Ecomorphological correlates in ten species of subtropical seagrass fishes :diet and microhabitat utilization
Philip J. Motta', Kari B. Clifton', Patricia Hernandez' & Bradley T . Eggold�'Department of Biology, University of South Florida, Tampa, FL ��6�0, U.S.A .� Wisconsin Department of Natural Resources, PO. Box 408, Plymouth, WI 5�07�, U.S.A .
Received 1 .10.199�
Accepted �0.11 .1994
Key words: Tampa Bay, Morphology, Feeding, Phylogeny, Convergence, Specialization
Synopsis
Ecomorphological correlates were sought among ten species of distantly related subtropical seagrass fishes .Morphomet ric data associated with feeding and microhabitat utilization were compared by principal compo-nents analysis, cluster analysis, and canonical correspondence analysis to dietary data . Morphology was gen-erally a poor predictor of diet except for a group of mid-water planktotrophic filter feeders . Separation of thespecies along morphological axes appears to be related more to microhabitat utilization resulting in threemajor groups: (1) a group of planktotrophic, mid-water fishes specialized for cruising and seeking out evasiveprey characterized by a compressed fusiform body, forked caudal fin, long, closely spaced gill rakers, short tointermediate length pectoral fin, pointed pectoral fin, large lateral eye, short head, and a terminal or sub-terminal mouth ; (�) slow swimming, less maneuverable epibenthic fishes that pick or suck their prey off thesubstrate . They are united by more rounded caudal and pectoral fins, and short or no gill rakers; and (�) agroup of more mobile and maneuverable epibenthic foragers characterized by a more compressed, sub-gib-bose body, long, pointed pectoral fins, forked caudal fins, large lateral eyes, subterminal mouth, and greaterjaw protrusibility. Cases of convergence in trophic and microhabitat utilization characters were apparent insome of the groups .
Introduction
Ecological morphology has as its major premisethat the ecology of an organism is related to its mor-phology. Whereas functional morphology is thestudy of form and function, ecological morphologyoverlaps functional morphology and emphasizesform in relation to biological role(s) . Althoughthere is no consensus on the definition of ecomor-phology, it may be defined as the study of the rela-tionship between environmental factors, both phys-ical and biotic, and form, such as to isolate the mu-
tual contribution of one to the other (Motta & Kotr-scha1199�) .
One of the purported advantages of ecologicalmorphology is its predictive power (Karr & James1975, Miles & Ricklefs 1984, Grossman 1986, Dou-glas 1987). Given that environments constrain mor-phology and ecology in parallel fashion, we shouldbe able to predict ecological patterns of individuals,populations, or species assemblages from theirmorphological characteristics (Wiens & Rotenber-ry 1980) .
Although there is no definitive protocol, eco-morphological studies generally seek patterns in
�8
the ecology or behavior of an organism or group oforganisms and try to relate them to patterns in form .In the initial step of such analyses, correlations areusually sought between morphological and ecolog-ical variables (e.g . Moyle & Senanayake 1984, Wik-ramanayake 1990, this study). This approach pro-vides predictability, but no causal explanation . Itleaves the investigator with uncertainty as to per-formance, optimality, and the constraints on the `fit'between form and biological role (Motta & Kotr-schal 199�) .
In the second step of the investigation, one notusually reached in many studies, the predictions aretested through experimentation or modeling in thelaboratory to determine the potential niche, andsecondly the effect of performance on actual pat-terns of resource use may be determined throughfield studies (i .e . the realized niche) (Wainwright1987, 1991) . An ontogenetic analysis of the eco-morphological relationship may be performed atthis stage (Galis 199�) .
In the third step, comparative phylogenetic ana-lyses are undertaken. Two different comparisons,
with different aims, have been applied . Most inves-
tigators advocate comparing the variation of astructure and its use within closely related taxa, usu-ally cogenera or cofamilials (Leisler 1980, Yamaoka198�, Felley 1984, Leisler & Winkler 1985, Yamao-ka, Hori & Kuratani 1986, Pounds 1988, Wainwright1988, Kotrschal 1989, Norton 1991); these may in-clude a comparison with a more distantly relatedoutgroup (Motta 1988, Losos 1990a) . It is assumedthat choosing closely related species will reduce therisk that coincidental differences will mask signif-icant patterns (Huey & Bennet 1986) . Findley &Black (198�) explicitly believe, and others implicitlyassume, that ecomorphological relationships maybe most detectable in closely related species thathave a long history of evolution and radiation in the
same region . This kind of comparative approach al-lows one to propose evolutionary scenarios and hy-potheses on the process of adaptation and mostreadily reveals parallel and divergent evolution(Motta & Kotrschal 199�) .
Convergence is difficult to study in small, moreclosely related groups . Because we usually havemore confidence in our phylogenetic groupings at
broader taxonomic scales, ecological (and morph-ological) convergences can be identified withgreater ease (Winemiller et al . 1995) . Therefore, an-other approach is the comparison of ecomorpho-logical relationships among guilds of more distantlyrelated organisms (Karr & James 1975, Ricklefs &Cox 1977, Gatz 1979a, b, Ricklefs & Travis 1980,Wiens & Rotenberry 1980, Miles & Ricklefs 1984,Moyle & Senanayake 1984, Watson & Balon 1984,Douglas 1987, Miles et al . 1987, Wikramanayake1990, Block et al . 1991, Wiens 1991a, b, Winemiller1991). This can be a powerful indicator of conver-gent evolution (e.g . Karr & James 1975, Wiens1991b) .
While there appear to be superficial ecomorph-ological correlations related to feeding in fishes, forexample, longer gill rakers are associated withplanktivory, shorter ones with carnivory ; longerguts with herbivory and shorter ones with carnivory(e.g . Chao & Musick 1977, Goldschmid et al . 1984) ;as well as correlations between fin and body shape,and locomotory abilities and microhabitat utiliza-tion (Keast & Webb 1966, Webb 1984), there are fewrelationships that document ecomorphological pat-terns in fishes beyond these . Gatz (1979a, b) foundextensive ecomorphological correlations based onfeeding and microhabitat separation ; Norton's(1991) work supported the ecomorphological hy-pothesis that dietary differences in cottids are inpart due to differences in relative mouth sizethrough the influence of mouth size on feeding per-formance. Furthermore, there is a strong relation-ship between selected morphological attributes andmicrohabitat exploitation in a tropical stream fishassemblage (Wikramanayake 1990) . Winemiller(1991) also identified ecomorphological divergenc-es within higher taxa of fishes from the same regionand convergences between phylogenetically diver-gent taxa from different regions .
On the contrary, Kotrschal (1989) concluded thatoral jaw morphology is not a reliable predictor offeeding ecology in �4 species of blennioid fishes .Similarly, Felley (1984) found that a priori morph-ological character sets derived from previous stud-ies of functional morphology and morphological-environmental associations could not be used topredict habitat use in cyprinids . An a-posteriori set
revealed by factor analysis for a subgroup of cypri-nids was more predictive. Because there is consid-erable amount of residual morphological variationin many communities that cannot be directly relat-ed to ecological variables, predictability of morph-ological relationships in one assemblage from thosein another is reduced (Strauss 1987) . Furthermore,Grossman (1986) believes that behavior may bemore important than morphology in determiningprey utilization in a rocky intertidal fish assem-blage. Motta. (1988) found that morphological struc-tures associated with feeding in butterflyfishes arecorrelated with how the fishes feed (e.g. suction,scraping, biting), rather than with what they feedon. Many of these studies that seek correlations be-tween morphology and diet or microhabitat utiliza-tion do not utilize multivariate techniques of directgradient analysis, that is, techniques that statistical-ly compare combinations of environmental andmorphological variables simultaneously. Conse-quently the interpretation of such studies can beweakened if the comparison between the environ-mental and morphological variables is subjective .
This study is significant in that it : (a) employs asuite of univariate and multivariate techniques, in-cluding direct gradient analysis of dietary andmorphological data ; (b) seeks ecomorphologicalpatterns among a taxonomically divergent group offishes to investigate whether morphology is a goodpredictor of diet ; and (c) determines if cases of evo-lutionary convergence can be identified in this feed-ing guild of seagrass inhabiting fishes . Utilizing tennumerically abundant and taxonomically divergentspecies of fishes from a subtropical seagrass habitat,this study addresses the following questions : (1) Ismorphological similarity among species reflectiveof dietary similarity? (�) Do these species exhibitconvergence in morphology? And, if so, (�) doesmorphological convergence appear to be related tofeeding and/or microhabitat utilization in this di-verse fish assemblage?
Materials and methods
Study site
Fishes were collected from a seagrass habitat in Bo-ca Ciega Bay, near the entrance to Lower TampaBay, Florida (�7°41'N, 8�°41'W), from May to Octo-ber, 1989 and 1990 . The Tampa Bay system is a large,shallow (< 4 m), subtropical estuary lined by limit-ed mangrove forests and salt marshes . Tampa Bay isapproximately 56 km long, 16 km wide at themouth, with �41 km of shoreline . Five major riversdischarge into Tampa Bay (Comp & Seaman 1985) .This nitrogen and phosphorus enriched estuarysupports seasonally high phytoplankton biomassand productivity. Seagrass beds, macroalgae, andbenthic microflora also contribute to the total pri-mary production of this system (Johansson et al .1985) .
