Environmental Biology of Fishes 26: 177-199. 1989.

@ 1989 Klu~'er Academic Publishers. Printl'd in thl' Netherlands

Ontogenetic diet shifts and resource partitioning among piscivorous fishes inthe Venezuelan llanos

Kirk O. WinemillerDepartment of Zoology, The University of Texas, Austin, TX 78712, U.S.A

Accepted 12.12.1988Received 18.8.1988

Key words: Caquetia, Charax. Diet breadth, Diffuse competition, Gymnotus, Hoplias. Resource overlap,Piranha, Pygocentrus, Rhamdia. Seasonality, Se"asalmus

Synopsis

Resource utilization by nine abundant piscivores from a diverse tropical fish assemblage was examined overthe course of a year. All nine species exhibited peak reproduction during the early wet season and a similarsequence of size-dependent shifts from a diet composed primarily of microcrustacea, to aquatic insects, andfinally fishes. Three piranha species specialized on fish firis, particularly at subadult size classes (SL30-80mm). Gradual dessication of the floodplain during the transition season was associated with fishgrowth, increased fish density, and decreased aquatic primary productivity and availability of invertebrateprey. Based on 118 resource categories, average pairwise diet overlap was low during all three seasons: wet,transition, and dry. Of 72 species pairings, only one pair of fin-nipping piranhas exhibited high overlapsimultaneously on three niche dimensions: food type, food size, and habitat. Adults of two species, agymnotid knifefish and pimelodid catfish, were largely nocturnal. Patterns of habitat utilization indicate thatpiranhas may restrict diurnal use of the open-water region by other piscivores. Collective diet overlap ofindividual piscivore species with the other eight feeding guild members and collective overlaps with the entirefish community each revealed two basic seasonal trends. Four species that showed an early switch topiscivory also showed a high degree of diet separation with both the guild and community at large on ayear-round basis. The five remaining species showed lowest collective diet overlaps during the transitionseason when availability of invertebrates was reduced and fish densities were maximal. Whereas predationmay playa role in habitat separation, diffuse competition for food resources during the approximatelyfour-month transition season probably is the principal factor yielding patterns of diet specialization.

bers provides the most straightforward means ofattaining a preliminary assessment of the relativeimportance of interactive processes within naturalcommunities. Whereas descriptive studies that re-veal patterns of resource subdivision are limited intheir abilities to demonstrate the mechanisms cre-ating observed relationships (Toft 1985), thought-ful comparisons and the weight of evidence derivedfrom numerous studies already have yielded muchinsight (Schoener 1974, Helfman 1978, Sale 1979,

Introduction

Most diverse taxonomic groups of organisms ex-hibit divergence in trophic biology (e .g., cichlidfishes, Greenwood 1981; ostariophysan fishes,Fink & Fink 1981). Whether or not ecomorpholog-ical divergence is random or adaptive in the contextof community interactions can be debated (Stronget al. 1979, Grant 1981). The study of resourceexploitation among syntopic feeding guild mem-

178

focuses on the piscivore guild as a potential unit forcompetitive biological interactions. Given the cli-mate and natural history of the Cano Maraca eco-system, several basic questions can be asked: (1)Do piscivorous fishes partition food and habitat,and if so, what is the influence of seasonal changesin their environment? (2) What are the relation-ships between ontogeny, resource partitioning,and seasonal changes in resource availability? (3)Given seasonal patterns of resource availabilityand exploitation by consumers, is there evidence ofinteractive effects on piscivore feeding behavior?(4) Do seasonal trends of pairwise species dietoverlaps follow patterns from collective diet over-laps within the piscivore guild or total fish commu-nity, and if not, which index offers a better assess-ment of the likelihood of interactive effects?

Methods

Study site

During every month of 1984, fishes were collectedfrom Calio Maraca, a swamp-creek of the RioApure-Orinoco drainage in the western llanos ofVenezuela (8° 52' 30" Lat. N; 69°27' 40" Long. W).The region studied (termed an 'estero') lies withinlow, flat terrain that experiences extensive sheetflooding during the wettest months (late May-Au-gust). During this time, a large region borderingthe creek channel is converted from exposed, sun-baked soil and thorn-scrub habitat into a produc-tive marsh with diverse aquatic vegetation dom-inated by Cyperus sp., Eichhornia diversifolia, andHeteranthera reniformis. During the driest months(Dec.-May), the aquatic habitat is reduced to anetwork of mud-bottom pools blanketed by Pistiastratiotes, Salvinia spp., and Lemna spp.. Dis-solved oxygen concentrations are reduced at thistime and many fishes rely on special respiratoryadaptations for survival. Annual rainfall in the re-gion was 1300 mm during 1984, with 860 mm fallingbetween June and September (Fig. 1). Based onchanges exhibited by stream physico-chemicalparameters associated with rainfall and their ef-fects on aquatic organisms (Winemiller 1987a,

Werner 1979. Toft 1985). In a recent review of theliterature on fishes. Ross (1986) observed greatestpartitioning on food dimensions. followed by hab-itat and time. Many studies treating two or moreniche dimensions have shown complementarity. in-dicating a potential effect of interspecific competi-tion. An interactive effect between competition.predation, and temporal shifts in abiotic environ-mental parameters has been suggested for somesystems (Werner & Hall 1977, Werner et al. 1983,Mittlebach 1984. and see Schoener 1974. Toft1985). Ross (1986) observed a general pattern ofstudies revealing less evidence of interactive effectson community structure in harsher environments.

Seasonal changes in resource availability and on-togenetic diet shifts can affect both predator-preyand competitive interactions among size-structur-ed fish populations. The present study exploresseasonal and size-related patterns of resource ex-ploitation within the piscivore guild of a diverseneotropical fish community. Although varyinggreatly in both duration and breadth of taxonomiccomparison, several recent studies have shownmarked resource partitioning among tropical fishes(Angermeier & Karr 1983, Power 1983, Moyle &Senanayake 1984, Schut et al. 1984, Watson &Balon 1984, Araujo-Lima et al. 1986, Hyslop 1986,Prejs & Prejs 1987, Nico & Taphorn 1988). Fishpopulations frequently are size structured due tosmall hatching or neonate size and continuousgrowth. Body size significantly affects the sizerange of food particles potentially available to con-sumers (Brooks & Dodson 1965. Werner 1974,Magnan & Fitzgerald 1984). Furthermore, size dis-tributions of fish populations directly influencepredation rates on different age classes and oftenthe spatial distribution of fishes (Fraser & Cern1982, Mittlebach 1984, Power 1984, Werner & Gil-liam 1984).

The study site, Calio Maraca, experiences highlyseasonal rainfall that is characteristic of many partsof central and southern Venezuela. Most fishesreproduce during the early and peak wet season, asituation that leads to temporal changes in averagebody size for most species (Winemiller, unpublish-ed). Although similar data were collected for allfishes that occurred at the site, this investigation

179

. . .. ....10. ,.. -0' A~, _0' .lu. .lul ... 00' Mo. 0..

Fig. 1. Average monthly rainfall (closed symbols) and maximum

midpool depth measured each month (open symbols) during thestudy period at Cano Maraca.

1987b), the annual cycle was divided into threeterms of equal duration: dry season (Jan.-Apr.),wet season (May-Aug.), and transition season(Sept.-Dec.). Dry season conditions gave way tothe wet season very abruptly with the commence-ment of heavy rains, whereas the wet season slowlygraded into the next dry season over a prolongedperiod (i.e., transition season). During the transi-tion season, fish densities increased and biotic in-teractions intensified within gradually diminishingaquatic habitats in the floodplain. Estimates of fishbiomass at Cano Maraca at the end of the transitionseason are among the highest densities recordedfor natural freshwater ecosystems (141-165g m-2,Winemiller 1987a). All available climatic and ec-ological information indicate that 1984 was not un-usual in the sequence or magnitude of events. Forexample, the rainy season sometimes begins earlieror later than the last week of May, yet the first rainsare typically heavy, persist for several months,slowly subside, and nearly cease all together forseveral months during the peak dry season.

three wet months, collecting effort expended wasapproximately equal during each sampling period.More total time was spent collecting during the firstthree months of the wet season due to the compara-tive inefficiency of adult fish captures in the newly-expanded aquatic environment. All aquatic hab-itats in the estero region were sampled for a mini-mum of six hours (habitats combined), and a col-lecting outing was terminated when two hours ofsampling yielded no additional species. Collectionsused for comparisons of relative densities weremade during either one or two days between the11th and 28th of each month. Additional collec-tions were made on other dates to supplement stan-dard samples of uncommon species. A completelisting of collecting dates and sample sizes for pre-served specimens appears in Winemiller (1987a).Collecting activities should not have affected nat-ural fish densities between sampling periods, be-cause (1) only a fraction of the total aquatic habitatwas sampled on a given date and (2) fishes couldmove freely between the estero region and thepennanent channel downstream (except duringMarch and April). All collected individuals werepreserved in 15% formalin rather than the usual10% to safeguard against decomposition of stom-ach contents. The largest fishes were cut on thelateral wall of the visceral cavity to hasten pene-tration of formalin. Following examination, a por-tion of the fishes collected was deposited with theMuseo de Ciencias Naturales de UNELLEZ, Gua-nare, Portuguesa, Venezuela. The remainder wasdeposited in the Natural History Collection of theTexas Memorial Museum, University of Texas,Austin, Texas.

