8073

• UCREFRP Catalog Number: 8073
• Author: Gatz, J.
• Title: Community Organization in Fishes as Indicated by Morphological Features
• USFW Year: 1979
• USFW - Doc Type: Ecology
• Copyright Material: YES

growth. The use of a pooled sample of fishes from all
three streams for species occurring in all streams
might also be regarded at best as a simplification, and
at worst as a method resulting in the loss of any in-
tere$ting ecological information, in that any local ad-
api?.ions would be obscured. I investigated this pos-
sibility of intraspecific variation in morphology and
found such variation to exist between populations
within different streams in only one feature, relative
body depth (measured as the ratio of maximum body
depth to standard length). For three different species
of sunfishes, the sample, measured from the faster-
flowing of the two streams compared, had a statisti-
cally significant lower mean value than the sample
from the slower stream. Because this same effect was
seen in all three species, and a high relative body depth
normally is taken to indicate a slow-water habitat pref-
erence (Nikolskii 1933, Aleev 1969), I assumed that
the observed differences probably involved adaptation
to local habitat conditions. Alternative explanations
based on phenomena such as character displacement
or the founder effect cannot be ruled out, but seem
less likely to me, especially considering the absence
of additional cases of intraspecific variation.
Morphological features studied included measures
of body shape and proportions, mouth size and posi-
tion, fin shape and location, and internal features such
as swim bladder volume, gut length, and red muscle
content. As a control for size effects, all length, sur-
face area and volume measurements were divided by
standard body length, surface area or volume, as ap-
propriate, and thereby converted to relative indices;
A multivariate analysis of the resultant set of data did
not identify size as a principal component. Altogether.
I studied 56 features on each species. and each feature
was selected for study based on some inferred ecolog-
ical significance. (See the Appendix for a list of these
features and Gatz [1979J for a complete description
and extensive interpretation.)

CommunilY slruclure

The information from the morphological analysis
was then used in two complementary ways in the anal-
ysis of niche relationships. First, the variance of the
data was employed, by using Newman-Kuels tests, to
identify boundaries of statistical significance. In this
case, I assumed statistical significance to be equivalent
to ecological significance. In other words, any mor-
phological feature shown to have statistically signifi-
cantly different mean values in two species was as-
sumed to indicate also that different optimal patterns
of resource utilization exist between the species, i.e..
that niche differentiation exists. Although this proce-
dure may, by virtue of type II statistical errors, have
resulted in missing some differentiating features be-
tween species, it seemed the most reasonable ap-
proach. To deny that significantly different amounts
of red muscle, for example, equip the species involved

for different habits and habitats is to deny that natural
selection has any significance.
The second way the data were utilized entailed the
use of species mean values and to a large degree ig-
nored variability. This approach supposes that a sub-
stantial separation could result between species if a
large number of features differed slightly, even though
none differed in a statistically significant manner.
The two different forms of analysis of overall com-
munity organization permitted by these two modes of
assessment follow.
In the first case, I borrowed the viewpoint of a nu-
merical taxonomist. I assumed that the wide variety
of features which I measured was in some sense a
random sample of all possible morphological features
related to the niche of the fishes. In other words. I
assumed that I had measured enough features related
to enough different aspects of each species' biology
so that most major aspects of the niche had been taken
into consideration. Therefore, the number of morpho-
logical features not significantly different between two
species when expressed as a percentage of the total
number of features measured was taken as an index
of percentage niche overlap of the two species. Com-
munity structure was, in this case, analyzed in terms
of percentage overlap between coexisting intrafamilial
species pairs.
In the second case, the raw morphological data were
subjected to further treatment before a measure of
niche spacing was undertaken. First, in order to
achieve a morphologically defined niche for each
species based on independent axes, a factor analysis
with orthogonal rotation wa~ performed. Nine factors
with eigenvalues greater than one were found which
together accounted for 79% of the total variance in the
data. Each such factor identifies one set of morpho-
logical features which show significant multivariate
trends in covariation with each other, but are inde-
pendent of any other sets of features. Thus, these in-
dependent dimensions are ideally suited for defining
a position for each species in N-dimensional (in this
case N = 9) space. The average factor scores (to three
digits) on each dimension for every species were used
to assign positions for all species in niche space as
defined by this morphological space. In other words,
the center of the niche of each species is taken to be
that point in this nine-dimensional space given by the
series of nine average factor scores.
Next, in order to determine the pattern of spacing
of these morphologically defined niches for coexisting
species, I calculated Euclidean distances according to
the equation:

