ECOLOGY OF IN THE AMAZON AND CONGO BASINS

TYSON R. ROBERTS

Abstract. Some relationships between the great a single group, the Ostariophysi. In con- diversity of fishes and physical and biological trast to tropical marine shore fishes, which aspects of the environment in the Amazon and have had only one principal center of radi- Congo basins are discussed. Aspects of physical ation—the Indo-Pacific— in environment considered are rainfall regimes; ostariophysans stabilizing influence of forests; seasonal fluctu- Africa, , and Eurasia have ations in water white le\el; water, clear water, radiated largely in isolation from one an- and black water; relative of Ama- accessibility other. Asia and Africa, of course, share zon and Congo basins to marine fishes; tidal elements. There is conditions in the lower Amazon; shoreline and many ostariophysan islands; river anastomoses and connections with some indication that the South American adjacent basins; stream captures; unique or and older x\frican ostariophysan elements peculiar causes of massi\e mortalitv' of biotypes: had a common origin. The fonnation of fishes. The main adaptive significance of parental the provided the opportunity care in Amazon and Congo fishes is e\identh for a remarkable radiation of related to reproduction in deoxygenated waters. Ostariophysi. The following topics are considered under bio- The Amazon and Congo basins have more logical interactions: fishes with brilliant coloration kinds of fishes than any other river basins or association of similarh' conspicuous markings; in the world, and both exhibit a high de- colored of fishes in mixed schools; fishes gree of endemism. of minute size: responses to predation; and par- One of the reasons titioning of food resources. Aliout one third of principal tropical the paper is devoted to discussion of trophic faunas are richer than temperate ones is adaptations and feeding habits. The topics con- that the\- ha\e not been as adversely af- sidered are predatory fishes; scale-eating chara- fected by glaciation. Pleistocene glaciation coids; fin-eating characoids; feeding habits of probably caused extinction or withdrawal "parasitic" trichomycterid ; parallelism in of marine fishes in lati- feeding habits of "weakl\- electrogenic" fishes in shore the higher the Amazon and Congo basins; plankton-feeding tudes, especially in the North Atlantic and habits of fishes; deposit feeders; feeding (Briggs, 1970). It is unlikely that fi.shes characins and other fishes in Amazonian rain- live more than a few miles inward imder forest streams. the ice caps. Freshwater fishes in most of North America and northern Eurasia were INTRODUCTION wiped out by Pleistocene glaciation. West- An overwhelming proportion of the spe- ern Europe now has a depauperate fish cies of fishes in continental fresh waters fauna of only about 60 species, all or al- are primarv' freshwater forms. This means most all derived from stocks that popu- that they are unable to live in salt water, lated the area in postglacial times. and have had a long history separate from It may be well to express here my suspi- that of marine fishes. The great majorit\' cion that the richness of the Congo and of these freshwater fishes are members of Amazon fish faunas is not necessarilv an-

Bull. Mus. Comp. Zoo)., 143(2) : 117-147, February, 1972 117 118 Bulletin Museum of Comparative Zoology, Vol. 143, No. 2

cient. Fishes have undergone considerable studied Mississippi, but the numbers diversification, with great increase in the known from the Amazon and Congo will number of species, in lakes less than five undoubtedly increase considerably as syste- million years old. I do not believe that matic studies continue. lakes represent the only ecological situation The limnology, seasonal changes and in which explosive adaptive radiations of some biological aspects of the rivers, lagos, fishes have occurred. It is conceivable that flooded forest (igapo) and streams the present Amazonian fish fauna, with its (igarapes) that constitute the main habi- large number of species, is the product of tats of Amazon fishes are described by only a few million years of evolution from Sioli (1964, 1967) and Fittkau (1964, an original stock of two or three hundred 1967). The ecology of fish habitats in the founder species. Some of the founders, of cuvette centrale of the Congo basin, similar course, would represent groups of con- in many respects to that in the Amazon, is siderable antiquity. described by Gosse (1963) and Matthes Draining two and a quarter million (1964). McConnell (1964) described the square miles, the Amazon basin is the ecological groupings of fishes and effects largest river basin in the world. Its mouth of the seasonal cycle on the fishes in the discharges an average of three to four Rupununi savanna of British Guiana. Many million cubic feet of water per second. The of her remarks are applicable to the Ama- Congo, discharging 1.4 X 10" cubic feet zon basin, especially to the savanna parts of it. if in per second from sliglitly over one and a Most, not all, of the species the half million square miles, is the second Rupununi are present in the Amazon. Mc- Connell 1969 reviewed factors con- largest. The vast area of these basins, with ( ) some abundant water and varied habitats, un- tributing to speciation in tropical fresh- water of doubtedly contributes to the large number fishes, and many her remarks of fish species in them. Habitats such as apply indirectly or directly to fishes in the and basins. streams with high gradients or streams Amazon Congo Knoppel (1970) in detail on the contents draining dry ground (igarapes de terra reported stomach of a number of fish from finua) are sometimes separated by hun- large species dreds of miles. Meandering creates a small rain forest streams near Manaus. regular succession of habitats in the main Myers (1947, 1949a) gave a general ac- count of Amazonian fishes and their courses of the big rivers. The high per- centage of the basins at base level, along ecology. Tlie main of this is divided with the relatively stable existence of so part paper into two sections. first section deals much aquatic habitat, favors the existence The with the interactions of fishes and of very large numbers of individuals, which physical in turn is conducive to the existence of aspects of environment in the Amazon and the second section with large numbers of species (Preston, 1962). Congo basins, interactions fishes in the As of 1967 (the last year for which the biological among Amazon and basins. The rest of Zoological Record has been issued) ap- Congo this introduction a brief sketch of proximately 1300 species of fishes had been provides recorded from the Amazon and 560 from the main groups of ostariophysans and the Congo (including the Lualaba River other fishes under consideration. Readers but not lakes Bangweolu and Moeru). The familiar with these groups may turn di- Mississippi basin, in comparison, with an rectly to the main part of the paper if they as wish. In drafts of this I tended to area ( 1,244,000 square miles ) almost paper large as that of the Congo basin, has only include extraneous notes about species oc- 250 species. It is unlikely that many species currence, etc., which might distract readers remain unrecorded in the relatively well- interested mainly in the discussion of ecol- Amazon- and Conc.o Fish Ecology • Roberts 119

it seems the ogy and adaptations. These notes, indicated ception, highl\ likely Weberian has increased the by arabic numerals in the text, are given apparatus op- fish at the end ot the paper. They are of con- portunitic\s for interactions between zoo- cern primarilx' to systematists and species. Characoids. Characoids or characins are geographers. mostly laterally-compressed, open-water fishes, active in the da\'time, many of them Ostariophysi—the Predominant or iridescent. Tlie\ usually have Fishes in Both Basins silvery jaw teeth, often of a highly complex nature In the Amazon 43 percent of tlie fishes (Roberts, 1967), and invariably lack bar- 3 are characoids, 39 percent siluroids, and bels. \\'\\\\ few exceptions, they are not 15 percent gymnotoids. In the Congo per- known to produce biologically significant cent are characoids, 23 percent sihu'oids, sounds (almost all fishes produce noises and 16 percent c>prinoids. All of these incidental to feeding and locomotion). Of fishes belong to the order Ostariophysi, the large groups of fishes inhabiting the which thus comprises 85 percent of the earth's fresh waters, characoids (as a Amazon's and 54 percent of the Congo's group) exhibit the least tolerance for salt fish fauna. Ostariophysi differ from all or brackish water. They occur onl\' in other fishes in the manner in wliich some of Africa and Central and South America. the neural arches and ribs of the first four Tlieir presence in Central America un- vertebrae are modified into an apparatus, doubtedly is the result of recent invasion, the Weberian apparatus, which conducts and it is unreasonable to believe that they vibrations from the swim bladder to the could have reached Africa and South inner ear. There is no precise understand- America by w^ay of Europe and North ing of how the Weberian apparatus affects America without leaving the least trace of sound (and pressure?) perception, nor is their passage (Myers, 1966). The only much known about the effects of sound on fossils identified with certainty as characins the behavior of ostariophysans in nature. It are African, or South or Central American the\' is however, that are 1960 . In the of exidence generally agreed, ( Weitzman, ) light "acoustic specialists," and that their world- for Continental Drift and for characoid wide in fresh waters is some- et al., Roberts, predominance antiquity ( Greenwood 1966; how linked with the Weberian apparatus. 1969) it is reasonable to think that chara- Experimental work with xarious ostario- cins were present in South America before sensi- physans indicates that their auditory it separated fully from Africa. tivity and range is greater than in many Gymnotoids. Gymnotoids, the so-called other fishes. Of considerable interest is electric eels, all have speciahzed electro- whether they have directional hearing, i. e., genie and electrosensory organs. The can locaUze the source of a sound by means quality of the electric discharges is char- of their auditory apparatus. Most fishes acteristic of each species. The gymnotoid apparently have nondirectional hearing, but that has received the most scientific at- directional hearing evidently occurs in tention, the famous electric eel (Electroph- several groups (see Moulton and Dixon, orus elcctriciis Linnaeus) studied by 1967: 195). The only freshwater fishes Alexander von Humboldt, Faraday, and thought to have directional hearing are others, has a ver\' powerful discharge. The Ostarioph\'si and perhaps monnyroids. discharges of other gymnotoids are too Kleerekoper and Chagnon (1954) and Delco weak for us to sense without the help of (1960) reported directional hearing in instruments and have only recently come to three species of North American minnows. to our attention. We are just beginning \Vhatever the nature of its role in per- appreciate the roles played by gymnotoid 120 Bulletin Museum of Comparative Zoology, Vol. 143, No. 2

electric organs in object location and in graphically than any other ostariophysans, intra- and interspecific behavior. Many of and their inter- and intrafamilial relation- the species have specialized trophic struc- ships are not well understood. Two of the tures, and at least some are highly active living families (Ariidae and Plotosidae) at night, hiding or even burying in sand are predominantly marine. Representatives during the day. Their mode of swimming of a South American endemic family, the is peculiar (Lissmann, 1958; Greenwood et Aspredinidae, occur along the Guiana coast al., 1966: 38.S-384). The geographical re- (Myers, 1960a). Several people are work- striction of gymnotoids to part of South ing on the morphology and classification and Central America indicates that the of catfishes on a worldwide basis, and group had a later origin. Their greatest much new information relevant to their 18 and 35 is in will be available diversity ( genera species ) zoogeographic history the Amazon basin. They are almost as well soon. The startling diversity of endemic represented as this in the Guianas and the catfishes in South America and the pres- basin. Outside these areas their ence in southern South America of the diversity declines markedly. A few species, only two species in the family Diplomysti- belonging to wide-ranging genera, occur dae, the most primitive of living catfishes, north as far as southern Central America indicate that catfishes, like characoids, (a single species reaching Guatemala) and were in South America long ago, certainly south to the Plata basin. They are almost in the Mesozoic. absent west of the Andes (a single wide- Cyprinoids. Minnows, the only group ranging species reaching coastal Ecuador), of cyprinoids in the Congo basin, more or and are absent in the numerous Atlantic less resemble characins except they have coast drainages between the mouths of the protrusible jaws and frequently one or two Rio Sao Francisco and the Rio Paraiba. pairs of small barbels, while jaw teeth and Gymnotoids were derived from characoids. adipose fin invariably are absent (most Siluroids. Siluroids or catfishes typically characins and catfishes have a rayless adi- are denser-than-water, bottom-dwelling pose fin behind the dorsal fin). The lower fishes with flattened bellies and nocturnal pharyngeal teeth, however, are highly habits. The dentition usually consists of modified and despite their lack of jaw teeth numerous simple conical elements in bands minnows exhibit considerable diversifica- of varying thickness. The barbels, usually tion in feeding structures, as shown by two or three pairs, sometimes (as in Matthes (1963) for African forms. The four are almost are diverse in Bagridae ) pairs, invariably cyprinoids perhaps more present (one exception in South America, North America and certainly much more none in Africa) and serve as tactile and diverse in Asia than they are in Africa. It is gustatory organs. In the African Mochoki- generally thought that they originated in dae (represented by 37 species in the Asia. However this may be, the number the barbels are branched. Congo) highly of minnow species in Africa is very high, In contrast to characins, catfishes are noisy. especially in Ethiopia, Southern Africa, and sound from various ana- They produce in the rockier headwaters of many of the tomical structures; the most familiar bigger tropical rivers, including the Congo. sounds are due to stridulation between spine and Of the 31 fami- pectoral girdle. Non-ostariophysan Primary and lies of catfishes Greenwood recognized by Secondary Freshwater Fishes et al. (1966), eight occur in Africa (three is rich in the number of them endemic) and 14 in South and South America not of fishes it all . of freshwater Central America ( but Ariidae endemic ) major groups Catfishes are more widely distributed geo- harbors. The only primary freshwater fishes Amazon and ('()nc;o Fish Ecology • Roheiis 121