Separate samples were taken on 27 and 28 Sep-tember, 25 and 26 October, 22 November, and IIDecember from three habitats: open water mid-pool (day and night), shallow pool edge, and densevegetation. The open water habitat was sampled ina large pool at the lower end of the estero with the20m, 12.7mm mesh seine. These large seine sam-ples were preserved in 15% formalin and kept sep-arate from small seine (2.5 m, 3.17mm mesh) sam-ples taken from the same pool. Because the largemesh seine allowed small fishes to pass through,and large fishes easily avoided the small seine,

Collections

Fishes were collected by seine (3.17mm un-stretched mesh. 2.5m length; 12.7mm mesh. 20mlength). experimental gillnet. and dipnet. On eachouting. an attempt was made to sample the entirefish community such that the sample for each spe-cies reflected its relative abundance and size struc-ture during each month. With the exception of

!~i..0

180

values obtained for the number of individuals ofeach species in the two open water samples werelater combined as one sample. Night samples weretaken from the same open water location with bothseines on 27 September and 25 October (1800-2400 h). The 2.5 m seine was used to sample openand sparsely vegetated, shallow shoreline habitatduring the morning of each outing. For the vegeta-tion habitat samples, the 2.5 m seine was rolled onpoles attached to each end until the total area ofexposed net was 4.2 m2. The weighted end of theseine was passed below dense masses of floatingvegetation and quickly brought to the surface onthe other side, enveloping a quantity of aquaticvegetation approximately equal to the area of theseine and entrapping all fishes contained therein.The vegetation was then hauled to the adjacentshoreline where fIShes were extracted and pre-served.

Diet analysis

for stomach content analysis. For certain piscivo-rous species, such as Hop/ias ma/abaricus and Py-gocenlrus nolalus, all available specimens were dis-sected. When more than 30 specimens were avail-able, a subsample was chosen for dissections thatincluded representatives spanning the total spec-trum of size classes in the collection. During thecourse of one year, 83 fish species were collectedfrom the estero. Each monthly sample producedbetween 40 and 60 species (Winemiller 1987a).This paper reports findings for the nine most abun-dant piscivorous species only: Hop/ias malabaricus(Erythrinidae), Charax gibbosus (Characidae),Pygocenlrus nolalus (Characidae), Serrasa/mus cf.irritans (Characidae), Se"asalmus cf. medini, Ser-rasalmus rhombeus, Gymnolus carapo (Gymnoti-dae), Rhamdia sp. (Pimelodidae), and Caqueliakraussii (Cichlidae). For the purposes of this study,piscivores were defined as fishes averaging at least50% fish prey in the diet of the largest adult sizeclass. Six piscivorous species were rare in the localecosystem and omitted from the analysis [i.e., Ac-eslrorhynchus microlepis (Characidae), Ageneio-sus brevifilis (Ageneiosidae), Crenicich/a saxalilis(Cichlidae), Hoplerylhrinus unilaeniaIUS (Erythri-

nidae), Pseudoplalysloma fascialum (Pimelodi-dae), and Slernopygus macrurus (Sternopygidae»).Additionally, four common species that consumedfishes in minor proportions were excluded from thecomparisons [Aequidens pulcher (Cichlidae), As-lronOIUS ocellalus (Cichlidae). Parauchenip'erusgalealus (Auchenipteridae), and Synbranchus mar-moralus (Synbranchidae) J.

Food items in the anterior half of the gut wereremoved, examined under a dissecting microscope,and identified at the species (fishes) or ordinal level(invertebrates). Broader food categories were notused for numerical analyses, because (I) lower tax-onomic categories generally represented one ormore renewable resources, and (2) lumping re-source states often inflates niche overlap values(Greene & Jaksic 1983). Prey items then were sort-ed, carefully blotted with a dry paper towel, andwhile still moist, measured in an appropriatelysized graduated cylinder by water displacement.The degree of resolution for very small items was0.005 mi. For volumes less than 0.005 ml, a value

All preserved specimens were identified, counted,and measured for standard length to the nearest0.1 mm (alliengtbs are reported hereafter as SL).Fisb species names were provided by Donald Ta-phorn of UNELLEZ and Francisco Mago-Lecciaof UCV (Caracas). and voucher specimens areavailable in the Museo de Historia Natural (UN-ELLEZ) and The Texas Natural History Collec-tion (TMM. UT, Austin). Many Venezuelan spe-cies could not be given tentative species names dueto the current state of taxonomic confusion. Conse-quently, a few species were designated as distinctforms by 'sp.'. Two piranhas were initially identi-fied as Se"asalmus eigenmanni and Serrasalmussp. (Taphorn & Lilyestrom 1985). The piranha sub-family (Serrasalminae) is currently under revision.According to William L. Fink (personal communi-cation). these species correspond to Se"asalmuscf. irritans and Se"asalmus d. medini, respective-Iy. Caquetia kraussii was formerly listed as Peteniakraussii (Mago-Leccia 1970. and see Kullander

1983).Whenever available, 30 specimens of each spe-

cies from each monthly collection were dissected

181

was estimated by spreading the item on a glass slideand comparing it to a similarly spread substanceknown to equal 0.005 mi. Except for highly frag-mented material. the size interval associated witheach identified prey item was recorded (10mm in-tervals measured along the longest particle axis).An ocular micrometer aided linear measurementsof small prey items. Diet breadths were calculatedby Simpson's measure (Simpson 1949). Asymmet-rical diet overlaps were calculated by MacArthur &Levins' (1967) formulas:

mirrors the feeding habits of the population atlarge. This study employed niche overlap (fp) as ameasure of similarity only, and not as a direct esti-mate of interspecific competition (a) in the mannerpreviously used by some investigators.

Following the method of Lawlor (1980), electiv-ities were used in place of Pi~ for an additional setof diet overlap measures adjusted for prey relativeabundances (hereafter referred to as diet similar.ities). Consumer electivities (ei,) were calculated as

Pi/e= Rb' '/I

(fJik = --'-

.L P;/'IAI~I -.

L plj2i-I

and

fit! a

.I: PtIJ~I-I

.r P1.2I-I

where Pij is the proportion of prey item i consumedbyspeciesj. andpik is the proportion of the ith preyitem consumed by species k. These measures re-flect differences in the magnitude of overlap be-tween two populations due to unequal nichebreadths. For example, one would expect dietoverlap of the more specialized bluegill sunfish.Lepomis macrochirus. on the green sunfish. L.cyanellus. to be less than that of the latter on theformer (Werner & Hall 1977). The asymmetricalmeasure has been viewed as desirable in the sensethat potential for exploitation competition for re-sources is more realistically portrayed as unequalfor species having different diet breadths (Slobod-chikoff & Schulz 1980). Whereas this and othermeasures of niche overlap present difficulties forstatistical inference when three or more species arecompared. they at least provide a simple. reliableindex of diet similarity. All PijS were based on thetotal volumes of the food items consumed during aspecified period. This method assumes that stom-ach contents from a large sample of individuals

where Rj is the relative proportion of food item i inthe environment, and pjj is the proportion of item;in the diet of consumer species j. The relative abun-dance of each prey item in the environment wasestimated from the column sums in the communityresource matrix (items summed across all fish spe-cies examined, both piscivorous and nonpiscivo-rous). Although it is labor intensive, this method ofcalculating Rj has the advantage of basing total preyavailability on the overall performance of the en-tire fish community. Assessment of prey R;s basedon independent sampling efforts risk (1) under-estimating availability when microhabitats avail-able to fish are undersampled, and (2) overestimat-ing availability when large numbers of prey that areactually invulnerable to predation are sampled(Wallace 1981). In effect, the method estimatesfood availability using fishes rather than humans assamplers of their environment. The method ismore reliable when based on diverse assemblagescontaining many ecomorphotypes, because anysingle species would have a relatively small influen-ce on estimates of availability.

