:>ecologia (1989) 81 :225-241 (k(:i)logia D Springer-Verlag 1989
Patterns of variation in life history among South American fishes in seasonalenvironments
Kirk O. WillelDiUer Departmentof Zoology, The University of Texas,Austin, TX 78712.USA s..mary. Ten traits related to life history theory weremea- demographic parameters (Steams 1983a; Reznick 1985; sured or estimatedfor 71 freshwaterfish speciesfrom two Peaseand Bull 1988). For example, models have incorpo- locations in the Venezuelanllanos. Multivariate statistics rated negativecorrelations betweencurrent investmentand and cluster analysis revealedthree basic endpoint patterns future investments in reproduction, juvenile survivorship bounding a two-dimensional continuum. A suite of attri- and clutch size. reproductive effort and adult survivorship, butesassociated with parental care and aseasonalreprodo<:- and mean generationtime and the intrinsic rate of increase. tion appeared to correspond to an equilibrium strategy. Alternative life history predictions can be testedby evaluat- A secondgroup of small fISheswas distinguishedby traits ing patterns of covariation among reproductive and demo- associatedwith rapid colonizing ability: early maturation. graphic parameters among heterospeciflCpopulations or continuous reproduction, and small clutches. The third conspecificdemes that have experienceddifferent environ- basic pattern was associatedwith synchronizedreprodo<:- mental conditions for long time periods. The presentstudy tion during the early wet season.high f~undity, absence constitutes one such test. characteriDng patterns of repro- of parental care. and breeding migrations. A subset of duction in tropical fishes with a seriesof dependentvari- mostly small fishes exhibiting little or no parental care. ables.The degreeof speciesclustering in multivariate space small clutches.and two to four month reproductiveseasons indicates the likelihood that certain combinations of traits was intermediatebetw~ the opportunistic (rapidly colon- will occur together in nature. Finally, I evaluate species izing) and seasonalstrategies. All ten life history variables clusters for associationwith phylogeny and ~Iogical fac- showedsignificant effectsof phylogeny.The cluster of spe- tors. ciescorresponding to the equilibrium group was dominated Tropical freshwaterfishes exhibit large diversity in mor- by siluriform fishes and perciforms of the Cichlidae. The phological, physiological. and ~logical attributes (Lowe- opportunistic cluster was dominated by cyprinodontiform McConnell 1987).Despite few comparative studiesof tropi- and characiform fishes. whereasthe seasonalcluster con- cal fish reproductive strategiesto date (Africa-Welcomme tained primarily characiform and siluriform fishes. Seven 1969; Central America-Kramer 1978; Asia-DeSilva et al. of nine traits weresignificantly correlatedwith body length. 1985),these diverse fish assemblagesprovide excellentsys- The three reproductive patterns are interpreted as being tems for evaluation of life history patterns. The great drain- adaptativewith respectto relative intensity and predictabi- age basinsof South America harbor the most diversefresh- lity of temporal and spatial variation in abiotic environmen- water fish assemblageson earth (ca. 2400described species). tal parameters.food availability. and predation pressure. Rather paradoxically, this ichthyofauna is derived from Key wonls:Llanos - Reproduction- Strategies- Tropical only a few basic stocks (speciesbelonging to the Characi- fishes-Venezuela formes. Siluriformes. and Cichlidae (Perciformes)comprise about 93% of all species,Lowe McConnell 1987).I investi- gated fish reproductive biology in the llanos of Venezuela during twelve co~tive months in 1984.The region has The diversity of methods by which organisms reprodtx:e an averageannual temperatureof 26.2° C and a highly sea- has long intrigued naturalists. One can argue that traits sonal pattern of rainfall, with 1189mm of precipitation oc- associatedwith reproduction should be subject to intense curring from June-Novemberand 178 mm from December- natural selection, as these directly affect the individual's May in 1984. Despite extreme changesin environmental geneticrepresentation within subsequentgenerations. Alter- conditions caused by seasonalrainfall. fishes exhibited a native modes of reproduction have obvious demographic wide rangeof reproductivecharacteristics in the llanos. This implications, sinceall populations experienceeither gradual report describesthese characteristics for 71 speciesand in- or periodic turn-overs.Alternative patternsof reproduction terprets the basic patterns within an ~logical context.