Ecological Divergence and Convergence in Freshwater Fishes

Home , Gasterosteiformes, Hydrocynus vittatus, Synodontis woosnami, Tropical fish

KIRK O. WINEMILLER . IOLOGISTSAND POUCYMAKERSARE RAONG the clock to document EcologicalDivergence and preservebiological diversity in the faceof a growing glob- and Convergencein al human population and alterationsof naturalhabitats. Given FreshwaterFishes the magnitude and rapid pace of changesto natural ecosys- , biologists are challengedwith the taSkof identifying essentialfeatures in the organizationof biological communi- Strong relationships between basic ties. Notwithstanding the inherent uniquenessof different geographical fonn and ecologicalfunction in fISh- regionsand the evolutionaryhistories of their flora and fauna,the recogni- es allow the comparative study of tion of commonfeatures among communities and the speciesthat comprise ecological relationships using mor- them is sometimespossible. This study contributesto our understandingof phological features. Within-fauna biodiversity by further examining the ecologicalcharacteristics of natural variation and between-fauna simi- freshwaterfish communitiesfrom around the world. larities of ecomorphological charac- One of ecology'sgreatest challenges is to explain patternsof regional teristics of freshwater fish assem- speciesdiversity as well as the ecological,morphological, and behavioral blagesfrom five regions around the variation among and within populations. Patternsof ecologicaland mor- world were compared and contrast- phological diversification are of interest to both ecologistsand phyloge- ed. Ecomorphologicalfeatures netic systematists.Systematists use interspecificvariation as the raw mate- allowed identification of a number rial for constructing phylogenies. Ecologists would like to know if of ecologically convergent spedes interspecificvariation in a given trait arisesrandomly during evolutionary from different regions. Based on a divergencefollowing speciation, or whether this variation is correlated relative scale of convergence,tropi- with ecological function. This difficult task is further complicated by cal fish assemblagesexhibited a unique events.Traits adaptivein a particular ecosystemmay be neutral or greater degree of ecological conver- maladaptive in a new environment. If a maladaptivetrait is genetically gence than temperate assemblages. correlatedwith one or more strongly adaptivetraits, then it might persist The ecomorphological comparison in the population for many generations.In one form or another, evolu- supports the view that biotic inter- tionary biologists and ecologistsseek to identify, interpret, and model pat- actions have a greater relative terns of functional (adaptive) variation.21Similarities in patterns of indi- influence on evolution in the tropics vidual, population, and community-levelvariation amongdifferent faunas compared with temperate regions. provide evidence that, in many cases,the same deterministic processes organizebiological systems. Biological convergenceis the independentevolution of the samefea- ture (physiological,morphological, ecological,etc.) within two divergent Figure1. phylogenetic lineages. Divergence is the evolution of dissimilar traits The body shape.posidon of fins. and between closely related lineages.Remarkable examples of morphological nwuth structure of afish influenceits eco- and ecological divergence are well-known in vinually all higher taxa. logicalperformance. Trout possessa gen- Examples from the independent adaptive radiation among marsupial eralited nwrphologythat allows them to mammals of Australia and eutherian mammals of the other continents usethe endre water columnand a wide variety offood items in habitats that con- appearin most introductory biology textbooks.Convergence of ecological tain reladvelyfew compedngspecies. niches is seenin rodents and marsupial mice; shrewsand marsupialinsec- PAlNTING BY WALTER A. WEBER Q NGS tivores; and eutherian canines and marsupial wolves. The bony fishes

308 NATIONAL GEOGRAPHIC RESEARCH &. EXPLORAllON 8(3):308-327; 1992 I am preparing a report on (Osteichthys) are a much older and speciosephylogenetic lineage and as patterns of resource utiliza- might be expected,exhibit numerous examplesof evolutionary convergence throughout the world (Figure 1). Within the Characiformes, tion and guild organization Curimata (=Steindachnerina) argentea is an epibenthic mud-feeder; among diversefish assem- Hydrocynusvittatus is a large, roving midwater piscivore;Metynnis argen- blagesin Central and teus is a midwater planktivore/granivore;and Charaddiumfasciatum is a South America, North small, bottom-resting invertebrate-picker.Within the PerCidae,Etheostoma chlorosomumis a small, bottom-resting invertebrate-picker; and Perca America and Africa. In the flavescensis an epibenthic and mid water insectivore/piscivore.A remark- interest of communicating able convergence of body form and ecological function has occurred with a diverse scientific between the phylogeneticallydistant Characidiumand Etheostomaspecies audience,I have attempted (cylindrical body with flattened ventrum, broad pectoral fins, pelvic fins to make the analysis fairly locatedanteriorly; subtenninal mouth, etc.). Despitethe number of instancesof remarkableconvergence that can be straightforward. This is identified within most large taxonomic groups (Figure 3), the conver- sometimesdifficult to gencephenomenon is difficult to study quantitatively;This is especiallyso achievewith numerical for small, closely related groupings. Our ability to recognizeevolutionary analysesaimed at revealing convergenceis entirely dependenton knowledgeof phylogeneticrelation- ships. Becausewe usually have more confidence in our phylogenetic temporal trends within groupings at broader taxonomic scales, ecological convergencecan be speciesassemblages con- identified with greaterease. Within smaller and more recent phylogenetic taining 30 to 70 species. lineages,it is often difficult to distinguish divergenttraits from convergent CORRESPONDENCEtraits. In fact, a great deal of the methodology of modern phylogenetics seeks to identify the suite of traits that is uniquely divergent versus the suite that is convergent.2Phylogenies are basedon the information con- tained in uniquely divergent (derived) characteristics.Convergent traits are seen as potentially confounding characteristicsin phylogentic con- structions. Among closely related taxa, the distinction between derived and convergent traits is sometimesunclear and only becomesapparent after phylogenetichypotheses have been generatedand tested. Here I present results from a comparative study of morphology and ecologyof freshwaterfish assemblagesfrom five widely separatedregions around the world. Study sitesspan a latitudinal gradientfrom Alaska (650 N) to the neotropics (90 N), and a tropical longitudinal gradient from Central America (830W) to Africa (230 E). Like many taxonomic groups, freshwater fIShesshow a strong latitudinal gradient in speciesrichness, with the highest regional diversity appearing in the neotropics. R. H. Lowe-McConneW5counted> 1300 freshwaterfish speciesfor the Amazon River basin comparedwith 192 for all of Europe.Since that time, new dis- coverieshave causedmost estimatesfor South American fIShesto more than double, while the record for European speciesremains essentially unchanged. Whereas climatic and biogeographicalhistory are probably the primary underlying agents that produce global patterns of regional diversity;the basicmechanisms that foster coexistenceof largenumbers of speciesin tropical faunasare open to debate.The extent to which assemblagesfrom different regions are organizedin the samemanner indicates the degreeto which detenninistic processesregulate the populations that comprisecommunities.33 Moreover, the relativeamount of ecologicalcon- vergenceamong taxa from different regionsshould indicate the strength of KIRK O. WINEM1U.ER, assistant professor. Departmentof Wildlife and Fisheries similar forms of directional selection in relation to random evolutionary Sciences.Texas A&M University. College changeand evolutionary stasisdue to geneticconstramts. Here I contrast Station.TX 77843-2258. patterns of morphological divergencewithin the five local fISh assem-