Ecological measures and statistics were comput-ed from the three seasonal resource matrices (9predators x 118 potential prey categories). Foranalysis of ontogenetic diet shifts, dissected speci-mens were grouped into either 10 or 20 mm in-tervals, depending on the total range of SLs record-ed for each species. The average proportion ofmicrocrustacea, aquatic insects, and fish prey itemswas determined for each fish size interval. Fishscales were omitted from this analysis, because

182

scales comprised a minor portion of diets and noneof the nine species are believed to rely on scales as asignificant food resource (sensu Roberts 1970, Sa-rima 1983). Most scales not associated with fleshwere probably taken during aggressive interactionsby these species. All statistical tests were computedby SAS (SAS Inst. 1985). A generalized linearmodel (GLM) procedure was used for calculationof ANOVA statistics for comparisons of data setswith large unequal sample sizes.

ta argentea. numerically dominated the estero

(Winemiller 1987a).Average size of all nine piscivores was greater

during the transition season than the wet seasondue primarily to growth of juveniles produced dur-ing June and July (Fig. 2). Hoplias. Charax, Pygo-centrus, Gymnotus, and Caquetia were extremelyabundant within shrinking aquatic habitats duringthe transition season. Populations of Hoplias,Charax, and Caquetia consisted of a mixture ofadults, subadults, and juvenile size classes duringthe transition season. Both adult and subadult sizeclasses of Rhamdia and Gymnotus were taken dur-ing the transition season, whereas the four piranhaspecies primarily consisted of subadults. Duringlate October, dissolved oxygen concentration ex-hibited a temporary reduction (mean dissolved0] = 0.73 ppm), which probably resulted from apulsed increase in total microbial metabolism asso-ciated with the death of aquatic macrophytes inshallow standing waters of the drying floodplain(Winemiller 1989). At this time, all four piranhaspecies were rare or absent from the estero regionof Cano Maraca. The four species reappeared onthe site after dissolved oxygen had increased withinthe stream channel (0] = 2.s-6.0ppm) within sixweeks. Young-of-the-year Pygocentrus (76-127 mm) were taken in large numbers during No-vember and December (Fig. 2).

Results

The piscivore guild of the dry season was dom-inated by adult Hop/ias ma/abaricus, followed byCaquetia kraussii (17-185 mm), adult Gymnotuscarapo. and adult Rhamdia sp. (Fig. 2). During theharshest months of the dry period (Feb.-earlyMay), Charax gibbosus and the four piranha spe-cies (Pygocentrus and Se"asalmus spp.) emigratedfrom the estero region to deeper aquatic habitatsdownstream. Collections made on Caito Maraca ata location 14km east (downstream) of the esteroduring the height of the dry season yielded severalindividuals of Charax. Se"asa/mus irritans, and S.medini. Small juveniles of Hop/ias and especiallyCaquetia were taken at various intervals during thefour month dry season, indicating that at least por-tions of the local populations were reproductivelyactive.

Charax, Pygocentrus, and S. irritans returned tothe estero within one week of the abrupt onset ofthe rainy season. All individuals of these speciestaken during June were adults with gonads in ad-vanced states of maturation. Eleven large malePygocentrus in black nuptial coloration were takenin shallow water among emergent stands of aquaticmacrophytes with one haul of the 2.5 m seine onJune 6. Each male had milt-laden testes, and theirstomachs contained either large cycloid scales orplant debris. Juveniles of all nine species were col-lected during July and August. During the wetseason, small omnivorous and herbivorous char-acids (especially Astyanax bimacuJarus. Markianageayi. Ctenobrycon spi/urus, and Odontosti/beDu/cher), and the detritivorous curimatid, Curima-

Ontogenetic diet shifts

Of m9 specimens from Cafio Maraca that weredissected for diet analysis, the nine major pisci-vores comprised 1980 (20%). Small juveniles of allnine species exploited aquatic microcrustacea andaquatic insects (Table 1, Fig. 3). Pygocentrusshowed two distinct ontogenetic shifts in basic dietcomponents: from primarily microcrustacea to aq-uatic insects (20-40 mm), and from aquatic insectsto fishes (4()...()() mm). Rhamdia experienced similardiet shifts between 2a-40 mm and ro-so mID re-spectively. A mixture of microcrustacea and aquat-ic insects dominated the diets of juvenile Hoplias«50mm), S. irritans «30mm), S. medini«40mm), S. rhombeus «40mm). GymnotILS

183

, .. A . 0 N 0Monlll

, M AM" ..Mon."

Fig. 2. Mean monthly standard lengths {or each piscivore species at Cafio Maraca. Error bars represent one standard deviation and

coUection sample sizes appear within each open bar.

184

equal portions of unidentified fish fins and wholeAstyanax bimaculatus. The largest S. medini speci-mens examined (71.6-80.6mm) corresponded tosubadult size classes. and all contained fins (thelargest specimen contained fins and associatedflesh of Hoplias). Between 40 and 50mm. Pygo-centrus shifted to a diet comprised primarily ofwhole fishes and sheared pieces of fish flesh. Awide variety of fish species were taken by Pygocen-Irus. but primarily Curimata argentea. small omni-vorous characids. and the large detritivorous char-aciform. Prochilodus mariae (Prochilodontidae).Adult Pygocentrus consumed portions of other ver-tebrates (turtles. birds. mammals) when fisheswere dispersed at low densities during the early wetseason. Hoplias also shifted to a diet comprised ofdiverse fISh species between 40 and 50 mm. butunlike Pygocentrus. Hoplias always swallowed itsprey whole. Hoplias is capable of consuming rela-

«I20mm). and Caquetia «40mm. Fig. 3).Ephemeroptera nymphs alone comprised 65% ofthe diet of juvenile Charax «30mm. Table I).Gymnotus smaller than 40 mm consumed 62Ofo chi-ronomid larvae (Diptera). whereas between 120and 240 mm. Ephemeroptera. Odonata nymphs.and chironomid larvae were predominant in Gym-notus stomachs. Juvenile Caquetia between 40 and70 mm preyed heavily upon a wide variety of aquat-ic insects (Table 2).

The shift to a diet comprised primarily of fishcomponents occurred between 20 and 30 mm for S.irritollS, and between 30 and 40 mm for S. mediniand S. rhombeus (Fig. 3). These narrow-bodiedpiranhas were fin specialists at subadult and adultsize classes (Tables 2.3). Fins comprised the majorfraction of the diet of S. rhombeus larger thanSOmm, although this sample consisted of only 2specimens. A 195 mm specimen contained nearly

T~ 1. Resource matrix of dominant rood items consumed by juvenile size classes or piscivorous fishes at Caito Maraca. Utilization isexpressed as a percentage of the total volume of the prey item based on all specimens. Nine fISh and 20 non-fish itenlS with a combinedvolume of 45.2 liters are not listed (HMA = Hopl;as molabaricus. CGI - Charax g;bbosus. PNO - Pygocenrrus nolalus. SIR =Serrasalmus ;rr;lans. SME = Serrasalmus medini. SRH - Serrasalmus rhombeus. GCA ~ Gymnotus carapo. RHA = Rhamd;a sp..CKR - Caqwlio trows;;).

HMA 001 PNO SIR SME SRH OCA RHA CKR

o.I.

17O.

I.71.2.

O.I.

I.I.

o.o.O.9.O.o.

37.6.4.4.O.1.1.2.

12.4.

1.

0.170.51

16.30

0.61

5.18

6.01 0.120.619.124.07

-0.24

11.960.072.290.400.020.03

17.~20.120.020.42

0.47O.~1.860.032.15

o.I.I.2.o.

65.4.9.7.O.O.O.

21.O.O.

9.61

O.M

4.50

1.~

1.~

0.0963.89

4.8410.666.92

16.73 20.

O.

1.

O.

2.