~ and the degreeto which these are s~ful in different environmentalsettings have formed the basisfor both theo- Methods retical and empirical researchunder the headinglife history (Steams 1976; Southwood 1977; Horn 1978; Sibley and Collections Calow 1986a).In the most basic terms, life history theory During every month of 1984, fishes were collected from involvesconstraints, or trade-offs, among reproductiveand two locations in the state of Ponuguesa, Venezuela. The 226 more faunistically diversesite, Cano Maraca, is a swamp- servedin 10% fonnalin and depositedin the MHN (UNEL- creek located in the western llanos region (8°52'30" Lat. LEZ) and the TNHC (TMM). N; 69°53'30" Long. W). The area studied (termed an "es- tero") lies within low, flat terrain that experiencesextensive The data sheet flooding during the wettest months (June-August). During this time, the creek's broad floodplain is converted Several attributes related to reproduction and population from exposed,sun-baked soil and thorn-scrub habitat into structure weremeasured directly and other life history traits a productive marsh. The aquatic habitat is reduced to a werecalculated or estimatedfrom these.All preservedspec- network of pools (averagedepth ca. 1.0 m) during the driest imens were identified and measured for standard length months(December-May). Dissolved oxygen concentrations (SL). When available, 30 specimensof each speciesfrom are reducedduring this time, and many fishesrely on special eachmonthly collection at both Calio Maraca and C. V 01- respiratory adaptationsfor survival (Winemiller 1988). can were dissected.The condition of the gonad was coded The other site, Cano Volcan (8°59'15" Lat. N; basedin its relative sizeand color using the following cate- 69°53'30" Long. W), is a third order stream in the lowest gories: clear (1), translucent(2), opaque(3), small (1), medi- tier of the Andean Piedmont. This narrow stream flows um/small (2), medium (3), medium/large (4), and large (5). through deciduousforest and contains both pool and rime The overall gonad code for each individual was equal to habitats. Cano Volcan differs from C. Maraca in having (color code+size code) divided by two, and as a result, a slightly steepergradient, more stabledry seasondischarge, values ranged from 1.0 to 4.0. Although the same basic and a more forested watershed.Cano Volcan experiences criteria were usedin coding male and female gonads,inter- frequent but brief flash floods during wet seasondown- sexual and interspecific differences in the size, color, and pours. Physico-chemicaldata for each site during the year texture of fully mature gonads were taken into account. studied are given in Winemiller (1987). Undevelopedovaries were tiny transparent, globular or la- Fishes were collected by dipnet, seine (3.2 mm mesh/ mellar structures. Following a gradual transition, fully-de- 2.5 m length; 12.7mm/20 m), experimental gillnet, and veloped fish ovaries were large opaque-yellow or orange hook and line. At eachsite, an attempt was made to sample structures having a grainy appearancedue to the presence the entire fish community such that the sample for each of mature oocytes. Mature ovaries are cylindrical or lie speciesreflected its relative abundanceand population size as sheetsalong the lateral walls of the abdominal cavity. structureduring eachmonth. The collecting effort expended Immature testesappeared as transparent threadlike or min- each month was approximately equal for each site, con- ute lamellar structures in contrast to the smooth, opaque, sisting of three or four hauls of the 20 m seine in open milky-white appearance of mature testes. Fully-mature water habitats, and alternateuse of the 2.5 m seine,dipnets, testeswere either cylindrical. lamellar. or multilobed (the and gillnets for a minimum of 6 h within open water, shore- latter condition occursin catfishesof the Pimelodidae).For line, vegetation,and peripheralpool habitats. Samplingwas severalspecies, gonad codeswere regressedagainst gonado- terminated when two hours of collecting yielded no addi- somatic indices to test their reliability as overall indices tional rare species.Collections used for comparisonsof rela- of gonadal development(Fig. 1). tive population densitieswere made during either one or Diameters of ten of the largest oocytes in each mature two days betweenthe 11th and 28th of each month. Addi- ovary were determined using a dissectingmicroscope and tional collections were made on other dates in order to an ocular micrometer. Oocyte sizeswithin each examined supplementthe numbersof uncommon speciestaken in the ovary wereclassified as extremely uniform, moderately uni- standard samples.Except for very abundant species(for form, moderately variable, or extremely variable. The which a subsamplerepresentative of the abundance rank number of separateoocyte size classespresent in each ma- in the collection was retained),all collectedindividuals were ture ovary was recorded whenever these were essentially preservedin 15% formalin to assurepreservation of viscera non-overlapping. The fat content of the body cavity was and stomach contents. Following examination, preserved coded using the following criteria: 1 = none, 1.5= traces of specimenswere deposited in the Museo de Historia Natural fat in connective tissue of the coelomic cavity lining, 2 = de la Universidad de los Llanos OccidentalesEsquiel Za- small amounts around viscera and traces in the connective mora, Guanare, Portuguesaand the Natural History Col- tissueof the coelomic cavity lining, 2.5 = moderatedeposits lection of the TexasMemorial Museum, University of Tex- around the viscera and a thin layer on the coelomic cavity as, Austin. lining, 3 = large deposits around the viscera and coelomic During March, June, July, September,and November cavity lining, but not filling the coelom, 3.5= very large of 1984,several collections were made at RepresaLes Maja- deposits filling the coelom. but not producing a visible guas, a 4250ha reservoir in western Portuguesa(9°40'00" bulge, and 4 = coelom packed with fat deposits producing Lat. N; 