in South America otlicr than (^starioplnsi through April or May. During the rest of arc a of Lepidosirenidae, two genera tlK> year the rain decreases somewhat. In of Osteoglossidae, and two genera of \irtually the entire Rio Branco basin, the Nandidae. The secondary- freshwater fishes Rio Negro basin north of Uaupes, and (sen^ti Myers, 1949b) consist of Cichlidae, most of Colombian Amazon, on the other Cyprinodontidae, Poecihidae, Galaxiidae, hand, the rainy season is generally from and Percichthyidae. All of these families March to November or December, or from of primar\- and secondary- fishes excepting May through August, and the dry season Galaxiidae and Percichthyidae are present (at least in the savanna area of the Rio in the Amazon basin, and all of them ex- Branco) is somewhat more pronounced. In cepting Poeciliidae and Percichthyidae also the Congo basin the great Ubanghi River occur in Africa. from north of the equator and the smaller The primary and secondar\' freshwater but numerous southern tributaries are in fish fauna of Africa is much more complex. high water at opposite times of the year; It can be broken down into three main thus the maximum and minimum average categories: 1) groups shared with South rates of flow of 75,000 and 50,000 cubic America—ostariophysans excluding Cy- meters per second at Kinshasa are close. prinidae, Lepidosirenidae, Osteoglossidae, Tlie water level in the is Nandidae, Cichlidae, Cyprinodontidae, and mainly determined by contributions from Galaxiidae; 2) groups shared with Asia, southern tributaries and run-off from melt- and which probably originated in Asia— ing snow in the Andes. Snow melts during perhaps complicated by the fact that they much of the year, and rainy seasons are may have been centered in India when prolonged (it may rain a lot during the India was closer to Africa than to Asia— so-called dry season). The bulk of these Notopteridae, Cyprinidae, Mastacembeli- contributions comes at somewhat different dae, Anabantidae, and Ophiocephalidae or times of the year. The relative difference Channidae; and 3) an unparalleled as- between maximum and minimum water semblage of archaic primary freshwater levels in the lower courses of the Amazon forms known only from Africa—PoKp- and Congo rivers, although considerable, teridae, Denticipitidae, Pantodontidae, is less than in any other river in the world. Phractolaemidae, Kneriidae, Mormyridae, The absence of a period of severe desic- and Gymnarchidae. Practically all of the cation has no doubt enhanced the evolution families reach their greatest diversity in of rich faunal diversitv. Africa in the Congo basin, excepting Forests and relative environmental sta- Nandidae, Denticipitidae, and Gymn- bility. Today forests cover a large part of archidae, which are unknown from there. the Congo and Amazon basins but this has The weakly electrogenic monnyroid fishes not always been so. During much of the Recent and Pleistocene the forests parallel gvmnotoids in manv ways (see pp. periods retreated and advanced in to 140-141).' response cyclical climatic changes. Their present lies somewhere between the INTERRELATIONS BETWEEN FISHES range greatest and least extents of the In the Ama- AND THE PHYSICAL ENVIRONMENT past. zon basin much of 's Roraima Ter- IN THE CONGO AND AMAZON BASINS ritory is "campo" or savanna. Great tongues Since the Amazon and Congo basins of campo extend between the southern straddle the equator, they have two rain- tributaries of the Middle and Lower fall regimes. Most of the Amazon basin lies Amazon. Montane forests and gallery south of the equator, and in this portion the forests along the great tributaries and the rainy season is generally from December main Congo and Amazon rixers were the 122 Bulletin Museum of Comparative Zoology, Vol. 143, ISIo. 2

portions of forest that would have suffered the local faunal composition. During any least from long dry periods. Certainly some intensive collecting in the Amazon or montane forests persisted, and perhaps the Congo, a good proportion of species ob- main gallery forests never disappeared. tained in a given locale are likely to be The forests undoubtedly mitigated the ad- either recently arrived or in the process of verse effects of dry periods on the fish moving away from habitats that became fauna. On a shorter time scale, they con- unfavorable. Thus a rotenone collection in tribute greatly to over-all wetness and an igarape may yield 70 or 80 species, but temperature stability in the basins. Forest only a third or less will be represented by soils imd swamps retain a vast quantity of fair numbers of specimens. Even when water, which is relinquished throughout such a collection comprises one or two the year. Transpiration keeps the air laden thousand specimens from a stretch of with moisture and lengthens the rainy igarape one mile long, it frequently hap- season. pens that a dozen or more species are rep- Seasonal fluctuations in water level. resented by a single specimen. The indi- Seasonal fluctuations in water level have cations are that most of these came from profound effects on feeding, reproduction populations established elsewhere, that and of fishes is dispersal ( Matthes, 1964; Mc- their presence largely due to chance, and Connell, 1964). During high water more that they will soon disappear unless ad- food is available, and many fishes move ditional individuals enter the igarape. into the flooded lands to feed and to However, the species composition of ad- reproduce. Growth is rapid and fishes are jacent igarapes that superficially look alike widely dispersed. As the waters go down, may differ markedly. Occasionally an food becomes scarcer for most fishes ex- igarape has numbers of a species that is cept predators. Losses to predation are absent or represented by a single specimen greatest during low water, when fishes are in collections from other igarapes nearby. least dispersed. The effects, however, are Investigation of the physical and biotic not nearly so great as in savanna rivers components might provide an explanation such as the Zambesi, in which the flo\\' in some instances, but I suspect that, in during the marked dry season is a small addition to subtle differences in environ- fraction of that during flood time, and ment, chance plays a major role. Tlie fact massive fish mortalities are an annual that many species are represented by event except in unusually wet years. The widely separated populations over the seasonal fluctuations mean that certain entire basins probably contributes to evo- habitats exist only part of the time, and lution and maintenance of species diversity. this must play a very important role in Certainly it is difficult to conceive of the fish distribution. It should also be noted rapid extinction resulting from biotic that these cyclic changes are more pre- factors of species dispersed in so broad a dictable than some other kinds of vari- manner. Furthennore, the pattern suggests ability (especially in the temperate zones), a favorable model for allopatric speciation. and allow adaptive responses to evolve, Physical and hiolog.ical nature of white thus increasing the effective environmental water, clear water, and black water rivers. heterogeneity—and increasing rather than Tlie big rivers of the Amazon basin are of decreasing the species diversity. three main types: white water, clear water, The forest streams may be flooded in the and black, water. These types are not abso- wet season or go down so much in the dry lute. Some rivers are intermediate between season that the fishes must either leave or white and clear, others between clear and accommodate to great changes. Such con- black; some rivers may be clear water in stantly changing conditions alter drastically the drv season and white water in the Amazon and Congo Fish Ecology • Roberts 123

rainy season. Ho\ve\er, the main Amazon merous species of mosquitoes, are com- river (known as the Solinioes abo\e monl\ attracted to l^oats plying a white Manans) is white water vear round, and water river. In the daytime one is liable to its largest tributar\ . tlie Rio Negro, is lilack be bitten by hordes of Simuliidae. On the water year round. Wliite water (its color other hand, during 25 days and nights actually tan or cafe-au-lait) is due to spent on the Rio Negro (above Manaus) at suspended earth; the white water rivers widely different times of the year, I failed arise in the Andes and are constanth' erod- to encounter a single biting insect (to my ing the alkuium from their steep banks. immense relief!) and noted that even Proceeding up the Solimoes from Manaus moths were relatively scarce. It may be one is impressed by evidence of landslides that the extremely acid conditions of the (terras caidas) from the steeper banks, black waters in tlie Rio Negro are lethal some of which must release tens of thou- for most aquatic insects. Perhaps largely sands of tons of earth into the ri\er. The because of the absence of insects, ter- resulting white \\^aters are relatively rich in restrial such as frogs and lizards salts compared with other Amazonian (relatively abundant along the Solimoes) waters (Gibbs, 1970). Because of the silt, were seldom encountered in my ramblings they are opaque, and light does not pene- along the shore. Fish populations are prob- trate more than two or three feet, usuall\- ably lower here than elsewhere, although muc^i less. Thus the\- do not support there is no data to support this. In the photosynthesizing plant life. The pH of black water and white water rivers most the white waters ranges from 6.2 to 7.2 of the food available for fishes must come (in part from SioH, 1967: 31-34). from terrestrial sources (Marlier, 1967) or Seen from a distance or from the air, the floating vegetation. There is probably \'er\^ black water rivers appear quite black. In little food for nonpredaceous fishes in the a glass the water is the color of weak tea main Rio Negro, which presumabh' find and is quite clear. The black water rivers most of their food in habitats peripheral to of Amazonas apparently originate in low- the main stream, especially during periods land areas with bleached, sandy podsol of high water. In the white water rivers soils (Klinge, 1967). These soils, and per- earth slides probably bring considerable haps decaying organic matter in the amounts of plants and lower life flooded forests, provide humic coloring that serve as food directlv into the substances which give the water its char- main channels, and production along the acteristic tint. Light is effectively cut off main watercourse by floating vegetation about three to six feet below the surface. probably is also greater. Men who have The mineral content is precipitation domi- learned to fish in the Rio Negro are some- nated (Gibbs, 1970) and so low that the times conceded to be the best fishennen water may be considered "distilled water in Amazonas. Some black water rixers are slightly polluted with organic compounds." known as "rios de fome" because there is \'irtuallv' no primary productivity occurs in little food to support the population. black water rivers. Furthennore, the black Extremely little human habitation is found waters are highlv acid, with /;H ranging along the main Rio Negro betw een Manaus from 3.8 to 4.9 (SioH, 1967: 34). and the mouth of the Rio Branco. The only The characteristics of the basins and the towns are small and usually located near water chemistry play profound roles in the mouths of clear or white water af- determining the distribution and abun- fluents. Along the Solimoes River habita- dance of animals in the Amazon basin. In tions occur ever\- quarter- or half-mile over the e\-enings an incredible \ariet\' of moths long stretches. Tlie "moradores" usualh' and dipterous insects, including nu- ha\e flower and vegetable gardens and 124 Bullefiu Miisctnn of Comparative Zoolofiy, Vol 143, No. 2

small feijao plantations. The Solimoes has Coniio basins to marine fishes. Most of the large towns and cities, but these are usually African continent is relatively high above some distance up black water rivers or just sea level, and marine fishes ascending downstream from a large black water out- rivers usually do not get very far inland let, which provides a degree of protection (Marlier, 1967). Most fishes in the cuvette from insects. centrale of the Congo are primary fresh- The clear water rivers are more diverse water fishes {scnsii Myers, 1949b). Two in origin and possess a broader range of families, Cyprinodontidae and Cichlidae, pH values—from 4.5 to 7.8—than black are secondary freshwater fishes that have and white water rivers, indicating that they undergone extensive radiations in fresh are chemically (and biologically) a hetero- water. The remaining secondary freshwater geneous assemblage. The only character fishes in the cuvette centrale belong to four they share is the relative lack of organic families—Clupeidae, Eleotridae, Centro- coloring materials and suspended matter pomidae, and Tetraodontidae—which are (Sioli, 1967: 3:3-34). better represented in salt water than in In the Congo basin the rivers can be fresh water. The particular groups in- similarly classified according to tlieir volved, however, entered the fresh waters waters. The southern tributaries are mostly of Africa a very long time ago, and must black water, and during part of the year have invaded the Congo basin via other the main Congo River is a deep-tinted river systems instead of directly from the brown, approaching black water. The sea. The clupeids are members of the Ubanghi River is white, however, when it Pellonulinae, also present in fresh water passes its flood waters to the Congo River. in tropical Asia.^ All but two or three of At this time contributions from the south- the eleven or twelve species of Pellonulinae ern black water tributaries are at their in the Congo basin appear to be endemic. lowest. The Ubanghi is the only true white The Eleotridae are represented by a single water tributary in the Congo basin; the small species, Kribia nana. This species, or rest are either black water or clear water. one closely related to it, is found in many The main Congo River never becomes rivers in West Africa. The genus is known sufficiently laden with silt to be considered only from fresh water. The Ccntropomidae a white water river. In the Congo River, as are represented by hates niloticus, present in the Amazon, food chains must originate in most of the bigger rivers of tropical largely on the land. The relatively recent Africa. Finally, the Tetraodontidae (with introduction and widespread establishment two species of Tetraodon in African rivers of Eichhornia in the Congo basin must outside the Congo basin) are represented have greatly increased the relative con- in the Congo by four species, three of them tribution of floating plants to food avail- endemic. Not a single sporadic marine able in the main river courses. There are invader has been reliably recorded from many varieties of floating plants in the the Congo River above the lower rapids.- Amazon basin, including Eichhornia, but In contrast to the Congo, the Amazon these are little in evidence along the main basin is relatively open to invaders from river channels. Along much of the Congo the sea. Fourteen families of fishes that River Eichhornia is extremely abundant. are predominantly marine have species Its roots offer haven and presumably food widelv distributed in the Amazon basin. to many small species of fishes and to the The elasmobranch families Carcharhinidae young of many larger species of catfishes, and Pristidae, each with a single species in characins, and mormyroids in Stanley Pool the Amazon, are probably sporadic in-