-1.04

1.72

5.664.. 1.8 0.49

14.2525.840.41

21.37

4.350.05

3.55

61.-

1.20

Prey category

Vegetative detritus

CopepodaOadocera

AmphipociaEubranchipodaOstracoda

EphemeropteraOdonata nymphsMisc. aq. HemipteraCorixidae (Hemiptera)

TrichopteraAq. Coleoptera (Iarv)Aq. Coleoptera (adlt)aaironomid larvae

Mosquito larv.(Diptera)Misc. aq. insectsFish finsMisc. fish

At'qllidt'tUPMCilia

HopliasRachovio

Coq~lia

68.238.635.823.613.887.497.21

4.19

J2B2,10,59

967949

7242

~45

os458S,2S294818388739045368417412

53

202740(1J264003787720a8782(M88

78

21

04

31

4S

185

s.lr888tnus thombeua

..-

..~-..-

'.0-

Uo;

u-OA-

u.

M.L:: ~ T - . .'-~~!'0 00 10 '. ,. -

~~(-)

GymnotJ8 ~

'20 '80 200 200 280 320

LengtllrMrval (mm)

. ~fI*IIct8 SII.

fO 80 120

L8ngtI kI-.V8J (mm)

s.r..tInu8lni18ns

.j

!

120.0 80

~ k8V8I (mm).-."'Aq.-.-

~ kt8usU....uuu

~20 .. 80 10 '00 120 '.. '.0L8ngtI ~ (mm)s ~

r ;.~+A4-..-

0.0 ~ ~:2::::::::::::::;?:!':. - : -0 20 40 " 10 100 120 140 ," ," 200 220 2-0LqfI J..-,y8I (mm)

Fig. 3. Volumetric proportions of microcrustacea. aquatic insects. and fish prey categories taken by piscivores grouped by length

interval.

u,u-0.0-

0.2-

0.0-0

186

upon small characids, especially Astyanax bimac-ulatus, Odontostilbe pulcher, and Ctenobrycon spi-lurus. Ctenobrycon spilurus and Curimata argenteawere the primary prey of large adult Gymnotus(>200mm). Rhamdia larger than 6Omm con-sumed mostly whole Curimata, but also fresh re-mains of Astyanax and Prochilodus that apparentlyresulted from frenzied attacks by schooling Pygo-centrus. The proportion of fish in stomachs washighly correlated with standard length (Table 4).

tively large prey items. For example, a 305 mmindividual had consumed a 95 mm specimen of theheavily armored, spined catfish, Hoplosternum lit-torale (Callichthyidae). A 288rnrn Hoplias con-tained a subadult Prochilodus measuring over athird its own body length. Cannibalism was observ-ed among all size classes of Hoplias.

Between 50 and 60 mm, Charax shifted fromfeeding primarily on aquatic insects to small char-acid fishes, particularly Odontostilbe pulcher. LikeCharax, Caquetia over 70mm preyed most heavily

Table 2. Partial resource matrix of dominant food items consumed by subadult size classes of piscivorous fishes at Cano Maraca. Sevenfish and 28 non-fISh items totalling 43.4 liters are not listed.

SRH GCA RHA CKRHMA CGI PNO SIR SMEPrey category

o.~

0.020.050.04

6.2510.352.52

14.890.360.450.260.19

0.720.189.59

40.16

7.56

.)"'"-,-

1.263.SJ

0.114.37

0.460.040.570.03

0.11

0.011.050.07

32.516.89

10.3717.19

4.375.S9

---

2".'J'

:;rc~~:

",~

...;'~

M.17

2.41

,..~

::-

.~-,ii,.':;;

! .,.

;~~

j,..;!!'";,

.-

.~

~

.

O.

M).

0

O.

8.

1.

6.

3.

7.

20.

S.

4.

O.

19.

O.

O.

3.

1.

7.

-

c"

...

0.010.21

7.484.760.570.040.10

0.070.26

10.52

3.584.37

3.732.01

25.795.878.313.58

9.67

5.16

.,.~

~~...,..

~:,..

-~~~

~;~J

100

c~;

~~. ~

~;~

-

O.

3.

S.

37.

20.

2.

~.'

1.

-0.011.240.79O.~7.722.022.41

Fine detritus

Aquatic vegetationClams (Mollusca)Oadocera

EubranchipodaShrimp (Decapoda)EphemeropteraOdonata nymphsMisc. aq. HemipteraCorixidae (Hemiptera)

Aq. Coleoptera (Iarv)Aq. Coleoptera (adlt)Chironomid larvaeMisc. aq. insects

OnhopteraFish finsMisc. fish fleshMisc. fish (whole)

CorydorasOdonloslilbe

AequidensGephyrocharaxHemigrammusPoeciliaCurimala

AstyanaxHopliosRachoviaT riportheusMarkianoCaqueliaCheirodontopsGymnotusHypostomusApislogramma

2.382.77

12.630.376.54

15.21

1.164.54

14.221.000.501.001.33

-9.93

-15.68

85

02

,'T7

.40

173S854389541207

0340897999

58999S

10

44OS

4682

,84

93,24

187

Tobit' 3. Panial resource matrix of dominant food items consumed by adult size classes of pisciyorous fishes at Cailo Maraca. Twentyfour fish and 32 non.fish items totalling 92.9 liters are not listed.

GCA RHA CKRCGI PNO SIR SME SRHHMAPrey category

5.457.277.271.191.884.353.463.464.15

0.8922.197.006.92

30.63

3.13

2.530.31

-

3.155i.S3

o.

-~~

.!6.

-

e

0.041.910.84

12.01

8.961.240.44S.202.00

3.641.20

26.009.369.88

0.010.15

0.160.260.06

O.OS3.07O.~4.16

20.194.43

25.236.172.24

~.67

0.861.01

12.13

0.4723.1837.13

3.56S.SS0.29

0.920.040.860.13

0.3114.462.220.072.302.34

22.643.747.910.14

S.:M>

S.SS

-2.17O.~

0.17

4.86

16.30

IS.tIJ

2.33

0.05

6.70

2.81

11.38

10.54

0.84

ff~%

~"t~..

76.19

:

n...

Vegetative detritus

Aquatic vegetationHard fruits, seedsFish ewSnails (Mollusca)

Shrimp (Decapoda)Odonata nymphs

OrthopteraArachnidaCtenoid scalesFish finsMisc. fish fleshMisc. fish (whole)OdontostiJbe

CtenobryconPimelodellaCurimata

AstyanaxProchiJidusRineloricaria

HopliasTriportheusRoeboides

HoplostemumNon-fish vertebrates

50--J7,jS..

;........

I.~

17.17

.,,;:..;::,'""" ..

~~

-

3.10

Table 4. Product moment correlations between the proportion

of fish in the diet and the standard length of piscivores contain-ing food (0 = P

188

... .-OT-DOoro.

c~ I.J0AU0~

8.,

1_1n 1 J 1_- .L _. m~.u t8.PI.Mf"'_U

food ~

Fig. 4. Relative seasonal availability of fish food resources asindicated by total volumes consumed by the complete CanoMaraca fish fauna (total volumes recorded in liters for eachseason appear in the key; AQ = aquatic. TR = terrestrial).

To some degree. seasonal differences in the fre-quency of empty stomachs were due to ontogeneticdiet shifts. Although no significant overall effect offish size on frequency of empty stomachs wasnoted. the largest fishes had significantly moreempty stomachs than the smallest size class (Table5). The feeding rate of large piscivores. such as

Hop/ias. often is relatively low due to the timerequired for digestion of massive prey items priorto the initiation of a new feeding bout. All of thespecimens of the largest size classes of the fin-nipping piranhas. S. irritans. S. medini. and S.rhombeus contained food (Table 5). Pygocentrusand Gymnotus in all three size classes appeared tobe frequent foragers. To some degree. frequencyof piscivorous feeding may be overestimated bystomach contents. because compact bones and stifffin rays are digested more slowly than soft tissues.Additionally. different digestion rates for predom-inantly juvenile versus adult prey probably affectsthe frequency of empty stomachs. Chitinous matterassociated with invertebrate feeding by smaller pis-civores passes through guts more slowly than ver-tebrate flesh.

Diet breadth among piscivores showed no signif-icant overall effect of season (ANOV A). Mean dietbreadth among piscivores was 4.97 during the wetseason. 4.72 during the transition season. and 3.89

Tllblt 5. Percent empty stomachs encountered during three seasons and within three piscivore size classes.