69000'00" Long. W). Fishes were collected from a bulge of the belly region. Mean standard lengths, ratio the reservoir by seine (3.2 mm/2.5 m), dipnet, and hook of immature to adult SL's, mean gonad codes, and mean and line. Snorkeling observationsprovided an additional fat codes were plotted for common speciesby month for meansfor determiningthe presenceof fishesin clear regions estimation of the duration of breeding seasons.Immatures of the reservoir.Two regionswere sampled: a narrow arm weredefined as standard lengthsfalling below the minimum of the lake directly below an inflow canal (from the Rio length for which a fully gravid individual was observed Cojedes),and the main body of the reservoir, both near among all collections of a given species. and offshore. Physico-chemicalconditions within the main Ten variables related to life history theory were either body of the reservoir are relatively constant, while the measured,coded, or estimated for 59 speciesfrom Caiio former site experiencesradial changesassociated with sea- Maraca and 12 speciesfrom Catio Volcan. sonal rainfall (e.g. turbidity, depth, and flow rate all in- 1) The estimate of annual population density fluctuation creaseduring the wet season).Voucher specimenswere pre- was equal to the coefficient of variation (CV) of the popula- 227
>< )( W W Q Q ~ ~ U U ~ ~ 0( C ~ ~ 0 0 II) In 0 Q g 0( C Z Z 0 0 ~ ~
Astyanaxb/msculatus Rhamdis.,,1 0'» o.a
>< 0.8 )( w UI 0 0 0.15 ~ 0.- ~ u u ~ i= C c 0.15 2 ~ 0.10 0 0 II) CI) 0 0.10 0 0 0 C C ~ Z Z 0 0.- 0 " " a.- 0.5 1.0 1.5 2.0 2.5 3.0 3.5 4.0 4.5 0.1 1.0 1.5 2.0 2.5 3.0 3.5 4.0 4.5 GONAD CODE GONAD CODE F"tg.1. Linear regressions of gonadosomatic index with gonad code for two tetras and two catfishes from the Venezuelan llanos. Statistics for each regression are as follows: Bryconamericus beta (females) r=0.76. F23.1= 31.1, P 10 20 30 40 50 10 70 10 20 30 40 50 60 70 :J .. 10 50 100 '10 200 250 .. "i ~ .. '0 i; '> ~ :a ~ .5 '> - :c 0 ..s 10 20 30 40 50 80 70 .. - $ 0 .0 ~ E OJ ~ .c z E ~ z 50 100 150 200 250 10 20 30 .0 10 80 10 50 100 150 200 250 SL Interval SL Interval Fig. 2. Standard length frequency histograms during each month for Hop/ios malabaricus(extended breeder) and Curimata argentea (cyclic breeder)at Cano Maraca 229 Roebolde. dayl Aequldens pulcher 10 20 30 40 10 10 20 30 40 50 ,. 20 30 40 50 12.. May I- 0'- 10 20 30 40 so ~ i; ~ " 'C i; "> ~ :c- 'C c "> - :c 0 .5 .. - . 0 .0 ~ E e ~ .c z E ~ z .., HI... e. 10 o. 1. 20 10 .0 10 SL Interval 10 a, 30 40 50 ., SL Interval Fig. 3. Standard length frequency histograms during each month for two fishes with extended breeding periods, Roeboidesday; and AeQuidensDutcher. at Caiio Maraca ~ 230 i ! u 1 1 & ~ c . . I , ~ .~ . I f, e- .» E ... ,,~-\,\ f,f' -( " ,f' . J I ; I 0---0 ,.0---0-- ,°" 100 8 -- , '. ! "01 "V . .0 I '0" ""/ uj...... J. 4.M.M"".80MD .. A . J J 0 . 0 -- Month ~ ~ .. i ~ <1 "0 ~ I i r"... - P"', , , , , °'. ~ 0.1 , , ... , "S , "'-0-_.0 '. i' '; ~ , c I ~ 0.. ! ; O.t - > " 0 ~: ~ 2 r. : ~ ... 0.. b, ... " I ~, 0.4 I .2 . '.V r i . .,.:::'~::X""-' , ~. J i .m s ... '. ~ 0.2' : - ~ . .. - 0.2 ,'b,6 . "20 ... , ...... --:$ ' " ...... ~. J'M.MJJ.80NO J'M.MJJ.80ND .,..4. Comparison of reproductiveMonth measuresof a multiple--spawningcharacid (Bryconomericw beta) and cicblid (Caquetia kraw.ril) with a seasonally-spawningcharacid (Astyonax bimacrdatus)and cichlid (CichlD.romaor~) from the Venezuelanllanos (ciORd symbobcorrespond to mean fat code-topponeb. and mean SL-bottom~b; opensymbols correspond to mean gonad code-toppane-b. and ratio of juvenilesto adJdts-bottompanels) A = specialp1ac:ement of zygotes(1) or zygotesand larvae 8) Dry season age distribution of each population was (2) coded as either O-No adults (applied only to two species B = brief period of nutritional contribution to larvae (2), of annual k.illifish), I-Skewed in favor of juvenile and suba- or long period of contribution to larvae or embryos dult size classes.2-Approximately even distribution of im- (4) mature and adult size classes,and 3-Skewed in favor of C = brief period of parental protection by one sex (1), or adult sizeclasses. both (2), or long period of parental protection by one 9) Wet season age distribution of each popUlation was sex (2), or both (4) coded as either 1, 2, or 3 as before. D-extremely long period of gestation (4) (applied only 10) The maximum standard length among all specimens to Potamotrygon) of each specieswas measuredto the nearest0.1 Mm. Maxi- mum length was usedin analysesrather than averagelength Parentalcare codesranged from 1 to 8 and, with few excep- of mature adults, since it better reflects genetic potential tions, most details of reproductive behavior of a species among fishesexhibiting indeterminategrowth. were documentedby field observationsof fishes collected For ten speciesthat occurred at both sites, data from at the sites,or the scientific and aquarium literature. the site with the highest popUlation density were used in 231 analyses.Not surprisingly, a few life history variables re- dardized (mean= 0, standard deviation = 1) prior to com- vealedinterregional differenceswithin a species(these pro- puting Euclidean distances.Clustering was performed by vide a basisfor comparisonsin a future report). The fresh- the average linkage method (Sokal and Michener 1958), water stingray, Potamotrygonorbignyi, was collected from where the distance betweentwo