In: Community and Evolutionary Ecology of North American Stream Fishes. William J.

Matthews and David C. Heins (eds.), University of Oklahoma Press, Normal OK. 1987. 310 pp.
17. The Importance of Phylogenetic Constraints in Comparisons of Morphological
Structure Among Fish Assemblages
Richard E. Strauss
Abstract
Morphological approaches to the study of community relationships are based on the premise that morphological
differences among organisms reflect in large part their ecological relationships and that morphological "space" can be
mapped onto ecological "space." Although morphological characteristics have been used to account for aspects of
foraging and habitat use, phylogenetic effects of form diversity on the repeatability of morphological patterns of variation
among communities are unknown. The morphological structures of several North and South American freshwater fish
assemblages were compared to assess the dependence of morphological congruence on taxonomic similarity. Correlations
of morphological patterns are high (> 0.7) for geographically proximate assemblages, but for comparisons between North
and South American assemblages the average correlation is 0.20. The density at which species are "packed" into
morphospace, and thus the average similarity between species, is independent of the number of species present; within
assemblages, therefore, total size-independent morphological variation is highly dependent on species diversity.
Hypotheses about patterns in natural communities, their
structure and composition, are at the center of current ecolo-
gical debate (Cody and Diamond, 1975; Wiens and Roten-
berry, 1980a; Anderson et aI., 1981; Strong, 1983; Salt,
1984; Strong et aI., 1984). Important questions are the extc.nt
to which competitive interactions pattern the observed ecolo-
gical and morphological relationships of species within com-
munities, and to what degree the ecological functions
(niches) of members can be predicted from their morpholo-
gical characteristics. The assumption of strong correspon-
dence between niche and morphology (that is, that morpho-
logical "space" maps directly onto ecological "space";
Hutchinson, 1968; Ricklefs and Travis, 1980) is critical to
many community studies. For example, several researchers
have compared patterns of morphological structure among
communities (Ricklefs and O'Rourke, 1975; Findley, 1973,
1976; Gatz, 1979b; Ricklefs and Travis, 1980; Ricklefs et
al., 1981) or expressed relative niche characteristics in terms
of morphological differences among species (Hespenheide,
1973; Gatz, 1979a) . Yet congruence between morphology
and ecology is seldom tested directly (Felley, 1984; Miles
and Ricklefs, 1984).
It is evident that morphological characteristics can be used
to predict aspects of foraging behavior and habitat use (Fla-
nagan, 1981; Gatz, 1979b, 1981; Findley and Black, 1983;
Miles and Ricklefs, 1984). However, a considerable amount
136
of residual morphological variation in many communities
cannot be related directly to ecological variables (Weins and
Rotenberry, 1980b; Felley, 1984), reducing the predictabil-
ity of morphological relationships in one assemblage from
those observed in another. An additional confounding factor
is faunal composition: if the historical pool of taxonomic
"morphotypes" has been very different for two assem-
blages-that is, if the communities being compared are com-
prised of species derived from sets of independent evolution-
ary lineages-then the resulting morphological patterns
could differ profoundly. How might morphological structure
be affected, for example, when characins are "substituted"
(in a historical sense) for cyprinids? Two possibilities are (1)
that ecological (e.g., hydrodynamic, foraging) factors are so
important in sorting out sets of coexisting species that the
major patterns of variation remain relatively stable among
communities, irrespective of the actual species present, or
(2) that differences in body form among major evolutionary
lineages are large enough to override small-scale, localized
ecological pressures, so that, for example, North American
stream-fish communities might be structured very differently
from corresponding South American communities. In the
latter case some degree of correspondence in major patterns
of morphological variation might be evident once phylogene-
tic constraints (taxonomic similarities) have been taken into
account.
The Importance of Phylogenetic Constraints in Comparisons of Morphological Structure 137
Table 17.1. Localities and Their Taxonomic Composition*
Number of Collection
Locality
Families Genera Species Individuals
No.
Area
t
Sites
1. Little Fishing Creek, PA 8 19
Pennsylvania
2. Yellow Breeches Creek, PA 6 15
Pennsylvania
3. Lake Opinicon, ON 8 14
Ontario
4. Big Sandy Creek, TX 8 16
Texas
5. Martis Creek, CA 5 9
California
6. Rio Parana, PY 9 31
Paraguay
7. Rio Aquidaban, PY 5 21
Paraguay
8. Composite 22 22
'Taxa are given in the Appendix.