Species Wet DryTransition Tolal sampl,

HopliasCharax

Pygoct'nlrusS. irrilansS. mt'diniS. rhombt'us

GymnolusRhamdia

Caqut'tia

51.3

31.8

0

4.2

0

3.3

7.4

18.6

.$.1

10.92.86.4004.31.105.5

54.281.8

0

4.829.313.8

389

29S

238

76

68

SO

24M

124

370

13.4Mean 3.4 30.6 F= 3.67.2 DF. P

189

50

40

30

20

10HMA CGI ~ SIR SME IRK ~ MIA CKR

SpecIH 0

500.2 0.4 0.8 0.8 1.0 1.2

Fig. 5. Average diet breadths of piscivores during the wet,transition, and dry seasons at Cano Maraca (HMA = Hopliasmalabaricus. CGI = Clulrax gibbosus. PNO = Pygocmlrus no-tatus, SIR = Se"QSalmus irritans, SME = Se"asalmus medini,SRH = Serrasalmus rlJombeus. GCA = Gymnotus carapo.RHA = RhamdiQ sp., CKR = Caquffla kraussii).

40

30-020

10

during the peak dry season. The three Se"asa/musspecies exhibited relatively narrow diet breadths asa result of their fin-nipping specialization. Dietbreadths showed a trend of lower values duringdrier periods for Serrasa/mus spp., Hop/ias, Char-ax. and Rhamdia (Fig. 5). Pygocentrus and Caque-ria exhibited trends of somewhat more generalizedfeeding during drier periods. Gymnotus was a moregeneralized feeder during the transition seasonthan either the wet or dry seasons. During thetransition season, Gymnotus (31-3 mm) took avariety of small fishes and aquatic invertebrates,especially Odonata nymphs and Hydracarina.Gymnotus was the principal predator of Ctenobry-con spi/urus, a small characin that is abundant inthe area throughout the year.

Average seasonal diet overlap among piscivoreswas low (O.I8-wet, 0.19-dry, O.25-transition; Fig.6). For example, the proportions of cells within thepi/-based overlap matrices with values less than 0.10were 0.53 (wet), 0.57 (transition), and 0.65 (dry).During the wet season, the largest overlaps werebetween Charax and S. medini (q> = 1.11,0.46), S.irritans and S. rhombeus (q> = 0.70, 0.60), andGymnotus and Rhamdia (q> = 0.34, 0.74). Overlaps(Pij-based) equal to or exceeding 1.0 were obtainedfor Hop/ias x Pygocentrus. Hop/ias x Gymnotus.Hop/ias x Rhamdia, S. irritans x S. medini. and S.irritans x S. rhombeus during the transition sea-son. Among 20 pairwise comparisons, only the

Fig. 6. Frequency histograms for piscivore diet overlaps duringthree seasons at Cano Maraca.

Gymnolus by Rhamdia pairing showed large diet-ary overlap during the dry season (qJ = 0.64,0.85).

When diet similarities were calculated usingelectivities in place of Pi!', little diet similarity wasobserved among piscivores during any season. Theproportions of cells with electivity-based similar-ities less than 0.10 were as follows: 0.89 (wet), 0.76(transition), and 0.75 (dry). Average seasonal dietsimilarities were 0.05 (wet), 0.08 (transition), and0.10 (dry). Based on electivities, the highest dietsimilarity obtained for the wet season was qJ, = 0.43

1~

-..E

g

UN

..

>-U~D-

C.U

~

...2~;Vs.

'0c0

i~L

Fig. 8. Relative proponions of individuals of each species takenfrom four microhabitat samples during four months of tbe tran-sition season at Cado Maraca (species codes are the same as in

Fig. 5).

Spec-Fig. 7. Mean prey sizes encountered in diets or CaJ\o Maracapiscivor~s during the transition season. Error ban represent on~standard deviation. th~ number or pr~y items measured appeanwithin each open bar. and means ror species joined by linesbelow the x axis were not significantly different at a = 0.05according to Duncan"s Multiple Range Test (species codes arethe same as in Fig. 5).

(Pygocenlrus x S. rhombeus). Fin specialists, S.irritans x S. rhombeus provided the only transitionseason similarities exceeding 0.31 (tp, = 0.83, 1.19),whereas Hoplias on Charax (tp, = 0.96) represent-ed the only pairing above 0.24 during the dry sea-son.

Prey size

A significant relationship exists between prey sizeand piscivore species (F = 65.51. 8 OF.P

191

Although some opportunity for microhabitat spe-cialization probably does exist within the turbidwaters of the creek during these periods (e.g., ben-thic versus midlevel or surface), each habitat yield-ed rather consistent mixtures of species in seinehauls.

If one considers three niche axes simultaneously(prey type, prey size, habitat), and assumes conser-vatively that overlaps greater than 0.50 are biolog-ically significant, then piscivores separated on atleast one axis in all but four comparisons during thetransition season (prey type overlap based on PiJ.Of 36 pairwise species comparisons, six separatedon all three axes, 18 separated on two axes, andeight separated on one axis. Of the pairings thatshow high overlap on all three dimensions, Hop/iasand Pygocentrus actually had equivocal overlap forhabitat utilization (cp = 0.32, 0.87), sharing only

open, midpool regions to any large extent. Patternsof foraging behavior of these species show strikingdifferences as well (i.e., solitary, sit-and-wait Ho-p/ias vs. gregarious, pursuit Pygocentrus). Otherlarge, three-dimensional overlap pairings corre-sponded to fin-nipping piranha species (S. irri-tans x S. medini and S. rhombeus; S. medini x S.rhombeus). If electivity-based diet similaritiesgreater than 0.50 are used as the criteria for signif-icant dietary overlap, then only the S. irritans andS. rhombeus pairing showed significant three-di-mensional overlap during the transition season.

.

HMA CGI PNO SIR SME SRH O.lO)possessed by each piscivore species during the wet, transition,and dry seasons at CaJ\o Maraca (species codes are the same asin Fig. 5).

terns such as tropical fish assemblages. Total dietoverlaps were calculated as the sum of all pairwisediet overlaps exhibited by niche neighbors duringeach of the three seasons. Two pools of potentialniche neighbors were analyzed, the nine-speciespiscivore guild and the entire Cano Maraca fishcommunity. Because not all species were presentand response to aquatic hypoxia apparently super-ceded feeding during extended portions of the peakdry season, only wet and transition season diffuseoverlap trends are presented.

Analyses of the number of niche neighbors andpatterns of collective dietary overlap revealed twoclasses of seasonal response by piscivores. For fivespecies, the number of niche neighbors and totaldietary overlap were greater during the wet seasonthan the transition (Fig. 9, 10). Extreme trophicspecialization by subadult S. medini during thetransition season was indicated by a large drop intotal diet overlap relative to the previous samplingperiod dominated by juveniles (CPr = 17.S-wet, 2.7-transition). The remaining four species (Hop/iDS,Pygocentrus, S. i"itans, S. rhombeus) had compar-atively few niche neighbors throughout the year(Fig. 9). Moreover, relatively low total diet overlap(~

192

15HopI/.. nl8/.urlcu.

- - 0- - ~Ut«TYaULDCo

~.>0

..'ij

..-0I-

10

0 0

8

T'8n8111on 15We'

10

Q..~.>0.~

.."0I-

0r'8n8ltlonWet

GynwlOlu8 Qrapo

S."...,- /nt,.,.15

A.~~>0

..:;