Our study site is a gently-sloping sandy beachleading to a shallow (0 .5 to �.0 m) Thalassia testudi-num dominated seagrass bed with interveningsandy patches . Dense mats of macroalgae dominat-ed by Gracilaria and Hypnea spp, were often pre-sent. Maximum tidal fluctuations resulted in a94 cm change of depth, although collections werenot made at very low tides .
Collection and data analysis
Thirty individuals each of ten species found in orabove the seagrass beds, including the water co-lumn were examined in this study (standard lengthranges of the specimens used indicated) : Chilomyc-terus schoepfi (striped burrfish, 1�6-�15 mm), Flor-idichthys carpio (goldspotted killifish, 51-79 mm),Lagodon rhomboides (pinfish, 1�0-154 mm), Euci-nostomus gula (silver jenny, 71-114 mm), Fundulussimilis (longnose killifish, 7�-11� mm), Harengulajaguana (scaled sardine, 84-104 mm), Syngnathusscovelli (gulf pipefish, 91-164 mm), Anchoa hepse-tus (striped anchovy, 90-108 mm), Menidia penin-sulae (tidewater silverside, 54-80 mm), and Ariusfelis (hardhead catfish, �80-�91 mm) (Fig . 1) . In allcases only sexually mature individuals were sam-pled. These species were used because they consti-
�9
40
C . achoepfI
14le
0
tute some of the most numerically abundant fishesin Florida and Tampa Bay seagrass habitats(Springer & Woodburn 1960, Livingston 1976,Brook 1977, Livingston 198�, Stoner 198�, Comp1985, Thayer et al . 1987) .
Collections were made in and above the seagrassbeds with a 0 .95 cm (square measure) beach seine,�.5 cm, � .75 cm, and 5.0 cm (square measure)monofilament gill nets, and a � m otter trawl with
A . hepsetue
M . pen Ins u I a e
F . carplo
F . simills
S . scovoil I
Fig. 1 . Profiles of ten species of Tampa Bay seagrass fishes investigated . Bar indicates 1 cm .
0.�� cm mesh cod end pulled by a power boat . Fish-es were dissected immediately after capture and theentire gut removed and preserved in 10% bufferedformalin with Rose Bengal . Fishes were frozen forsubsequent morphological analysis . A detailed de-scription of dietary analysis is reported in Motta etal. (1995) . In brief, for species with a distinct stom-ach, only prey items in the stomach were included .For species without a distinct stomach the anterior
TLf
SL
P pC
Fig. �. A representative species, Eucinostomus gula, with ten mensural and five coded morphometric variables indicated . SL = standardlength, HL = head length, BD = body depth, B W = body width, PL = pectoral fin length, MW = mouth width, ED = eye diameter, RL = gillraker length, RS = gill raker spacing, HLP = head length with jaw protruded, CS = caudal fin shape, PS = pectoral fin shape, EP = eyeposition, MP = position of open mouth, DT = dentition type .
1/� of the intestine was evacuated . Specimens withempty stomachs or anterior intestines were not uti-lized. Prey were identified under dissecting micro-scope to order in most cases, and an Index of Rela-tive Importance (IRI) (Pinkas et al . 1971), based onwet weight, frequency of occurrence, and numbers,calculated for each prey taxa . Hill's (197�) diversitynumbers were calculated : NO, the number of preytaxa; N1, the number of abundant prey taxa ; N�, thenumber of very abundant prey taxa; and E5, even-ness. IRI values were used to cluster the species us-ing the Bray . .Curtis percent dissimilarity index, andHorn's index. of overlap indicated the dietary over-lap among the species .
In our analysis 15 morphological characters wererepresented by either mensural or coded variables .Ten mensural variables reflecting feeding and hab-itat use (Gata 1979a) were taken on each specimen :standard length SL (tip of closed mouth to end oflast vertebra), head length HL (anterior tip of
BD
BW
41
closed mouth to posterior edge of opercle), headlength protruded HLP (anterior tip of protrudedjaw to posterior edge of opercle), body depth BD(depth at widest part of body), body width BW(width at thickest part of body), pectoral fin lengthPL (base of fin to tip of longest ray), mouth widthMW (at widest part with mouth fully open), eye dia-meter ED (diameter between fleshy orbits along ananterior-posterior axis), gill raker length RL (long-est raker on first gill arch), gill raker spacing RS(distance between rakers on first gill arch in the vi-cinity of the ceratobranchial-epibranchial border) .In addition, five coded variables were scored withinteger values for each species : caudal fin shape CS(rounded =1, truncate = �, emarginate = �, lunate =4, or forked = 5), pectoral fin shape PS (rounded = 1,intermediate = �, or pointed = �), eye position EP(lateral =1, slightly dorso-lateral = �, or dorsal = �),open mouth position MP (supraterminal =1, termi-nal = �, subterminal = �, inferior = 4, or ventral = 5),
4�
and dentition type DT (cardiform =1, villiform = �,canine = �, incisor = 4, reduced incisiform = 5,brush-like = 6, molariform = 7, fused = 8, tricuspid =9, or no teeth =10) (Fig . �) . Mensural variables were
measured with a ruler or vernier calipers to thenearest 0.1 mm. Gill raker length and spacing weremeasured to the nearest 0 .01 mm with an ocular mi-
crometer fitted for a dissecting microscope .
Principal components analysis,(PCA) was usedto identify patterns in morphological variationamong species. PCA is a method of breaking down
or partitioning a resemblance matrix into a set oforthogonal (perpendicular) axes or components .
Each PCA axis corresponds to an eigenvalue of thematrix. The eigenvalue is the variance accounted
for by that axis . The first few PCA axes represent
the largest percentage of the total variation that canbe explained (Ludwig & Reynolds 1988). PCA was
conducted on a correlation matrix. All continuous
variables were log o transformed in order to moreclosely approximate a normal distribution and re-
duce heteroscedasticity. Two analyses were per-formed to explore the relative merits of mixing
mensural and scored variables: (1) a PCA of the
mensural and coded variables together, and (�) onewith only the mensural variables . The PCA analys-es were performed on SAS using the Princomp pro-
cedure. To aid in data analysis, three dimensionalplots were constructed using the G�D procedure.
The mensural variables that loaded heavily onPCA axes � and �, that is, those that account for thegreatest variance in the data set, head length, headlength protruded, body depth, mouth width, pecto-ral length, eye diameter, gill raker spacing and gill
raker length, were logo and alternately square roottransformed and found not to be normally distrib-
uted in either case. Therefore, an a-priori Kruskal-Wallis test was performed on the untransformed da-ta (N = �0) to detect significant differences amongspecies. Each variable was found to differ signifi-cantly among species, therefore a-posteriori Mann-Whitney U tests were carried out for all possible
pairwise combinations (N = 45) of species . TheBonferonni technique was used to determine thesignificance level for the Mann-Whitney U tests(Gatz 1979). An alpha level of a = 0.05/45 = 0.001was used for each pairwise comparison . For each
variable, species that were not significantly differ-ent were joined by a line. To assess the degree of jawprotrusion relative to head length the following var-iable was calculated for each species ((HLP-HL/
SL)X100) (N = �0) .Cluster analysis was performed on all fifteen un-
transformed morphological variables and on the
dietary data of Motta et al . (1995) using the Bray-Curtis percent dissimilarity index with the flexible
strategy ((� = - 0.�5) to create three clusters (Lud-wig & Reynolds 1988). The first cluster was con-structed with all morphological variables . In orderto investigate the relative proportions of themorphological attributes, for example, the size ofthe pectoral fin relative to the size of the fish, a sec-ond cluster was generated with the mensural varia-bles expressed as a percentage of standard length .Mean values were used for the ten mensural varia-
bles. In both analyses, the five coded variables were
entered as integer values . The third cluster was onthe IRI dietary values for nine species (excluding
M. peninsulae for which data were not available)from Motta et al . (1995) . Clusters were constructedwith the Statistical Ecology package (CLUSTERprogram) of Ludwig & Reynolds (1988) . While we
utilized ratios for exploratory data analysis in onecluster analysis, we specifically avoided ratios in thestatistical analyses due to their inherent limitations(Atchley et al. 1976, Reist 1985, Jackson et al . 1990,
Jackson & Somers 1991) .Canonical correspondence analysis (CCA) and
detrended canonical correspondence analysis
(DCCA by second order polynomials) were per-formed on two sets of morphological and dietary
data for nine species . Menidia peninsulae was ex-cluded from this analysis because the only dietary
data available was qualitative data from the litera-ture. Canonical correspondence analysis is a multi-variate technique of direct gradient analysis that se-lects the linear combination of environmental vari-ables (diet) that maximizes the dispersion of thespecies (morphology) scores. It chooses the best
weights for the environmental variables to con-struct the first CCA axis . The second and furtherCCA axes also select linear combinations of envi-ronmental variables that maximize the dispersionof the species scores, but subject to the constraint of
being uncorrelated with previous CCA axes . Envi-ronmental variables may be quantitative or nomi-
nal (Jongman et al. 1987, ter Braak 1986,1987) . De-trended CCA is an efficient ordination techniquewhen species have bell-shaped response curveswith respect to environmental gradients (ter Braak1986). We performed CCA and DCCA utilizing theprogram CANOCO version � .1� (ter Braak 1988) .