310 RESEARCH &. EXPLORAnON 8(3); 1992 Curimata Hydrocynus Metynn;s Characidium Etheostoma ~

Figure 2. Phylogenyshowing evolutionary diver- gencr.within chtuadform and perdform fishes, as well as ecologicalconvergence betweenthe lebiasinidgenus of small, bottom-resting.invertebrate-pickers, Cbaracidium (C. fasciatum,South AmtTica),and the percidgenus of small, blages. A substantial literature illustrates the correlations between a num- bottom-resting.invertebrate-pickers, ber of basic morphological features of fishes and ecological performance Etheostoma(E. chlorosomum,North (summarized in P:W Webb21),and morphology has been used successful- America). ly as a surrogate for ecology.6-8.11,20.26.28.32In addition, I compare the relative degree of ecomorphological convergence between fish species from different regional faunas.

Methods

SAMPLE COLLECTION Each of five study sites was sampledon at least two separatedates, and most locations were visited numerous times as part of long-term field studies.29-31Only numerically dominant specieswere used for the morphological analysis. Dominant speciesare defined in this study as the most common speciesat a site that (by number of individuals), when summedby order of rank. formed 99% of all individual fIShescollected over the full samplii1gperiod. Rare specieswere defined as those com- prising the 1% tail of smallest relative abundance contributing to the numerical total for a site. Whereasthis criterion for assemblagemember- ship could exclude a few rare, but perhapsstrongly interactive, community elements,it was designedto eliminate most of the rare fugitive and transientspecies that were essentiallynon-interactive componentsof the coreassemblage (Table 1). Two types of lowland aquatic environments-streams and backwaters-were sampledin eachof the following study regions:Alaska (North America), Texas (North America), Costa Rica (Central America), Venezuela(South America), and Zambia (Africa). This paper compares assemblagepatterns of morphological and ecological divergenceamong

ECOMORPHOLOGY OF FRESHWATER FISHES : WINEMILLER 311 FAMYL Y (commonname) F AMIL Y (commonname)

SEWARD PENINsUlA. ALASKA WESTERNPROVINCE, ZAMBIA RIVERS BACKWATERS TOTAL CREEK BACKWATERS TOTAL spp. spp. spp. MORMYRJDAE (dcpb8nt aoscs) 1 4 4 SALMONIDAE (salmon.whitefishes) 6 2 6 HEPSErIDAE (African pike) 0 1 1 EsocIDAE (pikes) 0 1 1 CHARAOOAE (teuu) 1 2 3 UMBRIDAE (bI8ckIIshes) 0 1 1 CITHARIN1DAE (A&fcan danas) 2 0 2 CornDAE (sc:uIpiDS) 1 1 1 CYPRINIDAE (minnows) 6 6 8 GASTEROSTlIDAE (sdckIebIcks) 0 1 1 ScHIlBElDAE (calfishes) 0 1. 1 ClARIIDAE (W8!kiDgcaIfisba) 0 No. GENERA ~ 6 8 MOCHOKIDAE ~ cat6shcs) 0 2 ) No. DoMINANT SPEaES 7 6 10 CYPRINOOONTIDAE (killi6sbes) 4 4 4: No. RARESPEaES 0 0 0 ClOlUDAE (dchIids) 4 10 11 ANABANTIDAE (climbing perches) 0 1 1 NEWlON COUNIY, TEXAS MASTACEMBAE1.1DAE (spiny eels) 0 1 1 CREEK BAyOU TOTAL No. GENERA 8 19 22 No. DoMINANT SPECIES 18 33 39 PETROMYZONIDAE (lamprey) 0 1 1 No. RARESPECIES H 32 35 CoUPEIDAE (sb8ds) 0 1 1 ESOClDAE (pikes) 0 1 1 CYPRINIDAE (miDnOWS) 6 6 9 ESTADOPoRTUGUESA. VENEZUELA CATOSTOMlDAE (suckers) '1 2 '2 CREEK CAAo TOTAL ICTALURIDAE (at6sbcs) 0 2 2 APHREDODERIDAE (pirate perds) 1 1 1 ERYTHRlNIDAE(guavbI85) 1 1 1 CYPRlNOOONTIDAE (kiUIfishes) 1 1 2 CHARACIDIIDAE(5. Amer. danen) 0 1 1

POEOllIDAE (liftbearen) 1 I i.. .1 lEBlAS1N1DAECguaYiDas) 1 1 2 ATHERINIDAE (sIlYasjdcs) 0 1 1 CURlMAnDAE (curimatas) 0 1 1 CENTRARanoAE (sunfishes) 3 6 7 PROCHILOOONTIDAE(aJPOlOS) 0 1 1 PERaDAE (dartas) 1 3 3 CHARACIDAE(1eUIS) 6 12, 16 GASTEROPELECIDAE(haldletllshes) 0 1 1 No. GENERA 8 18 19 GYMNOnDAE (knIfe8shes) 0 1 1 No. DoMINANT SPEOES 13 26 31 STERNOPYGlDAE(knIfc6sbcs) 0 1 1 No. RARE SPECIES 0 0 0 PlMaODwAE (catBsba) 0 _3 3 AUCHENlPTERIDAE(catBsba) 0 1 1 TORTUGUEROPARK, COSTA RICA TRlQK)MYCIERlDAE(~cat&sbes) 0 1 1 AsPREDINlDAE(I.Djo catfishes) 0 i 1 CREEK CAAo TOtAL UwanHYIDAE (amIOredcalfBhes) 0 . "4 CHARAODAE (1£1ras) 1 3 3 LoRlCARllDAE (amIOredcadlshes) 2 5, 6 PIMB.oDlDAE (at6sbes) 1 0 1 CYPRlNOOONTIDAE(kIIIIftsbcs) 0 1 1 CYPRlNOOONTIDAE(kjIIIIIsba) '1 0 1. POEOUIDAE (II~) 1 1 1 POEauIDAE (ltYebams) 3 5 ~ ClanJDAE (dcbJIds) 2 5 6 ATHERINlDAE~) 0 1 ".Lc 1 SYNBRANOUDAE(swamp eels) 0 1 1 SYNGNATHlDAE(pipe6sbcs) 0 1 1 No. GENERA 12 40 46 POMADASYIDAEGOBIIDAE(8I)IIies) 1 2 2 SYNBRANCHIDAE<-DIP eck) 0 1 1 SoLElDAE (soles) 0 . 2 2 BonDDAE (I»undas) 0 1 1