'Sampled length of stream or surface area of lake.
To explore such questions in detail, I initiated a study of
morphological variation among freshwater fish assemblages
collected from a variety of climatic regions and having vary-
ing degrees of taxonomic similarity to one another. Specifi-
call y, the study was designed to answer the following ques-
tions: (l) What are the major patterns of variation, in-
dependent of ontogenetic effects on body form, among fishes
in stream assemblages? (2) Are patterns of variation con-
sistent among assemblages from different habitats? (3) Is
degree of concordance dependent on faunal composition? If
so, at what taxonomic level? (4) Is the overall level of
similarity, the degree to which species are "packed" into
morphological space, dependent on the number or composi-
tion of species present? And does packing density differ for
temperate versus subtropical habitats?
Materials and Methods
Localities and Taxonomic Composition
Localities were chosen to represent differing habitats and
taxonomic assemblages (table 17.1). Assemblages were rep-
resented by series of fish collections taken within a limited
region (less than 25 km linear extent) of the watershed.
Localities and associated collections were selected from
three different sources: personal records, selected literature
records, and museum expedition records.
The two Pennsylvania localities (PA) were extensively
sampled in 1978-79; all specimens were deposited in the
Pennsylvania State University research collections. Lake
Opinicon, Ontario (ON; Keast and Webb, 1966; Keast,
1978b), Big Sandy Creek, Texas (TX; Evans and Noble,
1979, stations 3-5), and Martis Creek, California (CA;
Moyle and Vondracek, 1985) had been surveyed by the
authors cited. On the basis of their species lists and de-
scriptions, I located collections in the University of Michigan
28 176 6 12 km
23 138 4 14 km
16 114 5 8 km
2
30 186 3 12 km
12 78 4 3 km
38 240 5 16 km
24 158 5 24 km
22 290
Museum of Zoology (UMMZ) that corresponded as nearly as
possible to the cited geographical locations and that had been
collected in similar habitat, especially lake versus stream
(similar studies that lacked corresponding UMMZ col-
lections were disregarded). These collections were consid-
ered to be representative of the described assemblages, under
the reasonable assumption that intraspecific geographic var-
iation is negligible in relation to ontogenetic and interspecific
variation. The two Paraguayan (PY) localities were selected
from a large survey conducted by the UMMZ Paraguayan
Expedition of 1979. All specimens examined were taken
directly from those collections.
Because the North American and the Paraguayan assem-
blages had virtually no faunal overlap, a composite sample
was assembled by pooling data on one species from each of
22 families represented in the study (see Appendix). Species
were chosen to maximize taxonomic overlap among locali-
ties. In several analyses of morphological structure this com-
posite was treated as an independent assemblage, though the
data comprising it were replicated from the other samples.
Measures of taxonomic similarity (S) among assemblages
were based on Dice's (= Sorenson's) index, S = 2N)(N, +
N 2)' where N, and N 2 are the numbers of taxa in two assem-
blages and Nc is the number held in common. The index
ranges from 0 to I and is linearly proportional to the propor-
tion of shared taxa (Cheetham and Hazel, 1969). Indices
were computed separately for species, genera, and families;
the three values were averaged to give a composite measure
of taxonomic similarity. This procedure diminishes the effect
of taking at face value the names and ranks assigned to taxa,
particularly for South American groups taxonomic treat-
ments of which have been uneven in scope and quality.
Cluster analyses of assemblages in relation to taxonomic
structure were performed on the l' s complements of similar-
ity values (D = 1 - S = [N, + N2 - 2NcJ/[N, + N2]) using
the UPGMA clustering algorithm (Sneath and Sokal, 1973).
138 Community and Evolutionary Ecology of North American Stream Fishes
Description of Morphology
The biological reliability of any morphological study is criti-
cally dependent on how form and form differences are quan-
tified and analyzed (Strauss and Bookstein, 1982; Bookstein
et a!., 1985). The geometric and statistical methods used in
this study were chosen (1) to provide a fairly complete
morphological description of each species, unbiased by prior
functional expectations; (2) to encompass sets of functionally
analogous anatomical reference points (= landmarks); (3) to
explicitly include information on ontogenetic (allometric)
variation; and (4) to allow forms to be archived and recon-
structed.
Species in each assemblage were represented by 5-12
specimens, deliberately chosen to give the largest size range
possible (from 1: 1. 5 to 1: 17, smallest:largest standard
lengths). All species were used except for several extreme
phenotypes: anguillids, synbranchids, gymnotids, and rham-
phichthyids. Data consisted entirely of mensural characters
(distances measured with respect to anatomical landmarks).