In the first data set, species data consisted of the15 mensural and coded morphometric variables . Toreduce colliriearity due to too many environmentalvariables, only IRI values for eighteen very abun-dant dietary taxa, as indicated by Hill's N�, wereutilized in the analysis . In the second comparison,species data consisted only of the morphometricvariables that were considered related to microhab-itat utilization (body depth, body width, pectoral finlength, eye diameter, caudal fin shape, pectoral finshape, eye position) and not to feeding per se . Asbefore, the environmental variables comprised theIRI values for the eighteen very abundant preytaxa .
Results
Principal components analysis-mensural variablesonly
The first two principal components had eigenvaluesgreater than one and accounted for 88% of the cu-mulative variance . Principal component 1 account-ed for 75 % of the variance, PC � accounted for 1� %,and PC � accounted for 7% (Table 1, Fig . �). Theeigenvalues of the first principal component werepositive and approximately equal in magnitude in-dicating a strong size vector (Strauss 1987) .
With increasing magnitude along principal com-ponent axis �, the species had longer, more closelyspaced gill rakers, longer pectoral fins, shorterheads, and larger eyes . Principal component � sep-arated fishes by decreasing spacing of the gill rak-ers, larger mouths, shorter pectoral fins, and de-creasing body depth (Table 1, Fig . �) .
The following species grouped close in morphos-pace : Menidia peninsulae, A . hepsetus, and H. ja-guana formed a single group ; F carpio and F similisformed a group ; Eucinostomus gula was positionedbetween F similis and L . rhomboides; Chilomycte-rus schoepfi, A. felis, and S. scovelli were separatedin morphospace (Fig . �) .
4�
Table 1. Principal components analysis for morphometric data on ten species of seagrass fishes (N =�0) . Eigenvalues for the first threeprincipal components on mensural variables only .
SL = standard length, HL = head length, BD = body depth, B W = body width, PL = pectoral fin length, MW = mouth width, ED = eyediameter, RL = gill raker length, RS = gill raker spacing, HLP = head length with jaw protruded .
Fig. �. Principal components analysis of the ten mensural morphometric variables for ten species of seagrass fishes (N = �0) . The variableswith larger eigenvalues (Table 1) indicated on each axis . Club = A . felis, Diamond = C. schoepfi, Star = L . rhomboides, Cross = F similis,Spade = S. scovelli, Heart = E. gula, Circle = A . hepsetus, Triangle = H. jaguana, Square = M. peninsulae, Cube = F carpio, BD = bodydepth, PL = pectoral fin length, MW = mouth width, ED = eye diameter, RL = gill raker length, RS = gill raker spacing, HL = head length,HLP = head length with jaw protruded .
Principal components analysis-mensural and coded
forked caudal fins and pointed pectoral fins at thevariables combined
other extreme, for example L. rhomboides . Mouthposition was supraterminal in S scovelli and gener-
The first three principal components had eigenva- ally subterminal at the other extreme (Table �, Fig .lues greater than one and account for 87% of the
4).cumulative variance. Principal component 1 ac-
Fishes with more rounded caudal and pectoralcounted for `i5 % of the variance, PC � for ��% and
fins, eye more dorsolateral, shorter gill rakers, morePC � for 10% (Table �) . Mensural variables on prin-
supraterminal mouth, loss of dentition, and increas-cipal component 1 were positive and approximately
ing head length scored high on PC�. Along princi-equal in magnitude indicating a strong size vector,
pal component axis � there was a shift towards moreThis was not generally the case for the coded varia-
dorsolateral eye position, more pointed pectoralbles. Dentition type was strong and negative indi-
fins, and forked caudal fins (Table �, Fig. 4) .cating the shift from villiform teeth, towards tricus-
Principal components analysis based on com-pid teeth (F carpio) and loss of dentition (S. scovel-
bined mensural and coded variables resulted inli) on PC axis 1. Caudal fin shape and pectoral fin
each species occupying less morphospace with gen-shape indicated a gradient from more rounded fins
erally less overlap among species . Menidia penin-(for example S. scovelli) at one extreme to more
sulae, H. jaguana, and A. hepsetus formed a tight
Table �. Principal components analysis for morphometric data on ten species of seagrass fishes (N = �0) . Eigenvalues for the first threeprincipal components on mensural and coded variables .
SL = standard length, HL = head length, BD = body depth, B W = body width, PL = pectoral fin length, MW = mouth width, ED = eyediameter, RL = gill raker length, RS = gill raker spacing, HLP = head length with jaw protruded, CS = caudal fin shape, PS = pectoral finshape, EP = eye position, MP = position of open mouth, DT = dentition type .
group, with E. gula and L. rhomboides clusteringnearby. Floridichthys carpio and F similis weregrouped very close together in morphospace withC. schoepfi nearby. Arius felis and S. scovelli wereeach morphologically distinct and separated fromthe other species (Fig. 4) .
Principal components analysis of the morphometricvariables resulted in a few species groups and somespecies separated in morphospace. Univariate anal-ysis of the morphometric variables that loadedheavily on either PCA � or � combined with the cod-ed variables resulted in the following charactercomplexes: (1) a group including H. jaguana, A .hepsetus and M. peninsulae characterized by a com-pressed fusiform body with forked caudal fin, long,closely spaced gill rakers, short to intermediatelength pectoral fin, pointed pectoral fin, relativelylarge (A. hepsetus andH. jaguana) lateral eye, shorthead, and terminal or subterminal mouth ; (�) Fun-dulus similis and F carpio which both had roundedcaudal and pectoral fins, short pectoral fins, smallbody size, small, lateral eyes, subterminal mouth,short gill rakers, and small mouth width and bodydepth. These species shared a few characters withC. schoepfi and S. scovelli: rounded caudal and pec-toral fins, and short or no (C. schoepfi) gill rakers ;and (�) Lagodon rhomboides and E. gula united bytheir sub-gibbose body shape (including large bodydepth), forked caudal fins, long, pointed pectoralfins, large, lateral eyes, and subterminal mouth . Ari-us felis was an outlier in morphospace, character-ized by being large bodied with a forked caudal fin,long, pointed pectoral fin, slightly dorsolateral eye,subterminal, wide mouth, long, and widely spacedgill rakers (Fig . 5) .
Of the species that protrude the upper jaw, L.
47
Fig. 4. Principal components analysis of the ten mensural and five coded morphometric variables combined for ten species of seagrassfishes (N = �0) . The variables with larger eigenvalues (Table �) indicated on each axis . Club = A . felis, Diamond = C. schoepfi, Star = L .rhomboides, Cross = F similis, Spade = S. scovelli, Heart = E. gula, Circle = A . hepsetus, Triangle = H. jaguana, Square = M. peninsulae,Cube = F carpio, RL = gill raker length, RS = gill raker spacing, HL = head length, HLP = head length with jaw protruded .
rhomboides and F carpio protruded their jaw �.7%of their standard length ; E. gula �.5%, and M. pen-insulae �.�% . Chilomycterus schoepfi (1.5%) and Fsimilis (0.�%) had slight protrusibility.
Morphometric clustering
Cluster analysis of the fifteen, untransformed mor-phometric variables (excluding standard length) re-sulted in three groups and one outlier species . Eu-cinostomus gula, H. jaguana, F similis, and A. hep-setus formed the first group, clustering at the 0 .11level . Fundulus carpio and M. peninsulae formed asecond group at the 0.1� level, and A. felis, C.schoepfi, and L. rhomboides formed a third groupat the 0 .�4 level . Syngnathus scovelli, the outlierspecies, was joined to the first two groups at the 0 .�9level (Fig . 6) .
Cluster analysis of the proportional measure-ments (mensural morphometric variables ex-pressed as percent standard length), and the coded
variables together resulted in a different arrange-ment of three groups and one outlier species . Lago-don rhomboides, E. gula and F carpio formed agroup at the 0.10 level ; F similis and C. schoepfiformed a second group at the 0.1� level; and H. ja-guana, M. peninsulae, A . hepsetus, and A. felisformed a third group at the 0 .14 level . Syngnathusscovelli was the outlier species joined to all threegroups at the 0 .86 level (Fig . 7) .
Dietary overlap and cluster analysis
Dietary overlap data (Table �) (Motta et al . 1995)and cluster analysis of the IRI data (Fig . 8) revealedsimilar species grouping . The nine species on whichdietary data were collected separated into threegroups: one with relatively high overlap comprisedof F carpio, H. jaguana, and A. hepsetus ; a second
48
J
Fig . 5 . Head length, head length protruded, mouth width, body depth, pectoral fin length, gill raker length, gill raker spacing, and eye
diameter of ten species of seagrass fishes . Mean ± one standard deviation indicated . Species that are not significantly different indicated
(N = �0, Mann-Whitney U-test, p = 0.05) .
group with intermediate overlap levels includes C .schoepfi, F similis, and E. gula . A third group in-cluding A. felis, S scovelli, and L. rhomboides hadrelatively little dietary overlap .
