7839

• UCREFRP Catalog Number: 7839
• Author: Douglas, M. E.
• Title: Analysis of Sexual Dimorphism in an Endangered Cyprinid Fish ‎(
• USFW Year: 1993
• Copyright Material: YES

DOUGLAS-SEXUAL DIMORPHISM IN GILA CYPHA

337

...

display, while a Zenith 1490 flat-screen monitor
displayed programing commands. The NEC was
harnessed to a Matrox PIP 640-B frame-grab-
ber board titted into the microcomputer for
image capture. Software to drive the system (i.e.,
to store, retrieve, and manipulate images and
to extract and save morphometric measure-
ments from these images) was produced by Bio-
Scan, Inc. (Everett, Washington) and marketed
under the name "Bio-Scan Optimas." The soft-
ware oJ\lerated within the Microsoft Windows
environment and was mouse driven.
Eighty adult G. cypha were filmed during 1988
ancl1989. Images were evaluated by first wiring
the camcorder to the frame-grabber board of
the microcomputer and then operating the
camcorder in VCR mode. When a focused im-
age wa$ displayed on the monitor, the camcor-
der /V~R was halted and the image saved onto
hard disk using the Optimas software. Each re-
corded image was thus a single "frame" selected
from the 1 O-sec series of frames recorded in the
field.

Extraction of morPhological data.-Of the 80 taped
images, 63 (29 male, 34 female) were of suitable
resolution for detailed morphological analysis.
In the first step, the 10-cm rule in the image
was digitized (in mm) to serve as a measurement
standard for that image. A truss (Strauss and
Bookstein, 1982) consisting of 53 individual
measures was then superimposed onto the im-
age (Fig. 1 A). During this process, special at-
tention was afforded the head and nuchal hump
(Fig. 1 C). Anatomical landmarks (Strauss and
Bookstein, 1982) from which to position the
truss were scant in the nuchal region. To com-
pensate. a series of computed "landmarks" were
produced. Eight of these were located on the
dorsal edge of each image from snout to origin
of the dorsal fin in the following manner. First,
three anatomical landmarks (i.e., pupil of eye,
nape of head, and origin of pectoral fin) were
identified. The horizontal coordinates of these
three landmarks (produced by the software)
were projected vertically to the dorsal edge of
the image and recorded. Then, four additional
points were positioned along the snout-to-dor-
sal origin to be equidistant between each of the
three projected points and the endpoints. For
example, a point equidistant between snout and
vertical-from-pupil was produced by subtract-
ing the x-coordinate of the former from that of
the latter and dividing by two. The distance
between the last two points in this series (i.e.,
between vertical-from-pectoral and origin-of-
dorsal) was considerably larger than the pre-
ceding distances. Because this region was

crucial for demarcating the nuchal hump. two
additional points were located between these
landmarks. effectively dividing this last segment
into quarters. Consequently, 11 total landmarks
(three bona fide and eight calculated) were po-
sitioned along this anterior dorsal margin:
Characters (9)-( 19) (Table 1: Fig. I A, 1 C) rep-
resent distances connecting each of these 11
points to the pectoral origin. Characters (20)-
(29) connected the same 11 points to one an-
other in a linear sequence from anterior (i.e.,
snout) to posterior (i.e., dorsal origin). The oth-
er 32 truss measures outlined the remainder of
the image and summarized various aspects of
head, body, and peduncle. plus fin lengths (Ta-
ble 1; Fig. lA, lC).
The resulting matrix of 63 fish by 53 dis-
tances was migrated to the ASU IBM-3090-5000
mainframe supercomputer for analysis using the
Numerical Taxonomy System of Multivariate
Statistical Programs (NT-SYS: F. J. Rohlf, J.
Kishpaugh, and D. Kirk, 1974, unpub!. tech.
rept., State University of New York, Stony
Brook), the Statistical Analysis System (SAS,
1985), and BioMedical Statistical Software
(BM DP; Dixon, 1990).

Data analysis.-AII data were first transformed
to base-l 0 logarithms, then grouped by sex and
tested for skewness and kurtosis. The variance/
covariance matrices for both sexes were then
tested for equality. Correlations were calculated
amongst characters over all individuals, regard-
less of sex. Arithmetic means and standard de-
viations were then calculated bv sex for each
character. To test each variable for sex-related
differences, an analysis of covariance (AN-
COV A) was performed, with the longest linear
distance (i.e., that between pectoral and pelvic
fins) serving as the covariate. This approach re-
moved variation in a particular measurement
correlated with overall body size. Mean mea-
surements for each sex were subsequently ad-
justed to a common body size, allowing a more
equitable comparison of measurements on the
two sexes.
Variables were then standardized on the basis
of pooled, within-sex standard deviations (Roh-
wer and Kilgore, 1973; Schnell et a!., 1985).
Standardization was accomplished by substract-
ing the value of a given character from the grand
mean of that character averaged over both sex-
es. This value was then divided by the pooled,
within-group standard deviation of the char-
acter (see Rohwer and Kilgore, 1973:160, fig.
2). This technique has advantages over other
standardization methods because more empha-
sis is given to those characters possessing a rel-