "small mouth nyererei" 115, 1l6, 284 - "black nigricans" 70, 69, 281 - "Ukerewe nyererei" 115, 1l6, 284 - "blue scraper"74, 73, 76, 77, 281 - "zebra nyererei" 102,104,105, 283 - "black tail giant scraper" 86, 88, 281 - "zebra Senga" 103, 105, 283 - "cross dresser" 90 - "zebra Miandere" 103, 105, 283 - "eastern blue scraper" 78, 80, 281 - "giant scraper" 83, 85, 281 Paralabidochromis lineage - "Jinja big scraper" 90 chilotes complex - "Juma scraper" 75, 77, 281 - chilotes 170, 169, 172, 287 - "kruising" 84 - "short head chilotes" 174,172, 287 - "Labeo scraper" 82, 84, 281 chromogynos complex - "large eye nigricans" 86, 88, 282 - chromogynos 175, 172, 287 - "large scale nigricans" 87 - "fleshy lips" 177,176, 287 - "long black" 79, 80, 281 - "long teeth" 177,176, 287 - nigricans 67, 68, 69, 281 - "pointed jaw chromogynos" 177, 287 - "pseudoblack", Neochromis? 87, 89, 282 "rockkribensis"/plagiodon complex - "red anal nigricans" 90, 92 - "blue short snout scraper" 166, 165, 287 - "red tail giant scraper" 86, 88, 281 - "elongate short snout scraper" 166, 165, 287 - "short head nigricans" 79, 80, 281 - plagiodon 158, 250, 249 - simotes 89 - "red short snout scraper" 163, 164, 165, 287 - "slender scraper" 89, 89 - "rockkribensis" 159, 28, 160, 161, 287 - "unicuspid scraper" 82, 84, 281 - "rockrnacula" 167,168, 287 - "velvet black" 70, 72, 73, 281 - "short snout scraper" 163, 165, 287 - "Vesi scraper" 75, 77, 281 "rockpicker"/victoriae complex - "yellow anal scraper" 78, 80, 281 - "blue rockpicker" 155,156, 157, 286 - "chessboard picker" 150, 152 Nyererei complex - "elongate rockpicker" 155, 156, 286 - "all black nyererei" 107, 109, 283 - "orange anal picker" 154, 156, 286