Dietary-morphometric correspondence
Canonical correspondence analysis (CCA) and de-trended CCA (DCCA) of the 15 morphological var-iables and dietary data indicated that the morph-ological variables were poorly related to the firstfour environmental (dietary) axes (eigenvalue =0.0�� on Axis 1, Table 4) . Only � % of the morph-
7060
E 50ta 40
-o �0
N :Z0-00
N
d
0 O N 07 c � 'ao , 0 N (V
>N 0a U 0N
U)
0
CO
a
ological variance could be accounted for by dietaryaxis 1 . Analysis of the seven morphological varia-bles related to microhabitat utilization revealed asimilarly poor relationship (eigenvalue = 0.055 onAxis 1, Table 4) . Eigenvalues of approximately 0.�and higher are quite common in ecological applica-tions (ter Braak 1988, Jongman et al . 1987), and or-dination axes with very low eigenvalues (< 0 .0�)should be discarded (ter Braak 1988) . Because ofthe poor relationship between morphological anddietary axes in this study, further analysis of the axeswas not warranted .
T
0�0
10 4I
I I 1 1 1 1
iN N .Z 0 0 0 N N Q)a) O Na
C .a ~_I!! :P O o , o L E N 7 >
0 0
0 0 . N N ot 0 N a c mE
U WN0O a c
OI
La 0i0'F_ U
100a
E 100
E 80
E
a 80
E
s 60
a)v
0 60
cC0
at
40 Q 400 C
S�0
a)
0 �0a)I
N . N N o 0 N 0 a)a a> 0a .a>E rn° w 7O 0
d ~ -O 0'N 0 0a
N E NW .~ C0 CLv d
(0 :r N N o U)a 00
O o 'E0 0a oo a,
N- U)N o
Q uV)0CL
oN
N ° EL O
U Q : I
J
C NE
a
U L
Discussion
Principal components analyses of the morphomet-ric data indicated that the species separated by a va-riety of characters associated with feeding and mi-crohabitat utilization . In general, the first principalcomponent was indicative of overall body size .Body size in these fishes is believed to be an impor-tant factor in niche separation (discussed below) .Thus, body size is not always a nuisance factor thatmust somehow be removed from data before `true'systematic or ecological relationships can be deter-mined (Douglas 1987) ; it may be an important fac-tor influencing species resource utilization . Whencoded variables were considered, variables indica-tive of trophic (dentition type) and microhabitat
_ 6cEE 50
40C
�000 �0UU)0 10
_
4 f
1
I
I
1
I
I .
I
I
I
l
e
g
E 4E
16
Fig. 5. Continued .
IIIII
O 0 0 0 U) 0 0 0 CC O
N
05 _ a .O
•
tOn O O Oa Q -O C W N U 0•
vt
O N
LU L<Li
Li
I
0
U)
a¢
I
0
a0
LU
S
If
•
_
7
h
49
(caudal and pectoral fin shape) differentiationloaded heavily. The second and third principal com-ponents were linked by trophic and microhabitatuse, being strongly influenced by gill raker lengthand spacing, caudal and pectoral fin shape, pectoralfin length, head length and head length protruded,eye diameter and position, mouth width and posi-tion, body depth, and dentition type .
Principal components analysis of the ten specieswas greatly affected by the mixing of mensural andcoded variables . The analysis of all 15 variables .re-sulted in the species groups clustering tighter inmorphospace and separating more from other spe-cies groups . Our analysis does not allow us to rec-ommend one particular approach over another, be-cause two factors differ between the approaches . In
0 U) 4- _O 0 0 O 0 U) .=U)
U a J O pO a ~ E N > >
¢OD U w ~
O. U) 0 0
a0 0c UE
00 li L v
s' cva W
U) 0 0 0) 0 00v-
N0
v~� a a CO
O0 O U) N 7¢ .~ W -0 U a
vC
0i p L L NE
¢a i
J 75
J
� .5E
� .00)c
1 .5U0Q_In
1 .0a
05 ∎
0 .0
50
o .o
0 .0
Morphological variables
` y ; rn > -n x m > n r •
Ne0
a
O 7a a-~
VV N
aOCN
N
�
a
mCw7a
f0Cw
V 0CLaN
Fig. 6. Cluster analysis of fifteen untransformed morphometricvariables for ten species of seagrass fishes . Clustering by theBray-Curtis percent dissimilarity index utilizing the flexiblestrategy (N = �0) .
the combined analysis additional morphometricvariables were considered (the coded variables) . Inaddition these variables had inherently less varia-bility being integer values . Perhaps instead of codedvariables, features such as eye position and caudalfin shape should be quantified by continuous valuesderived by multidimensional shape analysis. Thesecontinuous values would have more inherent varia-tion than the coded integer values.
In both analyses certain species groups were ap-parent. The first group was comprised ofH. jagua-na, A. hepsetus and M. peninsulae. A second consis-
0 .1
0 .0- r
Morphological variables
Percent standard length
-' m.
COID
0
-V0
�a.0aaN
wwCA)7w
I
TlDVaa
C0,
a
m
(00
0aV
N
�
N
NNO0<a
Fig. 7. Cluster analysis of fifteen untransformed morphometricvariables for ten species of seagrass fishes . The ten mensural var-iables expressed as a percent of standard length . Clustering bythe Bray-Curtis percent dissimilarity index utilizing the flexiblestrategy (N = �0) .
tent group included F similis and F carpio . Whenall characters were considered C . schoepfi and S.scovelli were closer to this group in morphospacethan to the other species. In both analyses, but moreso in the combined analysis, L. rhomboides and E.gula were close in morphospace. The catfish, A. fe-lis, was separated from the other species.
Cluster analysis based on all untransformed mor-
0 .8
0 .7
0 .e
0 .a
0 .8
0 .7
0 .5 0 .0
0 .4 0 .5
0 .� 0 .4
0 .� 0 .�
0 .1 0 .�
phometric variables grouped the species loosely bysize. Anchoa hepsetus and H. jaguana clustered to-gether (along with F similis and E. gula), and sep-arated from the smaller M. peninsulae . However,cluster analysis of all variables with mensural varia-bles expressed as a ratio of body length grouped thefishes primarily by shape. In this case, L. rhom-boides and E . gula again formed a tight group withF carpio at a higher level, and the group comprisedof H. jaguana, M. peninsulae, and A . hepsetus wasretained and joined to A. felis at a higher level .
Group 1 fishes
The group including H. jaguana, A. hepsetus andM.
peninsulae was characterized by common trophicand microhabitat morphometric characters : a com-pressed fusiform body with forked caudal fin, long,closely spaced gill rakers, short to intermediatelength pectoral fin, pointed pectoral fin, relatively
large (A . hepsetus and H. jaguana), lateral eye,short head, and terminal or subterminal mouth .This form is common to pelagic or mid-water inhab-iting fishes that are more specialized for cruisingand seeking evasive prey (Keast & Webb 1966,Aleev 1969, Chao & Musick 1977, Gatz 1979a, Webb1984, Wikramanayake 1990) . This morphology con-fers considerable mobility and maneuverability andis common to freshwater planktivores (Keast &Webb 1966). Relatively short head length, as foundin these species, was correlated with small prey sizein stream fishes (Gatz 1979a), and presumably withsmall prey (primarily copepods) in these species .
1 .s IRI cluster
0 .4
0 .�
0 .�
0 .I
0 .0
s0am
0
Y
V
a
x n
n
m a
r
e
0
a
a
Ca
sz
�
0•
m H �•
e 0'7•
m
04e61
I
51
Fig. 8. Dietary cluster, based on IRI values, for nine species ofseagrass fishes . Clustering by Bray-Curtis percent dissimilarityindex utilizing the flexible strategy (N = �0) (from Motta et al .1995) .
Watson & Balon (1984) identified four recurringecomorphological patterns among stream fishes inBorneo. The pelagic type was characterized by highcaudal fin aspect ratio, increased lateral compres-sion, peduncle compression and relative depth, alateral eye, and high pectoral fin aspect ratio. Intra-
•
m•
n
Table �. Horn's (1966) index of niche overlap based on IRI values for nine species of fishes in Tampa Bay, Florida (N = �0) (Motta et al .1995) .
