8238

• UCREFRP Catalog Number: 8238
• Author: Douglas, M. E., W. L. Minckley and B. D. DeMarais
• Title: Did Vicariance Mold Phenotypes of Western North American Fishes? Evidence From Gila River Cyprinids
• USFW Year: 1999
• USFW - Doc Type: Evolution
• Copyright Material: YES

240

MICHAEL E. DOUGLAS ET AL.

..

.

FIG. 2. Sketch maps of parts of Arizona-Sonora, southwestern North America, showing some place and region names used in the text
and general relationships of Tertiary basins and/or sedimentary rock units (indicated by diagonal or horizontal hatching) to present
streams of the Gila River system. (A) Oligocene Epoch; (B) Miocene; and (C) Pliocene (modified from Nations et al. 1982).

Measurements (to 0.1 mm) followed Hubbs and Lagler
(1974), except when described in parentheses; length of head,
middorsal head (occiput to snout), snout, postorbit, upper
jaw, predorsal, prepe1vic, pelvic fin, dorsal fin, anal fin, pec-
toral fin, caudal fin (hypural plate to tip of upper lobe), and
caudal peduncle; distance from pectoral insertion to pelvic
insertion, snout tip to isthmus, chin tip to isthmus, and fleshy
orbit; depth of head, body, and caudal peduncle (least); and
width of postorbital, head, fleshy interorbital, gape, and body
(greatest). Measurements are from DeMarais (1986), who
also tabulated means, listed collection data, and provided
additional information on ecology and distribution.

Data Analysis

Sheared PCA (Bookstein et al. 1985, as modified by Rohlf
and Bookstein 1987), was used to derive size-free shape
scores from 10gIO-transformed morphometric data. These
were converted to a matrix of pairwise taxonomic distances
(Sneath and SokaI1973), which was tested against the various
evolutionary models described below.

Pattern Analysis

Mantel (1967) devised a generalized, nonparametric re-
gression approach to matrix comparisons, where the sum of
cross-products of analogous cells of two matrices are com-
pared against an expected value calculated on the null hy-
pothesis of random permutations between rows/columns of
the second (the computation is actually the inner product of
each permutation within the first matrix). Because Mantel's
procedure evaluates all possible permutations of rows and
columns of the second matrix, it employs a generalized per-
mutational distribution to assess statistical significance.
Rows and columns of test matrices were permuted 1000 times
for each evaluation (Jackson and Somers 1989).
Because the test is nonparametric, distributional abnor-
malities of data collected from different sources (e.g., eco-

logical, morphological, geographical) are circumvented. In
addition, potential problems with comparing categorical ver-
sus continuous data (here, morphology vs. hypothesis ma-
trices) are negated as well. Sokal (1979) brought Mantel's
test to the attention of systematists and suggested applications
in geographic variation analysis (for reviews see Douglas and
Endler 1982; Manly 1985; Douglas and Matthews 1992;
Smouse and Long 1992).
Pairwise Mantel tests were used to evaluate a null hy-
pothesis which assumes no significant covariation when fish
morphologies are compared against ecological conditions or
hydrography. As in previous work (Douglas and Endler 1982;
Douglas and Matthews 1992), the Bonferroni technique (Har-
ris 1975, pp. 96-101) was applied to assign significance level
and establish a probability level for making a Type I error
(identifying a matrix comparison as significant when it is
not). Under the conservative Bonferroni technique, the prob-
ability level for our tests would be 0.05/n, where n is the
number of matrix comparisons.
Smouse et al. (1986) modified and extended the method-
ology so that three matrices could be contrasted simulta-
neously. This allows correlations and partial correlations to
be generated between matrices, which are then used to cal-
culate coefficients of multiple determination (Smouse et al.
1986; Smouse and Long 1992). We used three-way Mantel
comparisons to extend those two-way tests in which signif-
icant results were indicated between the morphological ma-
trix and two or more "response" matrices (as in Douglas and
Matthews 1992).

CONSTRUCTION OF ALTERNATIVE MODELS

Three scenarios were developed to explain the distribution
of morphological variation among sampling sites.

Model I: Ecophenotypy and Ecotypy

An ecophenotype represents distinctive individuals of a
species occupying contrasting habitats. Ecophenotypes pre-