•' observation . are ( personal ) vaders. Clupeidae (herrings) repre- Relative accessibility of Amazon and sented l)y five species, at least two en- Amazon and Congo Fish Ec:ology • Rohciis 125

dcmic; Engranlidae (ancliovies) by a of tropical rivers \\ ith the rising tide. Fishes do/cn species, al)out half endemic; Beloiii- apt to do this in the lower Amazon delta needlefishes three two area include of due ( ) by species, species Centropomidae, half-beaks Belonidae, Poma- endemic; Hemiramphidae ( ) by Mugilidae. Scia(>nidae, a single species, Hiiporamplms hrederi das\idae, Lutjanidae, Ariidae, Atherinidae, and ( Femandez-Yepez ) ( Collette, personal Carangidae, C>'lupeidae, Engraulidae, communication); Mugilidae (mullets) by a Dasyatidae. Tidal bores (pororoca) occur in of the delta area single siK'cies of Mu

as well as food for the adults and young of tributary of the Ubanghi), and some of many species. Calm and deep places down- these have connections between them- stream from sand banks and islands and selves. What is probably tlie greatest com- backwaters in which mud and organic plex of river anastomoses in the world oc- detritus accumulate provide particularly curs between the Solimoes and the Rio suitable habitats for bottom-feeding fishes Japura in the vicinity of Fonte Boa (Furo such as and Furo Auati-Parana of Laheo, Cifharinus, Tilnpia. Boia, ) , portions During the period of high water many of which were explored by the EPA expedi- the islands are partially flooded, thus ad- tion in 1968 (see Acknowledgements). The ding to the inundated areas accessible to Auati-Parana, a navigable furo connecting fishes for feeding and reproduction. In a the Japura and Solimoes and having nu- given section of the Middle Congo River merous connections with smaller tributaries, fish productivity probably bears a strong is 125 miles long. Such anastomoses per- relationship to the number and size of the haps play a role similar to that of islands in islands present in it. making the flooded forest more accessible The contributions of islands and shore- to fishes. They must also facilitate dispersal line outlined above for the Middle Congo of fishes. It seems likely (although the also apply to the Amazon. In the evening maps I have do not show any) that con- small fish characins many ( especially ) nections between the Rio Negro and move close to the shoreline in order to feed Solimoes exist some 500 miles above their and larger fish including predators and confluence. Such features frequently do catfishes probably come inshore from not show up on maps, largely because they deeper waters. In the daytime Curimatidae occur in relatively uninhabited and eco- and Hemiodontidae are to be found feed- nomically unimportant areas. Even with ing over sandy bottom near shore. Be- aerial sui-veys the smaller connections must ginning about 40 miles above its confluence be difficult to trace with certainty, and with the Solimoes and continuing some since in any case diey would not be navig- 350 miles upstream, the Rio Negro exhibits able, there has been little reason to map a multitude of islands that no other river in them. And of course maps cannot take into the world can rival. For this entire distance account all the minor changes in stream there are very few stretches without courses and the extent of flooding, which islands, and such stretches are never more varies so much from year to year. During than five miles long. Often there are two exceptionally wet years the interconnec- or three big islands abreast of each other, tions available for fish dispersal must be and a straight line across the Rio Negro greatly augmented. Since interconnections it its little or where attains greatest widths ( 15-18 such as the Auati-Parana have no miles) may intersect a dozen islands. High sovuces of their owoi, the direction of flow shoreline coefficient, however, cannot make in them presumably depends on the water up for the unproductivity of the waters of level in the rivers they connect. Thus if the Rio Negro, the shores of which must the Solimoes is higher tiian the Japura, be among the least productive in Ama- the Auati-Parana must flow towards the zonas. Japura, and vice versa; if the water level River anastomoses and stream captures. is similar in both rivers, it has little or no "River anastomoses" or interconnections current. If the level of both rivers should occur in the interior of the Congo basin as drop below that of the Auati-Parana, it well as in the Amazon. Tlius the Congo might be drained until only disconnected River above Coquilliatville has a series of pools are left. connections (Chenal de Bosesela, Chenal The Amazon basin has important con- etc. Giri a nections with other basins. The of de Nyoi, ) with the River ( large largest Amazon and Congo Fish Ecology • Roberts 127

these, the Canal de Casiqiiiare. links the ence in the Ogowe of a number of fish Rio Negro and Rio Orinoeo in \'ene/Aiela's elements characteristic of the (Jongo. Con- Amazonas Province. In Colombia the nections have been reported with the Nile Amazon basin reputedly is linked to the in the region of (iaramba; with the Chad Magdalena by the Japura, to the Guaviare basin by affluents of the Ubanghi, Gri- (which flows into the Orinoco) b\- the bingui and Ouham; and with the Zambesi to the Inirida also affluents of the Lualaba Uanpes, and ( flowing by (Bell-Cross, into the Orinoco) by the Guaniii (an af- 1965). As Gosse (1963: 152) noted, some

of the Rio . of these connections occur in fluent Negro ) The Mapuera h\drographic links the Rio Trombetas and Essequibo. In swamp\" areas, thus facilitating exchange Mato Grosso the Amazon basin supposed!) of certain fishes only. is linked to the Paraguai by the Tapajos The Congo and Amazon basins evidently and the Guapore. The Tocantins is linked ha\e been growing by stream capture at to tlie Sao Francisco. The Casiqniare is a the expense of adjacent basins. A ver\' good-sized waterwa\' throughout the year. important capture was that of the Lualaba, Some of the other connections probabU" are which probabl}' connected earlier with the broken during particularly dr>' years. Hase- Upper Nile. The capture apparently oc- 1912 find con- at a "les man ( ) was unable to any curred place named portes de nections between the Amazon basin and I'Enfer" (Poll, 1957: 60). Map study indi- the Rio Paraguai. It seems likelv. however, cates that southern tributaries of the Congo that within the recent past most of these River captured numerous headwaters of connections pennitted faunal exchanges. westward flowing rivers in Angola. Tliere The Casiqniare undoubtedly has provided is a theor\' that the Tocantins was a sepa- an easy route for exchange of fishes be- rate basin before it was captured by the tween the Orinoco and -\mazon; many spe- -Amazon. To my knowledge there are no cies have been recorded from it.^ reports of important Amazon or Congo The Orinoco basin and Guiana un- tributaries being captured by adjacent doubtedly have been areas of di\ersi- basins. If this is true, it represents a fication of the fish species independent of mechanism by which the Amazon and from the Amazon ( as evidenced by large num- Congo could have gained species bers of endemic g\mnotoids and doradid adjacent basins with little or no release of catfishes in the Orinoco and such peculiar species into them. forms as Lithoxus in the Guianas). Ex- Special hiofopes. The Congo and Ama- change between these basins and the .Ama- zon basins pro\ide instances of endemic zon basin has undoubtedly enriched the fishes that are restricted to the special fish faunas of all three. Despite existing biotopes in which they evidently origi- Amazonian connections, however, the fish nated. Several peculiar genera and species faunas of the Sao Francisco and Magda- (most of them catfishes and cichlids) lena retain marked peculiarities (a sur- known onlv from the lower rapids of the prisingly large number of endemic genera Congo (Poll. 1959. 1966; Rol)erts, 1968) of pimelodontids and species of characins are almost certainly endemic there and un- in the Sao Francisco; an extensive radiation likeh' to be found awa\' from the rapids of Astijanax-Mke fonns in the Magdalena). biotope. Two lakes in the central bowl of There also are some connections between the Congo appear to have endemic fishes: the Congo basin and river systems adjacent lac Tumba has an apparenth- endemic to it. Poll indicated an (1957) "occasional genus of {Clupeopctersius) , h\drographic confluences between the endemic species of catfish {Eutropius Ogowe and Congo" (without stating where tumbamis). and an endemic subspecies of they are) as an explanation for the pres- cichhd (TylocJiromis lateralis microdon). 128 BuUetin Museum of Cnmparativc Zoology, Vol. 143, No. 2

and lac Fwa has two endemic genera of moderate size. The paucity of this fauna Cichlidae (Cyclophan/nx and Neophaiynx) and other evidence suggests it is relatively and an endemic species of the cichlid genus recent (personal communication from G. Haplochromis (H. rhecrf)hihi.s) (Poll, 1957: R. Smith). The only highly distinctive 60). The peculiar characoid Paraphago, group that appears to have originated in known only from lac Leopold II at Kutu, the Andes is the Astroblepidae. In this is probably a relict, rather than an endemic family a single genus has speciated ex- form and may yet be found elsewhere."' tensively. Tlie Himalayas, on the other As noted by Marlier (1967), all of the hand, are the center of distribution for a present Amazonian lakes are shallow, very rich assemblage of highly modified hill young, and markedly dependent on the stream and mountain torrent fishes (pre- water level in the rivers with which they dominantly cyprinoids and siluroids). It is are linked; the fishes in them move in and unclear why the Andes have such a rel- out from the rivers. Although geologically atively poor fish fauna. old lakes are absent in the Congo and Although there are no important moun- Amazon basins, the temporary shallow tain ranges in the Congo basin, it is highly lakes have undoubtedly been a hydro- probable that several of the species of graphic feature for a fairly long time. It Barbus described by Poll (1967a) from the may be well to consider the possibility that elexated headwaters of tributary sb'cams many fishes of the Congo and Amazon are in Angola are localized endemics. In most essentially lacustrine fonns that must fre- of the big river systems of Africa the quently resort to the rivers to find suitable greatest representation of endemic Barbus habitats. (This is not to deny the existence species is to be found in rocky headwaters of many strictly riverine forms that seldom, in relative isolation from the richer low- if ever, enter lakes, such as cynodontids, lands fauna. various catfishes etc. Extreme environmental and gymnotoids, ) pliysical factors There is evidence that during wetter times and their effects on fish Ufa. McConnell in interiors of 1969 that the Pleistocene much of the ( ) hypothesized biotic pressures the Amazon basin (Marlier, 1967) and of are of far greater importance than climatic the Congo basin were covered by a large or physical factors in the evolution of tropi- lake or a series of lakes. One would expect cal freshwater fishes. There are, however, that such conditions had important conse- major physical environmental factors which quences for the evolution of fishes. have met with highly varied evolutionary The Amazon does not have any rapids responses. Concerning highly acid waters of magnitude comparable to those in the (black waters with a pH figure sometimes Congo. Nevertheless, four strange genera much lower than 5 are a major environ- of trichomycterid catfishes are known only mental feature of the Amazon and Congo from the Sao Gabriel rapids on the Rio basins), we have virtually no precise evi- Negro (Myers, 1944). It seems likely that dence of how the fishes have responded. exploration of other rapids such as those The blood of most temperate-water fishes of the Araguaia will divulge peculiar en- is unable to exchange Oo and CO2 at pH's demic forms. Amazonian headwaters in the as low as those in which many tropical Andes provide a great amount of high- freshwater fishes live. Some Amazonian gradient stream habitats. The principal fishes apparently occur only in white or fishes fomid in the highlands streams clear waters, perhaps because they are un- are astroblepids, loricariids (especially able to adapt physiologically to the acid Ancistrus and related forms), various conditions encountered in black waters. trichomycterids and small pimelodontids, Certainly many Amazonian fishes are char- and various Tetragonopterinae of small to acteristic of black waters, and some of Amazon and Congo Fish Ecology • Rohciis 129

them seem able to reproduce onl\- in water dry years when many igarapes and lagos so acid tliat it would kill iiian\- kinds of dry up completely and even big rivers may fishes. be reduced to a series of largc'ly discon- Attention should he called to the nected pools, the stagnation must be in- "friagens" or cold spells, which cause fish credible. At such times massive mortality mortalities in the Brazilian territory of Acre of fishes is apt to occur. Gadow (1909: and in other parts of Upper Amazonas. 447) reported a prolonged drought on the Bates (1892: 289) obser\'ed mortaHty of island of Marajo, in the mouth of the Ama- fry of different species of characins at Tefe zon, that caused swamps and lakes to dry that he ascribed to a very sudden and quite up to the extent that the alligators in them considerable drop in temperature caused migrated towards the rivers, and many winds. Reference cited thousands died in the by southerly ( by attempt. Spruce to 1: that the waters J. G. Myers, 1935: 20). According (1908, vol. 118) reported Geisler (1969), fish mortalities accompany- of the Tapajos are sometimes dammed by ing a friagem may be due not to a drop in the Amazon and are rendered stagnant for temperature, but to the uprising of water several weeks, during which time the water with little or no oxygen. is considered very unwholesome. (Refer- 1935: In The nonbiotic factors that have had the ences cited by J. G. Myers, 19.) most obvious, or at least best understood, the Amazon and Orinoco basin stagnation effects on the evolution of tropical fresh- of the waters seems to be associated with water fishes are deoxygenation and some kind of non-specific epizootic which drought. A variety of physiological, mor- kills fishes and other vertebrates (J. G. phological, and behavioral adaptations of Myers, 1935). South American and African freshwater In 1839 a drought occurred in the Rio fishes permit them to survive such con- Negro which brought on forest fires, and ditions. Many of these adaptations are well immense quantities of fishes killed by the known (see Carter and Beadle, 1931; Mc- heat and drought lav on the sandbanks Connell, 1964: 132-134) but some probably (Schomburgk, 1931: '181-182). In 1926 remain undiscovered. Lewis (1970) docu- during an exceptionally severe dry season mented the morphological and behavioral in the Rio Negro-Rio Branco area a great adaptations that pennit cyprinodontids in fire blazed for over a month destroying habitats othenvise totally deficient in Oi- vegetation along the lo\\'er Rio Negro, and to utilize the 02-charged water of the first it is said that fishes succumbed (Car\'alho, few millimeters immediately belo\\' the 1952). Fire probably has had little long- surface. tenn effect on the rain forest. Several circumstances lead to waters The occurrence of fires in the rain forest deficient in O- in the Amazon and Congo is an example of "less predictable" \'ari- basins. Black waters, with relatively low ability, and the fishes probably show little pH and reducing properties, are particul- adaptation to them. In savanna regions, arly subject to deoxygenation. Stagnation however, where fires are more regularh" occurs in receding flood \\aters when occurring, fishes may have evolved adapta- masses of newly drowiied vegetation rot; tions to them. The burro\\ing or estivating in lagos that become o\ergro\\ai with float- habits of lepidosirenids, Synhrauchiis, and ing vegetation; in the lower courses of other forms may well enhance the prob- some tributaries (such as the Purus and abilit\- of their survival in regions fre- Tapajos) when they are dammed at the quently subjected to fires. confluence by rising water in the main The role of fire in the ecology of South river; and in pools (including those in America undoubtedly has increased since of 1969: rapids ) during low water. In exceptionally the arrival man (see Sternberg, 130 BiiUetin Museum of Comparative Zoology, Vol. 143, No. 2