Species 1 � � 4 5 6 7 8 9
1 . A . fells 1 .000 0.064 0.190 0.006 0 .�87 0.065 0 .104 0.07� 0.015� . A. hepsetus 1 .000 0.�7� 0.�5� 0 .1�8 0.788 0 .5�9 0.7�5 0.�59� . C. schoepfi 1 .000 0.004 0 .059 0.175 0 .490 0.157 0.���4 . S. scovelli 1 .000 0 .1�9 0.�79 0 .06� 0.��5 0.1905 . L. rhomboides 1 .000 0.196 0 .1�6 0.168 0.1146 . F carpio 1 .000 0 .�61 0.796 0.�507 . F similis 1 .000 0.�5� 0.�848. H. jaguana 1 .000 0.�8�9 . E. gula 1 .000
5�
specific morphological polymorphism in bluegillsunfish has been found to be related to foraging be-havior and microhabitat utilization . Open water in-habitants had fusiform bodies and short fins, where-as littoral vegetation feeders, which are suited formaneuvering, had deep bodies, long pelvic and pec-toral fins, and pectoral fins attached in a posteriorposition (Ehlinger & Wilson 1988, Ehlinger &Gross 199�). Bluegill with longer pectoral fins con-sistently searched more slowly and spent more timein the vegetation habitat compared to bluegill withshorter pectoral fins (Ehlinger 1990) .BothH. jaguana andA. hepsetus were mid-water
planktivores, filter feeding primarily on copepods.They had high dietary overlap, and clustered closelyby diet with F carpio . Lucas (198�) found that Meni-dia peninsulae fed selectively on calanoid copepodsand barnacle cypris larvae as adults, whereas gravidfemales primarily fed on fish larvae and amphipods .Harengula jaguana, A . hepsetus and M. peninsulaehad relatively long, closely spaced gill rakers. Long,closely spaced gill rakers are associated with plank-tivorous filter feeders (Lagler et al . 196�, Chao &Musick 1977) although the extent to which gill rakerspacing determines filtration efficiency is not clear
Table 4 . (Top) Eigenvalues for canonical correspondence analy-sis and detrended canonical correspondence analysis of 15 men-sural and coded morphological variables with IRI values foreighteen very abundant prey taxa (Hill's N�) for nine species ofseagrass fishes (M. peninsulae excluded) ; (Bottom) eigenvaluesfor canonical correspondence analysis and detrended canonicalcorrespondence analysis of seven morphological variables asso-ciated with microhabitat utilization (body depth, body width,pectoral fin length, eye diameter, caudal fin shape, pectoral finshape, eye position) with IRI values for eighteen very abundantprey taxa (Hill's N�) for nine species of seagrass fishes (M. penin-sulae excluded) .
(see Sanderson & Cech 199�) . Harengula jaguanaand A. hepsetus lacked upper jaw protrusion as iscommon of filter-feeding fishes (Chao & Musick
1977) . Menidia peninsulae most likely picks its prey
with its small, protrusible mouth. Therefore,
morphological similarity in these species was relat-ed to dietary similarity, particularly between H. ja-guana andA. hepsetus .
Examining an intertidal fish assemblage off Cali-fornia, Grossman (1986) similarly found PC1 to be ageneral size-related component, and PC� to be
trophically linked. Principal component � wasstrongly influenced by number of gill rakers, mouthorientation, and eye position . Wikramanayake
(1990) found a very similar suite of characters asso-ciated with digestive efficiency, foraging behavior,and foraging position in a tropical stream fish as-semblage .
Group � fishes
Principal components analysis of all characters sep-arated F carpio, F similis, C. schoepfi, and S scovel-li from the other species . This group was character-ized by rounded caudal and pectoral fins, and shortor no (C. schoepfi) gill rakers . Fundulus similis andF carpio clustered closely in morphospace when all
variables were considered, and with S. scovelli whenonly mensural variables were considered . Fundulussimilis and F carpio were united by a suite of char-acters including small body size, rounded or trun-cate caudal fins, rounded pectoral fins, short pecto-ral fins, small and lateral eyes, subterminal mouth,short gill rakers, and small mouth width and bodydepth .
In these species, morphology was reflective oftheir epibenthic microhabitat utilization . Morpho-logically, F similis, F carpio, CC schoepfi, and S. sco-velli are typical of slow swimming, less maneuver-able, epibenthic fishes that are unable to sustainlong periods of high-speed swimming (Keast &Webb 1966, Aleev 1969, Webb 1984). Although theyoccupied similar epibenthic microhabitats, F similisand F carpio frequented shallow inshore waters,whereas C. schoepfi was found mostly in deeper,sand and seagrass benthic habitats (personal obser-
Diet axesEigenvalues
1 � � 4
CCA 0.0�� 0.0�7 0.01� 0.008DCCA 0.0�� 0.0�6 0.009 0.000
vation) . Syngnathus scovelli was primarily capturedamong seagrass blades .
Dietary similarity among these four species didnot correlate with morphological similarity . Florid-ichthys carpio primarily fed on copepods presum-ably by picking or suction, consequently, it clus-tered by diet with Group 1 fishes, although it did notgroup with those species in morphospace. AlthoughF similis and C. schoepfi grouped together by Indexof Relative Importance (along with E. gula), theirdiets were only superficially similar . Fundulus simi-lis fed primarily on small (0 .4�-� .00 mm length) bi-valves as weiil as eggs which were ingested intact,and C. schoepfi crushed large (operculum diameter�.0-�.5 mm) gastropods, bivalves (1-� cm length),and barnacles . Inclusion of bivalves, albeit very dif-ferent sizes, primarily accounted for the dietaryclustering of these species (Motta et al . 1995) . Al-though these species occupied similar microhabi-tats, particularly F carpio and F similis, there waslow dietary overlap among them . Syngnathus sco-velli primarily utilizes suction feeding on amphi-pods and shrimp and was an outlier species in termsof both diet and morphology. Its supraterminalmouth and dorsolateral eyes suit it for capturingprey above its body, and its lack of dentition is char-acteristic of suction feeding fishes (Suyehiro 194�,Lagler et al . 196�, Davis & Birdsong 197�, Alexan-der 1974, Motta 1985,1988) .
Group � fishes
Principal components analysis grouped E. gula andL. rhomboides closely in morphospace . The size in-dependent cluster analysis based on shape and pro-portional variables grouped L . rhomboides, E. gula,and F carpio. Eucinostomus gula and L. rhom-boides were united not only by their sub-gibbosebody shape, but also by their forked caudal fins,pointed pectoral fins, large eyes, lateral eye posi-tion, and subterminal mouth . Lagodon rhomboideshad a larger mouth, and longer and more closelyspaced gill rakers than E. gula . There was generallylittle dietary similarity and overlap between thespecies at this study site. Motta et al . (1995) found L.rhomboides to feed primarily on algae and tuni-
5�
cates. The pinfish is the numerically dominant spe-cies within Thalassia testudinum beds along the sub-tidal areas of the Gulf of Mexico (Hansen 1969) andan important predator on macrobenthic organismswithin these seagrass beds (Young & Young 1978) .This species undergoes several ontogenetic dietaryshifts which may correspond with food availability(Huh & Kitting 1985), although the data are con-flicting. An increased tendency towards carnivorywith growth was reported by Subrahmanyam &Drake (1975), and Carr & Adams (197�). This con-trasts with the ontogenetic shift to herbivory citedby Darnell (1958), Hansen (1969), Stoner (1980),and Stoner & Livingston (1984) .
Eucinostomus gula had a more diverse diet withfive very abundant prey items, polychaete worms,bivalves, cumaceans, amphipods, and gastropods .These findings are consistent with other studies .Copepods dominated the diet of smaller size classesand were gradually replaced by polychaetes as sizeincreased (Springer & Woodburn 1960, Carr &Adams 197�, Brook 1977, Livingston 1984) .
Both L. rhomboides and E. gula were epibenthicforagers over sandy substrates and within seagrassbeds (personal observations). Mojarras (Gerrei-dae) use their extremely protrusible mouth to biteor suck their benthic prey off the substrate (Cyrus &Blader 198�, personal observations) . Lagodonrhomboides either bites off pieces of seagrass andalgae with vertically opposed, straight-edged incisi-form teeth (Stoner & Livingston 1984), or, suctionor ram feeds on elusive prey at this size (K .E Liempersonal communication) . Protrusible upper jawsare typical of fishes that utilize either suction feed-ing, picking or biting during feeding (Motta 1984,1985, 1988, Liem 1980) . Lagodon rhomboides, E .gula, and F carpio all had the most protrusiblemouths of the species examined (� .7%, �.5%, �.7%of standard length, respectively) and all suck or bitetheir prey off the substrate .
Morphological similarity between L. rhomboidesand E. gula was not related to diet, but more to mi-crohabitat utilization and how they fed. Their morecompressed, sub-gibbose body, long, pointed pecto-ral fins, and forked caudal fins make them suitablefor greater maneuverability and speed (comparedto C. schoepfi, F similis, F carpio, and S. scovelli) as
54
they suck or bite their relatively large prey withinseagrass beds or over sandy substrates (Keast &Webb 1966, Gatz 1979a, Webb 1984) .
Outlier species
Syngnathus scovelli (discussed previously) and A .felis were separated from most of the other speciesin morphospace. Arius felis is large bodied with aforked caudal fin, long, pointed pectoral fin, slightlydorsolateral eye, subterminal wide mouth, longwidely spaced gill rakers, and ventral barbels . Its fu-siform body shape grouped it with the mid-waterplanktivorous species . However, the ictalurid bodyform is suited for bottom feeding (Keast & Webb1966). The importance of crabs and tunicates in itsdiet clustered it with L. rhomboides which also con-sumed tunicates although dietary overlap was lowbetween the species.
Correspondence between morphology and diet
Canonical correspondence analysis is a powerfultool for direct measurement of the association be-tween environmental and species data . We foundan overall poor correlation between the morpho-logical variables under investigation and diet, indi-cating that the species distributions (morphology)did not differ much along the environmental gra-dient (dietary gradient) (ter Braak 1986) . Evenwhen morphological characters associated with mi-crohabitat were considered (body depth, bodywidth, pectoral length, eye diameter, caudal shape,pectoral shape, and eye position) there was poor as-sociation with diet .