'i"0I-

T'8naltlon5

~~~~~~048~~~~~~ ~~-~~~~~~-~~..0

TrlnlillenWet

s.n.sa/mus ,ho~s

Do~.>0

i'i

..0I-

TransitionWe'

Fig. 10. Total diet overlap exhibited by each piscivore species during different seasons at Caiio Maraca. Total community diet overlaps

are represented by open symbols. and total piscivore guild overlaps are plotted as closed symbols.

193

during the transition season, both within the pisci-vore guild and the community at large (Fig. 10). Inthe cases of Gymnotus, Caquetia. and Serrasa/musmedini. collective guild overlaps were low relativeto overall community overlaps during both the wetand transition seasons. Rhamdia exhibited a slightincrease in total guild overlap between the wet andtransition seasons, although overall communityoverlap declined.

Werner & Hall 1979. Machado-Allison & Garcia1986. Keast 1985b). Morphological changes associ-ated with greater trophic specialization in the formof piscivory should reduce the efficiency of mi-croinvertebrate feeding. Following this reasoning.Griffiths (1975) recognized a dynamic trade-off be-tween maximizing prey number versus prey energyin the feeding strategies exhibited by piscivorousyellow perch. Perca flavescens, during ontogeny.

Discussion Resource partitioning

Food resource partitioning was widespread withinthe Cafio Maraca piscivore guild, particularly dur-ing the transition season. Given the high concord-ance between ontogenetic diet shifts, prey avail-ability, and seasonal rainfall, can high levels ofresource partitioning be interpreted in relation tobiological interactions, especially interspecificcompetition? Three general, alternative explana-tions could account for observed patterns of re-source exploitation. First, resource exploitationcould occur at random. Yet sufficient subdivisionof available resources was observed among thenine piscivores as to render this null hypothesisunlikely. For example, average seasonal diet over-laps were considerably lower (q> = 0.18 - 0.25)than those obtained for series of randomly con-structed 100species communities (q> = ca. 0.75,Pianka et al. 1979). Second, contemporary patternsof niche partitioning could be founded in genetical-ly-based morphological and behavioral traits thatevolved in response to past environmental condi-tions that differed from current conditions (sensuConnell 1980). This second hypothesis contrastssharply with a third possible explanation. that con-temporary patterns of resource subdivision areadaptive in the context of present day biotic inter-actions. Interspecific competition is foremostamong biotic mechanisms cited in studies of com-munity resource partitioning, although the role ofpredation has recently received greater attention,particularly with respect to habitat selection (Fras-er & Cerri 1982, Sib 1982, Werner et al. 1983,Mittlebach 1984, Power 1984. Power et al. 1985.Hobson & Chess 1986).

Ontogeny and diet

All nine sympatric, tropical piscivores showed dis-tinct shifts from invertebrate feeding by small juve-niles to primarily fish feeding by subadult and adultsize classes. Predominance of fin nipping by sub-adult and adult size classes of serrasalmine pira-nhas is consistent with earlier findings (Goulding1980, Machado & Garcia 1986, Northcote et al.1986, 1987, Sazima & Zamprogno 1986, Nico &Taphorn 1988). Three primary factors probably areinvolved in producing size-related patterns of feed-ing among piscivores. First, juvenile fishes are con-strained by their small size to exploit relativelysmall food particles (Werner & Gilliam 1984, Keast1985a, 1985b). Only following a period of initialgrowth can whole-fish swallowers switch to largerfish prey. Second, availability of small zooplanktonand benthic invertebrates (e.g., aquatic insect lar-vae) was much greater during the wet season com-pared to either the transition or dry seasons atCaiio Maraca (Fig. 4). Observed switching frominvertebrate to fish prey largely coincided withchanging food availabilities as indicated by feedingperformance of the entire fish assemblage. Where-as total aquatic primary production and aquaticinvertebrate availability declined during the gradu-al desiccation of the floodplain, fish densities in-creased markedly. Jackson (1961) made a similarobservation for African fishes in seasonal aquatichabitats. Third, many predatory fishes exhibitchanges in relative body proportions and other an-atomical traits during growth that are associatedwith greater feeding specialization (Werner 1974,

194

In recent reviews of resource partitioning in low-er vertebrates, Toft (1985) and Ross (1986) in-dicated that causal factors responsible for differ-ences in resource use are rarely known. MacNally(1983) and others have argued that descriptive fieldstudies alone cannot demonstrate competition.Whereas field experiments potentially can demon-strate interspecific competition (Schoener 1983),they are beset with problems of interpretation aswell. Positive, negative, or even a complete lack ofpopulation response to experimental manipula-tions potentially are confounded by effects of in-direct interactions (Bender et al. 1984, Tilman1987) or measurement of inappropriate variables(Diamond 1983, MacNally 1983). Even though de-scriptive studies cannot demonstrate competitiondirectly, they often provide strong circumstantialevidence with the potential to support or reject thehypothesis of interspecific competition. The infer-ential value of this circumstantial evidence dependsupon the strength of comparisons based on inter-population or temporal and spatial environmentalvariation.

In this study, trophic specializations were associ-ated with lower diffuse diet overlap during thetransition season when fish population densitieswere highest and availability of invertebrate preywas reduced. Due to the general pattern of in-terspecific synchrony in reproduction with the on-set of the wet season, ontogenetic shifts towardmore specialized piscivore diets largely corre-sponded with the progression of the transition sea-son and fish growth. These observations are consis-tent with McKaye & Marsh's (1983) hypothesis thatAfrican cichlid fishes switch from more opportunis-tic, generalized diets to more specialized feedingduring periods of relative resource scarcity (i.e.,following the 'niche overlap hypothesis' of Pianka1974). Morphological adaptations for specializedfeeding may be selectively advantageous during alimited period during the annual cycle of ecologicalevents. At Caito Maraca, four species that switchedto piscivory at small sizes also exhibited low collec-tive diet overlaps throughout the year (Fig. 10).

In summarizing work involving tropical fish com-munities, Lowe-McConnell (1987) suggested thatcompetition is potentially most important during

the late wet and dry seasons in floodplain ecosys-tems. During this 'crunch period', resources be-come increasingly limited (especially resources re-lated to space) as local stocks of aquatic organismsbecome denser within shrinking bodies of water.Biotic interactions, especially fish predation, in-tensify at this time, but are piscivores necessarilyresource-limited? Whereas apparent availability offish prey may be high during the transition season,niche segregation and trophic specialization amongpiscivores should OCCl'r for two reasons. First, fol-lowing optimal foraging theory (Krebs 1978, Pyke1984), scarcity of invertebrate prey and temporaryabundance of prey fishes would favor exploitationof the most profitable forms by predators, based ontheir morphology and innate components of feed-ing behavior. Profitability would be expressed herein terms of the ratio of energy gained by maximiz-ing the rate of prey capture, divided by the totalenergy cost associated with searching, pursuit,handling, digestion, and assimilation. Evolution-ary divergence of prey escape tactics increases re-quirements for specialization on a restricted num-ber of prey species (Rand 1967). Second, the brev-ity of the season of high fish density, together withthe high risk of mortality due to predation and dryseason dessication/anoxia, conveys a tremendousfitness advantage to individuals that maximize theirintake of temporarily abundant prey, then survivethe ensuing dry season. A disproportionate frac-tion of the next generation would be founded byefficient piscivores that are larger and more fecund(or more effective in providing parental care) thanless efficient conspecifics. In effect, a major frac-tion of the annual caloric intake of tIanos piscivoresis concentrated into the approximately four-monthtransition season. Relative to other seasons,growth and visceral fat deposition is maximal formost species during the transition season (personal

observation).Given the selective premium that is placed on