418-422), and it may be that fires of such and South America. Whereas an over- magnitude as those of 1839 and 1926 would whehning proportion of the rich endemic not have occurred along the Rio Negro cichlid fauna of these three lakes practice unless set by man. In the Rio Uraricoeira oral brooding, such behavior is far less area near the Ilha Maraca, I accompanied common in riverine Cichlidae in Africa, a fazendeiro who was setting the grasslands and in South America, where it has been of his fazenda ablaze toward the end of the reported only in some species of Geopha- dry season (in April 1969) in order that g,iis. The explanation for this remarkable the tough grass on which his cattle could behavioral dichotomy probably lies in rel- not feed would be replaced by tender atively simple considerations. In the still, green shoots when the rain came. clear littoral waters of the lakes the eggs 1945 are less Beebe ( ) reported an astonishing and young readily dispersed and variety of fishes from a small "all but dried far more susceptible to visually oriented up mud-hole" in northeastern , predators. Under riverine conditions the Some 34 species, comprising six catfishes eggs are likely to be separated from the (a trichomycterid, three callichthyids and parents by current; moreover, in the black two loricariids), fifteen characins, a gym- waters and white waters of many tropical notid (Hypopomus), four cyprinodontoids, rivers, the eggs and young would be ex- six cichlids, Polycentnis sclwmhiirii^kii, and tremely difficult to detect visually by Synhmnchtis marmorottis, were taken from predators, and perhaps even more im- malodorous mud and decayed vegetation portant, by the parents themselves, covered by damp slime (but no free water) In tropical riverine fishes of Africa, in what was left of a drying pool that had South America, and Asia, parental care been "almost unswimmable" slime for occurs mainly in fishes in which adults weeks. He estimated that in another week spend at least part of the time in swamps or ten days without water all would have or other oxygen-deficient habitats; many perished. Of the 15 characins, two are air- are capable of air-breathing. Tliese fishes, breathers. Copeina and some of the others if they reproduce in such habitats, are perhaps utilize oxygenated water immedi- usually obliged to care for the young. In ately below the surface in a manner similar African and South American lungfishes a to that of cyprinodonts. Copeina is char- nest is constructed that is subsequently acteristic of swampy places and stagnant guarded and aerated by the male, backwaters. But it is difficult to imagine Notopteridae, osteoglossoids, and Panto- how Astyanax, Creagrutus, Moenkhausia, dontidae are air-breathers and guard the Farag,omatcs, Pristella and Serra.salmus young. Gymnarchiis, which is probably could have survived in such a habitat for capable of air-breathing, makes a ^oating as in is long as they did. Most of the species nest of plants dense swamps; ,it the foimd in this Venezuelan mud-hole occur only mormyroid knowai to guard the in the Amazon basin. young. Young Gymnarchus have external Significance of parental care in Amazon gills with numerous fine filaments and a and Congo fishes. McConnell (1969) stated highly vascularized, enlarged yolk sac, both that many tropical freshwater fishes have structures playing an important role in gas some form of parental care and implied exchanges. No odier mormyroids have that biotic factors such as predator pressure such structures. Parental care might be are of far greater importance than climatic expected of Polyptenis; none has been re- factors in the evolution of such behavior, corded. In Asia the largest group of fresh- This seems to apply to cichlid fishes in vvater fishes with parental care is the lakes Victoria, Tanganyika and Nyasa but air-breathing anabantoids. These fishes not to the riverine fishes of tropical Africa typically build a floating froth nest of air Amazon and Co.\t;o Fisii Ecology • Robert,') 131

bubbles ill which the eggs are deposited. literature and additional references see The bubbles nia\' proxide a suppl\- of Breder and Rosen, 1966: 163-165). The oxygen. Usually one or both parents sta\ main events seem clear enough. The eggs near the nest to protect and maintain it. are actually deposited out of the water. A Froth nests have been attributed to the pair about to spawn station themselves at African characoid Hcpsetus (which is not the surface beneath an overhanging object known to be an air-breather, lout may lea\'e (presumably the underside of a leaf in its young in habitats likely to be oxygen- nature) that is about three or four inches deficient). Nesting habits are also ascribed above the water. With bodies close to- to characoids of the family Erythrinidae. gether, the pair leap and momentarily Er\thrinids tend to enter swampy regions, adhere to the object. After repeated trials, and Ho))lenjthrinii.s is evidently capable of the actual spawning begins. Half a dozen air-breathing. Callichtliys and Hoplo- to a dozen eggs are deposited with each stermim. South American air-breathing leap, until sixty or more eggs are left lying catfishes, construct froth nests, whereas close together (but not on top of one the related Corydoras, which probably are another) in a clump of spawai resembling not air-breathers, generally scatter their that of a snail. Afterwards the male re- eggs amidst plants. In most characoids, turns every twenty to thirty minutes and although courtship and selection of spawn- splashes the eggs with water by making ing site may be highly complex, parental \igorous fin and body movements. Be- behavior probably ends once the eggs are tween splashings the male is stationed at deposited. There are no records of parental a spot some distance from the eggs. The care in g\mnotoids, excepting the uncon- eggs hatch in two or three days, and the finned report (Du Bois-Reymond. 1882) fry drop into tlie water and spend the next that Elecfwphonis (the electric eel) prac- few^ days near the surface (summarized tices oral brooding. In Elcctwphonis a from Stoye, 1935). This fish occurs through- highly convoluted oral epithelium facili- out much of the Amazon in backwaters and tates air-breathing (Carter, 1935). (There swampy places where deoxygenation is is also reason to think that Gymnotiis. likely. This fact, as w^ell as its small size, which is related to Electrophorus, takes the habit of frequently resting at the surface, young into its mouth.) It would appear and somewhat c>prinodontlike appearance that the great majority of fishes in the suggest that it can utilize the oxygen-rich Amazon and Congo have no parental care. layer of water at the air-water interface. As out pointed by McConnell (1969: 6.3- The aquarists' accomits of its spawning, Amazonian fishes in 64) many engage up- parental care, and behavior of its fry are streai]^ spawning migrations (piracema), suggestive of adaptations to permit repro- producing extremely large numbers of eggs duction by fishes in habitats almost totally per female, all or most of these eggs being deprived of oxygen.*' laid at one time at the start of, or early in, the season. In all such fishes rainy parental BIOLOGICAL INTERACTIONS OF care is very probably nonexistent. None of FISHES IN THE AMAZON AND CONGO the fishes with modifications for breathing air participate in piracema. Apart from predator-prey and host-para- A remarkable exception to the general site relationships, with one exception there rule that non-air-breathing characoids lack are no recorded instances of biological parental care is provided by CopelUi, the interactions between Amazon or Congo spawning and parental care of \\hich have fishes and animals of other classes. The been the subject of several reports by exception is commensalism of chironomid aquarists (for detailed extracts from the larvae attached to Loricariidae and Astro- 132 Bulletin Museum of CoDipamtivc Zoology, Vol. 143, No. 2

blcpidae in the Amazon recorded by dus, all but two of them distinctively or

Freihofer and Neil 1967 . Insects even ( ) aside, strikingly marked. Equally good ex- invertebrates apparently are of minor amples could be drawn from the Charac- consequence in the Amazon and Congo idae, the cyprinid genera Labeo and river systems. Almost all of the marine Barbtis, and the anabantoid genus Cteno- groups with which tropical reef fishes dis- poma. The major exceptions to the trend play so many complex symbiotic and com- in Africa are the uniformly drab or mensal relationships are absent, and noth- cryptically colored mormyrids, in which ing has taken their place. The relatively nonvisual sensory structures are highly scarce molluscs and crustaceans exhibit specialized and the eyes are reduced. In little diversity compared to those of tropi- South America the gymnotoids constitute cal reefs. Excepting parasites, then, almost a similar exception. These bright colors the only animals with which the fishes can and striking markings are evidently meant interact are other fishes.' to advertize the presence of their possessor. Conspicuoushj marked Congo and Ama- The bright colors, unlike those of many zon fishes. Africa and South America are gaudy tropical birds and insects, are not famous for brightly colored and strikingly confined to mature males, but generally patterned fishes, most of which regularly appear at an early age in all individuals. find their \^'ay into the aquarium fish trade. If the fishes were distasteful, venomous, or Such fishes are not evenly distributed harmful in some way, we could make an among the river systems of the tsvo conti- hypothesis that their coloration warns away nents. In general, more brightly colored predators. At least for the majority of fishes occur in forest rivers than in savanna characoids, however (which provide most rivers; and more in black water or clear of the best examples of brilliantly colored water rivers than in white water rivers. Tlie fishes both in the Amazon and in the highest proportion of brightly colored Congo), there is no indication whatever fishes occurs in the river systems and in the that they are inedible or dangerous in any fish groups with the highest numbers of way to predators. It would seem, although species. The great majority of brightly there are no data available to support this, colored fishes in the aquarium trade of that the brightly colored characins are just Africa come from the Congo. The most as subject to predation as drably colored brightly colored of all freshwater fishes are ones, if not even more. In this respect the some of the small Amazonian Characidae. colors advertizing their presence are prob- The trend toward more distinctively ably a disadvantage. Many of the small, marked representatives in the Congo and brightly colored Amazonian characins form Amazon basins affects almost all groups of schools. This is tiiie of the most brilliant fishes in Africa and South America. To characins of all, the neon tetras and car- cite just two examples from Africa: the cat- dinal tetras. Most of these brightly colored fish genus Synodontis is represented by 15 characins live in black water or clear water species in the Volta basin and 37 species igarapcs, a habitat that can be highly in the Congo. None of the Volta species temporary (see p. 122). Thus popu- can match the bright colors or contrasting lations frequently are split up or dispersed, patterns of the Congo species S. angeliciis, and species recognition and schooling S. ornatus, S. ornatipinnis, S. flavotaeniatus, habits are probably important means of S. decorus, S. nummifer, and S. notattis. reconstituting populations. Again, the Volta has three species of the Association of species of Amazonian characoid genus Distichodus, all with in- characoids in mixed schools. Myers distinct vertical bars and drab colors, while (1980b), Gery (1960), and McConnell the Congo has eleven species of Disticho- (1969) recounted examples of generically Amazon and C()xc;o I'isii Ec{)loc;y • Roberts 133

distinct but similar appc^arina; Soutli Amer- minor), one of the smallest needlefishes

of Belonion 1966b ) ; two ican characoids that form mixed scliools [ apodion Collette, tsvo or more species. M\ers {ibid., p. 207) tiny species of Eleotridae {Mierophihjpnus reported that tlie small characoids Crea- Myers, 1927); and minute catfishes in gruditc maxillaris and Creanruiiis pJia.sma, several subfamilies of , a which look very similar, were collected to- number of them smaller than an\- other the which gether (it is miknown whether they were catfishes. Excepting needlefish, schooling together). He suggested that is verv slender, everv one of these is less they appear to fonn an instance of Batesian than an inch long when fully adult. All mimicry, with Creagruditc the model and excepting the catfishes are secondary fresh- Creagrutus the mimic. Re-examination of water fishes and belong to groups whose fishes identified as Cicagntdite from the presence in the Amazon can be thought of same area (upper Orinoco-upper Rio as "marginal." Large poeciliids are absent Xegro) revealed among them another in the Amazon except for three good-sized species of Creagrutus, which appears to species of Poecilia that haxe penetrated no be C. tnelanzoniis (Myers and Roberts, further inland than Para. Poecilia minor 1967). All three species have a blackish itself seems confined to the Lower Amazon. crescentic humeral blotch and arc closely It is known from only two collections in more than 100 within similar Ger\' 1960 ( appearance. ( ) sug- separated by years ) gested that similar cheirodontines and 100 miles or so upstream from Obidos. tetragonopterines that school together both Poecilia scalpridens, the only other benefit from the association, and thus the poeciliid in the interior of the Amazon, and mimicry is Miillerian. Ger}' {ibid., p. 37) not much larger than P. minor, is known labelled the schools "protective associ- from a few localities in the Middle and ations" but did not identif\' the nature of Lower Amazon. Fluviplujlax is a phyleti- the protection the association supposedly cally isolated fonn widespread in the Its provides. As noted by McConnell ( ibid., -Amazon basin. relationships evidently p. 130) the facts are not available to show lie either with Procatopodinae, a subfamily what ad\antages the fishes gain from such otherwise restricted to Africa, or with it Fundulinae found in North and mimicry. Perhaps pemiits small numbers ( mainly rather with Ri\u- of isolated indi\ iduals of two ( or more ) Central America) than to a for the other of species fonn nucleus aggregation linae ( only subfamily Cyprino- with increased chances that breeding popu- dontidae in the Amazon basin) (Roberts, lations eventually will be reconstituted. 1970a). Large gobioids occur in fresh Moynihan (196S) discussed several in- water in man\- places throughout the stances of mimicry that seem to facilitate tropics where primar\' freshwater fishes are flocking in neotropical mountain birds. poorly represented. Although a number of Amazon cnid Congo fishes of minute size. large gobiids and eleotrids have been re- Consideration of the fishes in the Amazon corded from the mouth of the Amazon, the and Congo that have minute body size as tw^o minute species of MicrophiUipnus are adults indicates that this is primarily a the only gobioids known from its interior. response to biotic pressures. In the Ama- A third species of Microphihjpnus, perhaps zon, where biological interactions among the smallest one, occurs in the Orinoco fishes perhaps are greater than anywhere basin (Myers, 1927). The Congo eleotrid, else in the world, we find more minute Kribia nana, w hile considerably larger than freshwater fishes than anywhere else, in- Micro])hilii))nus, is nexertheless a ver\' cluding: the smallest oviparous and the small fish. It is the only gobioid in the smallest viviparous cyprinodonts in the interior of the Congo. Until recently the world (Fluviplujlax jjygtnaeiis and Foecilia cyprinodont Aplocheilichtliys mijersi was 134 Bulletin Museum of Comparaficc Zoology, Vol. 143, No. 2