The only group in which there was some congru-ence in morphology and diet was in the plankto-
trophic, mid-water group 1 fishes : H. jaguana, A .hepsetus, and M. peninsulae . These species consis-tently grouped together in morphology, and hadhigh dietary overlap . However, in this group of tenspecies dietary similarity was not necessarily pre-dictive of morphological similarity, for example, Fcarpio had a high dietary overlap and clustered byIRI with the above three species, yet was morph-
ologically quite distinct . Similarly, morphologicalsimilarity was not necessarily predictive of dietarysimilarity, as exemplified by the L. rhomboides - E.gula, and F carpio - F similis species groups whichhad similar morphologies yet low dietary overlap .
The general lack of correlation between mor-phology and diet is not surprising as morphologymay not only be correlated with what a fish feedson, but also with feeding behavior (e.g. suction, bit-ing) or microhabitat utilization . Our study indicatesthat the `fit' is not simply between morphology anddiet, but also between morphology and microhab-itat utilization . Quantitative assessment of micro-habitat utilization in these seagrass fishes wouldhave allowed us to test the association betweenmorphology and habitat use .
Gatz (1979a), Moyle & Senanayake (1984), Dou-glas (1987), Wikramanayake (1990) and Winemiller(1991) have found that morphological diversifica-tion and specialization in stream fishes was associ-ated with dietary and/or microhabitat specializa-tion. Niche compression, or specialization in streamfish communities occurred primarily in relation tohabitat selection, in the case of rainforest fishes ofnorth Borneo by vertical stratification, and second-arily in preference for food resources (Watson &Baton 1984) .
Grossman (1986) found that morphological simi-larity was a poor predictor of dietary similarity in anintertidal fish assemblage . Similar to our findingswith such species as E. gula and L. rhomboides, hefound that in some cases species which were quitesimilar morphologically frequently possessed verydifferent diets and vice versa. Pacific and WesternAtlantic butterflyfishes exhibited cases of conver-gence, divergence, and parallelism in jaw morphol-
ogy; and jaw and head morphology was correlatedwith how these fishes feed, rather than with whatthey consume (Motta 1985,1988) .
The inconsistencies in ecomorphological studiesattempting to relate morphology to dietary prefer-ence are to be expected owing to the numerous fac-tors that can influence the relationship . A variety ofbehavioral, ecological, physiological, and morpho-logical constraints can confound ecomorphologicalrelationships . These constraints can be evolution-ary (historial) or current. Current constraints may
be ecological (e .g . environmental instability, re-source availability, competition), behavioral (e .g .behavioral flexibility), physiological (e.g. sensorylimitations, nutritional requirements), or morpho-logical (e .g . structural and spatial limitations oncombining functionally relevant forms, phenotypicplasticity) (Motta & Kotrschal 199�) .
One ecological constraint that can affect eco-morphological relationships in a study such as this isresource availability over the species range. Thediet of each species should be examined throughoutits range and over a long enough time period toavoid localized variability . These limitations areusually too difficult to address and only a few stud-ies have addressed them (see Grossman 1986) . Sim-ilarly, how and where an organism feeds must be ex-amined over its spatial and temporal range beforeecomorphological patterns can be ascertained .There is also evidence to believe that resource par-titioning (Ross 1986), and consequently ecomorph-ological patterns relating to resource use, will notnecessarily be similar in different ecosystems suchas marine or freshwater habitats, streams, lakes,coral reefs, subtropical seagrass beds and the like .
Furthermore, various analytical methodologieswill affect the putative correlations : the choice ofmorphological characters ; as we have demonstrat-ed, lumping of mensural and coded variables signif-icantly changes the principal components analysis,utilizing ratios or raw measurements, and the meth-od of dietary analysis, for example, whether onepresents the data volumetrically, as dried or wetweights, as percent frequency of occurrence, or assome cumulative index such as the Index of Rela-tive Importance. There is, therefore, no surprise tous that there is so much variability and lack of con-cordance among the ecomorphological studies offish feeding . Perhaps we should seek ecosystem-specific patterns, and standardized methods or atleast compare the different methodologies. Evenwhen correlations are found between diet or micro-habitat use and morphology, we lack the certaintyto ascertain causal relationships . Ecomorphologi-cal studies must then proceed to the more difficultstep, performance testing of the form-functioncomplex, for example, the feeding efficiency of vari-ous gill raker designs .
Phylogenetic patterns
It has been argued that examining more distantlyrelated species, rather than those with a high degreeof taxonomic relatedness, decreases the probabilityof detecting ecomorphological patterns becausechoosing closely related species will reduce the riskthat coincidental differences will mask significantpatterns (Huey & Bennet 1986), and closely relatedspecies that have a long history of evolution and ra-diation in the same region are more likely to haveecomorphological relationships that are detectable(Findley & Black 198�). However, re-occurringecomorphological relationships among more dis-tantly related taxa provide powerful evidence forconvergence (Karr & James 1975, Wiens 1991b, Lo-sos 1990a, b, Motta & Kotrschal 199�, Winemiller etal . 1995) and reduce the probability that the eco-morphological pattern is a chance event, but ratherone shaped by evolutionary forces related to thatparticular ecomorphological relationship .
In order to ascertain that different faunas haveevolved similar patterns of niche diversification inresponse to similar environmental factors one musthave a phylogeny to distinguish between conver-gence and parallelism and to identify historical de-sign constraints among contemporary taxa . Fur-thermore, one must have reasonably comparableunits and scales among : (1) heritable morphologicaltraits that reflect ecologically relevant functions, (�)faunas and taxa, and (�) regions and physical andbiotic environments (Winemiller et al . 1995) . It maythen be possible to rigorously test evolutionary eco-morphological hypotheses as has been outlined byFelsenstein (1985), Losos (1990a, b), Winemiller(1991), Winemiller et al . (1995), Wesneat (1995) .With the availability of a phylogeny of this broadtaxonomic group, and morphological charactersthat are ecologically relevant, we can at least makegeneral statements about evolutionary conver-gence and parallelism in this group of seagrass fish-es.
These ten species of generally distantly relatedfishes formed a variety of groups that clustered inmorphospace, indicating convergence and parallel-ism in form . The only good correlation betweendietary similarity and morphological similarity oc-
55
56
curred for the group of mid-water, plankton-feed-
ing fishes : A. hepsetus, H. jaguana, and M. penin-sulae . The former two species are more closely re-lated (Fig. 9), are closer to the more primitive tele-ost body form than the other species, and share theancestral characters of forked caudal fins, lateraleyes, and intermediate length pectoral fins (JamesAlbert personal communication) . Similarity in thecharacters between these clupeomorph fishes and
' rotoosnthoptorsll'
Ptomlltormoo
U10pItormoo
Myotophltormos
Pa n Os at II opt p ryg II
slenlformss
Os.Nroptotformois
-== =---o/posulformss
Fig. 9. Partial phylogeny of the Teleostei based on Lauder & Liem (198�) with the ten species of seagrass fishes under investigation
indicated .
the atherinoid silverside is most likely due to evolu-
tionary convergence.The epibenthic, highly mobile and maneuverable
benthic-feeding perciform species E. gula and L.rhomboides are relatively closely related (Fig . 9)
and presumably share derived [sub-gibbose bodyform with long, pointed pectoral fins, subterminalmouth, protrusible mouth (Schaeffer & Rosen1961)] and ancestral characters (forked caudal fins
and lateral eyes). Close phylogenetic relatedness
most likely indicates parallel evolution in manycharacters .
As a group, the epibenthic cyprinodontids E sim-ilis and F ca rpio show convergence in some charac-ters with C. schoepfi and S. scovelli, notably in finshape (more rounded pectoral and caudal fins) andreduction in gill raker size . Ecomorphological con-vergence in these seagrass fishes is related both tomicrohabita t and trophic utilization, as has beenidentified in other groups. Winemiller (1991) simi-larly identified convergence in characters associat-ed with diet and micro-habitat in five regional as-semblages of fresh water fishes .
In this assemblage of ten species of seagrass fish-es, relatively few microhabitat and trophic charac-ters accounted for most of the variance in morphol-ogy. Although Felley (1984) cautioned that eco-
morphological associations shown from one groupof species may not be relevant to other groups, wenote that ecomorphological studies encompassing
freshwater and marine fishes, including the worksof Gatz (1979a), Felley (1984), Moyle & Sen-anayake (1984), Watson & Balon (1984), Grossman(1986), Mott:a (1988), Wikramanayake (1990),
Winemiller (1991), and this study, have found that :
(1) ecomorphological associations, when they oc-cur, are related primarily to microhabitat utilizationand feeding ; and (�) relatively few morphologicalcharacters account for most of association withecology. They include, not in any particular order ofimportance: body size and shape, gill raker lengthand spacing, mouth orientation and size, eye posi-tion and size, gut length, shape and size of the pecto-ral and caudal fin, head length, tooth shape and size,caudal peduncle shape and size, degree of jaw pro-trusion, and presence/absence of barbels . Eco-
morphological studies should perhaps focus onthese morphological characters when evolutionary
patterns are sought .In summary, this assemblage of ten distantly re-
lated species of subtropical seagrass fishes demon-strated generally poor correspondence betweenmorphology and diet . Morphological similarity wasonly reflective of dietary similarity in the guild ofmid-water, planktotrophic fishes . Morphologicalsimilarity in most of the remaining fishes was appar-
57
ently reflective of microhabitat utilization and feed-ing behavior. In general, three groups segregated
out: (1) a group of mid-water fishes specialized forcruising and seeking out evasive prey; (�) slowswimming, less maneuverable epibenthic fishesthat picked or sucked their prey off the substrate ;(�) and a group of more mobile and maneuverableepibenthic foragers. Within this assemblage, casesof convergence in trophic and microhabitat utiliza-tion characters were apparent in some of thegroups. Future ecomorphological studies on fishes,and seagrass fishes in particular, should examinecorrelations between morphology and diet, micro-habitat utilization and foraging behavior .