clusters is the averagedis- other sites in the llanos (estado Apure) and included in tance betweenobservations and/or clusters.Canonical dis- the cluster analysis,since it has very distinctive life history criminant function analyseswere performed to explore pat- characteristics(e.g. an invariant clutch size of two, vivipar- terns of within classvariation relative to betweenclass vari- ity, and a long gestation period). With the exception of ation among life history variables. The type of wet season Prochi/odusmariae (both sites) and Brycon whitei (C. Vol- age distributions (3 states),dry seasonage distributions (4 can), and the freshwater stingray, only speciesknown to states)and parental care (6 states)were used as class vari- be reproducingat the two siteswere included in the analysis. ableswith the other nine traits as dependentvariables. The Immature yearlings, non-gravid, and gravid adults of Pro- remaining sevenlife history variablescontained ten or more chi/odus and Brycon were collected, but spawning took states, and were thus treated as dependent variables for place at remote downstream locations. At the beginning the purposesof canonical discriminant function analysis. of the rainy season,long distancemigrations (adults down- stream,immatures upstream)are a conspicuousfeature of Results the ecology of these species(Lilyestrom 1983; Lilyestrom and Taphom 1983). If one or two variables could not be Univariatestatistics estimatedfor a specieswith confidence,the value assigned Comparedwith temperatezone assemblages(Mabon 1984; to its closestphylogenetic relative was used as an estimate Wooton 1984), the fishes of both Caiio Maraca and C. of the unknown character (This was performed for only Volcan exhibited an extremely wide range of life history 21 of 720, or about 3% of the observations in the data attributes (Appendix 1). Large interspecific variation in fe- matrix). Becausethe reproductive biology of the genus is cundity, body size, and the timing of reproduction was ap- distinct and well documented (Wourms 1972; L.G. Nico parent within most orders and families. For example,both and D.C. Taphom pers.comm.), the annual killifish, Ptero- aseasonaland seasonally-spawningspecies were observed /ebias hoignei, was included in the data set, even though in the specioseCharacidae as well as the comparatively only two individuals were collectedat Cano Maraca during species-poorCichlidae (Fig. 4). Diameter of mature oocytes the wet season. ranged from a low of 0.45 mm in Curimata argentea to a high of 4.00 rom in Ancistrus triradiatus (i.e., excluding Analyses 35.00mm reported for Potamotrygon by Thorson et aI. 1983). Forty seven species,including Curimata argentea, A principal components analysis was computed from the Astyanax bimacuJatus.Brycon whitei, and Eigenmanniavir- correlation matrix derived from 71 fish species and all 10 escens,exhibited virtually no parental care following spawn- life history variables in the original data set (Appendix 1). ing. Cichlids and loricariid catfishesexhibited highest levels Principal components analysis is an ordination method that of parental care. allows a multidimensional swarm of correlated data points lntercorrelations among the ten life history variables to be viewed within two or three new orthogonal dimen- appear in Table 1. Significant large positive correlations sions. When present, intrinsic patterns within the multidi- (r~ -0.5, P<0.01) were obtained for generationtime with mensional swarm will emerge, generally within a plot of log SL, log fecundity with log SL, and egg sizewith parental the first two or three components (i.e. independent axes care. Significant large negative correlations (r ~ 0.5, P < derived from the original dependent variables; Pielou 1984). 0.01) were obtained for generation time with length of The strength of each component is reflected by its asso- breeding season,reproductive bouts per year and wet sea- ciated eigenvalue, with values greater than 1.0 being gener- son agedistribution; log fecundity with wet seasonage dis- ally accepted as explaining a significant proportion of the tribution; and log SL with wet season age distribution. variance in the multidimensional data set. Correlations be- Twenty six life history variable correlations were negative tween the first two principal component scores for each and 19 were positive or zero. When size was factored out speciesand the original life history variables for each species of the analysis using residuals from the log length regres- were computed to provide an index of the relative impor- sion, signsand relative magnitudesof correlations remained tance of each variable in the ordination of species into re- essentiallyunchanged (Table 1). productive strategies. In accordance with recent considerations of the effect Multivariate statistics of allometry on life history traits (Calder 1984; Dunham and Miles 1985; Gittleman 1986), additional PCA's were When all ten variables were included in the data matrix computed on modified data matrices with 1) log-trans- (raw scoresfor 8 variables plus log fecundity and log SL), formed SL and fecundity and 2) with nine dependent vari- the flISt three principal components reduced variation by ables corrected for SL (residuals from log SL linear regres- 76 percent (Table 2). The result was very similar when SL sion). To test the hypothesis that patterns of