No. GENERA 11 23 25 No. DoMINANT SPEaES J3 30 32 No. RARESPEaES 9 28 1

For . complete list of the domI1Iant spedtS at each site and their IaXOIIOmIc aIIlliaIions. see WInemillcr on

312 RESEARCH& EXPLORAnON 8(3); 1992 backwater sites onl~ A statistical analysis of patterns of ecomorphological I am not presenting a dispersion in both backwater and channel assemblages is presented in detailed analysis of the Winemiller.32 All of the backwater environments had soft substrates , abundant aquatic vegetation, and minimal or no flow during periods of niche. We used nets to low precipitation. A bayou assemblage(Big Cow Bayou) within the Sabine sample within-habitat types River drainage of Texas was sampled inJune 1953 andjune 1986 by Clark (open-water vs vegetation Hubbs and colleagues of the University of Texas at Austin. I sampled a mats) and pool zones swamp (Caflo Maraca) assemblage in the western llanos region (Rio Apure drainage) of estado Portuguesa, Venezuela, throughout 1984 and a (midpool vs pool edge). I swamp/backwater (Caflo Agua Fda Viejo) assemblage of Tonuguero directly observed and noted National Park in the Caribbean lowlands of Costa Rica throughout 1985 the height at which most (Rio Tortuguero drainage). From May to December 1989, fishes were sam- individuals of each spedes pled from backwaters of the upper Zambezi River floodplain (Barotse Plain) in Zambia's Western Province. During June and July 1990, fISh swam in the water column. assemblageswere sampled from the Nuikluk and Fish River dminages of These records allow me Alaska's Sewald Peninsula. Using concerted efforts to ensure that common to categorize spedes by species were well represented and that few rare species escaped capture, predominant habitat with fIShes were collected by nets and hook and line. Adult size-classes of salmon (Oncorhynchusketa and o. gorbuscha) were collected in Alaska but no difficulty. are not included in this analysis because they had entered the stteams for CORRESAJNDENCE spawning and nonnally feed in the marine environment As opposed to adult salmon, which enter streams only to spawn and die, adult char (Salvt'linus malma) and juvenile sahnon can be considered normal components of coastal stream food-webs. Juvenile Oncomynchuswere obselVed in the stteams, but were not collected becauseof their special protected status. I made morphological measwements on juvenile specimens of o. gorbuscha and O. kisutch from British Columbia (Canada) s~ MORPHOLOGICAL TRAITS Morphological measurements were made on specimens deposited in the Texas Natural History Collection of the Texas Memorial Museum, Austin. Three specimens corresponding to adult size-classes (except for Oncorhynchus spp.) were measured (to nearest 0.1 mm with vernier calipers, and to the nearest 1 mm with a clear plastic ruler for measures >130.0 mm), and values were later recorded as arithmetic means for each trait and each species. The following morphological features were measured or estimated (detailed explanation of measurements appear in Wmemiller32): 1) standard length, 2) maximum body depth, 3) maximum body width, 4) caudal peduncle length, 5) caudal peduncle depth, 6) caudal peduncle width, 7) body depth below midline, 8) head length, 9) head depth, 10) eye position, 11) eye diameter, 12) mouth position, 13) mouth width, 14) mouth height, 15) snout length shut, 16) snout length open, 17) dorsal fin height, 18) dorsal fin length, 19) pectoral fin length, 20) pectoral fin height, 21) caudal fin length, 22) caudal fin height, 23) pelvic fin length, 24) anal fin height, 25) anal fin length, 26) pigment code, 27) tooth shape, 28) gill raker, 29) gut length, 30) swim bladder length. Body length in the multivariate morphological analysis was the maximum standard length recorded for each species at each site (Figure 6). Other linear distance measurements were convened to ratios for use as components of body, head. and fin shape. Ratios employed in the analysis were chosen based on earlier functional interpretations7.27and were standardized to yield equal variances before use in comparisons. Standardized ratios served as relatively size-independent dimensions of shape that have

ECOMORPHOLOGYOF FRESHWATERFISHES : WINEMILLER 313

Several studies havefailed straightforward functional interpretations (e.g., relative body height and to detect a strong relation- width influence a fish'scapability to remain stablein the midwater column and to turn sharply;versus its ability to rest on the substratein an upright ship between mo1phologi- position or slither through densemats of vegetation).Distances in the ver- cal traits and ecological tical dimension were convenedto componentsof shapeusing body depth performance in fishes and as the denominator (e.g.,mouth height/bodydepth). Body width was the birds. Becauseresources denominator for ratios involving pedunclewidth and mouth width; head length was the denominator for the ratio of eye diameter and snout vary over time, measures length; and head depth was the denominator for eye position. Snout of average resource utiliza- length open was divided by snout length to provide a measureof relative tion recorded over a long mouth protrusibility. The remaininghorizontal linear measurementswere period can result in under- divided by standardlength for useas componentsof shape. estimation of ecological DATA ANALYSIS Data were transformed to adjust for the influence of differential scale on specialization. distancemeasures. Calculations were performedon transformedmorpho- CORRESPONDENCE logical data sets approximating a normal distribution, with a mean of 0 and a standarddeviation of 1. Principal componentsanalysis was used for inter-assemblagecomparisons of species,morphological variation, as well as for identification of patternsof covarianceamong morphological char- acteristicsrelating to ecology.Principal componentsanalysis is a form of multivariate analysisthat producesindependent orthogonal axes (principal axes) from a multivariate cluster of data points. The first severalcom- ponents (usually the first three or four) identified in the analysisusually model a major portion of the variation amongoriginal variables.Principal componentsanalysis was performedon the 160-speciesdata set basedon the correlation matrix of standardizedmorphological variables. For the entire 160-speciesdata set, a matrix of ecomorphologicaldis- similarities (Euclidean distances)between speciespairs was calculated basedon the formula: n Djk = ~L (Xj(i) - XkO»2] 1/2 1=1 where n is the number of attributes,and XjO) and XkO) are standardized values of the sameattribute for the speciesj and k. From each matrix of Euclidean distances,I ranked eachspecies' nearest neighbors from closest (NN1) to most distant (NN1S9). I alsoranked eachspecies' nearest neighbors iti terms of their phylogeneticaffiliations. CONVERGENCE TEST No definitive phylogeneticscheme spans the taxonomicbreadth of species employedin this study;Figures 2&4 illustrate a hypothesizedphylogenet- ic relationship for the taxa in the dataset derivedfrom the 10 assemblages (i.e., both backwater and channel assemblagesfrom each of the five regions). The relationships of higher taxa are based primarily on J. S. Nelson,is except for families within the Characiformes and Gymnotiformes, which follow F: Mago17for neotropical families and J. Gery9 for African characifonns. Assignment of genera within families and specieswithin generafollows Mago,17D. S. Lee and coworkers,Hand G. Bell-Crossand J. L. Minshull.i For a simple test of the hypothesis of ecomorphologicalconvergence at the specieslevel, I calculateda convergence index for the first- through fifth-nearest ecomorphologicalneigh- bors for each of the 160 species.The convergenceindex was equal to the