Specimens were photographed in lateral aspect, with
accompanying metric scale, with the use of small insect-
mounting pins to locate midsagittal landmarks (e.g., anus,
supraoccipital crest) that could not easily be seen from lateral
view. Landmarks (fig. 17.1) were marked on each photo-
graph and digitized as Cartesian coordinates. Euclidean dis-
tances were subsequently computed between pairs of scaled
landmark coordinates; distances were selected to form' 'trus-
ses" (fig. 17.1 A; Strauss and Bookstein, 1982) and addition-
al combinations of measurements in lateral projection (fig.
17.1 B). Auxiliary measurements of body width, taken with
digital calipers into a computer file, were used to approxi-
mate a triangulation in dorsal projection (fig. 17 .IC); bilater-
al measurements were averaged before analysis. Body-width
measurements and distances across the abdomen were
selected to be minimally affected by variation in volume of
gas bladder, gut, or ovaries. The resulting data set adequately
describes the major features of external body form, including
body proportions and relative sizes and positions of mouth,
eye, head, and fins. It omits information about complex fin
shapes; degree of separation of dorsal fins; adipose fins,
barbels, and other specialized appendages; and internal anat-
omy.
Morphological Space
Analyses are based on the positions of specimens in a multi-
dimensional morphological hyperspace (morpho space) the
axes of which are logarithms of the 54 mensural characters.
The logarithmic transformation linearizes allometries, stan-
dardizes variance, and produces a scale-invariant covariance
matrix (Bookstein et a!., 1985).
Ontogenetic size variation was explicitly included as a
major factor by an examinination of secondary size-
independent patterns of variation with respect to multivariate
size factors. This procedure takes into consideration the
substantial changes in form that occur during growth. It also
negates the necessity of arbitrarily limiting size ranges to
"adults. "
The full morphospace was reduced by means of principal
component analysis (PCA) of the covariance matrix (Book-
stein et a!., 1985). Use of the covariance matrix preserves
allometries and leaves the geometric space undistorted, so
that Euclidean distances based on original log measurements
and on principal component scores are identical. Because of
the large size ranges within and among species, the first
component was in all cases a strong size vector, accounting
for> 72% of the total variance within assemblages and>
97% of the variance within species, with consistently posi-
tive loadings. The first five components always accounted
for> 98% of the variance among species; thus the sixth and
subsequent components were disregarded for multivariate
comparisons (but not for calculation of nearest-neighbor
distances). For size-free comparisons among assemblages,
sheared principal components (Humphries et a!., 1981;
Bookstein et al., 1985) were computed by regressing out the
pooled within-group size factor (PC 1) while maintaining t?e
group centroids. Sheared components are ver;: stable d.lS-
criminant axes that are unaffected by differences III mean size
among groups.
A
c ~ ~
i _ - ----- ---------
Fig. 17.1. The set of 56 mens ural characters used for morphoI?etric
analyses. Anatomical landmarks are indicated by open Circles;
distance measures, by dotted lines. A: Midsagittal truss measures.
B: Auxiliary measures in lateral projection. C: Auxiliary measures
in dorsal projection.
The Importance of Phylogenetic Constraints in Comparisons of Morphological Structure 139
Assessment of pairwise congruence of morphological var-
iation among assemblages was based on shared principal
component analyses. For each pairwise comparison the data
for two assemblages were pooled, PCs were extracted, and
PC I was sheared (regressed) from components 2-5. Within
this reduced size-free space, loadings can be represented as
vector correlations (directional cosines; Wright, 1954), es-
timated for each mensural character by its correlations with
projection scores across individuals (fig. 17.3). The correla-
tion between assemblages for each character is the cosine of
the angle between corresponding vectors. Mean character
correlation was estimated by standardizing all character cor-
relations between assemblages with arc sine transformations
(Sokal and Rohlf, 1981) and reconverting the mean value
across characters with an inverse arc sine transformation.
Species-packing analyses were based on PCAs performed
separately for each assemblage. Size-independent nearest-
neighbor distances (NNDs) were computed within the re-
duced (n - I, for n characters) PC space from which the first
component had been sheared. Distributions of NNDs were
compared against random models using the method of Don-
nelly (1978) and Sinclair (1985).