interspecific and fecundity were untransformed (first three PC's mod- variation among life history traits have a phylogenetic basis elled 70% of total variation). The first axis was influenced (Dunham and Miles 1985), a nested analysis of variance more or lessequally by sevenvariables (population fluctua- was performed on each of nine life history variables, using tion, egg size,and parental care excluded).The secondaxis family nested within order and log length as a covariate. was derived primarily from egg size,parental care, and log A cluster analysis was performed using Euclidean dis- length. Population fluctuation, length of breeding season, tances based on all 10 attributes among 72 species (stingray dry and wet seasonsize distributions were also significantly included). Each of the ten life history variables was stan- correlated with PC2 (Table 3). The third principal compo- 232 Table I. Correlation matrix of absolute life history variables for fIShesfrom two sites in the Venezuelanllanos. Values in paralthe1e8 are for residualsof regressionwith log SL Life history Fluctu- Genera- Length Bouts Parental Dry Wet trait ation tion breeding per care size size time season year distr. distr. Fluctuation x ( -0.09 : -0.29)** (0.13) (-0.01) (-0.22).. (-0.36).* (-0.09) (-0.09) Generationtime -0.04 x :-0.63)** (-0.S3)** (0.41)** (-0.16) (-0.18) (0.53)** (-0.30).* Lenlth breedingseason -0.30-- 0.6S** )( (0.39)** (-0.25)* (0.27). (0.50).. (-0.45)** (0.41) Bouts per Year 0.10 -0.62** O.~" )( (-0.49)** (0.31).. (0.11) (-0.75).. (0.31)*. Fecundity 0.01 0.62** -0.3S** -0.59** )( (-0.49)** (-0.33)** (0.45)*. (-0.44).. Egg size -0.18- O.IS* 0.12 0.08 -O.OS x (0.62).. (-0.34).. (0.27). Parentalcare -0.34-- -0.01 0.41.* 0.00 -0.09 0.65** x (-0.38)** (0:11». Dry sizedistr. -0.07 0.49** -0.46.* -0.73.* 0.43.. -0.26. -0.34** x (-0.10) Wet sizedistr. -0.10 -0.S7** 0.46.. 0.46** -0.67 -0.16 -0.05 -0.15 x Maximum length -0.04 -0.S7 -0.2S. -0.37.. 0.61.. 0.47** 0.25. 0.11 -0.70.. Mean val~ 1.06 10.69 3.78 2.80 5.269~,d 1.39 1.94 2.67 1.46 .P T~ 3. Correlations of rant two principal component axes with the original life history variables tire history variable PC1 PC2 Absolute data Fluctuation 0.07 -0.37.. - CumuJati~ v8riaIa Generation time 0.86.. 0.09 Length breedingseason -0.69.. 0.36.. PC! 4.01 0.401 Bouts per year -0.79.. 0.09 PC2 2.29 0.630 Log fecundity 0.82.. 0.05 PC3 1.30 0.760 Egg size 0.05 0.85.. Parentalcare -0.12 0.86.. Variable PCI PC2 PC3 Dry ~ distr. 0.63.. -0.43. Wet ~distr. -0.77.. -0.25. Fluctuation 0.034 -0.244 0.692 Log length 0.69.. 0.54.. Generationtime 0.430 0.061 0.692 p Fluctuation -0.073 -0.611 0.270 these speciestended to exhibit comparatively small-scale Generationtime -0.364 0.281 0.390 fluctuations in local population density over the course of length breedingseason -0.381 0.127 -0.546 the year. A fourteen speciescluster around the first apex Bouts per year -0.3WJ -0.366 0.090 was dominated by catfishes (Siluriformes) and cichlids Fecundity -0.368 0.071 0.090 (Fig. 6). Egg size 0.325 0.325 0.497 The secondapex resulted from low scoreson PC1 and Parentalcare 0.314 0.472 0.151 intermediate or low scores on PC2 (Fig- 5). This region Dry si2I:distr. -0.394 0.243 -O.1S1 of two-dimensional component space is charactcrlud by Wet si2l:distr. 0.276 0.061 -0.227 little or no parental care, prolonged breeding seasons,re- ~ 233 5 4 3 2 C\I U c. 0 -2 -3 -7 -5 -3 -1 1 3 5 PC 1 Fig. 5. Plot of Venezuelanfish speciesscores on the first two princi- pal components(analysis derived from raw life history data with log fecundity and log SL). Symbolsgroup speciesinto three broad categoriesbased on a cluster analysis.Top ellipse boundsan equi- librium strategy,left eclipsebounds an opportunistic strategy,and right ellipse bounds a seasonalreproductive strategy. The small ellipse identifies fishes of small body size that exhibit a strategy intermediate betweenextreme seasonalreproduction and oppor- tunism 0 Opportunistic; 8 Seasonal; d Equilibrium 5 4 3 2 N U Fig. 7. Cluster diagram of Venezuelan fishes based on Euclidean D.. distances computed from ten standardized life history attributes. 0 Root mean square distances between species is 4.47. Species number codes are given in Appendix 1. Explanation of three pat- terns appears in text -2 dominated dry seasonpopulations, juvenile-dominatedwet seasonpopulations, long generation times, intermediateto -3 high fecundities, small oocytes, and intermediate or large -7 -5 -3 -1 1 3 5 body size. For the most part, specieslocated near the third PC1 apex exhibited large local population fluctuations. The 48 Fig. 6. Plot of tropical fish PCA scoresas in Fig. 5, but with sym- specieslocated in the region of the third apex belonged bols identifying speciesby order. Again, the top apex corresponds to orders Characiformes and Siluriformes exclusively to the equilibrium strategy, left apex to opportunism, and right (Fig. 6). Seventeencharaciform and siluriform fishes were apex to a seasonalreproductive strategy 0 Characiforms; . Siluri- intermediatebetween the secondand third apices(interme- forms; t. Perciforms; A Cyprinodontiforms; 0 Synbranchiforms diate and low PC1 and low PC2 scores;Fig. 5). Theseinter- mediate fishes were all comparatively small, with pro- peatedbouts of reproduction, an evensize distribution dur- longed, yet distinctly seasonalreproduction. Egg sizeswere ing the wet season(for somespecies