highly efficient foraging by subadult and adult pis-civores during the transition season, it appears like-ly that observed trophic specializations are adap-tive in reducing the negative impact of diffuse com-petition within this rich fish assemblage. For com-petition to occur, resources must be in limited~

195

mass in stomach contents was estimated fromlength/mass relationships. Both of these factors in-crease risks of sampling error. Third, and mostseriously, the taxonomic identities of the eight fishspecies are confused (see Taphorn & Lilyestrom1985 for listing of regional ichthyofauna). The onlyunconfused species name, Roeboides dayi (Stein-dachner), is perhaps the most suspect of the eight.Prejs & Prejs (1987) found 99.9 and 100% vegeta-tion and detritus in the stomachs of R. dayi duringthe dry season, whereas fish scales comprised 76(N = 121) and 25% (N = 49) of the diet of R. dayiat Calio Maraca during the transition and dry sea-sons respectively. Scales formed 24 (wet) and 15%(dry) of the diet of R. dayi by volume at CalioVolcan, a small piedmont stream (N = 313, Wine-miller 1987a). All other Roeboides species, forwhich data are available, specialize on scales inaddition to aquatic insects (Sazima 1983, Wine-miller 1987a). Most of the fishes referred to as 'R.dayi' by Prejs & Prejs (1987) were probably Cre-nobrycon spilurus (Valenciennes), a morpholog-ically similar, omnivorous characid that is extreme-ly common in shallow, productive environmentsthroughout the llanos.

Some temperate zone studies also have demon-strated greater diet segregation among fishes dur-ing periods of reduced food availability (Nilsson1955, Harrington & Harrington 1961, Thorman1982). In addition to diet segregation, interspecificsize and time related differences in habitat use havebeen demonstrated for both temperate and tropicalfish assemblages (Werner & Hall 1977, Moyle &Senanayake 1984, and see Ross 1986 for others).Habitat segregation by Calio Maraca piscivoresprobably was as much a function of predator avoid-ance as food resource segregation. Fishes used aq-uatic vegetation as the principal form of cover foravoidance of predation by dense schools of red-bellied piranhas, Pygocentrus noralus. Additional-Iy, and perhaps to a lesser extent, smaller fin-nip-ping Serrasalmus species posed a threat to diurnal,open-water dwellers. Where it is common, Pygo-centrus may have a pervasive effect on the spatialstructuring of fish communities, since only highlycryptic, heavily armored, or fast schooling fisheswere generally taken from deep open-water re-

supply and contested by two or more populations.The presence of numerous heterospecific pisci-vores during the transition season of high fish den-sity could restrict foraging to the most profitableprey types. Reductions in pairwise diet overlapduring periods of reduced food availability havebeen shown in several earlier studies of fish com-munities (see Ross 1986). Though based on limitedsampling, reduced diet overlap among fishes wasreported by Zaret & Rand (1971) and Greenfield etal. (1983) during the dry season in Central Amer-ican streams. Alternatively, lowe-McConnell(1964), Knoppel (1970), and Goulding (1980)stressed high dietary overlap in South Americanfish communities, although no attempts were madeto study trophic guilds quantitatively over time.The findings of the current study contrast sharplywith those of Knoppel (1970), who interpreted hisdata as indicating a perplexing lack of correspond-ence between morphological characteristics anddiets of fishes from five stream collections in cen-tral Amazonia. Also in opposition to the currentfindings was Knoppel's (p. 346) contention that'stomach contents according to age of fishes did notmerit close consideration'. Quite the contrary,Knoppel probably did not examine stomachs andsize classes closely enough to discern the obvious

patterns.Recently, Prejs & Prejs (1987) observed higher

dietary overlap during the dry season among small,vegetation-dwelling fishes in seasonal pools of theVenezuelan low llanos. These authors interpretedpatterns of diet overlap as ecologically unimpor-tant, because their data showed reduced feedingrates or a switch to detritivory during the dry sea-son (apparently, the dynamic transition season ofthe llanos was not sampled). Lower availability ofinsect prey and increased predation pressure werebelieved to be the primary factors accounting forreduced dry season feeding; however, results ob-tained by Prejs & Prejs (1987) must be interpretedwith caution for several reasons. First, their com-munity subset of eight small, vegetation-dwellingfishes does not represent a trophic guild, but rathera diverse subset of small vegetation-dwelling fish-es. Secondly, total sample sizes for stomach con-tent analysis were small (N = 9-30), and prey bio-

196

the Mato Grosso region of western Brazil (I. Sazi-ma, personal communication). Further field andlaboratory studies are needed to sort out the nicherelations of fin-nipping piranhas more accurately.

Conclusions

High fish species diversity in the neotropics is asso-ciated with impressive morphological and ecolog-ical diversification (Roberts 1972, Fink & Fink1979, Goulding 1980, Sazima 1986, Lowe-McCon-nell 1987). Coexistence of more than 100 fish spe-cies in some llanos creeks (Taphorn & Lilyestrom1985) may be due in part to relaxed interspecificcompetition during horizontal flooding and thebloom of primary production that occurs during thewet season. Diets of five of the nine piscivoresexamined overlapped more with the rest of the fishcommunity during this period than any other.Large diffuse overlap during the wet season largelyresulted from heavy exploitation of abundant mi-crocrustacea and aquatic insect larvae by juveniles.Lower collective overlap during the transition sea-son probably was derived from complex interac-tions among several factors, including: (1) reducedavailability of juvenile food resources, (2) fishgrowth and ecomorphological trade-offs favoringdiet shifts, (3) increased fish densities and diffusecompetition favoring maximal rates of prey exploi-tation, (4) differential prey profitability due totrade-offs associated with prey escape tactics, and(5) the threat of predation by piranhas restrictingdiurnal foraging by other piscivores in open-watermicrohabitats. In highly diverse communities, dif-fuse overlap with numerous niche neighbors repre-sents a greater potential negative effect than isolat-ed pairwise interactions.

gions of Cano Maraca and other ilanos streamsduring daylight. Nocturnal predators (e.g. Rham-dia and Gymnotw) leave the dense mats of aquaticvegetation to forage in the open water habitat un-der cover of darkness. Approximately 80 man-hours of nocturnal hook-and-line fishing at astream in the low llanos of Apure state (location ofCano Maporal given in Nico & Taphorn 1988)yielded nine species of large piscivores (unpublish-ed data). Diurnal hook-and-line fishing over thesame period yielded 13 large piscivorous species,yet only three species occurred in both samples(e.g. Pygocentrus notatus, Pseudoplalystoma las-ciatus, and Plagioscion squamosissimw). Sixty sev-en percent of the nocturnal sample was comprisedof siluriforms (catfishes and knifefishes), whereas92% of the diurnal sample belonged to the ordersCharaciformes (piranhas, tetras, and relatedforms) and Perciformes (cichlids and one fresh-water drum). Furthermore, most diurnal piscivoreswere taken near cover at the stream margin, fur.ther supporting the hypothesis that Pygocentrusrestricts diurnal use of the open-water, midpoolhabitat by other piscivores.

The three fin-nipping piranhas exhibited thehighest similarities in diet and habitat niche para-meters at Cano Maraca. As might be expected dueto genetic and morphological similarities, closelyrelated species frequently exhibit relatively lessecological segregation than more distant taxa.Lack of diet segregation among fin-nipping pira-nhas, especially Serrasalmw irritans and S. rhom-beus subadults, was due in part to my inability toidentify fin fragments at the species level in mostinstances. Records of the incidence of fin nips onpreserved Cano Maraca specimens clearly indicatethat not all fishes are equally vulnerable to Serra-salmw attacks (unpublished data). Preliminarytests of aquarium-housed juveniles and subadultssuggest that Serrasalmw species possess distinctpatterns of foraging behavior associated with eitherstealth in open water (S. irritans; see Nico & Ta-phorn 1988 for a brief description of field observa-tion), persistent chasing with frequent remaneu-vering (S. medini), or use of cover for ambushing(S. rhombew). Similar behavioral patterns wereobserved among three syntopic piranha species in

Acknowledgements

I thank D.C. Taphom for his support and hospital-ity in Venezuela. The field assistance of D.C. Ta-phom, L. Nico, A. Barbarino, and N. Greig wasgreatly appreciated. I thank P. Urriola, R. Schar-gel, F. Mago-Leccia, and C. Hubbs for assistance