Results
Similarity in Faunal Composition Among Assemblages
The two PA and two PY assemblages were initially intended
to be pairs of "replicates" sampled from within major drain-
age basins. The PA samples are similar at all three taxonomic
levels (table 17.2), except for the presence of ictalurids and a
cyprinodont at Little Fishing Creek and for several sub-
stitutions of congeneric species. The Paraguayan assem-
blages are much less similar to one another than is the PA pair
owing to the presence of a diverse collection of catfishes at
the RIO Parana locality. The ON, TX, and CA groups share
intermediate numbers of taxa at all taxonomic levels. The CA
assemblage, which is small and has a high degree of endem-
ism (Moyle and Vondracek, 1985), is the most dissimilar
among the North American assemblages. As expected, there
is almost no faunal overlap between the Paraguayan and the
North American groups; Rio Parana and Little Sandy Creek
A
J
I
I
L
B
I
L
I
1.0 .8 .6 .4 .2
Mean Taxonomic Difference
I
0
PA
PA
ON
TX
CA
PY
PY
PA
PA
ON
TX
CA
Composite
PY
PY
Fig. 17.2. Dendrograms of assemblages clustered by taxonomic
dissimilarity, excluding (A) and including (B) the composite sam-
pie. Locality abbreviations are given in table 17.1.
Table 17.2. Numbers of Species, Genera, and Families Shared Among Assemblages (Above Diagonal), and Associated Measures of
Taxonomic Similarity (Below Diagonal)
Locality
1 2 3 4 5 6 7 8
PA PA ON TX CA PY PY Composite
1. PA 17, 12,5 7, 8, 6 3, 10,7 0,4,4 0, 0, 0 0,0,0 6,7,8
2. PA 0.70 5, 6, 3 2, 6, 4 1, 5, 5 0, 0, 0 0,0,0 5,6,6
3. ON 0.52 0.39 3, 8, 6 0, 1,2 0, 0, 0 0,0,0 6,8,8
4. TX 0.52 0.35 0.47 1, 1, 3 0,0, 1 0,0,0 1,7,8
5. CA 0.30 0.46 0.13 0.20 0,0,0 0,0,0 1,4,5
6. PY 0.00 0.00 0.00 0.04 0.00 9, 13,4 8,8,9
7. PY 0.00 0.00 0.00 0.00 0.00 0.45 5, 5, 5
8. Composite 0.37 0.32 0.43 0.31 0.23 0.38 0.27
140 Community and Evolutionary Ecology of North American Stream Fishes
Fig.17.3. A: Scatterplot of projections of all A
specimens within the plane of the second and ~ ~ - - - - - - - - = - : ' ~ . . . ,
third sheared (size-independent) principal 1"X
B
components, based on 54 mensural charac-
ters. Convex polygons enclose all points
within corresponding assemblages; pairs of 1'0
U
assemblages from P A and PY have been com- a...
bined. B: Selected character vectors portray- "0
ing correlations of characters with the corre- ~
o
sponding principal components of panel A. C1)
.s::.
(J')
body
depth
head
length
predorsal length
~ : : : : : : : - - I
gape
preanal I
length
body length
Sheared PC2 Correlation with PC2
possess a single but different poeciliid each. Dendograms of
the assemblages (fig. 17.2) provide an overall assessment of
taxonomic similarity.
Congruence of Morphological Distribution
All assemblages overlap considerably in morphological
space, and no pairwise combinations of localities could be
discriminated (fig. 17.3). However, species are not distrib-
uted evenly or randomly within assemblages; observed dis-
tributions of nearest-neighbor distances indicate that species
are significantly clustered, with NNDs both longer and short-
er than expected based on random models.
Comparisons of assemblages in the sheared PC space of
fig. 17.3 are meaningful only to the extent that the within-
assemblage size vectors (within-group PC 1 axes) are parallel
within the full morphological space (Humphries et a1.,
1981). Pairwise correlations among size vectors are all >
0.90 (table 17.3), indicating that PC 1 as a multivariate
measure of body size is highly consistent across species and
assemblages.
Mean character vectors within the sheared PC space (fig.
17. 3B) characterize the major size-independent trends in
body form within and among assemblages. At a given body
size species vary from elongated forms (lower right of fig.
17. 3B) with relatively wide body, posteriorly displaced me-
dial fins, and wide oral gape, to deep-bodied, narrow forms
(upper left). Size of the head and associated cephalic features
(snout length, eye diameter, posterior displacement of pec-
toral fin, etc.) are more or less independent of this morpho-
cline. These and other associated trends are generally con-
sistent with the major shape variations observed by Flanagan
(1981) for marine reef fishes. Note that these are average
character correlations across all assemblages; as described
below, the trends within any particular assemblage deviate
from these to some extent.