during the dry season small, fecundity intermediate, and parental care absent or as well), short generation time, relatively small clutches, weakly developed in this group that included Corydoras small oocytes,small body size,and intermediatepopulation species,Ochmacanthus altern us, and OdontostilbeDulcher. fluctuations. Characiformesand Cyprinodontiformes were predominant among the nine speciesgrouped near the sec- Cluster analysis ond apex (Fig. 6). Clustering of 72 speciesbased on Euclidean distancesre- The third apexwas composed of specieswith high scores sulted in three major groupings plus two single species on PC1 and intermediate or low scores on PC2 (Fig. 5). branches (Fig. 7). The freshwater stingray, Potamotrygon Speciesin this region of principal component two-space orbignyi, was included in this analysis(estimates for several were characterizedby very little or no parental care, short Potamotrygon life history parameters were based on the breedingseason, few bouts of reproduction per year, adult- closely related P. morro using data from Thorson et al. 234 1983). The same three basic life history strategies were T.~ 4. Results of nestedANDY A for family within order with fonned by three large clusters,with Potamotrygonand the log length as a covariate characid, Brycon whitei, representingextreme examplesof two of thesebasic suites of characteristics. Trait Classvariable Type I SS Potamotrygoncould be consideredan extreme case of F p the life history strategythat correspondsto long generation time, well-developedparental care, large oocytes,and low Population Order 4.38 0.003 fecundity. The same 14 species(5 families, 3 orders) that Fluctuation comprised the ftrst apex from PCA produced this cluster Family (order) 2.42 0.009 2.36 0.01 (Figs. 5 and 7). Likewise,Brycon representsan extremecase Lollength 0.01 0.974 of the strategyexhibited by the 48 speciescluster (19 fami- Generation Order 6.98 0.0001 4.S7 0.0018 0.6S lies, 2 orders) characterizedby seasonalreproduction, high time or intermediatefecundity, no parental care. and generation Family (order') 1.53 0.12 1.24 0.27 times from 6 to 48 months. This large cluster of seasonally Log length 24.83 0.0001 reproducing fishes contained two subunits that corre- LCDIth Order 5.SO 0.0005 5.96 0.(XX)2 0.50 spondedto 29 large specieswith high fecundities(e.g. Pro- seuoD chi/odusmariae. Rhamdia spp., and Rhamphichthysmarmor- Family (order) 0.98 0.49 0.91 0.S7 atus) and 19 small specieswith much lower fecundities(e.g. Log length 3.49 0.07 Characidiumspp., Microglanis iheringi, and Ochmacanthus Bouts per Order 20.38 0.(xx)1 59.07 0.(xx)1 0.78 alternus).An eight speciescluster (4 families, 3 orders) cor- year respondedto the short generationtime, low fecundity, mul- Family (order') 3.54 0.CMM»3 3.51 0.CMM»3 tiple-brood strategyof the PCA (Figs. 5 and 7). Log length 6.13 0.02 Log Order 24.41 0.CMM»117.32 0.0001 0.88 Creagrutussp. and Hemigrammussp. we~ grouped with fecundity aseasonally-reproducing,fast-maturing speciesby PCA, yet Family (order) 6.38 0.0001 4.01 0.0001 placedwithin the small body sizesubunit of the seasonally- Log length 104.31 0.0001 reproducing group by averageclustering (Figs. 5 and 7). Egg size Order 6.68 0.0001 7.98 O.lml 0.77 Gephyrocharaxvalenciae was placed within the multiple- Family (order) 4.72 0.0001 4.12 O.lml brood, fast-maturing group by cluster analysis, yet only Log length 48.99 0.0001 slightly nearer to small-sized,seasonally-spawning species Parental care Order 72.97 0.0001 67.96 0.0001 0.91 by the fIrSt two principal components.If one examinesthe Family (order) 4.59 0.0001 4.58 0.0001 original life history variables (Appendix I), it is clear that Log length 27.64 0.0001 all three groups grade into one another as a continuum Dry season Order 18.37 0.0001 11.37 0.0001 0.76 of reproductivestrategies. size distribution Family (order) 3.36 0.0005 3.37 0.0005 Log length 0.24 0.62 Analysisof variance Wet season Order 5.86 0.0003 2.39 0.OS2 0.74 The composition of the three groups clustered near the size apicesof Figure 5 is associatedto somedegree with phylo- distribution Family (order) 2.92 0.002 2.13 0.021 geny. For example,five of the speciesin the first apexgroup Log length 55.78 0.0001 (high PC2 scores)are cichlids (Perciformes),seven are cat- Log length Order 4.10 0.0035 fishes(Siluriformes), one is a characiform (Hoplias malabar- Family (order) 1.31 0.225 icus), and one is a synbranchiform eel (Synbranchusmar- moratus). Likewise, the sa:ond group (low PCl scores)is dominated by characiforms and cyprinodontiforms (two patterns of covariation among life history variables could killifishes and one livebearer).The third group (high PCl have been affected by allometry (Steams 1983a; Dunham scores)is composedentirely of tetras, catfishes,and knife- and Miles 1985). Log length was significantly correlated fishes(Characiformes and Siluriformes).On the other hand, with sevenof nine life history variables (Table 1). Again, membersof the most diverse order, Characiformes, were results from nested ANOV A with log SL as a covariate found in all three groups. Results of nestedANOV A show indicate that length influenced the pattern of variation in that order had a significant effect on all ten life history only one of 18 tests for phylogenetic affects on life history variableswhen length was not included as a covariate (Type variables(Table 