316 RESEARCH 6r. EXPLORATION 8(3); 1992 A2natha Petromvzontiformes PeIlomyzontidae (1/1)

Mormyridae (4/4) Gnathostomata - Osteichthyes Clupeidae (1/1) Salmonidae (4/5) Salmoniformes Esocide(1/2) Cyprinodonlidae (1/6)- Cvprinodontiformes Aplocheilidae (2/2) Umbridae (1/1) Percoosiformes Poeclliidae (&17> Atheriniformes Aphredoderldae (1/1) Atherinidae (2/2) Gasterosteiformes I! CvDI'inifDl'Ines Cyprinidae (4/17) UIosDnidae (2/2) Syngnathidae(1/1) SvnRnathiformes OIaraddae (2W22) Curlmatidae (1/1) Svnbranchiformes Characlformes Gasterope\ecidae(1/1) Scorpaeniformes lebiasinidae (3/3) Prochilodontidae (1/1) Cidlarinidae (1/2) Cenb'an:hidae(J/7) PelCklae(3/3) GymfX)tidae (1/1) POOIadasyidae(1/1) StemapYlidae (1/1) Lutjanidae (1/1) PefCiformes Centroponidae (1/1) ktaluridae (1/2) Ochlidae (11/34) PImelodidae (3/4) Anabantidae (1/1) Siluriformes Auchenipteridae (1/1) Eleotridae (3/4) Trichomycteridae (1/1) Aspredinidae (1/1) MastacembelidaeGobiidae (2/2) (1/1) - Callichthyidae (2/4) loricariidae (~) dariidae (1/1) Pleuronectifonnes Schllbeidae (1/1) Mochokidae (1/2) SoIeIdae (2/2) -

number of closely related species that were actually less similar morpho- Figure4. logically to the target species than its morphological nearest neighbor (N) PhylogenyfoT 160fish speciesfrom Alaska, Texas,Costa ~ Vmez:uela, and Zmnbia divided by the total number of possible species pairings (P). P was equal wed in the morphologicalanalysis of ecolog- to 158 for the first nearest neighbor, 157 for the second, and so on. The ical divergenceand convergence(higher convergence index (NIP) was equal to 0 if the nearest neighbor was actu- divisionsbased on NelsonIB). ally the most closely related taxon, and the index equaled 1.0 if the nearest neighbor was the most distantly related taxon in the entire data set. BecauseI did not distinguish degrees of relatednessbetween specieswithin a genus or genera within a family, any bias due to lack of true phylogenetic information was in the direction of no convergence. As a result, the convergence test is very conservative and slightly underestimates the magnitude and frequency of ecological convergences. ECOLOGICAL DATA Information on feeding habits of ttopical fishes came from earlier long- term studies at the Venezuelan, Costa Rican, and Zambian sites.29-321ndrefer- - Wrdn Diet was characterized by volumetric analysis of stomach con- tents based on large samples of formalin-preserved specimens. Those data are summarized here in the form of basic ttophic niches: algivore/detriti- vore, omnivore, invertebrate-feeder, invertebrate/fish-feeder, piscivore. The characterizations of basic trophic niches for Texan and Alaskan fIShes were based in part on information in K. D. Carlander.3.4

Results

ASSEMBLAGE COMPOSITION Each of the study regions is dominated by different families (Table 1). Salmonids(salmon, grayling, whitefish) are the major componentsof the

ECOMORPHOLOGYOF FRESHWATERFISHES: WINEMILLER 317 0.60

0.40

0.20

0.00

0.15

0.10

0.05

0.00

0.16 ~, Figure 5. Distribudon of spedesnumerical 0.12 relative abundanceby rank for five lowland backwaterand .~ 0.08 "to floodplainfish assemblages. "ii 0.04 rC~tiiiWI;i ~ VI .!! 0.00 ~~-;." ::1ig~

V)~ 0.20 :"~':~,'J~{;:!;:,!; ~ilt.~~:j;,: ""C'-""""'-;;;-

-', 0.15 ".u,,~.

0.10

0.05

0.00 0.12 ~~

0.08

0;04

0.00 Species Abundance Rank Alaskanassemblages, and cyprinids (minnows) and centrarchids(sunfishes) dominate Texan assemblages.In the tropics, cichlids and poeciliids Give-bearers)dominate CostaRican assemblages; characids (tetras, piranhas) and catfishes(silurifonns) fonn the largestcomponent of Venezuelan assemblages;and cyprinids and cichlids dominateZambian assemblages. Within a habitat type (channelor backwater),assemblages show a strong latitudinal gradient in speciesdiversity with the highest diversity occur- ring in Venezuela.32Richness at the genusand family levels also showed strong latitudinal gradients among the five regions,but diversity at the ordina1levelwas not associatedwith latitude (Alaska-5 orders,Texas-9 orders,Costa Rica-7 orders,Zambia-6 orders,Venezue1a-6 orders). In a1llD assemblages,distribution of relativeabundance among species was strongly skewedin favor of uncommonspecies (Figure 5). In terms of both individuals and biomass,relatively few speciestended to dominate eachfish assemblage.With few exceptions,numerically dominant species

318 RESEARCH & EXPLORAnON 8(3); 1992

~ 400

200

Figure6. Di.stribudonof spcdes standard lengths by rank for five lowland backwater tmd floodpl4in fish GSStmblagtS (tach spedes standard length is the maximum -=~ ,,=--,- . Iii recorded among all individu- "~~Et6(X) als colltcted at a silt).