the smallest fish known from the Congo of many catfishes (especially in the fami- basm. note that at least some of lies Doradidae and Loricariidae and the We may ) ; these little fishes (Fluvipliylax, Poecilia protective dorsal and pectoral fin spines of Kribia most catfishes. In catfishes these minor, and ) apparently reproduce some all year round; probably they all do. Their spines are very sharp and venomous, as in size may permit them to utilize food re- the Amazonian caratai, Centromochhis sources unexploited by adults of other hechclii, and other auchenipterids. In other fishes. And it may place them below the catfishes the spines are stout and can be size threshold for attack by most predace- locked in erect position. Centrochir crocodili fishes. of the Rio is called ous ( Humboldt ) Magdalena The best African example of a fish group "mata-caiman" after the wounds it inflicts with an essentially marginal distribution, on crocodiles attempting to swallow it except for unusually small species in the (Eigenmann, 1922: 47). It is said in Ghana midst of the richest lowland faunas, is pro- that crocodiles sometimes are killed trying vided by the Kneriidae. The species of to swallow AiicJienoglanis occidentalis. Kneria and Parakneria, some of which are Many catfishes in the Amazon and Congo 80-150 millimeters long, are now knowai basins are equally equipped with pectoral from high gradient streams around virtually and dorsal spines, and once they grow to a the entire periphery of the Congo basin certain size they presumably are seldom (Poll, 1966; 1969, map on p. 360) but have preyed upon by predators that swallow yet to be found in the cuvette centrale or their prey whole. The spines do not pro- in the main rapids of the Congo River, tect them from piranhas, which bite out where one might expect ecological con- chunks rather than swallow their prey ditions would be suitable for them. Tlie whole, or from attack by various kinds of kneriid Grosseichthys p^ahonensis- Gery candirii (see pp. 136-138). On the other the electric eel of the Amazon EJec- (1964), only 18-20 millimeters in standard hand, ( length and very slender, was discovered in trophorus clectricus) and the electric cat- 1964 in in fishes of Africa be en- l:)y Gery forest streams Gabon ( MoJopterurus ) may and liy myself in forest streams in the tirely exempt from predation. Malapterurus western part of the cuvette centrale. This only an inch or so long are capable of is now the smallest known species of fish producing a jolting shock (personal ob- in servation . the Congo basin.^ ) Adaptive responses to predation. McCon- Fartitioniuii, of food resources, and nell 1969 of Another obvious ( ) noted the high number trophic adaptations. aspect predatory fishes in the Amazon (which she of biological interaction or accommodation felt was relatively higher than in African in Congo and Amazon fishes is the par- fresh waters) and suggested that this per- titioning of food resources. Food-place, mitted more prey species to co-exist. (The food-time, and food-type partitioning are converse true. The basic may also be ) Of adaptations probably equally important. displayed by Amazon and Congo fishes dichotomy of ostariophysans into nocturnal that lessen the toll of predation for their and diurnal groups (siluroids and gvm- possessors, reference should be made to notoids vs. cliaracoids and cyprinoids) is the alarm substance and fright reaction in mentioned in the introduction. The habitat Ostariophysi, Kneriidae, and Phractolaemi- preferences of many of the fishes in the dae (Pfeiffer, 1963, 1967); the abihty of Amazon and Congo are in many instances gymnotoids to withstand mutilation (Ellis, closely linked with feeding habits. Many 1913); the cryptic body form and color- of the fishes have highly modified trophic ation of such fishes as Farlowcllo: the structures, some of which are quite heavy (and frequently spmy) body annor peculiar. Interestingly enough, for some Amazon and Congo Fish Ecology • Roberts 135

of the most peculiar types of feeding we survival, even in habitats where cata- have examples of convergence between strophic (i.e., nonbiological) causes of ex- fishes of the Amazon and Congo basins. It tinction tend to be minimal, has been said that riverine fishes of tropical The main categories of food in the big Africa and South America exhibit less tropical rivers are relatively few, namely, trophic specialization than cichlids in lakes 1) other fishes; 2) insects, both acjuatic Victoria, Tangan\ika, and Nyasa and that and terrestrial, and aquatic insect larvae; this is due to relative 3 fruits leaves largely year-round ) higher plants, including and availability of food in the lakes in contrast fallen into the water and roots growing to marked seasonal abundance and dearth out into the river from the banks, as well of food in the rivers. Tlie observation and as some aquatic plants; and 4) mud or the reasoning, it seems to me, while they earth, including interstitial organisms, dead may be relevant in savanna rivers having organic matter, and possibly bacteria. Each extreme seasonal vicissitudes, do not apply of these categories seems to be highly to the Congo and Amazon rivers. In any partitioned among the fishes and to have event, it should be remembered that fishes evoked various kinds of trophic, be- are capable of fasting for considerable havioral, and morphological adaptation. It periods, provided they get an adequate should be noted that some lake cichlids share of the resources available during that are highly specialized for feeding on times of abundance. A more important plankton and molluscs would find little or reason for the relative paucity of fishes none of their predilected food in rivers of with highly specialized trophic structures the Amazon or Congo basins. in savanna rivers is probably faunal im- Predatory fishes. It would appear that poverishment resulting from drought. The nowhere else on earth have fishes evolved diversity of feeding habits of fishes in the as many manners of preying on other fishes Amazon and Congo is perhaps as great as as in the Amazon. Excluding fin-eaters and one could expect given the nature of the scale-eaters, in the Amazon over 40 species food available. Were it not for partitioning of characoids alone are primarily or ex- of food resources and concomitant evo- clusively piscivorous. The more voracious lution of specialized feeding behavior, the species of Scrrmalmus (known as piranha astounding number of fishes present in chata in Brazil) bite out chunks from the Congo and Amazon probably would be larger fishes. Gery (1963: 615-616) noted unable to coexist. Partitioning of food that for each species of Serrasalmiis in the resources apparently has proceeded to the voracious subgenera Vygocentrus and Tad- point where utilization and cycling of energy dyella there is a geographically correspond- and materials are very efficient. The upper ing species in the less specialized subgenus limits of partitioning presumably are de- Serrasohjiii.s. He hypothesized a parallel termined by complicated factors such as evolution of sympatric species in which the the variety of foods and their relative less aggressive Serrasalmiis (known in availability in space and time; and the Brazil as pirambebas) benefited from as- capacity of fishes to exist (e.g., by fasting sociation with Py(i,ocenfriis and Taddyella. or facultative feeding) when such re- Several of the most archaic fishes of the sources are unavailable, and to reproduce Amazon and Congo basins are rapacious when they become available. Food re- predators. Some species of Polypterus (e.

sources cannot be partioned indefinitely, g., P. seneg^aJus) are insectivorous, but the because a point would be reached where two largest species—P. endJichcri and P. individual parcels of energy and materials coniiicus—are piscivorous. The African are too small to support populations big osteoglossoid Ilcterotis is a filter-feeder, enough to have sufficient probability of but the Amazonian Osteoglossum is 136 Bulletin Museum of Comparative Zoology, Vol. 143, No. 2

piscivorous (feeding mainly on characins), eaters. The small Microstomatichthyohortis as is Arapaima fligas (the largest osteo- katangae feeds on aquatic insect larvae. glossoid), which sometimes feeds on Gavialocharax (from Cameroun), with its Osteoglossum. The African Hepsetus, wonderfully elongated jaws, probably is a which appears to be the most primitive fin-eater, as are the species of Belonophago characoid (Roberts, 1969), is a voracious (from the Congo). As with scale-eaters, piscivore, as are a number of phyletically they attack fishes much larger than them- isolated (and perhaps primitive) South selves. The fin-eater diet, rich in bone American characoids.'* This should not minerals, probably was prerequisite for the come as much of a surprise, for a remark- development of the exceedingly hard, able number of survivors of archiac fish platelike dermal armor (derived from groups throughout the world—viz., lam- scales) present in Belonophago and Fhago. preys, sharks, chimaeras, gar-fishes, bow- Daget (1967) reported that Ichthyhorus fin, tarpon, and pikes—are extremely hesse hesse in the Chad basin is a fin-eater, effective piscivores. Far from inhabiting while Ichthyhorus hesse congolensis in the situations geographically or ecologically Lualaba seems to have readopted second- isolated and open to relatively few orga- arily the diet of a micropredator. Daget nisms (as do survivors of some archaic (ibid., p. 142) suggested that the differ- invertebrate groups), these modern pre- ence in feeding habits between the two daceous representatives of archaic fish subspecies might be explained thus: in groups are frequently dominant forms in Katanga the "ecological niche" of fin-eaters shallow seas and especially in lowland is occupied by Phago and EugnathicJithys, rivers e. the while less well ( g., Congo ) where the fish Ichthyhorus, probably fauna is exceptionally rich. adapted to fin-eating, has its diet restricted Scale-eoting and fin-eating characoids. to small fishes and invertebrates; in the South America has several scale-eating Chad basin, on the other hand, Ichthi/honis characoids (Roberts, 1970b). Four genera has no competitors (no other ichthyborids and about ten species are present in the occur there) and occupies only the "eco- niche" of fin-eaters. In Amazon ( scale-eating has yet to be verified logical the Amazon for all of them). Eating scales appears to be the only known fin-eaters are in the genus of major importance in the mode of living Serrasahnus (e.g., S. eJongatus), and they of these fishes. Several of the species have tend to utilize other foods to a considerable been observed actually feeding on scales degree. from other fishes, and scales generally Feeding hahits of the Amazonian cat- predominate in their stomach contents. fishes known as candiru. Candiru is an Their teeth are obviously specialized for Amerindian name for certain catfishes removing scales. They generally attack which attack other fishes and, occasionally, fishes larger than themselves, which pre- man. (For a delightful account of candiru attacks on see 1930. sumably usually escape without being man, Gudger, ) Very killed. Although the Congo is rich in little is recorded concerning their feed- characoids, no scale-eaters have been re- ing habits. About 30 species have been ported there. The only other freshwater described from the Amazon basin, repre- scale-eating fishes are cichlids from lakes senting two unrelated families of cat- Nyasa, Tanganyika, and Victoria. fishes, Cetopsidae and Trichomycteridae. The Congo, on the other hand, has a The trichomycterid candirus belong to remarkable group of fin-eating characoids, four subfamilies, Pareiodontinae, Stego- all in the family Ichthyboridae (Matthes, philinae, , and Tridentinae. 1961). Most but not all members of this During fieldwork with EPA in October and family will probably be implicated as fin- November 1968 many candiru species were Amazon and Conco I<^ish Ecoloc.v • Roberta 13"

collected, some of them in considerable the\- leave a nearly perfect circular scar. numbers, and obsenations were made on Parciodon or Hemiceiopsis eight inches their feeding behavior. Most were col- long leave a scar about a half inch in lected with a fine-meshed wire basket diameter. Stomachs of these fishes fre- (open at one end) baited with fresh fish quently contain a half dozen or more round heads, or with a fine-meshed dipnet and a chunks of flesh or \iscera. There is no fresh fish head suspended in the water on evidence that either Parciodon or Hemice- a string. Large catfish heads worked best. topsis enter the gill chambers or suck the Lichtenstein blood or fluids of other fishes. coecutiens ( ) gets body at least a foot long and tw o inches in diam- Tlie majority of candiru belong to tricho- eter, and is the largest candiru in Ama- mycterid subfamilies, , Tri- zonas. On several occasions I saw this dentinae, and Vandelliinae. These range species taken in open water in midday, on in size from less than an inch long and hooks baited with pieces of fresh fish. Its extremely slender to perhaps six inches back is deep sky-blue, its belly milk white. long. In Stegophilinae and Tridentinae, It was never taken at night or b\- means of which are closely related, the mouth is my fish heads. Hemicetopsis candiru wide, and teeth m both jaw^s are very (Spix) reaches about eight inches in length numerous and arranged in several rows. and its whole bod\' has a pinkish or reddish In \'andelliinae the mouth is relatively hue, especially in somewhat smaller speci- narrow and the teeth are few and in only mens. The largest specimens tend to ha\e one or two rows in both jaws. According a more greyish cast. It is apparently active to Eigenmann (1918), Reinhardt in 1858 only after dark. At first none of this species was the first to record that a species of were collected. Then one evening, finding candiru {Steii,0})hilus insidiosus Reinhardt, m)'self without fresh material for bait, I from the Rio das \>lhas, Rio Sao Francisco tried some old fish heads from pre\ious basin) enters the gill chambers of other fishing that had been tied to the boat, fishes. Eigenmann (1918) recorded similar and got a fair number of Hemicetopsis behavior for a species of \'andelliinae candiru and virtually no other species. {Branchioica heiionii from the Rio Parana). Earlier it in the EPA fieldwork a number of ( had been noted that, general, During fish heads were most efficacious as candiru \'andelliinae and Stegophilinae were ob- \ fresh. tained from the chambers of bait w hile cry ) Later on, fishing directly gill was carried out with fresh and rotten fish fishes caught on hook and line, usually heads simultaneously, with the finding that large catfishes, and their stomachs were Hemicetopsis came only to rotten baits. In frequently gorged with blood. Stegophili- this it differs from all other candiru. nae and \'andelliinae were readily caught Parciodon microps Kner reaches about by using fresh fish head baits. Pseudosiego- eight inches long and its bod\ is unifonnly philus nemurus (Giinther) were observed grey. It was more commonly captured than on onl\- one occasion, at midday, when our any other candiru w^th fish heads used as boatmen were unusually successful in bait. On several occasions 50 to 100 were catching dourada and filhote (catfishes of the hook and captured in a baited wire basket within an genus Brachyplatystonia ) by hour or two and on one occasion about 300 line. One specimen dropped out of the of a after it were caught in less than an hour. They gill chamber dourada was came mainly in the first hour or two after landed and was positively identified. The nightfall and seemed to prefer fresh baits. Pscudostegophiltis appeared in numbers Hemicetopsis candiru and Parciodon and were darting about near the surface, microps feed by biting out circular chimks. even jumping clear of the water in evident If they are unsuccessful in removing a bite. excitation. Other than on this occasion, 138 Bulletin Museum of Comparative Zoology, Vol. 143, No. 2