TABLE I. Percentage overlap in morphology for all species within families in Easl Prong Lillie Yadkin (mean =

CYPRINIDAE
I 2 3 4 5 6
1. Clinostomus funduloides 100 62 70 79 75 55
2. Hybop,.;. hYP,';nolus 100 64 57 64 52
3. No('omis ll~plocephalus 100 73 84 62
4. Notropis analos/anus 100 73 55
5. Nutrop;s chi/ilicus 100 68
6. Phoxinus areas 100
7. SemOlilus alromacularus
ICT ALURIDAE
I 2
I. JC'laJurus nebu/osus 100 50
2. ^,otufUS ;nsign;s 100
CENTRARCHIDAE
1 2 3
I. Lepomis aurilUS 100 77 66
2. Lepomis cJanel/us 100 70
3. L('pomis macrochirus 100
PERCIDAE
1 2
I. Erheosloma flabellare 100 52
2. ElheoSfoma olmsted; 100

factor k: and N = nine dimensions. The standardized
means are calculated according to the equation:
x',.. = (XI.. - X.l/SO.

where Xi.. = measured mean value for species i in fac-
tor k: x. = mean value for all x,.. for factor k, and
SO. = standard deviation of factor k. The use of this
formula has the effect of making the set of all species
means for each factor fit a standardized normal curve,
i e a normal curve with a mean of zero and a standard
de~'iation of one. The resultant array of Euclidean dis-
tances between coexisting species was taken as a sec-
ond measure of community structure.
The results of these analyses were analyzed in more
than one way. First, the results of both the percentage
overlap studies and Euclidean distance studies in each
of the three streams were compared with each other
for the emergence of a repeating pattern. Second, the
distribution of Euclidean distances between species
was compared to a distribution of distances which
would obtain ifall resources were used randomly. This
stochastic model which permitted direct testing for the
existence of "nonrandom assemblages of interacting
species" was developed as follows. First, a series of
"random niches" was generated using a random num-
bers table to assign a three-digit position for each hy-
pothetical species on each of nine independent dimen-
sions. Each of the nine random numbers assigned to
every hypothetical species was analogous to a mean
factor score for a real species on one of the nine factor-
analysis axes. Note that because each factor of the
factor analysis identified an independent suite of mor-
phological features, there need be no necessary rela-

O,.! = ..j r (x'J.k - X'J..)2
.-,

where 0,., = Euclidean distance between species i
and speciesj; x',.. = standardized mean for species i in
factor k; x'J,. = standardized mean for species j in

tionship between the values found on one d
and those found on any other. This is signi!
cause it means that the hypothetical species E
were not composed of a randomly chosen in
combination of morphological features. Ratl
generated point in nine-dimensional space rc
the location of the center of a niche for a hn
species whose mean resource utilization in e,
niche dimensions has been randomly chosen.
sets of random niches were generated and 1
acters were standardized as above for re;
scores. Finally, the distribution of Euclidean I
between all pairs of hypothetical species w;
mined for each of the 10 sets of random niche
these resulting distance distributions which w
pared with the distance distributions of reai
using Kolomogorov-Smimov goodness of .
tests and a randomization test (Sokal and Rol

RESULTS

The percentage niche overlaps for all possil
familial comparisons of sympatric species are
Tables 1-3. An important point to notice is t
the ranges of percentage overlaps and their l
(x = 64-66%) are nearly the same in all three
irrespective of the number of species present
fail to identify any significant differences
streams. The direct implication of this resul
the overall range of morphologies seen is a
function of the number of species present. E
ence, then. this broader morphological range
that more resources are being utilized and mo