~C.!'II}i~"" l~;I: ;;;J~1i: ".J:.)r_;~c" ;;iJii~~.E::":~.~ " 300 :J;~:,,~:I,~:~ I :. "~',!t.-c )~i~?:;JI; :!~*':;~g:::! rj]~~~~~~~ !if, 10 . 8t1 600 1~~:,;~f:;~;':Mar3~ venezuel~;:' I .'t~ '".t , J,:.:I ";ic~1 I I300 I 0 Specieslength Rank

fed near the base of the aquatic food web (e.g., detritivores, algivores, gr.miVOl'ts, omnivores, planktivOl'ts). The predatory nonhero pike (Esax ludus) was the most abundant species in the Alaskan backwater sample. More than for other sites, the relative abundance figures in the Alaskan backwater sample were probably biased to some extent by the collecting methods (dip-netting in shallows, hook and line in depths). The Texas bayou sample was dominated by minnows (Notropis venustus, N. texanus, N. volucellus), mosquitofish (Gambusia affinis), and longear sunfish (Lepomis megalotis). A tetra (Astyanax fasciatus), predatory cichlid (Cichlasoma loiselld), eleotrid (EleotTis amblyopsis), and two algivorous live-bearers (Phallichthys amatt'S, Poedlia gilli) were most numerous in the Costa Rican backwater sample. The Zambian floodplain sample was dominated by a minnow (Barbus paludinosus), two catfishes (Schilbe mystus, Synodontis woosnami), a killifish (Aplochdlichthys johnstoni) , and two her- bivorous cichlids (1ilapia rmdalli, 1: spamnanO.The most numerousfish-

ECOMORPHOLOGYOF fRESHWATERfISHES : WlNEMIliER 319

~ l!;;~~~~;,~£ ~ ~J;;~

Figure7. Results of principal components I analysis based on 30 morphological characters and 160 fish spedes from five lowland backwater fish ;{~ assemblages.The first PC axis modeled 24% of the total morphological varitJtion (eigenvalue=7.03) and the second axis modeled 14% (eigenvalue: 4.18). Vabal interpretations of the dominant variable loadings appear ;:f;'*'l~ belowthe axis labels. *1~!~:

-8~, " .,,"4""'",."".~ " ... ~B~i~~;;ti, ~.".,4:' -10 -s 0 5 10 In highly variable es in the Venezuelanswamp included two small characids(Ctenobrycon environments, the advan- spilurus, Odontostilbe pulcher), a mud-feedingcurimatid (Curimata argentea), a small callichthyid catfish (Corydorashabrosus), and an omnivorous tages of a mo1phological cichlid (Aequidenspulcher). specialization may only be ECOLOGICAL DIVERGENCE perceived during a very The shapeof species-lengthdistributions (maximum standardlength) was limited period when similar among backwatersites (Figure 6). Small and intermediate-sized resourcesare less abun- fishes dominated all five assemblages.The averagesize of fIShesdid not dant, population density differ significantly among backwater assemblages,except that Alaskan is higher, or competitors fishes were larger than fishes from the other four regions.32The largest speciesin the comparativestudy was the northern pike from the Alaskan are more abundant assemblage(750 mm). Most of the smallestadult fishes were characids, CORRESPONDENCE minnows, and killifishes from the Zambianand Venezuelansystems. The African minnow (Barbushaasianus) attaineda maximum size of 17.3 mm; the largest specimen of the African killifISh (Aplocheilichthyshutereaui) was 20.0 mm; and the largest specimen of the benthic citharinid (Hemigrammocharaxmachadoi) was 25.0 mm. Among Venezuelanfishes, the tetra (Creagrutussp.) reachesonly 28.0 mm; the catfISh (Corydoras habrosus)reaches 25.0 mm; and the dwarf cichlid (Apistogrammahoignei) attains 30.0 mm. Most thesesmall fishes have shon life spans ranging from severalmonths to just over a year. In most instances,these diminu- tive fishesspawn multiple clutchesof only a few to >100 small eggsdur- ing the annual spawning period. The largestfIShes at all five sites have batch fecundities of tens or hundreds of thousands of eggs and syn- chronousspawning during relativelyshon seasons.29 Figure 7 summarizesresults from principal componentsanalysis of ecomorphological variation in the five backwater fish assemblages.For an extensivestatistical analysis of variation in morphologicaltraits associated with ecological performance,see Winemiller.32 Assemblages containing fewer speciesat higher latitudes show less morphologicalvariation (i.e., their speciescluster near the centerof multivariatemorphological space in

320 RESEAROi& EXPLORAnON 8(3); 1992

~ Small arthropods, seeds, filamentous algae

H~ Ceratolxanchia

Hemigrammus

-', V\

Fishscaies Fishes ExfXkNJ 0Iara.- Genycharax

V Shearedflesh ~a.sa/mus

Aquatic plants Alpe, sponges, Invertebrates Schizodon L~us ..", ") ~

\ ';1" "::j ~ ~;/

Figure 8. Examplesof adaptivedivugence in tooth rrIO7phorogyamong South Ammam charaddand anostomidfishes (Charad- formes). DRAWINGSBASED ON GtRy (1977)

ECOMORPHOLOGY OF FRESHWATER FISHES : WlNEMIUER. 321 Thoracocharax stellatus Gephyrocharax va/enciae

Figure9. Adaptive divergences in body form relative to distribution in the water column and swimming style in small. omnivorous, pool-dwelling charadd and gasteropelecidfishes Aphyocharax alburnus (Charadformes) of South America. Astyanax bimaculatus

Popte/la orbicularis

Creagrutus sp.