Patterns of Species Packing
The average morphological "density" of species within
assemblages, as indicated by mean nearest -neighbor distance
(NND), is not a function of number of species (fig. 17.4A).
Species in diverse assemblages are neither more nor less
similar than species in depauperate ones. The slight nonsig-
nificant trend of decreasing distance with increasing number
in fig. 17.4A is due to the influence of the two Paraguayan
assemblages, which have slightly smaller (but not signifi-
cantly different) mean NNDs from those of the North Amer-
ican assemblages. However, when total occupied morpho-
logical space (that is, total size-independent variance sub-
sumed) is examined as a function of diversity (fig. 17.4B)
there is a significant positive relationship, indicating that the
total space occupied (total variance) increases in proportion
to the number of species present. Again, the Paraguayan
species are relatively more similar to one another than are the
North American species. For example, the Rio Parana
assemblage, with 38 species, has approximately the same
total variance as a PA assemblage having 28 species; and the
Rio Aquidaban, with 24 species. has about the same variance
as Lake Opinicon, with only 16 species.
Table 17.3. Pairwise Correlations Among Size Vectors (Within-Assemblage PC 1) (Above Diagonal) and Among Sets of Character Vectors
(Below Diagonal)
I 2 3 4 5 6 7
Locality
PA PA ON TX CA PY PY
1. PA 0.99 0.97 0.96 0.96 0.93 0.93
2. PA 0.77 0.97 0.96 0.96 0.95 0.95
3. ON 0.44 0.42 0.98 0.96 0.94 0.94
4. TX 0.63 0.49 0.73 0.96 0.93 0.94
5. CA 0.37 0.71 0.28 0.24 0.91 0.91
6. PY 0.32 0.25 0.19 0.20 0.07 0.98
7. PY 0.28 0.17 0.16 0.34 0.26 0.44
The Importance of Phylogenetic Constraints in Comparisons of Morphological Structure 141
Q)
.5 u
c
0
--
(/')
(5
4
....
0
..0
..c
(J'l .3
'(i)
Z
I
-- .2
(/')
Q)
....
0
Q)
Z
.1
c
0
Q)
Z
. 8
Q)
u
c
o
..... 7
~
g .6
(J'l
o
o
..c .5
0..
....
o
Z .4
A
TX
PA

CA
---ON

.---
p'A---py-
py 0
0
r = -0.29
p= 0.26
10 20 30 40
Number of Species
B
TX

PA /.
PA py
~ . 0
ON P1
CA/ r= 0.66
/. p=0.05
10 20 30 40
Num ber of Species
Fig. 17.4. Scatterplots of (A) mean nearest-neighbor distances and
(B) total size-independent morphological variance as functions of
number of species within each assemblage. Variance is expressed as
a percentage of total size-independent variance among all assem-
blages. Solid circles = North American localities; open circles =
South American localities. The expected grand-mean NND for
randomly distributed points would be 0.51; observed values are
significantly less than expected, indicating aggregation.
Distributions of NNDs are significantly nonrandom (P
0.01) for all assemblages and indicate clustering within
morphospace with respect to random models. This
nonrandomness could be due to phylogenetic patterning, in
that confamilial and congeneric species might be expected to
aggregate and to be distant from less closely related species
(Gatz, 1979a). To examine this possibility, NND distribu-
tions were determined for cyprinids in the two PA assem-
blages and for characids in the two PY assemblages; of the
families studied, these were the only ones having sufficient
numbers of species for comparison. In all four cases NND
distributions were not significantly different from random
(though somewhat overdispersed), indicating that the mor-
phospace aggregations do correspond to taxonomic group-
ings at the family level.
Congruence of Morphological Structure
Mean character correlations, estimated among all possible
pairs of assemblages (table 17.3), range from 0.77 between
the two PA localities to a nonsignificant 0.07 between the CA
and Rio Parana (PY) assemblages. The high congruence
between the PA "replicates" is not surprising; if two such
localities had exactly the same species composition, their
patterns of morphological variation should be identical ex-
cept for minor differences owing to geographic variation.