dusk and the first hour or so after nightfall (identified in the field as Ochmacantlms appeared to be the time of greatest activity reinhardti). The fishennan indicated that for Vandelliinae and Stegophilinae. Pseuclo- many such candiru had been clinging to steg,ophilws, with its handsome golden-hued the piramcu as he hauled it into the canoe, body and dark cross bars, is the most color- Upon examining the gill chamber of the ful candiru, which further suggests that it pirarucu, I could not detect any indication is diurnal more than other species. of damage to the tough gill filaments and Kelley and Atz (1964) iDublished a photo- strongly doubt that the Ochmacanthus graph of VandeUio attacking goldfish in an obtained any blood from the gill filaments aquarium and reported that after gorging or elsewhere. Perhaps they were feeding on blood the settled to the bot- on mucus, which is abundant on pirarucu. torn of the aquarium. Vandelliinae, except Ochrnaconthus observed on sandy bottoms for their eyes and viscera, are generally in relatively clear water were always light transparent in life. Plectrochilus has a colored. I suspect they can change color longitudinal black mark on the caudal to match the substrate, whether it be light peduncle and fin. Vandelliinae are slender sandy bottom or the dark body of a fishes but are capable of considerable ab- pirarucu. dominal expansion to receive blood. There Mr. Jonathan Baskin, who is stvidying is no evidence that either Vandelliinae or the family Trichomycteridae, called to my Stegophilinae spend protracted periods in attention the scale-eating habit of the the gill chambers of another fish; perhaps stegophiline Apomatocews alleni Eigen- members of both subfamilies gorge them- mann. The mouth of this species is capa- selves fairly soon after entering and then ble of being everted to form a discoid swim out. Haseman (1911: 315) stated sucker about twice as wide as the head, that Vandellia and other trichomycterids and is provided with numerous bands of lie buried in sand bars. When disturbed teeth. The evidence of scale-eating is pro- the Vandellia "rise like a flash and bury vided by examination of an alizarin prepa- themselves again in an instant, leaving a ration and radiographs of two specimens small round hole where re-entered." 105 and 111 as 109804 they ( mm, catalogued no. He reported collecting four hundred Stego- in the fish collection of the Academy of philtis buried in the sand in an igarape Natural Sciences of Philadelphia), the ali- near Santarem; however, his remark that mentary canals of which are partially filled these were "minute and could be seen by with scales about 3 millimeters long, the millions in this locality, where the The food habits of the Tridentinae are water was too shallow to pennit the pres- unknown. All are small (usually an inch ence of large fishes of any kind" indicates or less in length). We collected several they may have been Tridentinae instead species of this subfamily by using rotenone of StegO})hilus. McConnell (1964: 116) re- in shallow, forested backwaters wth deeply ported Trichomyctenis (Pygidium), minute tinted, clear water at localities along the trichomycterids (Tridentinae ?), and Rio Solimoes between the Rio Purus and Hemicetopsis amidst a large variety of Rio 19a, but no observations were made on other fishes (mostly catfishes) hiding in their behavior. crevices and hollows in logs in a creek of Amazonian fishes uitJi trophic structures the Rupununi River in British Guiana. for straining minute organisms from mid- On one occasion I watched a fishemian icater. Production of phytoplankton is return from spearing a pirarucu (Arapaima practically lacking in most Amazonian gigas, the largest scaled fish in Amazonas). waters and zooplankton is often absent. In In the bottom of his canoe were several black waters and white waters this is specimens of a very dark stegophiline probably due mainly to the low penetra- Amazon- and Congo Fish Ecology • Roberts 139

tion of light. The big black water rivers most to the symphysis of the lower jaw, with \\ide mouth-bays have \ery little so that virtually the entire floor of the primar\' production of ph\toplankton, oropharyngeal cavity is lined with gill partly because the water is highly acid and rakers. The gill rakers on all of the gill extremely poor in inorganic ions and arches are elongate and exceedingly num- nutrients. An insignificant amount of erous. There are no less than three verv phytoplankton, mainly of desmids, and distinct species of HypophihaJmns in the more surprisingly, a certain amount of Amazon. At least one of them gets to be zooplankton may be present. The principal two feet long. In Brazil they are called biotopes in which phytoplankton develops mapara. The ventrally directed eyes and are the mouth-ba\s of clear water affluents, laterally compressed body with its compli- such as the Tapajos and the Xingu, and the cated system of criss-crossing lateral line shore lagoons or lagos of white water canals make the genus readily recogniz- rivers. In some places veritable plankton able. The trophic structures are highly blooms occur. The shore lagoons are often distinctive. The opening of the mouth is favored fisliing grounds. Fishes are scarce large and its roof is smooth. Most of the in the mouth-bays, however, and the main gill rakers are borne on the elongate lower consumption of the phytoplankton pro- limbs of the gill arches. A 300-millimeter duced in them ma\^ happen in the white specimen of H. edentatus has about 240 water rivers into which they flow (Sioli, gill rakers are borne on the elongate lower 1968). of the gill rakers tapers off at either end A number of Amazonian fishes have of the gill arch, but most of them are trophic stnictures that could pennit them extremely long, about 15-17 millimeters. to utiHze plankton. Bohlke (1953) de- Rakers on succeeding arches are almost as scribed a minute (25-30 mm) herringlike numerous as those on preceding arches, characid from the upper Rio Negro, and only slightly shorter. The first two Thrissohiycon pectinifer, with "otter- arches bear rakers on the leading edge board" maxillaries in a nearly toothless only; the full length of the trailing edge of these arches a mouth and about 25 long gill rakers on the supports strong mem- lower limb of the first gill arch. Bohlke branous flap the height of which is about inferred that it is an open-water, schooling one-half of the length of the gill rakers. planktophage. Amazonian clupeids and The third and fourth arches bear rakers on engraulids tend to be predators, but Ceten- both leading and trailing edges, and no graulis juruemis Boulenger, with about flap. The fifth arch bears rakers on its

40 long, finely denticulate gill rakers on leading (free) edge. The mouth, branchial the lower limb of the first gill arch, is prob- arches, and gill covers can be held in such ably planktophagous. The Amazonian a position that all water leaving the gills fishes with the most highly modified ap- must first pass through a fine sieve formed for the rakers observed of paratus straining minute organisms by ( by manipulation from the water are the catfishes of the preserved specimens). The tips of the gill genus Hypophthalmiis and the characin rakers of the trailing and leading edges

Anochis elongatus and one or t\vo of its of the third, fourth, and fifth gill arches close relatives. In these presumably plank- meet to fonn a AAAAA -shaped sieving tophagous catfishes and characins the mechanism. For this to be completely ef- mouth is toothless and the gill slits are fective all the water must exit through the extremely long. The gill membranes are third and fourth gill slits. The membranous free from the isthmus. The gill openings flaps on the trailing edges of the first and and gill arches extend anteroventrally al- second gill arches presumably prexent 140 Bulletin Museum of Comparative Zoology, Vol. 143, No. 2

water from exiting through the first and dependent evolution of weakly electro- second gill slits. Ilypophthalmus form large genie freshwater fishes in Africa and South schools and extensive undergo migrations. America is a particularly striking example They are one of the most important food of parallelism because of its novelty and fishes in the lower Tocantins and are the pervasiveness of its effects. Some 18 among the fifteen or so commonest species genera and 35 species of gymnotoids have in the fish market at Manaus. been recorded from the Amazon basin. The highly streamlined Anodus elon- Undescribed Amazonian species are pres- gafiis appears to have an even more perfect ent in most important museum collections

straining mechanism than Hijpophthalmus . and the group is in great need of syste- Both leading and trailing edges of its first matic revision. Ten genera and 93 species four gill arches bear rakers, and the fifth of mormyroids have been recorded from arch bears rakers on its leading (free) the Congo basin. It is believed that all edge. The upper limbs of the arches are mormyroids and gymnotoids possess both about two-thirds as long as the lower limbs electrogenic and electi-osensory faculties, and bear correspondingly fewer rakers. and this has been verified for at least one The number of rakers is only slightly re- species in almost every one of the genera. duced on successive arches. The first arch In both groups it is apparent that virtually of a 200-millimeter specimen bears 80 + all aspects of the morphology and behavior 110 rakers, most of which are 10 or 11 have become specialized and integrated millimeters long. The dorsal edge of each with the electric faculties. Certain aspects raker bears two rows of about 100 or more remain uninvestigated. For instance, it is tiny denticles approximately 0.2-0.3 milli- known that electrosignalling functions in meters long and 0.1 millimeter apart. The territorial and aggressive behavior in denticles of adjacent rakers mesh to fonn gymnotoids and monnyroids, but its (pre- an exceedingly fine sieve. The stomach is sumably important) role in sexual behavior thin walled and the intestine relatively has not been described in either group. straight. In the specimens examined by The parallels in habitat selection, mode me, collected by the Thayer Expedition, of locomotion, and feeding habits between the stomachs are empty. A related form, gymnotoids and mormyroids are intimately Eigenmannina melanopogon, from the up- bound up with their electric faculties, al- per Amazon has exceedingly numerous gill though this has not been fully appreciated rakers, and is also presumably planktopha- and much remains to be learned concern- gous.^" ing it. It is my belief that elucidation of In the Congo basin none of the catfishes the interrelations between electric behavior or larger characins are planktophagous. and feeding habits in mormyroids and The small characin Chipeopctersius schon- gymnotoids will contribute materially to tedeni Pellegrin in Lake Tumba is a pelagic understanding the evolutionary history and planktophage (Matthes, 1964: 43, figs, b perhaps the very origin of both groups. The and d, pi. 1), as are some endemic Congo strongest evidence that electric behavior species of Pellonulinae. has profoundly affected feeding habits in Porallelism in the feeding habits of mormyroids and gymnotoids lies in the mormyroids and gymnotoids. The nature repeated development in both groups of of the electric faculties and their biological highly peculiar and remarkably similar in significance gymnotoid and mormyroid trophic structures, e. g., diverse types of fishes is now under intensive investigation; elongated tubular mouths with weak jaws evidently a great deal remains to be dis- and feeble dentition. These structures evi- covered and understood (the best general dently pennit efficient exploitation of a review is still Lissmann, 1958). This in- rich bottom fauna of small worms and Amazon and Conco Fish Ec-olocy • Roberts 141

\\'()rmlikc insect larvae (e.g., enchytraeids able amounts of substrate with their food. ti.shes serious efforts have been made to de- and chirononiid lai\ ae ) which other No at all. For termine what of the material can use only marginally or not ( part ingested food habits of gymnotoids, see Ellis, 1913. these fishes digest. Obviously this will be Little comprehensive information is avail- difficult to work out. The food may be in able on mormyroid food habits. For food virtually unrecognizable condition before habits of Congo mormyroids, see Matthes, ingestion, and is liable to be taken in with 1964. Obser\ations of mine and a few pub- all sorts of iionnutritional material. Never- lished accounts confirm that some monny- theless, partitioning of food resources prob- roid species feed heavily on chironomid ably occurs in these bottom feeders. At larvae.) It seems likely that several factors first glance, the manner of ingesting quan- enter into this relationship between electric tities of bottom material might seem to faculties and feeding habits. The nocturnal preclude fine selection of food resources, beha\ior of some gymnotoids and mormy- but I expect this contention would dis- roids may mean that they are feeding appear if we had more infonnation about actively when such wonnlike prey is most habitat selection and substrate preference susceptible to predation. But this in itself of the bottom feeders. seems insufficient to account for the suc- One of the main reasons African chara- cessful exploitation evidenced. Two further coids are less diverse than those in South possibilities merit consideration. The first America appears to be that in Africa other is that mormyroids and gymnotoids are fish groups pre-empted certain major food able to locate such minute prey electro- resources. Mormyroids and cyprinoids, to sensorily. I think this highly likely. The cite what appear to be the two most im- second is that the weak electric emanations portant examples, appear to have largely of gymnotoids and momiyroids affect these or entirely taken over bottom feeding prey in a manner that makes them more niches that might otherwise have been susceptible to predation. Either of these available to characoids. None of the last two possibilities (or both acting to- African characoids, with the possible ex- gether) would, in my opinion, go a long ception of Citharinus, have bottom feeding way to\\'ards explaining the evolution of habits comparable to the Prochilodontidae almost all of the more peculiar trophic and toothless Curimatidae of South Amer- modifications exhibited by mormyroids and ica. In the Congo the cyprinoid genus gymnotoids. I would go even further, and Laheo, which parallels the family Pro- suggest that the interrelation between chilodontidae in certain respects, is repre- electrical faculties and feeding played a sented by at least 22 species, most of them decisive role in the initial divergence of endemic. Much of the present diversity of the gymnotoids and momiyroids from non- African and South American characoids is electrically specialized ancestors. probably due to relatively late radiations, Bottom fccd'mii^ fishes with generalized and in Africa these radiations occurred in trophic structures. Roughly a third of competition for food resources with a Amazonian and Congo fishes are bottom greater number of other fish groups. feeders. Mormyroids and gymnotoids, and Trophic specialization evidentK' plaxed Chilodontidae and Hemiodontidae among a major role in the adaptive radiation of the characoids, are highly selective in re- characoids. The variety of characoid teeth, moving food items from substrate. Most from simple conical elements to highly bottom feeders, however, including mem- unusual forms of multicuspid teeth, is bers of the large South American characoid unparalleled by any other living group of family Curimatidae, cyprinids of the genus vertebrates except mammals. The peculiar Laheo, and many catfishes, ingest consider- modes of tooth formation and tooth re- 142 Bulletin Museum of Comparative Zoology, Vol. 143, No. 2