dwellers have fairly symmetrical dorsal and ventral proffies and terminal mouths (e.g., Asytanaxbimaculatus). Fusifonn fishes (e.g., Aphyocharax albumus)are proficient in dashingafter food particlesthat lie severalbody lengths away; however, they suffer a trade-off in positional stability and ability to perfonn sharp turning maneuvers.In contrast,deep-bodied fishes (Metynnis aJgenteus)can turn sharply and position themselvesin the water column with great accuracy;However, the deep-bodymorphology carrieswith it a trade-off in proficiency for rapid straight-line swimming. Extreme forms of the deep-bodiedmorphology are conspicuousin pelagic fishes of tropical freshwaters,but are essentiallynonexistent in pelagic freshwaterfishes of the temperatezone. Most of the deep-bodiedmidwa- ter fishes of the tropics are planktivores or fruit- and seed-eaters. Examplesamong the neotropical Characidaeinclude the giant pacus(e.g., Colossomamacropomum), smaller silver dollars (Mylossoma,Metynnis, Myleus spp.), and tetras (Tetragonopterus,Gymnocorymbus spp.). During certain times of the year,these fishes competefor seeds,fruits, and terres- trial arthropods that drop or are blown into the water. There is very little aquatic primary and secondaryproduction in the nutrient-poor aquatic systems inhabited by many of these fiShes. In such systems, I have observedlarge schools of Metynnissp. frantically snatching up tiny adult midgesthat were trappedon the water'ssurface during the previousnighL Thesefeeding frenzies are intenseand short-lived as the midgesare deplet- ed in a period of several minutes. Similarly, the robust, deep bodies of schooling piranhas (Pygocentrusspp.) are well adaptedfor quick maneu- vering at close range during feeding frenzies.In contrast, the long cylindrical bodies of pike-like predatorsare efficient for swimming by meansof brief, rapid bursts along a straight-line trajectory; I placed each of the speciesof the five backwaterassemblages into five generaltrophic categories(detritivores/algivores, omnivores, invertebrate- feeders,piscivores, invertebrate-feederstpiscivores) based on dietary data (volumetric proportional usageof general food categories).Species were also assignedto five habitat categories(benthic, epibenthic,midwater, surface, vegetation) based on the field collection records'of fishes captured from different habitatsand pool zones.The relative frequencyof speciesin

322 RESEAROi 67.EXPLORATION 8(3); 1992 the five n-ophic categories showed some general similarities across regions (Figure 10). Except for the codominance of invertebrate-feeders and omnivores at the Venezuelan site, invenebrate-feeding was the most common n-ophic guild irrespective of the region. Algivores/detritivores, pisci- VOTes,and species that, as adults, consume a mixtUre of invenebrates and fishes were also present in all regional assemblages. Nevertheless, there were significant differences in the relative proportions of trophic categories represented in regional fISh assemblages (X2 = 34.78, df- 16; p< 0.005). Omnivores (diets consisting of a mixture of 8lgae, seeds, fruits, detritus, insects, crustacea) were just as abundant as invertebrate-feeders in the Venezuelan assemblageyet were entirely absent from the two temperate zone assemblages. Fish assemblages from higher latitudes were more heavily skewed in favor of invertebrate-feeders, whereas n-ophic categories tended to be more evenly apportioned with decreasing latitude (Figure 10). Habitat categories exhibited fewer interfaunal differences than n-ophic categories (x2 - 12.62, dj-16, p= 0.70). Again, the Venezuelan backwater assemblage tended to have more even distribution of species within the five categories than the high-latitude assemblages.The majority of AJaskan fishes were midwater forms, and none were surface-dwellers. Benthicfishes dominated the Venezuelan swamp assemblage. ECOLOGICAL CONVERGENCE Accolding to the operational definition of ecomorphological convergence (convergenceindex values ranging from 0 to 1.0), convergencewas seen in all five regional assemblagesat low frequencies (Figure 11) and was usually not extreme. Most convergence index values were small (0< index value

figure 10.

i,.~,I'j"i,c-~;"f)![;~"" frequndes of spedesassociat- ed with Jive general trophic ~I' .. ,,'.0 .,w,)_c.,. . ~ categories(top) andJive habitat ~~~.1::~;;;'"!i~.. ~ii!1 G) "';f,"'I"~"i;" ,.. 0 6 ~ categories(bottom) for back- ;J'J!,.~;;;'1~"""',r" c ;r;~.1!","~"#.""",, water fish assemblagesin five biotic regions. ":i:'fQ0:4 j

.. :;~ yt 0.2

;,;;,. ".. -- - Fishes ~;; 0.0 . Alaska,6 species 0Im'- Algae/delt. . Texas,26 species Trophic Category . CostaRica. 30 species 0.6 . Zambia,33 species . Venezuela,43 species .~\ G.4

-"~:~. I.-:iti~- Mi6NaIeI Surface HabitatCategory

ECOMORPHOlOGY OF FRESHWATER FISHES : WINEMIllER 323 The ability to shift from <0.5), indicating that for the majority of species, the nearest ecomorpho- a restricted diet to one logical neighbors were actually the most closely related taxa (Figure 3). Yet several species were actually most similar in form to quite distantly that incorporates several related taxa (index value >0.5). Eel-like fishes (Ichthyomyzon gaga, alternative foods is obvi- Synbranchus mannoratus, Afromastact:mbt:lusfrmatus) exhi;bited extreme ously adaptive when these convergence (index value -1.0). The spiny eel, Afromastacembelus alternative resourcesare (Perciformes), is genetically more similar to perches, sunfIShes, and cichlids than to other eel-like fishes. The extremely elongate African catfish, seasonally abundant. Clarias theodorae,is more similar in morphology and ecology to the eel- CORRESPONDENCE like fishes than it is to most other ~tfishes. A taxonomically diverse group of small, surface-feeding insectivores are ecological equivalents (e.g., Alfaro cultTatus [PoeciliidaeI, J\plochdlichthysjohnstoni [Cyprinodontidae I, Gephyrocharaxvalendat: [Characidae]). A number of instances of extreme ecomorphological convetgence were identified even at the family level. Within the family Cichlidae, the large predatory guapote (Cichlasoma dovii) was more similar to predatory African cichlids (Serranochromismacrocephalus and S. robustus) than to more closely related neotropical Cichlasoma species (Figure 3). Similarly; the predatory, vegetation-dwelling cichlid Cichlasoma loise-lId (Costa Rica) was more similar to the African cichlid Htmichromis elongatus and the Nonh American sunfish Lepomis cyandlus than to most other Central American Cichlasoma.Within the Cyprinidae, several African Barbus sp. (e.g., B. paludinosus) matched most closely with North American Notropis sp. (e.g., N. atrocaudalis) with similar ecological niches. Numerous other examples of pairwise ecomorphological convergence could be cited among the data that yield the nonzero values in Figure 11. Among Alaskan fishes, extreme instances of convetgence with Species from other faunas involved the sit-and-wait/stalking piscivore Esox lucius (Figure 3). In general, high convergence index values (>0.5) were more frequent among tropical speciesthan temperate species (Figure 11). Discussion ...... Comparisonsamong five regionalfish assemblagesyielded patterns of latitudinal taxonomic diversity that are consistentwith most earlier studies involving large groups.e.g.,nGiven the greaternumber of speciesin the tropics, it is perhapsnot surprisingthat greatermorphological and ecological diversity is observedin tropical fIShesthan in temperatefishes. In a separateanalysis based on the samespecies, I concluded32that the statisti- calproperties of this greaterecomorphological variation in tropical assem- blagessupport an interpretationof evolution in responseto deterministic forces. Competition for food resources,either directly or indirectly via microhabitat selection,is probablya major sourceof natural selection on ecomorphological characteristicsin the tropics. Patterns of divergence within the Texan fIShassemblage were generally concordant with those observedin tropical fIShassemblages, which suggeststhat deterministic forcesare alsoat work in species-richtemperate faunas. Comparedwith tropicalfaunas, the lesserdegree of taxonomic and ecological diversificationin temperatefaunas most likely results from greater rates of speciesextirpations in temperateregions on a geological time scale.Most biogeographersagree that the habitat chang"esassociated with the glacial epochswere more severeat higher latitudes. In addition, the