302 - "pseudorockpicker" 148,148, 286 290 - "red pseudorockpicker" 150, 149, 286 - "blue and orange fins", ? ? 252, 253 - "rockpicker" 154, 156, 286 - "blue insectivore", ? 247, 248 - "sky blue picker" 158, 160, 286 - "blue microdon", Lipochromis 223, 225, 289 - "southern pseudorockpicker" 150, 149, 286 - "blue obliquidens", Haplochromis 231, 232, - victoriae, "rockpicker" complex ? 147 290 - "yellow rockpicker" 151, 152, 286 - "blue sharp snout", Psammochromis 186, 185, - "Zue rockpicker" 153, 153, 286 288 - brownae, "Astatotilapia" 242 "Pseudonigricans" complex - "brown narrow snout, ? 246, 245, 290 - "black pseudonigricans" 130, 132, 285 - cavifrons, Harpagochromis 210, 209, 288 - "black and yellow pseudonigricans" 135, 136, - "citrus", ? 255 285 - cryptodon, Lipochromis 221, 220, 224, 289 - "black Ukerewe" 142, 144, 285 - "deep water rock sheller", Ptyochromis 195, - "blue pseudonigricans" 133, 132, 285 196, 289 - "Gana pseudonigricans" 141, 141, 285 - diplotaenia, ? 227 - "long snout pseudonigricans" 142, 144, 285 - "duck snout", ? 246, 248 - "large eye pseudonigricans" 131, 132, 285 - guiarti, Harpagochromis 207, 208 - "orange pseudonigricans" 139, 137, 285 - howesi, Harpagochromis 211, 212, 288 - "pseudoblue" 142, 144, 285 - "incurved dorsal head profile", - "red pseudonigricans" 134, 133, 285 "Astatotilapia" 239, 240, 290 - "scraper pseudonigricans" 139, 141, 285 - "kribensis", ? 254, 256 - "Ukerewepseudonigricans" 141, 141, 285 - "large brownae", "Astatotilapia" 242, 244, 290 - "yellow-blue" 144 - lividus, Haplochromis 231, 232 - "yellow chin pseudonigricans" 136, 137, 285 - "longurius", ? 255 - "macula", Haplochromis 254, 253 Xystichromis - macrognathus, Prognathochromis 255 - "carp" 98, 97, 100, 281 - "matumbi hunter", Lipochromis 223, 225, 289 - "copper black" 91, 92, 93, 281 - megalops, "Astatotilapia" 255 - "large eye black" 95, 96, 281 - melanopterus, Lipochromis 219, 216, 217, 289 - nuchisquamulatus 91 - microdon, Lipochromis 223 - "orange carp" 98, 100, 281 - nubila, Astatotilapia 238, 240, 290 - "red anal blue" 95, 96, 281 - "nyererei paedophage", ? 226, 228, 289 - "red carp" ("pink giant") 99, 100, 281 - obliquidens, Haplochromis 230 - "short scraper" 95, 96, 281 - "orange belly", ? 246, 248, 290 - "Uganda blue scraper" 95 - "orange chest silvery scraper", Haplochromis 235, 237, 290 other lineages - "orange rock hunter", Harpagochromis 214, - aelocephalus, Psammochromis 185, 184, 185, 212, 288 288 Oreochromis niloticus 258, 261 - alluaudi, Astatoreochromis 255, 257 Oreochromis leucostictus 258, 261 - altigenis, Harpagochromis 255, 257 Oreochromis esculentus 262, 261 - "argens", ? 255, 256 Oreochromis (Nyasalapia) variabilis 259, 260 - barbarae, "Astatotilapia" 243, 244 - "pa1e egg dumrny", ? 246, 248, 290 - bicolor, Macropleurodus 250, 249 - perrieri, Prognathochromis 255 - "big blue hunter", Harpagochromis 210, 209, - piceatus, "Astatotilapia" 255, 257 288 - plagiodon, Paralabidochromis 158, 250, 249 - "big teeth", ? 255 - "purple rocker",? 247 - "black and red fins", Haplochromis 254 - "purple yellow", Haplochromis 234, 233, 290 - "black cave", "Astatotilapia"242, 241, 290 - "red back scraper", Haplochromis 235, 8, - "black long snout", "Astatotilapia" 241, 240, 236, 290

303 - "red giant sheller", Ptyochromis 198, 200, 289 - "red piebald", ? 243, 245 - "red rock sheller", Ptyochromis ? 199, 200, 289 - "red zebra", Psammochromis 187, 188, 288 - riponianus, Psammochromis 182, 180, 288 - "rock riponianus", Psammochromis 183, 184, 288 - "Ruti-Psammo", Psammochromis 186, 185, 288 - sauvagei, Ptyochromis 191, 192, 289 - saxicola, Psammochromis 182, 181, 288 - serranus, Harpagochromis 207, 208, 288 - "shovel mouth", ? 247, 248 - "small blue zebra", ? 251, 252 - "Sozihe crusher", ? 203, 205, 289 - "stone", Labrochromis 202, 204, 289 - "striped crusher", Psammochromis 186, 188, 288 - "striped rock sheller", Ptyochromis 195, 196, 289 - tanaos, ? 227, 229 - thereuterion, ? 227, 228 - "thick skin", ? 251, 252 Tilapia (Coptodon) rendalli 259, 261 Tilapia (Coptodon) zillii 259 - "two stripe yellow green", "Astatotilapia" 255 - "velvet black cryptodon", Lipochromis 222, 224, 289 - "yellow giant", ? 190, 189, 288 - xenognathus, Ptyochromis 194, 193, 289 - "Zue crusher", ? 203, 204, 289 - "Zue sheller", Ptyochromisl98, 200, 289

304 4 2