Acknowledgements
This study was funded in part by a University ofSouth Florida President's Research Award toP.J.M. We would like to thank Rebecca Wilcox,Robert Windheuser and all field volunteers fortheir assistance in making the research possible .
References cited
Aleev, Y.G. 1969 . Function and gross morphology in fish . Akad .Sci . USSR, Sevastopol Biol . Stn ., Israel Program for Sci .Transl., Jerusalem . �68 pp.
Alexander, R.McN. 1974. Functional design in fishes . Hutchin-son University Library, London . 160 pp .
Atchley, WR., C.T. Gaskins & D . Anderson. 1976. Statisticalproperties of ratios . I. Empirical results . Syst . Zool . �5 : 1�7-148 .
Block, W.M., L .A . Brennan & R.J . Gutierrez. 1991 . Ecomorph-ological relationships of a guild of ground-foraging birds innorthern California, USA . Occologia 87 :449-458.
Brook, I.M. 1977. Trophic relationships in a seagrass community(Thalassia testudinum), in Card Sound, Florida . Fish diets inrelation to macrobenthic and cryptic faunal abundance . Trans.Amer. Fish . Soc. 106 : �19-��9 .
Carr,W.E.S.& C.A. Adams . 197� . Food habits of juvenile marinefishes occupying seagrass beds in the estuarine zone nearCrystal River, Florida . Trans . Amer . Fish . Soc . 10� : 511-540 .
Chao, L.N. & J .A. Musick . 1977 . Life history, feeding habits, andfunctional morphology of juvenile sciaenid fishes in the YorkRiver Estuary, Virginia . U.S. Fish . Bull . 75 : 657-70� .
Comp, G.S. 1985. A survey of the distribution and migration ofthe fishes in Tampa Bay. pp. �9�-4�5. In : S.F Treat, J .L . Simon,
58
R.R. Lewis & R .L. Whitman, Jr . (eds) Tampa basis . Bay AreaScientific Information Symposium, Bellwether Press, Edina .
Comp, G .S. & W. Seaman, Jr . 1985 . Estuarine habitat and fisheryresources of Florida. pp. ��7-4�5 . In: W. Seaman (ed.) FloridaAquatic Habitat and Fishery Resources, American FisheriesSociety, Eustis .
Cyrus, D.P. & S.J.M. Blaber . 198� . Mouthpart structure and func-tion and the feeding mechanisms of Gerres (Teleostei) . S. Aft.J . Zoo! . 17 :117-1�1 .
Davis, W.P. & R .S . Birdsong . 197� . Coral reef fishes which foragein the water column. Helgol. wiss . Meeresunters. �4 : �9�-�06.
Darnell, R.M. 1958 . Food habits of fishes and larger inverte-brates of Lake Pontchartrain, Louisiana, an estuarine commu-nity. Publ . Inst . Mar. Sci. Univ. Tex. 5 : �5�-416.
Douglas, M.E. 1987 . An ecomorphological analysis of nichepacking and niche dispersion in stream-fish clades . pp. 144-
149 . In: W.S. Mathews & D.C. Heins (eds) Community andEvolutionary Ecology of North American Stream Fishes, Uni-versity of Oklahoma Press, Norman .
Ehlinger, T.J . 1990. Habitat choice and phenotype-limited feed-ing efficiency in bluegill : individual differences and trophic po-lymorphism. Ecology 71 : 886-896 .
Ehlinger, T.J . & M .R. Gross . 199� . Ecomorphology of alterna-tive male reproductive strategies . 7�nd Ann . Meeting Amer.Soc . Ichthyol . Herp., University of Illinois, Champaign-Urba-na (abstract) .
Ehlinger, T.J. & D.S. Wilson. 1988 . Complex foraging polymor-phism in bluegill sunfish . Proc. Natl. Acad. Sci . 85 :1878-188� .
Felley, J .D . 1984. Multivariate identification of morphological-environmental relationships within the cyprinidae (Pisces) .Copeia 1984: 44�-455 .
Felsenstein, J . 1985. Phylogenies and the comparative method .Amer. Nat. 1�5 : 1-15 .
Findley, J .S. & H . Black . 198� . Morphological and dietary struc-turing of a Zambian insectivorous bat community. Ecology 64 :6�5-6�0 .
Galis, F.199� . Interactions between the pharyngeal jaw appara-tus, feeding behavior and ontogeny in the cichlid fish, Haplo-chromis piceatus. A study of morphological constraints in evo-lutionary ecology. J . Exp . Zool. �67 :1�7-154 .
Gatz, A .J . Jr . 1979a. Ecological morphology of freshwaterstream fishes. Tulane Studies in Zoology and Botany �1 : 91-1�4 .
Gatz, A .J . Jr . 1979b. Community organization in fishes as indicat-
ed by morphological features . Ecology 60 : 711-718.
Goldschmid, A ., K . Kotrschal & P. Wirtz . 1984 . Food and gutlength of 14 Adriatic bleniid fish (Bleniidae; Percomorpha ; Te-leostei) . Zool . Anz. �1�:145-150 .
Grossman, G.D. 1986. Food resource partitioning in a rocky in-tertidal fish assemblage . J . Zool . Lond. 1 : �17-�55 .
Hansen, D.J . 1969. Food, growth, migration, and abundance ofpinfish Lagodon rhomboides, and Atlantic croaker, Micropo-gon undulates, near Pensacola, Flor . U.S. Fish . Wildl . Serv.,Fish . Bull. 68:1�5-146.
Hill, M.O.197� . Diversity and evenness: an unifying notation andits consequences. Ecology 54: 4�7-4��.
Huey, R .B. & A.F. Bennett . 1986 . A comparative approach tofield and laboratory studies in evolutionary biology. pp. 8�-98 .In : M.E. Feder & G.V. Lauder (eds) Predator-Prey Relation-ships : Perspectives and Approaches from the Study of LowerVertebrates, University of Chicago Press, Chicago .
Huh, S.H. & C .L . Kitting. 1985 . Trophic relationships among con-centrated populations of small fishes in seagrass meadows . J.exp . mar. Biol . Ecol . 9�:�9-4� .
Jackson, D.A ., H.H. Harvey & K.M. Somers. 1990. Ratios in aq-uatic sciences : statistical shortcomings with mean depth andthe morphoedaphic index . Can . J . Fish . Aquat . Sci . 47 :1788-1795 .
Jackson, D .A. & K.M. Somers . 1991 . The spectre of `spurious'correlations . Oecologia 86 :147-151.
Johansson, J.O.R., K .A. Steidenger & D.C. Carpenter . 1985 . Pri-mary production in Tampa Bay : a review, pp. �79-�98 . In: S.F.Treat, J .L . Simon, R.R. Lewis & R .L. Whitman, Jr. (eds) Tam-pa basis. Bay Area Scientific Information Symposium, Bell-wether Press, Edina.
Jongman, R.H.G ., C .J .F. ter Braak & O.F.R. van Tongeren (eds) .1987 . Data analysis in community and landscape ecology . Pu-doc, Wageningen . �98 pp.
Karr, J.R. & F.C . James. 1975 . Ecomorphological configurationsand convergent evolution in species and communities . pp.�58-�91 . In: M.L. Cody & J.M. Diamond (eds) Ecology andEvolution of Communities, Belknap Press, Cambridge .
Keast, A . & D . Webb. 1966 . Mouth and body form relative tofeeding ecology in the fish fauna of a small lake, Lake Opinion,Ontario . Journ. Fish . Res. Board Can . �� :1845-1874 .
Kotrschal, K . 1989 . Trophic ecomorphology in eastern Pacificblennioid fishes: character transformation of oral jaws and as-sociated change of their biological role . Env. Biol. Fish . �4:199-�18 .
Lagler, K .E, J .E. Bardach & R.R. Miller . 196� . Ichthyology. JohnWiley and Sons, New York . 497 pp.
Lauder, G. V. & K .F. Liem . 198�. The evolution and interrelation-ships of the Actinopterygian fishes . Bull . Mus. Comp. Zool.150:95-197 .
Leisler, B. 1980 . Morphological aspects of ecological specializa-tion in bird genera . Okol . Vogel . � :199-��0 .