324 RESEAROI&: EXPLORAnON 8(3); 1992 Figure 11. Frequenciesof spedespairs showingvarious degreesof ecomorphologicalconvergence basedon the relative convergence index. N~ero convergence values (valuesare groupedat intervals ~n the x-axis) indicates that at least one closephylogenetic relative was lesssimilar eco- morphologicallythan the ecomor- phologicalnearest neighbor. Convergencevalues >0.2 indicate that >20%of closephylogenetic relativeswere lesssimilar eco- morphologicallythan the nearest neighbor(convergence index definedin text).

~ c "t: "to 0- ~ .~ ~ VI - 0

. NN1 . NN2 . NN3 . NN4 . NNS

seasonalchanges in environmentalconditions are more extremeat higher latitudesand altitudes,but the ecologicaland evolutionaryimplications of these proximate fluctuations can be debated. Seasonalinterruptions of density-dependentenvironmental conditions occur in tropical as well as temperatesettings.29,JO All natural communities are influenced by both biotic density-dependentand abiotic density-independentfolCes, but the relative influence of each varies greatly from one geographicalsetting to the next. The analysisof fISh ecomorphologicalpatterns32 supports the

ECOMORPHOLOGY OF FRESHWATER FISHES : WINEMILLER 325 Just as mostfishes continue general view that biotic interactions have a greater relative influence on to grow throughout their evolution in the tropics than in temperate and polar regions.s Ecomorphological convergence among phylogenetically divergent taxa lives, ecological perfor- provides powerful evidence for determinism in the evolution of species mancefrequently exhibits traits. If two divergent taxa can be shown to have evolved similar solu- major changes over the tions for the same problem in different regions, we gain a measure of con- lifespan. Larval fishes gen- fidence in our inferences of adaptive functions and that at some level, a set erally feed on zooplankton of universal rules exists for the structuring of natural communities. Reports of convergent traits (e.g., fISh electroreceptio~12.15lizard toe and other microscopic fringes16) and ecologically equivalent species (e.g., fishes,24birds,1O.19.2S organisms. Wefind the lizards2J) abound in the literature. The phylogenetic foundations for greatest morphological hypotheses of convetgence are now receiving greater attention and integra- diversity and ecological tion with ecological research.2Morphological variation among homologous features at a micro-scale (osteology, musculature, ligaments, etc.) provides diversity among adult size systematists with evidence of phylogenetic relationships. When subjected classes,and that is why I to intense scrutin~ many of these small-scale changes in morphology can restricted my comparative be shown to cause significant shifts in ecological function; e.g., G. v: analysis to adult fishes. Lauder13 showed how small morphological changes in neuromuscular physiology result in differencesin feeding performance in sunfishes. CORRESPONDENCE The simple convergenceindex used in this study was designed to identify convergence in basic morphological features related to ecological function. This method requires knowledge of phylogenetic relationships. In practice this is never actually the case, because any phylogeny represents only one of several possible hypotheses of relationship. Evolutionary convergences are recognized when similarities in basic form and function emerge from different (analogous) structural components at the micro- scale. Parallel evolution occurs when two divetgent lineages independendy evolve the same features in response to similar selective environments. The convergence index cannot distinguish between parallel evolution and evolutionary stasis (lack of morppological differentiation in two independent lineages). As new morphological and biochemical data produce larger and more detailed phylogenies, new methods can be developed to test a variety of evolutionary ecological hypotheses. This study identified a number of convetgent speciespairs from different geographical regions. The analysis identified many more instances of partial or weak convetgencesand numerous species that showed no ecomorphological convergencewith speciesfrom other assemblages.At the level of the species assemblage,morphological diversification was similar among the three regions of intermediate species diversity. The Alaskan fish assemblage showed very litde interspecific variation in ecomorphology, and the speciose Venezuelanassemblage showed greater variation than other sites. As a function of the physics of movement through a fluid medium, basic body shape strongly influences ecological performance. In fishes, variation in body form correlates with variation in swimming behavior and partitioning of space within aquatic habitats. Feeding ecology is correlated with morphological features such as tooth shape, mouth size, and mouth orientation. Tropical fishes from species-richassemblages display the greatestvariety of ecological roles. Finall~ some morphological traits are associatedwith defense against predation (e.g., stout spines, cryptic coloration, mimicry) and reproductive biology (fin filaments, flashy coloration, nuchal humps). Tropical freshwater fishes appear to exhibit far greater diversification than temperate fishes in these features as well.