placement characteristic of characoids ap- many other forms reported on by Knoppel parently arose early in their history and are precisely what one would expect from provided the main moiphological variations the morphology of their trophic structures. that led to the evolution of diverse feeding Tlie numerous, sharp conical teeth of habits. These have been described and Hoplias are clearly those of a piscivore, discussed elsewhere 1967 . that ( Roberts, ) and Knoppel found adult Hoplias Feeding habits of fishes in .small Arna- ingested only fishes (ibid., p. 272). The zonian rain-forest streams. The stomach fan-shaped teeth of Poecilobrycon and contents of 49 fish species from three rain- Iguanodectes, with numerous small cusps, forest streams near Manaus were reported are adapted to feeding on filamentous on 1970 . His con- by Knoppel ( ) main algae, and Knoppel found that considerable clusions (pp. 343-346) were: amounts of filamentous algae were ingested by Iguanodectes and Poecilobrycon. Denti- 1. Terrestrial insects (especially ants), tion of similar appearance occurs in many aquatic insect larvae (especially Ephem- small characids, and it is predictable that eroptera and Trichoptera), and vegetable filamentous algae is an important food item remains are major items in the stomach for at least some of them. in- contents. Concerning testinal ratios, adult Curimatidae have 2. Most species exhibit considerable vari- extremely convoluted intestines, and this ability in the items ingested, and stomach corresponds with their habit of ingesting contents of various families are relatively amounts of fine a small uniform. large detritus, only ? of is ( ) portion which nutritional. In six 3. Stomach contents of the same species specimens of Curimatus spilurus? from collected at different times of the year 26.0 to 42.3 millimeters in standard length (May, July, and November) are generally Knoppel found "sand (40%) and detritus similar. (54%) in all stomachs"; in one stomach 4. The fishes find their food in the whole there was plant matter (ibid., p. 276). living space, even those species that appear While juvenile C. spilurus have "only a to be adapted to certain zones in the few concentric coils on the stomach sac" stream. {ibid., p. 335), I find that a 100-millimeter 5. Distinct specialists in food ingestion specimen has 21 concentric intestinal coils, are not present in the forest streams some partly folded over on themselves, the studied. entire mass filled with mud and occupying 6. Neither the structure of the snout and a space considerably greater than the space denture, nor the morphological structure occupied by the stomach. The intestines of the alimentary canal, nor even the in- are also extremely convoluted in Curimatus testinal ratio can be used as indices to the latior. Knoppel found that the stomach of feeding habits. a 151-millimeter specimen of this species The last three conclusions are too sweep- contained "gray-black, very fine sand, and ing even as broad generalizations, and are some larger grains" (ibid., p. 276). One not entirely in accord with statements in can also learn to recognize the morphology the main body of Knoppel's paper. Thus of trophic structures indicative of a mixed Hclogenes and Pijrrhiilina "picked up food diet. Characid omnivores, for example, have of tri- to nearly exclusively at the surface .. . yet usually moderate numbers most fishes find their food at the bottom" quinqui-cuspid teedi.^^ (ibid., p. 341). The stomach contents of The generalization about the relative Hoplias, Bryconops, Poecilobrycon, Igiiano- lack of narrow trophic specialists in such dectes, PyrrkuUna, Steatogenys, Eigen- small Amazonian streams is partly valid. mannia, Curimatus, Chilodns, Cichla, and Most of the highly specialized predatory Amazon a>jd Congo Fish Ecology • Rolwiis 143

characoids, such as piranhas and scale- NOTES eaters, are entirely absent. The commonest 1. Poll (1964) described Congothrissa from the which piscivore is probably Uoplkis, cinettc centrale and designated for it a new swallows its prey whole. Many nonpre- family, Congothrissidae. Its relationships are dis- et and van den daceous characoids with highly specialized cussed by Poll al., (1965) Thys Audenaerde It should be re- structures do not occur in small (1969). probably trophic ferred to the Pellonulinae. streams. Hemiodontidae, Proehilodontidae, 2. The record of Arius africanus Giinther from of are ab- and most genera Anostomidae Stanleyville (Nichols and Griscom, 1917: 716) is sent. On the other hand, few fishes are probably a mistake. The specimen upon which it is based should be in the American Museum of more "distinct specialists in food ingestion" Natural History. A recent search for it there was than the leaf fish Monocirrhiis polija- unsuccessful and it is presumed lost. one of the studied cant]}us, species by 3. Specimens of sharks and sawfishes from the Kn()ppel, which feeds exclusively on small Amazon River have yet to be examined by persons them. The Amazonian shark live fish (see Liem, 1970). Rhamphichthyid competent to identify is Carcharhinus leucas, the same spe- gymnotoids also have specialized means of presumably cies that occurs in Lake Nicaragua. A shark that ingesting their small prey. probably was this species was photographed at Iquitos (Myers, 1952). Both Pristis pectinatus ACKNOWLEDGMENTS and P. perrotteti might be expected, at least in the lower Amazon. The Museum of Comparative Thanks to Dr. P. E. Director Vanzolini, of Zoology possesses a watercolor by J. Burkhardt of the Museu de of the Zoologia University a specimen of Pristis perrotteti taken by tlie of Sao Paulo, I was able to participate in Thayer Expedition at Para on August 14, 1865, fieldwork of the Expedi9ao Permanente da with tlie pencilled comments "]e possede un morceau du bee de ce de Monte Alegre" Amazonia on the Rio Solimoes from Sep- poisson and "M. le Col. Michille m'a dit qu'on en avait tember 20 to November 5, 1968. This pro- pris de'le Rio Madeira, au-dessus de Borba." fishes in a vided an opportunity to see 4. Gnathodolus bidens Myers (1927), a remark- great variety of habitats. EPA is a con- able anostomid, is known onl>' from the type collected at the Orinoco-Casiquiare tinuing joint effort of the Museu de Zool- specimens bifurcation. ogia, Instituto Nacional de Pesquisas da 5. Poll (1959: 160-161) considered the possi- Amazonia, and Museu Goeldi. It is fi- Ijihty of endemic fish species in Stanley Pool and nanced by the Funda9ao de Amparo a concluded that diere probably are none. Of 13 known from Stan- Pesquisa of the state of Sao Paulo. The rich species he listed that were only five were subsequently recorded from collection of fishes being assembled is kept ley Pool, Yangambi (Poll and Gosse, 1963). at the Museu de Zoologia, under the 6. The fish aquarists have observed depositing of Sr. Heraldo A. Britski, to curatorship its eggs out of water is definitely not Copcina provide the basis for a systematic revision arnnldi Regan, but is probably either Copella of the Amazon fish fauna. nattereri Steindachner or Copella compta Myers communication from G. S. Myers). Neal I wish to thank M. Pierre Brichard for (personal Foster has called my attention to tlie account of me to with the helping get acquainted "Copeina arnoldi" spawning on a leaf above the fishes of Stanley Pool, and Sr. Willy water in their aquarium by Nieuwenhuizen (1964: Schwarz for providing me with an op- 156-159), which is illustrated by excellent photo- The fish in the are Copella. portunity to visit the lower Rio Negro and graphs. photographs 7. Most of the main rivers and streams in the Rio Jauaperi. Many colleagues have shared Amazonian lowland are relatively poor in num- fishes from the with me infonnation about bers of kinds and of indi\iduals of molluscs and Amazon and Congo. In this respect I must crustaceans. This is also true of the (^ongo. The particularly thank Prof. George S. Myers. paucity of these two groups is particularly striking when compared to their richness and abundance in Very helpful comments on drafts of this lowland streams of portions of Southeast Asia. paper were provided by Dr. Gerald R. Afjuatic leeches also seem to be more abundant R. and Prof. Smith, Dr. Robert Miller, in Southeast Asia. Perhaps molluscs, crustaceans, Myers. and leeches would be less abundant in Southeast 144 Bulletin Museum of Comparative Zoology, Vol. 143, No. 2

six small Asia if the dominant ostariophysans there were consists of four to relatively and widely characoids instead of cyprinoids. G. R. Smith spaced tricuspid teeth, the inner row of eight to informs me that shrimp are fairly abundant in 16 multicuspid teeth; in the lower jaw the outer some Amazonian headwaters. This may be related row consists of eight to 16 multicuspid teeth, tlie to the relative paucity of their fish fauna. Many inner row of two relatively small conical teeth molluscs and crustaceans presumably would find near the symphysis of the jaws. Occlusion ap- l:)lack water habitats acidic and too poor in calcium parently occurs only between the inner row of and white water habitats too silty. teeth in the upper jaw and the outer row in the 8. Grasscichthtjs is not the only minute kneriid lower jaw. Usually the number of cusps in these in Africa. Cromcria nilotica, which attains a maxi- teeth is from three to five, but there are forms mum standard length of about 28 millimeters, oc- with as many as 20 cusps on each tooth. Some of curs in the main courses of the White Nile, Upper the South American forms have conical teeth on Niger, Mayo-Kebbi, and Volta rivers. It evidently the maxillary but these are usually small and ap- in Be- spends much of its time buried the sand. parently with little or no functional significance; cause of its larval appearance, it was suggested at the African forms invariably lack maxillary teetli. time that Cwmerki be the of an the dentition in most of the one might young Poll ( 1967b ) figured kneriids have undescribed species, but no otlier Congo species. In very general terms, die spe- ever been reported from any of the basins in cies with few cusps probably tend towards in- it lives. which sectivory; those with intermediate numbers of 9. The characoid is remarkable in Agoniates cusps to mixed insectivory-herbivory; and those on otlier fishes when it is very small. On feeding with the most cusps to herbivory. October 15, 1968, five specimens about 1-1 ¥2 inches long were caught in the Rio Solimoes above Florianopohs. The stomach of each one was bulg- LITERATURE CITED ing with two or three cichUd young 5-8 mm long. H. W. 1892. The Naturalist on the River Young Agoniates have rapacious dentition similar Bates, Amazon. New Lxxxix 395 to that of adults. At this size most predaceous York, Appleton. + characoids (e.g., Hydrocijnus) are probably in- pp. sectivorous. Agoniates grows to about 150 mm. Beebe, W. 1945. Vertebrate fauna of a tropical 10. Anodus and Eigenmannina, constituting the dry season mud-hole. Zoologica (New York), 2 subfamily Anodontinae, are placed in the family 30(Pt. 2): 81-88, pis. Curinatidae by Eigenmann and Eigenmann Bell-Cross, G. 1965. Movement of fish across (1889), Regan (1911), and Greenwood et al, the Congo-Zambesi watershed in the Mwini- (1966). This relationship, if true, is of consider- lunga District of Nortliern Rhodesia. Proc. able interest because in other Curimatidae gill Central African Sci. Med. Congress 1963. reduced. All or almost rakers are absent or greatly BoHLKE, J. 1953. A minute new herring4ike all of the species are deposit feeders with highly characid fish genus adapted for plankton convoluted intestines. Observations of mine indi- feeding, from the Rio Negro. Stanford Ichth. cate that Anodus is closely related to Hemiodtis. Bull., 5(1): 168-170. of the famihes of The relationships Hemiodontidae, Branner, J. C. 1884. The 'pororoca,' or bore, Curimatidae, and Prochilodontidae to each other the Amazon. Science (Cambridge, Mass.), 4 and to other characoid families are worth study. (95): 488-492. small- to medium-sized Ama- 11. Most of the Breder, C. M., and D. E. Rosen, 1966. Modes 140 zonian Tetragonopterinae (roughly species) of Reproduction in Fishes. Garden City, New feed on whatever comes in their probably way York, Natural History Press. 941 pp. from in- (McConnell, 1969), benefiting greatly of North Briggs J. C. 1970. A faunal history the sects into the water or emerging from falling Atlantic Ocean. Syst. Zool., 19(1): 19-34. aquatic larvae. The medium- to small-sized African Carter, G. S. 1935. Respiratory adaptations of characids allied to Alestes (about 40 species in the fishes of the forest water, witli descrip- the Congo) probably have similar feeding habits. tions of the accessory respiratory organs of In a remarkable instance of the com- parallelism, Electrophoms electricus (Linn.) and Ple- dentition of many small species of South plex costomus plccostomus (Linn.). J. Linn. Soc. American a high Tetragonopterinae (comprising (Zool.), 39(265): 219-233. of species in the five largest genera of The proportion Carter, G. S., and L. C. Beadle. 1931. Characidae) is practically identical to that in fauna of the swamps of the Paraguayan small species of African Alestiinae. In these small Chaco in relation to its environment. — II. Linn. characids of both continents, there are typically Respiratorv adaptations in the fishes. J. 19-23. two rows of teeth in the upper jaw and two rows Soc. (Zooi.), 37(252): .327-367, pis. G. M. 1952. Notas de ao in the lower jaw; in the upper jaw the outer row Carvalho, J. viagem Amazon and Congo Fish Ecology • Roberts 145