However, there is no prior expectation for correlations be-
tween North American and South American assemblages,
which actually range from 0.07 (CA-PY) to 0.34 (TX-PY)
with a mean correlation of 0.20. Such low correlations seem
to indicate that phylogenetic differences among component
species are more important than ecological constraints in
determining overall patterns of morphological structure .
Moreover, the effect of taxonomic similarity is clinal
rather than discrete. The linear relationship between mean
character correlation and taxonomic similarity (fig. 17.5) is
highlysignificant(r = 0.82,t = 7.3,df= 26,P0.0l).
Thus assemblages having intermediate levels of taxonomic
similarity are also intermediate in morphological con-
gruence. The regression function, when extrapolated to the
case of identical faunal composition (S = I), gives a pre-
c
0

+-
0


Q)
....
~



0
U





Q)
.5 +-
()
0
~

r = 0.82


0
...c:

U

c I
0
Q)
I
~

0
0 .5
Mean Taxonomic Similarity
Fig. 17.5. Scatterplot of morphological congruence as a function of
taxonomic similarity for all pairs of assemblages, including the
composite. Each point represents one assemblage pair. The regres-
sion line assumes independent residuals; observed residual correla-
tions were statistically negligible.
142 Community and Evolutionary Ecology of North American Stream Fishes
dicted mean character correlation not significantly different
from the expected 1. The predicted mean correlation of 0.20
for complete taxonomic dissimilarity is a measure of the
amount of congruence expected from ecological constraints
acting alone.
Discussion
Morphological approaches to descriptive analyses of com-
relationships are based on the premise that morpholo-
gical features of organisms reflect in large part their ecologi-
cal relationships, or at least that morphological analyses may
reveal patterns of structure that require explanation in the
context of ecological and evolutionary theory (Ricklefs and
Travis, 1980; Douglas, chap. 18). The concept of morpho-
logical structure within a community or assemblage has had a
dual context in the literature, being used in some cases to
indicate the spatial patterns of species within the morpho-
space, particularly in terms of distances between species and
overlap among groups of species, and in other cases to
indicate patterns of variation among morphological charac-
ters. Because convergence in body form and other features
seems to be widespread among fishes in both freshwater and
marine habitats (at least at a gross morphological level), it
might seem likely that the ecological interactions that sort out
coexisting combinations of species may structure assem-
blages in predictable ways, even if a one-for-one replace-
ment of species in different habitats does not occur (Schlos-
ser, 1982a).
The results of this study, however, suggest that the ability
to predict patterns of variation among communities is limited
by the degree to which the component species have been
derived from a common evolutionary and biogeographic
background. Thus a serious methodological problem con-
cerns the scale of sampling. Population sizes and degrees of
ecological overlap among species may vary widely both
spatially and temporally, and the detection of repeatable
patterns may be a critical function of the scale on which
ecological processes act, particularly in relation to magni-
tudes of gene flow among populations inhabiting different
communities and experiencing varying combinations of
predators, prey, competitors, and physical-habitat character-
istics. Nevertheless, these results should serve as a caveat for
the comparative study of ecomorphological patterns. When
the communities chosen for study lie within the same
biogeographic or faunal region, a high degree of con-
cordance is to be expected (e.g., Gatz, 1979a, 1981).
Schoener (1974) originally suggested that the total amount
of space" occupied by an assemblage of coexisting
species should enlarge as the number of extant species in-
creases, because additional species may utilize previously
unused aspects of the available ecological resources. This
hypothesis has been tested indirectly by a number of in-
vestigators using morphological analysis (e.g., Karr and
James, 1975; Findley, 1973, 1976; Ricklefs and O'Rourke,
1975; Gatz, 1979a; Ricklefs and Travis, 1980; Ricklefs et
al., 1981; Findley and Black, 1983). Their results, and those
of this study, have generally supported Schoener's hypothe-
sis that the "core" of the community space, consisting of
relatively unspecialized species, is ecologically saturated
and that species are added to a community in secondary or
novel directions. An alternative explanation is that the eco-
logical conditions of less diverse communities are favorable
only to generalist species (Ricklefs and Travis, 1980);
however, the continuous relationship between total morpho-
logical variance and species number, demonstrated in this
study, would argue against a hypothesis invoking response to
harsh or variable environments as a general controlling
mechanism. A possible mechanism for the assemblage of
freshwater fishes into specific habitats is the sorting out from
the regional source pool of species (both specialized and
un specialized) that can coexist in the short term, with minor
subsequent modification following from competitive ecolog-
ical interaction.