326 RESEAROI61 EXPLORAllON 8(3); 1992 ACKNOWLEDGMENTS REFERENCES Numerouspe.sons assisted in the field 1. Bell-Cross,G.; &: Minshull,J. L: 18. Nelson,j. S.: Fishesof thc world, 2nd ed. john Wiley &. Sons,New York; collections,and specialthanks go to TheFishes of zimbabwe. National L KeIso-Wmt:mi1ler,D.C. Taphom, LG. Museumsand Monuments of Zimbabwe, 1984. 19. Niemi. G.j.: Pattemsofmorpho- Nico, A. Barbarino,E. Urbina,J. Masinja, Harare,Zimbabwe; 1988. G. Milini, Mr. Sinda.W. Ritter, and 2. Brooks, D. R; &: McLennan, D. A.: logical evolution in bird generaof New World and Old World peatlands.Ecology M. Kelso.L Kelso-Wmemiller has my PhylogOlY,Ecology, and Behavior.Univer- tremendousgratitUde for assistingin mea- sity of ChicagoPress, Chicago, IL; 1991. 66:1215-1228;1985. 20. Page,L. M.; &. Swofford,D. L.: surementsof-many fish specimens.I 3. Carlander, K. D.: Handbookof thank A. Brenkert for assistancein data FreshwaterFisheries Biology, vol. 1. Iowa Morphologicalcorrelates of ecological management.InstitUtional suppon abroad StateUniversity Press,Ames; 1969. specializationin daners. Environmmtal Biologyof Fishes11:139-159; 1984. was provided by D.C. Taphom of Ia 4. Carlander, K. D.: Handbookof Universidadde los Uanos Occidentalesin FreshwaterFisheries Biology, vol. 2. Iowa 21. Pagel,M. D.; &. Harvey,P. R.: Recentdevelopments in the analysisof Venezuela.H. Haug and E. Chamorro of el StateUniversity Press,Ames; 1977. Serviciode parquesNacionales de Costa 5. Dobzhansky,T.: Evolution in comparativedata. TheQuarterly ~cw of Biology63:413-440; 1988. Rica,and E. Muyangaand G. Milindi of the tropics. AmericanSdcntist the Departmentof Fisheriesof Zambia. 38:209-221; 1950. 22. ~ E. R: latitudinalgradi- entsin speciesdiversity. Ammcan Collecting and fishing pennits were 6. Douglas,M. E.: An ecomorphologi- obtainedfrom Ia DirecciOnAdministaci6n cal analysisof niche packingand niche dis- Naturalist 100:33-46;1966. y Desarollopesquero de Ia Republicade persionin streamfish clades. Matthews, 23. ~ E. R: Someintercontinen- tal comparisonsof desenlizards. National Venezuela,el Serviciode Parques W.J.; &: Heins,D. C, editors:Community Nacionalesde CostaRica, the Depamnent cmdEvolutionary Ecology of North Amerialn GeographicResearch 1:490-504; 1985. of Fisheriesand National Commissionof StreamFishes, 141-149. University of M. Sazima,I.: Similaritiesin feeding behaviorbetween some marine and fresh- DevelopmentPlanning of the Republicof OklahomaPress, Norman; 1987. Zambia,and the AlaskaDepanment of 7. Gatt, A.J.Jr.: Ecologicalmor- water fishesin tWotropical communities. JournalofFish Biology29:53-65; 1986. Fish and Game.Fieldwork was funded by phology of freshwaterstreaD1 fishes. the National GeographicSociety, Tinker TulaneStudies in Zoologycmd Botany 25. Schluter, D.: Testsfor similarity and convergenceof finch communities. Foundation,NSF DissertationGrant, and 21:91-124; 1979. a Fulbright Grant for International 8. Gat%.A.J.Jr: Community organiza- Ecology67:1073-1085; 1986. 26. Watson, D.j.; &. Balon.E. K.: Exchangeof Scholars.Oak RidgeNational tion in fishesas indicatedby morphologi- Laboratoryis managedby Manin Marietta cal featUres.Ecology 60:711-718; 1979. Ecomorphologicalanalysis of fish tax- EnergySystems, Inc. under contract DE- 9. Gtry.J.: Characoids of the World. acenesin rainforeststreams of northern AC05-840R21400with the U.S. TFH Publications, Neptune, Nj; 1977. Borneo.Journal ofFish Biology25:371- 10. Karr.J. R; &:James,F. C.: £co- 384; 1984. Departmentof Energy. morphologicalconfigurations and conver- 27. Webb, P. W.: Bodyform,locomo- gent evolution in speciesand communi- tion and foragingin aquaticvertebrates. ties. Cody, M. L &: Diamond.J. M., AmcricanZoologist M:I07-120; 1984. editors: Ecologycmd Evolution of Commun- 28. Wikmmanayake,E. D.: ides.258-291. BelknapPress of Harvard Ecomorphologyand biogeographyof a University, Cambridge.MA; 1975. tropical streamfish assemblage:Evolution 11. Keast.A.; &: Webb. D.: Mouth of assemblagestrUcture. Ecology and body form relative to feedingecology 71:1756-1764;1990. in the fish fauna of a small lake, Lake 29. Wmemiller, K. 0.: Patternsof Opinicon, Ontario. Journal of the variation in life history amongSouth FisheriesResearch Board of Clmada Americanfishes in seasonalenvironments. 23:1845-1874; 1966. Oceologla81:225-Ml; 1989. 12. Kirschbaum.F.: Reproductionof 30. Wmemiller, K. 0.: Spatialand weakly electricteleosts: Just anotherexam- temporalvariation in tropical 6sh trophic ple of convergentdevelopment? Environ- networks. EcologicalMonographs mentalBiology of Fishes10:3-14; 1984. 60:331-367;1990. 13.Lauder, G. V.: Neuromuscular~t- 31. Wmemiller, K. 0.: Comparative ternsand the origin of trophicspeda1izatiom ecologyof Serranochromisspecies in fishes.SdDIa 219:1235-1237;1983. (Teleostei:achlidae) in the Upper 14. Lee. D. S.; GUbert,C. R; &: ZambeziRiver. Journalof FishBiology Hocutt, C. H.: Atlas of North Amcrfam 39:617-639; 1991. Freshwater Fishes. North Carolina State 32. WmemUler,K. 0.: Museum of NatUral History. Raleigh; 1980. Ecomorphologicaldiversification in low- 15.l.owe-McConnell, R H.: Fish land freshwater6sh assemblagesfrom five Communitiesin Tropical Freshwatm: Their biotic regions. EcologyMonographs DIstribution,Ecology, cmd Evolution. 61:343-365;1991. LongmanPress,London; 1975. 33. Winemiller, K. 0.; &. Pianka, 16. Luke. C.: Convergentevolution of E. R: Organizationin naturalassem- lizard toe fringes. BiologicalJournalof the blagesof desenlizards and tropical6shes. Linne4nSocidy 27:1-16; 1986. EcologicalMonographs 60:27-55; 1990. 17. Mago, F.: Lista de IosFeces de V~ Oficina Nacional de Pesca, Ministerio de AgricultUI2 y Cria, Caracas, Venezuela;1970.

ECOMORPHOLOGY OF FRESHWATER FISHES : WINEMILLER 327