. nouvelle famille de Rio Negro. Miis. Nac. Rio de Janeiro ( not 1964. Une poissons seen; reference cited by Sternberg, 1969). dulca(iuicoles africains: les Griisseichthyidae. CoLLETTE, B. B. 1966a. A review of the venom- C. R. Acad. Sci. Paris, 259: 4805-4807. ons toadfishes, subfamily Thalassophryni- GiBBS, R. J. 1970. Mechanisms controlling world nae. Copeia, 1966(4): 846-864. water chemistry. Science, 170: 1088-1090. P. 1963. Le milieu et I'ecolo- . 1966b. Bclouion, a new genus of fresh- GossE, J. aquatique water needlefishes from South America. gie des poissons dans la region de Yangaml)i. American Mus. Novitates, No. 2274: 1-22. Ann. Mus. Roy. Africjue Centrale (Zool.), No. 116: 113-270, pis. 1-10. Daget, J. 1967. Le genre Ichthyhonis ( Poissons, P. D. E. S. U. Wkitz- Characifonnes). Cah. ORSTOM (ser. hy- Greenwood, H., Rosex, AXD G. S. Myers. 1966. drobiol.), 1(1-4): 139-154. MAX, Pliyletic studies of teleostean fishes, with a provisional Delco, E. a., Jr. 1960. Sound discrimination by classification of living forms. Bull. Amer. males of two cyprinid fishes. Te.xas J. Sci., 339-455. 12: 48-54. Mus. Nat. Hist, 131(4): Gudger, E. W. 19.30. The candirii, the only verte- Du Bois-Reymond, E. 1882. Ueber die Fort- brate parasite of man. New York, Paul B. pflanzung des Zitteraales (Gtjmnoius clcc- Hoeber, Inc. xvii + 120 pp. ( reprinted with tricus). Archiv fiir Anatomic und Physiologic additions and corrections, from Amer. J. (Leipzig), Physiologische Abtheilung, 1882: Vol. 8, Nos. 1 and 2). 76-80. Surgery, Hasemax, J. D. 1911. Descriptions of some new Eigenmann, C. H. 1918. The Pygidiidac, a fam- species of fLshes and miscellaneous notes on of Soutli American catfishes. Mem. Car- ily others obtained during the expedition of the negie Mus., 7(5): 259-398, pis. 36-56. Carnegie Museum to central South America. . 1922. The fishes of western South Amer- Ann. Carnegie Mus., 7(3-4): 315-328, pis. ica, part I. Mem. Carnegie Mus., 9(1): 1- 46-52. 346, 35 pis., 1 map. . 1912. Some factors of geographical dis- revision -, AND R. S. Eigexmaxx. 1889. A tribution in Soutli America. Ann. New York of the edentulous Curimatinae. Ann. New Acad. Sci., 22: 1-112. York Acad. 1-32. Sci., 4(18): Kelley, W. E., and J. W. Atz. 1964. A pygidiid Ellis, M. M. 1913. The gymnotid eels of tropical catfish that can suck blood from goldfish. America. Mem. Carnegie Mus., 6(3): 109- Copeia, 1964(4): 702-704. 204, pis. 15-23. Kleerekoper, H., axd E. C. Chagxox. 1954.

FiTTKAU, E. J. 1964. Remarks on limnology of Hearing in fish, with special reference to

Central-American rain-forest streams. Ver. Sernotihis atwmaculatus atwmaculatus ( Mit-

Intemat. Verein. Limnol., 15: 1092-1096. chill). J. Fish. Res. Bd. Canada, 11: 130- 152. . 1967. On the ecology of Amazonian rain- forest streams. Atas do Simposio sobre a Klinge, H. 1967. Podzol soils: a source of black-

: in Atas do Biota Amazonica, 3 ( Limnologia ) 97-108. water rivers Amazonia. Simposio

sobre a Biota 3 : Freihofer, W. C, axd E. H. Neil. 1967. Com- Amazonica, ( Limnologia ) mensalism between midge larvae (Diptera: 117-125. Chironomidae) and catfishes of the families Knoppel, H. 1970. Food of Central Amazonian Astroblepidae and Loricariidae. Copeia, fishes, contribution to the nutrient-ecology of 1967(1): 39-45. Amazonian rain-forest-streams. Amazoniana 257-352. Gadow, H. 1909. Amphibia and Reptiles. Cam- (Kiel), 2(3): bridge Natural History (London, Macmillan), Lewis, W. M., Jr. 1970. Moiphological adapta- Vol. 8. viii + 668 pp. tions of cyprinodontoids for inhabiting oxygen 319-326. Geisler, R. 1969. Untersuchungen iiber den deficient waters. Copeia, 1970(2): functional anat- Sauerstoffgehalt, den biochemischen Sauer- LiEM, K. F. 1970. Comparative stoffl:)edarf und den Sauerstoff\erbrauch \'on omy of the Nandidae (Pisces: Teleostei). Fischen in einem tropischen Schwarzwasser Fieldiana (Zool.), 56: 1-166. (Rio Negro, Amazonien, Brasilien). Arch. LissMAXX, H. W. 1958. On the function and evo- Hydrobiol., 6(3): 307-325. lution of electric organs in fish. J. Exp. Biol., 156-191. Gery, J. 1960. New Chcirodontinae from French 35(1): Guiana. Sencken])ergiana Biol., 41(1 and Marlier, G. 1967. Hydrobiology in the Amazon 2): 15-39, pi. 2. region. Atas do Simposio sobre a Biota Ama- 3 1-7. . 1963. Contributions a I'etude des poissons zonica, (Limnologia): characoides - 27: Systematique et exolution M.A.TTHES, H. 1961. Feeding habit of some de quelques piranhas (Serrasahmis). \'ie et Central African freshwater fishes. Nature Milieu, 14(3): 597-617. (London), 192(4797): 78-80. 146 Bulletin Museum of Comparative Zoology, Vol. 143, No. 2

— . 1963. A comparative study of the feeding fauna of Central America. Copeia, 1966(4): mechanisms of some African Cyprinidae 766-773.

(Pisces, Cypriniformes ) . Bijdr. Dierk., 33: -, AND T. R. Roberts. 1967. Note on the 3-35, 12 pis. dentition of Creagrudite maxillaris, a characid 1964. Les poissons du lac Tuml^a et de fish from the Upper Orinoco-Upper Rio la region d'Ikela. Ann. Mus. Roy. Afriqiie Negro system. Stanford Ichth. Bull., 8(4): Centrale (Zool.), No. 126: 1-204, 2 maps, 248-249. 6 pis. Myers, J. G. 1935. Epizootics among fishes and

McCoNNELL, R. H. 1964. The fishes of the reptiles on the Amazon and Orinoco. J. Rupununi savanna district of British Guiana, Animal Ecol., 4(1): 17-21.

South America, Part I. Ecological groupings Nichols, J. T., and L. Griscom. 1917. Fresh- of fish species and effects of the seasonal water fishes of the Congo basin obtained by cycle on the fish. J. Linn. Soc. (Zool.), 45 the American Museum Congo Expedition, (304): 103-144. 1909-1915. Bull. American Mus. Nat. Hist.,

. 1969. Speciation in tropical freshwater 37: 653-756, pis. 64-83. fishes. Biol. J. Linn. Soc. 1: 51-75. Nieuwenhuizen, a. van den. 1964. Tropical MoYNiHAN, M. 1968. Social mimicry; character Aquarium Fish, Their Habits and Breeding convergence versus character displacement. Behavior. Princeton, New Jersey, Van No- Evolution, 22(2): 315-331. strand. 200 pp. MouLTON, J. M., AND R. H. DixoN. 1967. Di- Pfeiffer, W. 1963. Vergleichende Unter- rectional hearing in fishes, hi Marine Bio- suchungen iiber die Schreckreaktion und den

acoustics, 2 : 187-232. Schreckstoff der Ostariophysen. Z. vergl. Myers, G. S. 1927. Descriptions of new South Physiol., 47: 111-147.

American fresh-water fishes collected by Dr. . 1967. Sclireckreaktion und Schreckstoff- Carl Ternetz. Bull. Mus. Comp. Zool., 68 zellen bei Kneriidae und Phractolaemidae

(3): 105-135. ( Isospondyli ) . ( Pisces ) . Naturwissenschaf-

. 1944. Two extraordinary new blind ten, 54(7): 177. nematognath fishes from the Rio Negro, Poll, M. 1957. Les genres des poissons d'eau representing a new subfamily of Pygidiidae, douce de I'Afrique. Brussels, La Direction de with a rearrangement of the genera of the rAgriculturc, des Forets et de I'filevage. 191 family, and illustrations of some previously pp.

described genera and species from Vene- . 1959. Recherches sur la faune ichthyo- zuela and Brazil. Proc. California Acad. Sci., logique de la region du Stanley Pool. Ann. 23(40): 591-602, pis. 52-56. Mus. Roy. Congo Beige, Tervuren, ser. oct. 1 . 1947. The Amazon and its fishes. Part 1. (Zool.), 71: 75-174, pis. 12-25, map.

The river. The Aquarium lournal, 18(3): . 1964. Une famille dulcicole nouvelle de 4-9; Part 2. The fishes. Ibid, 18(4): 13-20; poissons africains: les Congotlirissidae. Acad. Part 3. Amazonian aquarium fishes. Ibid, Roy. Sci. Outre-Mer, sci. nat. med., (N. S.), 18(5): 6-13; Part 4. The fish in its environ- 15(2): 1-40, 8 pis.

ment. Ibid, 18(7): 8-19, 34. . 1966. Genre et espece nouveaux de fleuve en de . 1949a. The Amazon and its fishes. Part Bagridae du Congo region 5. A monograph on the piranha. The Aquar- Leopoldville. Rev. Zool. Bot. Afr., 74(3 and ium Journal, 20(2): 52-61 and 20(3): 4): 425-428.

76-85. . 1967a. Contribution a la faune ichthyo-

. 1949b. Salt-tolerance of freshwater fish logique de I'Angola. Museu de Dundo Pubh- groups in relation to zoogeographical prob- cagdes Culturais, No. 75: 1-381.

lems. Bijdr. Dierk. (Leiden), 28: 315-322. . 1967b. Revision des Characidae nains

. 1952. Sharks and sawfishes in the Ama- africains. Ann. Mus. Roy. Afrique Centrale. zon. Copeia, 1952(4): 268-269. No. 162: 1-158.

. a des . 1960a. The genera and ecological ge- 1969. Contribution la connaissance ography of the South American banjo cat- Parakneria. Rev. Zool. Bot. Africaines, 80 fishes, family Aspredinidae. Stanford Ichtii. (3-4): 359-368. 132-139. a Bull., 7(4): , AND J. P. Gosse. 1963. Contribution

. 1960b. The South American characid I'etude systematique de la faune ichthyologi- genera Exodon, Gnathoplax, and Roeboexo- que du Congo Central. Ann. Mus. Roy. Afri- don, with notes on the ecology and que Centrale, No. 116: 43-110, 4 pis. fishes. 7 of characid Stanford Ichth. Bull., -, P. J. P. Whitehead, and A. J. Hopson. (4): 206-211. 1965. A new genus and species of clupeoid

. 1966. Derivation of the freshwater fish fish from West Africa. Acad. Roy. Belgique, Amazon- ant> Congo Ftsii Ecology • Roberts 147

sci. (5th sen), 51(3): 277-292. Sc:nw.\ssMANN, H. O. 1967. Orientation of Ama- Preston, F. W. 1962. The canonical distribution zonian fishes to the equatorial sun. Atas do of commonness and raiit\': Part I, Ecology, Simposio sobre a Biota Amazonica, 3 43: 185-215. Part II, Ecology, 43: 410-4.32. (Limnologia); 201-220. Regan, C. T. 1911. The classification of the SioLi, H. 1964. General features of the limno- teleostean fishes of the order Ostariophysi. — logy of Amazonia. Verb. Internat. Verein. 1. Cvprinoidea. Ann. Mag. Nat. Hist., ser. Limnol., 15: 1053-1058.

8,8:' 13-32, pi. 2. . 1967. Studies in Amazonian waters. Atas Roberts, T. R. 1967. Tooth formation and re- do Simposio sobre a Biota Amazonica, 3

in characoid fishes. Stanford Ichth. : placement ( Limnologia ) 9-50.

Bnll., 8(4); 231-247. . 1968. Principal biotopes of primary pro-

. 1968. Wieoglani.s dendrophorus and ductitni in the waters of Amazonia. In Misra, ZaircichiJiys zonatus, bagrid catfishes from R., and B. Copal (eds.), Proc. Symp. Recent the lower rapids of the Congo Ri\er. Ichthyo- Adv. Trop. Ecol., 1968: 591-600. logica, the Aquarium Journal, 39(3-4): Spruce, R. 1908. Notes of a botanist on the

119-131. Amazon and Andes ... during the years

. 1969. Osteology and relationships of 1849-64. (Edited by A. R. Wallace.) Lon- characoid fishes, particularb' tlie genera Hep- don, Macmillan. \'ol. I. 486 pp., 1 map. settis, Salminu.s, HopJias, Ctenolucius, and Sternberg, H. O'R. 1969. Man and environ- Accstrorfu/ncluis. Proc. California Acad. Sci., mental change in South America. In Fittkau, 36(15): 391-500. E. J. et al. (ed.), Biogeography and Ecology

. 1970a. Description, osteolog>' and re- in South America. The Hague, Junk. Vol. 1. lationships of the Amazonian c\"prinodont Pp. 413-445. fish FluvipJiylax pygmacus (Myers and Car- Stoye, F. H. 1935. Tropical Fishes for the valho). Breviora, No. 347: 1-28. Home, Their Care and Propagation (2nd 1970b. Scale-eating American characoid ed.). New York, Carl Mertens. 284 pp. fishes, with special reference to Probolodus TiiYs VAN den Audenaerde, D. F. E. 1969. De- heterostomits. Proc. California Acad. Sci., 38 scription of a new genus and species of clupe- (20): 383-390. oid fish from Sierra Leone. Rew Zool. Bot. ScHOMBURGK, R. H. 1931. Travels in Guiana and Africaines 80(3 and 4): 385-390.

on the Orinoco during the years 1835-39. . .. Weitzman, S. H. 1960. The systematic position

(edited by O. A. Schombiirgk ) . Georgetown, of Piton's presumed characid fishes from the Walter E. Roth translation (not seen; refer- Eocene of central France. Stanford Ichth. ence cited by J. G. Myers, 1935). BuU., 7(4): 114-123.