These findings are complementary to those of Mayden
(chap. 26) in demonstrating that an understanding of the
history of a community is an important prerequisite to assess-
ing the degree to which morphological features have been
molded by competition or other ecological interactions (see
also Brooks, 1985). In the absence of detailed knowledge
about the phylogenetic relationships of component taxa and
the geological history of the regions under study, it is impos-
sible to judge the extent to which extant communities have
resulted from recent coevolutionary processes versus the
chance assemblage of historically available species. In such
cases assessment of faunal similarity will at least provide a
rough estimate of the amount of morphological congruence
expected among assemblages in the absence of adaptive
modifications.
In summary, the hydrodynamic and competitive con-
straints imposed by the aquatic environment do not necessar-
ily lead to morphological similarity among assemblages by
way of evolutionary convergence, at least with regard to
patterns of variation among species. The teleost "morpho-
type" seems to be sufficiently flexible that morphological
structure need not be highly constrained in similar ways in
different habitats. Instead, each characteristic body fOlm
exploits the aquatic environment somewhat differently in
terms of its own morphological and ecological design, and
resultant differences among species (in magnitude and direc-
tion) are highly variable. Thus phylogenetic patterns of form
diversity are responsible both for consistency among differ-
ent fish assemblages within the same biogeographic regions
and for noncongruence among assemblages having little
faunal similarity. I
IThe University of Michigan Morphometries Study Group served as the
forum in which much of my philosophy about biological form and geometric
morphometry was developed, and I am grateful to the participants: F. L.
Bookstein, B. Chernoff, R. L. Elder, J. M. Humphries, and G. R. Smith. I
thank E. L. Cooperfor use of the PSU collections, and R. R. Miller, G. R.
Smith, and W. L. Fink for use of the UMMZcollections. Critical comments
by M. E. Douglas, M. A. Houck, and G. D. Schnell substantially improved
the manuscript. National Science Foundation grant BSR -8307719 provided
financial support.
LITERATURE CITED
Anderson, G. R. V., A. H. Ehrlich, P. R. Ehrlich, J. D. Roughgarden, B. C. Russell, and F. H.
Talbot. 1981. The community structure of coral reef fishes. Am. Nat. 117:476-495.
Bookstein, F. L., B. Chemoff, R. L. Elder, J. M. Humphries, G. R. Smith, and R. E. Strauss. 1985.
Morphometrics in Evolutionary Biology. Spec. Publ. No. 15, Acad. Natl. Sci. Phila., PA.
Brooks, D. R. 1985. Historical ecology: a new approach to studying the evolution of ecological
associations. Ann. Missouri Bot. Garden 72:660-680.
Cheetham, A. H., and J. E. Hazel. 1969. Binary (presence-absence) similarity coefficients.
J. PaleontoI. 43:1130-1136.
Cody, M. L., and J. M. Diamond (ede.). 1975. Ecology and Evolution of Communities. Belknap
Press, Cambridge, MA.
Donnelly, K. 1978. Simulations to determine the variance and edge-effect of total nearest neighbor
distance. In: I. Hodder (ed.), Simulation Methods in Archaeology, pp. 91-95. Cambridge Univ.
Press, London.
Douglas, M. E. 1987. An ecomorphological analysis of niche packing and niche dispersion in stream-
fish clades. In: W. J. Matthews and D. C. Heins (ede.), Community and Evolutionary Ecology of
North American Stream Fishes, pp. 144-149. University of Oklahoma Press, Norman.
Evans, J. W., and R. L. Noble. 1979. The longitudinal distribution of fishes in an east Texas stream.
Am. MidI. Natur. 101:333-343.
Felley, J. D. 1984. Multivariate identification of morphological-environmental relationships within
the Cyprinidae (Pisces). Copeia 1984:442-455.
Findley, J. S. 1973. Phenetic packing as a measure of faunal diversity. Am. Nat. 107:580-584.
Findley, J. S. 1976. The structure of bat communities. Am. Nat. 110:129-139.
Findley, J. S., and H. Black. 1983. Morphological and dietary structuring of a Zambian
insectivorous bat community. Ecology 64:625-630.
Flanagan, C. A. 1981. Fourier Morphometrics of Reef Fishes of the Gulf of California. Ph.D.
Thesis, Univ. Arizona.
Gatz, A. J. 1979a. Community organization in fishes as indicated by morphological features.
Ecology 60:711-718.
Gatz, A. J. 1979b. Ecological morphology of freshwater stream fishes. Tulane Stud. Zool. Bot.