4). Whereasthe potential for a high fecun- I SS,Table 4). When log length was included as a covariate dity might be expected to exhibit an allometric pattern, (Type III SS). the main effect of order on wet seasonsize larger fishesdid not always exhibit higher fecundities than smaller fishes (e.g. Parauchenipterusgo/eatus. Hypostomus distribution became marginally insignificant (Table 4). Family significantly affectedall life history variablesexcept argus. Loricarichthys typus, Pterygoplichthysmultiradiatus, generation time. length of reproductive season, and log Crenicichiageay,'). length (Table 4). Length as a covariate did not alter the fraction of family level tests (family nested within order) Canonicaldiscriminant function that attained statistical significance(P Pielou EC (1984)The interpretation of ecologicaldata. John Wi- StearnsSC (1983b) A natural experimentin life-history evolution: ley & Sons,New York field data on the introduction of mosquitofish(Gambusia af.fm- ReznickD, Endler JA (1982)The impact of predation on life histo- its) to Hawaii. Evolution 37:601-617 ry evolution in Trinidadian guppies(PoeciJia reticuJata). Evolu- Thorson TB. LanghammerJK, Oetinger M( (1983) Reproduction tion 36:160-177 sand developmentof the South American freshwaterstingrays, RofTDA (1984)The evolution of life history parametersin teleosts. PotamotrygoncircuJoris and P. morro. Env Bioi Fish 9:3-24 Can J Fish Aquat Sci 41 : 989-1(MX) Tinkle DW, Wilbur HM, Tilley SO (1970) Evolutionary strategies RofT DA (1986) Predicting body size with life history models. in lizard reproduction. Evolution 24:55-74 Biosci36:316-323 Welcomme RL (1969) The biology and ecology of the fishes of Sibley R. Calow P (1986a) Physiologicalecology of animals: an a small tropical stream.J ZooI158:485-529 evolutionary approach.Blackwell Scientific Publ.. Oxford Winemiller KO (1987) Tests of ecomorphologicaJand community Sibley R. Calow P (1986b) Why breedingearlier is always worth- level convergenceamong neotropical fish assemblages.Ph.D. while. J Theor Bioi 123:311-319 dissertn.,University of Texas,Austin Sokal RR. MichenerCD (1958)A statisticalmethod for evaluating Winemiller KO (1989) Developmentof dermal lip protuberances systematicrelationships. Univ Kansassa Bull 38: 1409-1438 for aquatic surface respiration in South American characid SouthwoodTRE (1977)Habitat. the templet for ecologicalstrate- fishes.Copeia 1989:382-390 gies?J Anim EcoI46:337-365 Wootton RJ (1984) Introduction: strategiesand tactics in fLShre- Southwood TRE (1988) Tactics. strategiesand templates.Oikos production. In: Potts OW, Wootton RJ (cds) Fish reproduc- 52:3-18 tion: strategiesand tactics. Academic Press,New Yort, Lon- StearnsSC (1976) Life-history tactics: a review of the ideas. Q don, pp 1-12 Rev Bioi 51:3-47 Wourms IP (1972) Developmental biology of annual fishes III. StearnsSC (1977) The evolution of life-history traits: a critique pre-embryonic and embryonic diapause of variable duration of the theory and a review of the data. Ann Rev Ecol Syst in the eggsof annual fLShes.I Exp Zool 182:389-414 8:145-171 StearnsSC (1983a) The influenceof si~ and phylogenyon patterns of covariationamong life-history traits in mammals.Oikos (Co- penhagen)41 : 173-187 Submitted February 2, 1989{ AcceptedMay 30,1989 order Myliobatifonnes family Potamotrygonidae 1. Potamotrygonorbignyi 0.46 ): ) 43 11 35.00 8 450 2 4 Equilibrium order Characifonnes family Erythrinidae 2. Hop/iasma/abaricus 0.63 2 12 ~ 4+ 2462 2.00 .2 352 521 3 Equilibrium 3. Hop/erythrinus 1.78 3 12 2; t+ 6024 1.50 7. 213 13 1 Seasonal unitaeniatus family Lebiasinidae 4. Characidiumsp. 1.16 3 2 11 4 2 154 0.65 28 234 .. Seasonal 5. Lebiasinaerythrinoides 0.46 3 r U 2 1 2688 1.30 121 332 2 Seasonal 6. Pyrrhu/ina/ugubris 1.27 3 2 12 3 1; 82 1.00 35 353 .. Seasonal family Anostomidae 7. Leporinusfriderici 1.54 3 12 1-2 28950 1.00 252 71 Seasonal 8. Schizodonisognathus 1.54 3 12 1-2 28950 1.00 2S5 91 Seasonal family Prochilodontidae 9. Prochi/odusmariae 0.91 3 24 ~ 28950 1.00 2W 363 3 Seasonal family Curimatidac 10. Curimataargentea 1.17 .) 1 12 2:-c~ 2 3528 0.45 92 1671 3 Seasonal family Characidac 11.Aphyocharax a/bumus 0.90 y 2 11 4+ 617 0.65 l 37 SO1 .. Seasonal 12.Astyanax integer 1.00 .3 t 12 2 8400 1.00 t 92 35 2 Seasonal 13. Astyanaxbimacu/atus 0.97 .3 .1 12 2 1+ 4287 0.90 t 91 965 3 Seasonal 14.Astyanax metae 0.97 3 12 .. 1+ 9528 1.00 1 106 185 2 Seasonal 15.Astyanax superbus 1.00 3 12 .. 2 800 0.65 t 64 10 2 Seasonal 16.Brycon whitei 1.00 3 J 24 .. 1 171545 1.50 t 313 14 ~ Seasonal 17. Bryconamericusbeta 0.61 2- Z 4 12 6+-- 796 0.95 t 44 1341 3 Opportunistic 18. Bryconamericus 1.15 3 2 6 6 3 243 0.85 1 39 289 2 Seasonal deuterodonoides 19. Charaxgibbosus 1.10 3 2 6 6 4 280 1.00 101 324 1 Seasonal 20. Cheirodontopsgeayi 1.39 3 2 12 2 t 1108 0.75 28 112 1 Seasonal 21. Corynopomariisei 0.80 2 2 4 10 4+** 135 0.85 43 323 2 Opportunistic 22. Creagrutussp. 0.91 2 2 4 6 2 94 0.80 28 754 2 Opport./Seas. 23. CtenobrycOllspihlrus 0.69 :) t 11 2-3 3 755 0.75 S2 1530 . Seasonal 24. Gephyrocharax 0.85 3 2 4 S S+** 734 0.7~ 37 819 3 Opport./Seas. valenciae 240 Species F1\1<:t. Dry Wet Gen- Season Bouts Length N Sita. Stratcay distr distr. erat. 25. Hemigrammus.fp. 1.23 2 4 7-1 4+ .. 26. Markiana geayi 1.07 t 12 2 1 27. Odonto.ftiJbe,uk_r 1.06 2 4 S 3+.. 28. Pygocentrusnotatus 1.19 t 12 3 2 29. Roeboide.Jdayi 0.84 2 3 7-12 6+.. 