197

in obtaining permits and visas. The Direcci6n Ad-ministracion y Desarrollo Pesquero of the Repub-lic of Venezuela provided a collecting permit. I amgrateful to D. Urriola for kindly allowing me towork on his property. This study was derived froma doctoral dissertation submitted to the Depart-ment of Zoology of the University of Texas inpartial fulfillment of the requirements for thePh.D. I thank my advisors E.R. Pianka and C.Hubbs for their support. Funds were provided by aresearch grant from the National Geographic So-ciety and an NSF predoctoral research grant.

References cites

African lakes. Cornell University Press. Ithaca. 8J9 lIP.Griffiths. D. 1975. Prey availability and the food of predators.

Ecology 56: 1209-1214.Goulding. M. 1980. The fishes and the forest: explorations in

Amazonian natural history. University of California Press.Los Angeles. 21M) pp.

Harrington. R.W. & E.S. Harrington. 1961. Food selectionamong fishes invading a high subtropical salt marsh: fromonset of ~ng through the progress of a mosquito brood.Ecology 42: 646-«16.

Helfman. G.S. 1978. Patterns of community structure in fishes:summary and overview. Env. Bioi. Fish. 3: 129-148.

Hobson. E.S. & J.R. Chess. 1986. Relationships among fishesand their prey in a nearshore sand community off southernCalifornia. Env. Bioi. Fish. 3: 201-226.

Hyslop. E.J. 1986. The food habits of four small-sized species ofMormyridae from the floodplain pools of the Sokoto-Rimariver basin. Nigeria. J. Fish Bioi. 28: 147-151.

Jackson. P.B.N. 1%1. The impact of predation. especially bythe tiger-fish (Hydrocyon vinalllS Cast.) on African fresh-water fishes. PrOto Zool. Soc. Lond. 136: M3-622.

Keast. A. 1985a. The piscivore feeding guild of fishes in a smallfreshwater ecosystem. Env. Bioi. Fish. 12: 119-129.

Keast. A. 1985b. Development of dietary specializations in asummer community of juvenile fishes. Env. Bioi. Fish. 13:

Angermeier, P.L. '" J.R. Karr. 1983. FISh communities alongenvironmental gradients in a system of tropical streams. Env.Bioi. Fish. 9: 117-135.

Araujo-Ulna. C.A.R.M., L.P.S. Portugal '" E.G. Ferreira.1986. FISh-macrophyte relationship in the Anavilhanas Ar-chipelago. a black water system in Central Amazon. J. FishBioi. 29: 1-11.

Bender, E.A.. T.J. Case & M.E. Gilpin. 1984. Perturbationexperiments in community ecology: theory and practice.Ecology 65: 1-13.

Brooks. J.L. &. S.J. Dodson. 1965. Predation. body size andcompostition of plankton. Science 150: 28-35.

Bullock. T.H. 1982. Electroreception. Ann. Rev. Neurosci. 5:121-170.

Connell, J.H. 191KJ. Diversity and the coevolution of compet-itors, or the ghost of competition past. Oikos 35: 131-138.

Diamond. J.M. 1983. Laboratory, field and natural experi-ments. Nature 304: 586-587.

Fink. S.V. '" W.L. Fink. 1981. Interrelationships of the

198

diel activity. food habits and spatial distribution of juvenilecreek chub. ~moli(us atromacu(alus. Env. Bioi. Fish. 11:301-307.

Mago-Leccia. F. 1970. Preliminary studies on the ecok>gy offishes in the ilanos. Acta Bioi. Venez. 7: 71-102. (In Spanish).

McKaye. K.R. & A. Marsh. 1983. Food switching by two spe-cialized algae-scraping cichlid fIShes in Lake Malawi. Africa.Oecologia 56: 245-248.

Mittlebach. G.G. 1984. Predation and resource partitioning ontwo sunfishes (Centrarchidae). Ecology 65: 499-513.

Moyle. P.B. &. F.R. Senanayake. 1984. Resource pamtioningamong the fIShes of rainforest streams in Sri Lanka. J. ofZool. London 202: 195-223.

Nico. L.G. &. D.C. Tapbom. 1988. Food habits of piranhas inthe low llanos of Venezuela. Biotropica 20: 311-321.

Nilsson. N.A. 1955. Studies on the feeding habits of trout andchar in Donh Swedish lakes. Rep. Inst. Freshw. Res. Drott-ningholm 36: 163-225.

Nonhcote. T.G.. R.G. Nonhcote &. M.S. Arcifa. 1986. Differ-ential cropping of the caudal fin lobes of prey fisbes by thepiranha. S~"a.salmus spilopltUra Kner. Hydrobiologia 141:199-205.

Nonhcote. T.G.. M.S. Arcifa & O. Froehlich. 1987. Fm-feedingby the piranha (S~~aImus spilopl~ura Kner): the croppingof a novel renewable resource. Proc. 5th Congr. Europ. Ich-thyol. Stockholm 1985: 133-143.

Pianka. E.R. 1974. Nidle overlap and diffuse competition.Proc. Nat. Acad. Sci.. USA 71: 2141-2145.

Pianka. E.R. 19m. Guild structure in desen lizards. Oikos 35:194-201.

Pianka. E.R.. R.B. Huey &. L.R. Lawlor. 1979. Niche segre-gation in desen lizards. pp. 67-115./n: D.J. Horn. R. Mitch-eU. &. G.R. Stairs (ed.) Analysis of Ecological Systems. OhioState University Press. Columbus.

Power. M.E. 1983. Grazing responses of tropical freshwaterfIShes to different scales of variation in their food. Env. Bioi.FISh. 9: 25-37.

Power. M.E. 1984. Depth distribution of armoured catfish:predator-induced resource avoidance? Ecology 65: 523-528.

Power. M.E.. W.J. Matthews &. A.J. Stewart. 1985. Grazingminnows. piscivorous bass and stream algae: dynamics of astrong interaction. Ecology 66: 1448-1456.

Prejs. A. & K. Prejs. 1987. Feeding of tropical freshwaterfishes: seasonality in resource availability and resource use.Oecologia 71: 397-404.

Pyke. G.H. 1984. Optimal foraging theory: a critical review.Ann. Rev. Ecol. Syst. 15: 523-575.

Rand. A.S. 1967. Predator-prey interactions and the evolutionof aspect diversity. Atas Simp. Biota Amazonn 5: 73-83.

Roberts. T.R. 1970. ScaIe-eating American characoid fishes.with special reference to Probolodus Mteroslomus. Proc. Cal.Acad. sa. 38: 383-~.

Robens. T.R.lm. Ecology of fishes in the Amazon and Congobasins. Bull. Mus. Comp. Zool.. Harvard 143: 117-147.

Ross. S. T. 1986. Resource panitioning in fish assemblages: areview of field studies. Copeia 1986: 352-388.

Sale. P.G. 1979. Habitat panitioning and competition in fIShcommunities. pp. 323-331. In: H.E. Oepper (ed.) Predator-Prey Systems in Fisheries Management. Spon Fishing Inst..Washington. D.C.

Sazima. I. 1983. Scale-eating in characoids and other fishes.Env. Bioi. Fish. 9: 87-101.

Sazima. I. 1986. Similarities in feeding behavior between somemarine and freshwater fIShes in two tropical communities. J.Fish Bioi. 29: 53-65.

Sazima. I. & C. Zamprogno. 1985. Use of water hyacinths asshelter. foraging place. and transpon by young piranhas.~muGlmus .Jpilop/~ura. Env. Bioi. Fish. 12: 237-240.

Schoener. T.W. 1974. Resource partitioning in ecological com-muniajes. Science 18S: 27-39.

Sdtoener. T.W. 1983. Field experiments on interspecific compe-tition. Amer. Nat. 122: 240-285.

Schut. J.. S.S. deSilva & K. Konmulder. 1984. Habitat, associ-ations and competiajon of eight Barbus (= Punlius) species(Cyprinidae) indigenous to Sri Lanka. Netherlands J. ?ro.34: 159-181.

Sih. A. 1982. Foraging strategies and the avoidance of predationby an aquatic insect. Nolon«ta hoffmanni. Ecology 63: 786-796.

Simpson. E.H. 1949. Measurement of diversity. Nature 163:688.

Siobodcbikoff. C.N. .It W.C. Schulz. I~. Measures of nicheoverlap. Ecology 61: IOSI-IOS5.

Strong, D.R.. A. Szyska & D. Simberloff. 1979. Tests of com-munity-wide character displacement against null hypotheses.Evolution 33: 897-913.

Taphom. D.C. .It C.G. Ulyestrom. 1985. Fishes of the Fernan-do Corrales module. Results of the ichthyological investiga-tion of project CONICIT.PIMA-18. Rev. Cie. Tccn.. UN-ELLEZ 3: 55-8S.

Thorman, S. 1982. Niche dynamics and resource panitioning ina fish guild inhabiting a shallow estuary on the Swedish westcoast. Oikos 39: 32-39.

Tilman, D. 1987. The imponance of mechanisms of interspecificcompetition. Amer. Nat. 129: 769-774.

Toft. C.A. 1985. Resource panitioning in amphibians and rep-tiles. Copeia 1985: 1-21.

Wallace. R.K. 1981. An assessment of diet-overlap indexes.Trans. Amer. Fish. Soc. 110: 72-76.

Watson. D.J. .It E.K. Balon. 1984. Ecomorphological analysisof fISh tax

199

Werner. E.E.. J.F. Gilliam. D.J. HaU & G.G. Mittlebach.1983. An experimental test of the effects of predation risk onhabitat use in fish. Ecology 64: 1540-1548.

Werner. E.E. & D.J. Hall. 1977. Competition and habitat shiftin tWo sunfishes (Centrarchidae). Ecology 58: 869-876.

Werner. E.E. " D.J. Hall. 1979. Foraging efficiency and hab-itat switching in competing sunfishes. Ecology ti>: 256-264.

Winemiller, K.O. 1987a. Tests of ecomorphological and com-munity level convergence among neotropical rlsh assemblag-es. Ph.D. Dissertation, University of Texas. Austin. 370 pp.

Winemiller. K.O. 198Th. Feeding and reproductive biology ofthe currito. Hoplostt'mum littoralt'o in the Venezuelan ilanoswith comments on the possible function of the enlarged malepectoral spines. Env. Bioi. Fish. 20: 219-227.

WinemiUcr. K.O. 1989. Development of derrnallip protuber-ances for aquatic surface respiration in South American char-acid fishes. Copeia (in press).

Zaret, T.M. & A.S. Rand. 1971. Competition in tropical streamfishes: SUPf'On for the competitive exclusion principle. Eco-~ 52: 336-342.