21:91-124.
Gatz, A. J. 1981. Morphologically inferred niche differentiation in stream fishes. Am. MidI. Nat.
106:10-21.
Hespenheide, H. A. 1973. Ecological inferences from morphological data. Ann. Rev. Ecol. Syst.
4:213-229.
Humphries, J. M., F. L. Bookstein, B. Chernoff, G. R. Smith, and S. G. Poss. 1982. Multivariate
discrimination by shape in relation to size. Syst. Zool. 30:291-308.
Hutchinson, G. E. 1968. When are species necessary? In: R. C. Lewontin (ed.), Population Biology
and Evolution, pp. 177-186. Syracuse Univ. Press, NY.
Karr, J. R., and F. C. James. 1975. Ecomorphological configurations and convergent evolution in
species and communities. In: M. L. Cody and J. M. Diamond (eds.), Ecology and Evolution of
Communities, pp. 258-291. Belknap Press, Cambridge, MA.
Keast, A., and D. Webb. 1966. Mouth and body form relative to feeding ecology in the fish fauna of
a small lake, Lake Opinicon, Ontario. J. Fish. Res. Bd. Canada 23:1845-1874.
Keast, A. 1978. Trophic and spatial interrelationships in the fish species of an Ontario temperate
lake. Environ. BioI. Fish. 3:7-31.
,
Strauss 2
Mayden, R. L. 1987. Historical ecology and North American Highland fishes: a research program in
community ecology. In: W. J. Matthews and D. C. Heins (eds.), Community and Evolutionary
Ecology of North American Stream Fishes, pp. 210-222. University of Oklahoma Press, Norman.
Miles, D. B., and R. E. Ricklefs. 1984. The correlation between ecology and morphology in
deciduous forest passerine birds. Ecology
Moyle, P. B., and B. Vondracek. 1985. Persistence and structure of the fish assemblage in a small
California stream. Ecology 66: 1-13.
Ricklefs, R. E., D. Cochran, and E. R. Pianka. 1981. A morphological analysis of the structure of
communities of lizards in desert habitats. Ecology 62:1474-1483.
Ricklefs, R. E., and K. O'Rourke. 1975. Aspect diversity in moths: a temperate-tropical
comparison. Evolution 29:313-324.
Ricklefs, R. E., and J. Travis. 1980. A morphological approach to the study of avian community
organization. Auk 97:321-338.
Sale, P. 1977. Maintenance of high diversity in coral reef fish communities. Am. Nat. 111:337-359.
Salt, G. W. (ed.). 1984. Ecology and Evolutionary Biology: a Round Table on Research. Univ. of
Chicago Press, IL.
Schoener, T. W. 1974. Resource partitioning in ecological communities. Science 185:27-39.
Schlosser, I. J. 1982. Fish community structure and function along two habitat gradients in a
headwater stream. Ecol. Monogr. 52:395-414.
Sinclair, D. F. 1985. On tests of spatial randomness using mean nearest neighbor distance. Ecology
66:1084-1085.
Sneath, P. H. A., and R. R. Sokal. 1973. Numerical Taxonomy. W. H. Freeman, San Francisco, CA.
Sokal, R. R., and F. J. Rohlf. 1981. Biometry, !nd ed. Freeman, San Francisco, CA.
Strauss, R. E., and F. L. Bookstein. 1982. The truss: body form reconstructions in morphometries.
Syst. Zool. 31:113-135.
Strong, D. R. 1983. Natural variability and the manifold mechanisms of ecological communities.
Am. Nat. 122:636-660.
Strong, D. R., D. Simberloff, L. G. Abele, and A. B. Thistle (eds.). 1984. Ecological Communities:
Conceptual Issues and the Evidence. Princeton Univ. Press, Princeton, NJ.
Wiens, J. A. 1980. Patterns of morphology and ecology in grassland and shrubsteppe bird
populations. Ecol. Monogr. 50:287-308.
Wiens, J. A., and J. R. Rotenberry. 1980. Bird community structure in cold shrub deserts:
competition Of chaos? Acta XVII Congr. Internat. Ornithol. (Berlin): 1063-1070.
Wright, S. 1954. The interpretation of multivariate systems. In: O. Kempthorne, T. A. Bancroft,
J. W. Gowen, and J. L. Lush (eds.), Statistics and Mathematics in Biology, pp. 11-33. Iowa
State Univ. Press, Ames; IA.