30. Serra.sa/musirritan.s 1.60 t 12 3 3 31. Serra.JOlmlLfmedini 1.92 t 12 3 3 32. Serra.salmusrhombeus 1.21 I 12 3 33. TetragOllopterus 1.34 t 12 1-2 1 argenteus 34. Triportheus.fp. 1.35 12 3175 .80 120 560 Seasonal family Gasteroplecidae 35. Thoracocharax.fteJlatus 0.89 11 2 755 0.80 42 351 Seasonal order Silurifonnes suborderGymnoloidei family Apleronotidae 36. Adontosternarchus 1.82 12 323 .35 tl5 ~ Seasonal deVtDlanzi family Gymnolidae 37. Gymnotuscarapo 0.66 12 3 567 2.00 297 362 Equilibrium family Hypopomidae 38. Hypopomussp. 0.95 12 2 3 323 1M 68 Seasonal family Rhamphichlhyidae 39. Rhamphichthys 1.56 12 t(XX) 3S 392 2J Seasonal mannoratus family Slemopygidae 40. EigenmanniDvire.fcerrs 1.44 323 1.35 in 232 Seasonal suborderSiluroidei family Auchenipleridae 41. Entomocorus gameroi 1.60 2 12 2 2+ 238 0.85 42 56 Seasonal 42. Parauc-nipterus 0.86 I 12 2 ., TSO 1.90 ItS 222 Seasonal gaJeatus family Aspredinidae 43. BIInocephohl.f.fp. 0.68 2 12 4+ 178 .00 58 652 Seasonal family Pimelodidae 44. Microg/anisiheringi 0.63 .2 12 3 3+ 649 1.10 38 292 Seasonal 45. PimeJodeJJasp.l 1.33 1. 12 t t 31SO 0.55 68 9 Seasonal 46. PimeJodeJIasp.2 1.57 12 2 2 4313 0.75 80 3Ot Seasonal 47. PimeJodeJJasp.3 0.63 12 2 2 1587 0.85 68 208 1 Seasonal 48. Rhamdia.fp.l 0.82 12 2 t 11- 1.10 134 122 1 Seasonal 49. Rhamdiasp.2 0.99 12 2 t 11S8S 1.0S 200 36 2 Seasonal family Ageneiosidae so. AgeneiO.fUSvittatus 2.17 750 I.~ 130 19 Seasonal family TrichomYCleridae 51. Ocltmacanthus aJterm4.f 0.83 12 3 2+ 238 38 4~ Seasonal family Callichlhyidae 52.Corydora.s - 1.27 12 2 1+ 152 1.45 44 281 Seasonal 53.Corydara.s 0.93 12 1 3+ 143 1.25 56 210 Seasonal ,yeptentirionaJis 54. Corydara.shabrO.fUS 0.86 2 12 2-4 3+ 12 0.70 2 25 477 Seasonal 55. HopJO.JtemumJittorak 0.83 2 12 2 2+ 2684 2.00 . 218 138 Equilibrium family Loricariidae 56. Ancistrussp. 0.65 2 12 4-5 4+ 48 4.00 4 68 114 ., Equilibrium 57. Hypoptopoma.fp. 1.08 2 12 2 4+ 52 1.6 00 43 1 Seasonal 58. Hypo.stomusargus 0.40 I 12 3-4 3+ 289 3.25 4 190 230 .1 Equilibrium 59. Loricarichthy.ftypus 0.85 1 12 3-4 5+ 421 3.05 4 224 233 .. Equilibrium 60. OtocincJussp. 0.88 1 8 4 4 52 O.~ 26 395 I Seasonal 61. PterygopJichthy.f 0.87 2 12 3 3+ 763 3.50 4 233 138 1 Equilibrium muJtiradiatus 62. RineJoricaria 0.62 12 .5 3+ 255 1.70 113 328 Equilibrium caraca.serrsis order Cyprinodontifonnes family Cyprinodontidae 63. Pterolebw hoigMi 1.97 2 1 4-S 12+.. 8 1.30 33 2 Opportunistic 64. Rachot1iDmacldipillnis 1.97 2 1 S 12+.. 8 l.3S 32 126 Opportunistic family Poeciliidae 241 65. PoeciliQreticulatQ 0.97 J , 2 11 4+ 19 1.90 4 26 1489 3 Opportunistic order Perciformes family Cichlidae 66. Aequidenspulcher 0.55 . 1 I ~10 }+- 871 1.75 6 89 1012 J Equilibrium 67. Apistogrammahoignei 0.95 ~ 1- 4 6-12 2. 59 0.65 4 30 220 Opponunistic ., 68. AstroMtus oce/latus 1.04 2. I 14 6 ~+ 2301 2.40 7 181 106 Equilibrium 69. Caquetaiakraussii 0.74 1 1 8 11-12 3+ 3702 1.85 6 230 550 Equilibrium 70. Cichlasomaorinocense 0.51 1 t 12 6 1+ 1287 1.85 6 112 301 Equilibrium 71. Crenicichlageayi 0.54 f 1 12 6 2+ 230 1.80 6 100 106 !}i: Equilibrium order Synbranchiformes family Synbranchidae 72. Synbranchus 1.08 2 12 1 2+ ISO .50 2 330 1.5 j Equilibrium marmoratus The reproductive strategy assignedto each speciesfor the Chi Square analysis appears in the last column. (*1 =Cano Maraca, 2= C. Volcan, 3 = both, 4 = collected from other sites in the llanos; ** conservativelyestimated from distributions of oocyte size classes and duration of breedingseason - actual value is probably many times higher - somespecies may evenspawn daily) Appendix2. Reproductivestrategies assigned to speciesinhabiting lentic and lotic habitats at represaLas Majaguas - Species Lentic Lotic Species Lentic Lotic family Potamotrygonidae SerrasaJmusirritans Seasonal Potamotrygon orbignyi Equilibrium Equilibrium SerrasaJmusrhombeu.f Seasonal Seasonal family Erythrinidae family Auchenipteridae Hoplias malabaricus Equilibrium Parauchenipterus gaJeatus Seasonal family Lebisasinidae family Pimelodidae Py"hulina lugubris Seasonal. Seasonal PseudopJatystomafasciatum Seasonal Seasonal family Prochilodontidae family Loricariidae Prochilodus mariae Seasonal Seasonal Hypostomus argus Equilibrium Equilibrium family Anostomidae Loricaria sp. Equilibrium Leporinus friderici Seasonal RineJoricaria caracasensis Seasonal Schizodon isognathus Seasonal family Poeciliidae family Characidae PoeciJia reticuJata Opportunistic. Opportunistic Astyanax bimaculatus Seasonal* Seasonal family Cichlidae Astyanax sp. Seasonal Aequidens diadema Equilibrium Bryconamericus beta Opportunistic Astonotus ocelJatus Equilibrium Bryconamericus deutero- Seasonal Caquetia kraussii Equilibrium Equilibrium dontonoides CichJa ocelJaris Equilibrium Bryconamericus sp. Seasonal CichJa temensis Equilibrium Colossoma macropomum Seasonal Seasonal Cichlasoma orinocense Equilibrium Creagrutus -'p. cf beni Seasonal Cichlasoma psittacum Equilibrium Gephyrocharax valenciae Opportunistic CichJasoma severum Equilibrium Opportunistic * Hemigrammus marginatus Crenicichla geayi Equilibrium Opportunistic * Hemigrammus sp. Crenicichla Jugubris Equilibrium Metynnis sp. cf orinocense Seasonal Geophagusjurupari Equilibrium Paragoniates alburnus Seasonal Geophagus surinamensis Equilibrium Pygocentrus notatus Seasonal family Synbranchidae Roeboidesdoyi Opportunistic. Opportunistic Synbranchus marmoratus s~~ Et.wl. S...e." .inhabitin~ shallow shoreline habitats only