Leisler, B. & H . Winkler . 1985. Ecomorphology. pp . 155-186 . In :R.F. Johnston (ed.) Current Ornithology, Plenum Press, NewYork .
Liem, K .F.1980 . Adaptive significance of intra and interspecificdifferences in the feeding repertoires of cichlid fishes . Amer.Zool. �0: �95-�14 .
Livingston, R .J. 1976. Diurnal and seasonal fluctuations of orga-nisms in a north Florida estuary. Estuarine and Coastal Ma-rine Science 4: �7�-400 .
Livingston, R.J . 198� . Trophic organization of fishes in a coastalseagrass system. Mar . Ecol . Prog . Set . 7 :1-1�.
Livingston, R .J. 1984 . Trophic response of fishes to habitat varia-bility in coastal seagrass systems. Ecology 65:1�58-1�75 .
Losos, J.B . 1990a . Ecomorphology, performance capability, andscaling of West IndianAnolis lizards : an evolutionary analysis .Ecol. Monogr . 60 : �69-�88.
Losos, J.B. 1990b. The evolution of form and function : morphol-ogy and locomotor performance in West Indian Anolis lizards .Evolution 44: 1189-1�0� .
Lucas, J.R . 198�. Feeding ecology of the gulf silverside, Menidiapeninsulae, near Crystal River, Florida, with notes on its lifehistory. Estuaries 5 :1�8-144.
Ludwig, J.A . & J.E Reynolds. 1988. Statistical ecology. John Wi-ley and Sons, New York. ��7 pp.
Miles, B.M ., R. E . Ricklefs & J . Travis . 1987. Concordance of eco-morphological relationships in three assemblages of passerinebirds . Amer. Nat. 1�9:�47-�64.
Miles, D.B . & R.E . Ricklefs . 1984. The correlation between ecol-ogy and morphology in deciduous forest passerine birds . Ecol-ogy 65 :16�9-1640.
Motta, P.J . 1984. Mechanics and functions of jaw protrusion inteleost fishes : a review. Copeia 1984 :1-18 .
Motta, P.J. 1985 . Functional morphology of the head of Hawaiianand Mid-Pacific butterflyfishes (Perciformes, Chaetodonti-dae) . Env. Biol. Fish . 1� : �5�-�76 .
Motta, PJ. 1988 . Functional morphology of the feeding appara-tus of ten species of Pacific butterflyfishes (Perciformes, Chae-todontidae): and ecomorphological approach . Env. Biol . Fish .�� : �9-67 .
Motta, P.J ., K.B. Clifton, E Hernandez, B .T. Eggold, S.D. Giorda-no & R . Wilcox. 1995. Feeding relationships among nine spe-cies of seagrass fishes of Tampa Bay, Florida . Bull . Mar. Sci . 56 :185-�00 .
Moyle, PB. & F.R. Senanayake . 1984. Resource partitioningamong the fishes of rainforest streams in Sri Lanka . J . Zool .Lond. �0�:195-��� .
Norton, S.F. 1991 . Capture success and diet of cottid fishes : therole of predator morphology and attack kinematics . Ecology7�:1807-1819 .
Pinkas, L. . M .S . Oliphant & I .L .K . Iverson . 1971 . Food habits ofalbacore, bluefin tuna and bonito in California Waters . Calif.Fish Game . 15� : 1-105 .
Pounds, J.A . 1988 . Ecomorphology, locomotion, and microhab-itat structure : patterns in a tropical mainland Anolis commu-nity. Ecol . Monog. 58: �99-��0.
Reist, J.D. 1985 . An empirical evaluation of several univariatemethods that adjust for size variation in morphometric data .Can . J . Zool. 6� : 14�9-14�9 .
Ricklefs, R .E. & O.W. Cox . 1977 . Morphological similarity andecological overlap among passerine birds on St . Kitts, BritishWest Indies. Oikos �9: 60-66 .
Ricklefs, R .E . & .J . Travis . 1980 . A morphological approach tothe study of avian community organization. The Auk 97: ��1-��8 .
Ross, S .T. 1986. Resource partitioning in fish assemblages : a re-view of field studies. Copeia 1986: �5�-�88 .
Sanderson, S.L . & J.J . Cech, Jr . 199� . Energetic cost of suspen-sion feeding versus particulate feeding by juvenile Sacramen-to blackfish . Trans . Amer. Fish . Soc. 1�1:149-157 .
59
Schaeffer, B . & D.E. Rosen . 1961. Major adaptive levels in theevolution of the actinopterygian feeding mechanism . Amer.Zool . 1 : 187-�04 .
Springer, V.G. & K .D. Woodburn . 1960 . An ecological study ofthe fishes of Tampa Bay area . Fla . St . Board Conser. Prof. Pap.1 :1-104 .
Stoner, A .W 1980. Feeding ecology of Lagodon rhomboides(Pisces, Sparidae): variation and functional response . U.S.Fish . Bull . 78:��7-�5� .
Stoner, A.W. 198� . Distribution of fishes in seagrass meadows :role of macrophyte biomass and species composition . U.S.Fish . Bull . 81: 8�7-846.
Strauss, R.E . 1987 . The importance of phylogenetic constraintsin comparisons of morphological structure among fish assem-blages . pp . 1�6-14�. In: W.J. Mathews & D.C . Heins (eds) Evo-lutionary Ecology of North American Stream Fishes, Univer-sity of Oklahoma Press, Norman .
Subrahmanyam, C.B. & S.H. Drake. 1975 . Studies on the animalcommunities in two north Florida salt marshes. Part III . Sea-sonal fluctuations of fish and macroinvertebrates . Bull. Mar.Sci. �5 : 445-465 .
Suyehiro, Y. 194�. A study on the digestive system and feedinghabits of fish. Japan. J . Zool . 10 : 1-�01 .
ter Braak, C .J .F. 1986 . Canonical correspondence analysis : a neweigenvector technique for multivariate direct gradient analy-sis . Ecology 67 :1167-1179.
ter Braak, C .J.F. 1987 . The analysis of vegetation-environmentrelationships by canonical correspondence analysis . Vegetatio69: 69-77.
ter Braak, C .J .E 1988 . CANOCO- a FORTRAN program for ca-nonical community ordination by correspondence analysis,principal components analysis and redundancy analysis (ver-sion �.1�) . Report LWA-88-0�, Agricultural MathematicsGroup, Wageningen . 95 pp.
Thayer, G.W., D.R. Colby & W.F. Hettler, Jr. 1987 . Utilization ofthe red mangrove prop root habitat by fishes in south Florida .Mar. Ecol . Prog. Ser. �5 : �5-�8 .
Wainwright, PC . 1987 . Biochemical limits to ecological perform-ance: mollusc-crushing by the Caribbean hogfish, Lachnolai-mus maximus (Labridae) . J . Zoo]. Lond . �1� : �8�-�97 .
Wainwright, PC. 1988 . Morphology and ecology: functional basisof feeding constraints in Caribbean labrid fishes . Ecology 69:6�5-645.
Wainwright, PC . 1991 . Ecomorphology: experimental functionalanatomy for ecological problems. Amer . Zool. �1 : 680-69� .
Watson, D .J . & E.K. Balon . 1984. Ecomorphological analysis offish taxocenes in rainforest streams of northern Borneo . J.Fish Biol . �5 : �71-�84 .
Webb, P W.1984. Body form, locomotion and foraging in aquaticvertebrates. Amer . Zool. �4:107-1�0 .
Wesneat, M.W. 1995 . Phylogenetic systematics and biomechan-ics in ecomorphology. Env. Biol . Fish . (in press) .
60
Wiens, J .A. 1991a . Ecological similarity of shrub-desert avifau-nas of Australia and North America . Ecology 7� : 479-495 .
Wiens, J .A . 1991b. Ecomorphological comparisons of the shrub-desert avifaunas of Australia and North America . Oikos 60:55-6� .
Wiens, J . A . & J .T. Rotenberry. 1980. Patterns of morphology andecology in grassland and shrubsteppe bird populations . Ecol .Monogr . 50: �87-�08.
Wikramanayake, E .D. 1990. Ecomorphology and biogeographyof a tropical stream fish assemblage : evolution of assemblagestructure. Ecology 71 :1756-1764 .
Winemiller, K .O. 1991 . Ecomorphological diversification in low-land freshwater fish assemblages from five biotic regions .Ecol. Monogr. 61 : �4�-�65 .
Winemiller, K.O ., L.C . Kelso-Winemiller & A .L. Brenkert . 1995 .Ecomorphological diversification and convergence in fluvialcichlid fishes. Env. Biol. Fish . 44 : ��5-�61 .
Yamaoka, K . 198�. Feeding behaviour and dental morphology ofalgae scraping cichlids (Pisces: Teleostei) in Lake Tanganyika .African Study Monographs 4 : 77-89.
Yamaoka, K ., M . Hori & S . Kuratani . 1986 . Ecomorphology offeeding in `goby-like' cichlid fish in Lake Tanganyika . Physiol .Ecol. Japan . �� :17-�9 .
Young, D.K. & M.W. Young. 1978 . Regulation of species densi-ties of seagrass-associated macrobenthos: evidence from fieldexperiments in the Indian River estuary. Fla . J. Mar . Res . �6 :569-59� .
