<br />excluding direct measurement of the nuchal hump,
<br />is shown in Figure 2. Component I reflects gener-
<br />al size; II is especially ~orrelated with numbers
<br />of vertebrae, dorsal rays, anal rays, and gill
<br />rakers. Figure 3 shows the plot of components II
<br />and III of the same analysis. Component III is
<br />uncorrelated with II, but is highly correlated
<br />with counts of fin rays and gill rakers as well as
<br />pectoral length and eye diameter. In the combi-
<br />nation of the three axes, which represent summa-
<br />ries of the three major trends in the original 34
<br />characters, robusta, elegans, and cypha are dis-
<br />criminated with almost no overlap.
<br />
<br />Because the status of cypha has been most con-
<br />troversial, it was analyzed separately with
<br />robusta and elegans. The discrimination of
<br />robusta and cypha, when these are considered with-
<br />out elegans (Figure 4), and of elegans and cypha
<br />alone (Figure 4), shows complete separation for
<br />the latter and only a one-specimen overlap between
<br />robusta and cypha. This fish (UMMZ 181281, spec.
<br />2) falls into the robusta cluster because it has
<br />9 anal rays (a character of robusta weighted
<br />strongly by this multivariate analysis) and sever-
<br />al other traits that lean toward robusta; it is
<br />noteworthy that this is the only specimen that
<br />did not fall into its own cluster and, when dis-
<br />criminated with key characteristics such as
<br />nuchal-hump development (which yielded a 9.2
<br />ratio--see Table 1), it is typical of cypha.
<br />Moreover, when treated with the other two species
<br />together, (Figures 2-3) it also was identified
<br />with cypha. We cannot be sure, however, that the
<br />several traits aligning it closely with robusta
<br />may not indicate that some robusta genes were
<br />present in this specimen.
<br />
<br />Thus, although some of the individuals falling
<br />at or within the borders of clusters (Figures 2-
<br />5) could possibly be interpreted as hybrids, the
<br />clusters are quite distinct, indicating genetic
<br />differentiation and strong reproductive isolation
<br />among the three populations in the big-river hab-
<br />itat. The differentiation and isolation are
<br />probably facilitated by different ecological roles
<br />and habitat preferences within the complex big-
<br />river habitat, although insufficient data are
<br />available to be certain at this time.
<br />
<br />\ The characters that contribute most to the
<br />ove clustering are dorsal- and anal-ray number,
<br />'ll-raker and vertebral-number, and depth and
<br />length of the caudal peduncle. Gila robusta and
<br />cypha separated primarily on counts of fin rays,
<br />vertebrae, lateral-line scales, gill-rakers, and
<br />post-anal length (II). Gila cypha and elegans
<br />separated on correlation patterns dominated by
<br />the same characters, excluding pelvic rays and
<br />post-anal length, but including snout, eye, and
<br />caudal peduncle dimensions. These characters
<br />can be used as key characters, but it is interest-
<br />ing that when used alone they do not provide as
<br />complete discrimination as do all 34 traits.
<br />
<br />Figures 2 and 3 show the position of eleven
<br />specimens of Gila robusta from the Virgin River,
<br />plotted with the clusters of robusta, elegans,
<br />and cypha to demonstrate the nature of overlap
<br />of peripheral populations. The Virgin River pop-
<br />ulation is clearly a single variable population
<br />with individuals that span some of the variation
<br />shown by robusta, elegans, and cypha. None of
<br />the Virgin River specimens have distinct humps,
<br />yet their body proportions may be similar to
<br />those of members of the other populations in
<br />analyses excluding the hump character. This is
<br />interpreted as indicating that environmental con-
<br />ditions in the medium-sized tributaries are
<br />selecting for some kind of variable average of
<br />the three morphotypes present in the large-river
<br />habitats, but that the range of heterogeneity of
<br />habitat in the medium-sized river is not
<br />
<br />sufficient to support three separate species.
<br />The populations in the medium-sized rivers, for
<br />example, the Virgin and the San Juan, are inter-
<br />preted as adapted to their local environments,
<br />and not as intergrades (in the introgressed
<br />sense), though limited introgression over the past
<br />tens of thousands of years cannot be ruled out.
<br />When gill-raker number is considered (Table 2)"
<br />the Virgin River chub separates well from popula-
<br />tions of Gila robusta inhabiting the main river.
<br />
<br />The results of the above analyses suggest that
<br />three populations in the main Colorado River are
<br />morphologically segregated and are behaving as
<br />reproductively isolated species. Exceptional
<br />circumstances exist in peripheral tributaries,
<br />for example, the Virgin River, as mentioned above,
<br />and possibly in disturbed habitats, such as arti-
<br />ficial Lake powell. These will be mentioned
<br />again in the discussion.
<br />
<br />The search for key characters to discriminate
<br />the above populations involving univariate analy-
<br />sis of a larger sample of 261 individuals. The
<br />development of the nuchal hump, as expressed by
<br />means of a special ratio (see below), was analyzed
<br />in 72 individuals.
<br />
<br />One of the impediments to key discrimination
<br />of these fishes is the variability and lack of
<br />complete discriminating power of the nuchal-hump.
<br />and caudal-peduncle characters. These traits are
<br />obvious and striking and one would like to be
<br />able to use them to discriminate the populations,
<br />but difficulties in quantification of the nuchal-
<br />hump characters have heretofore prevented its
<br />effective use. Previous multivariate analyses
<br />have included subjective scores for nuchal-hump
<br />development, raising the question of subjective
<br />influence on the final results,
<br />
<br />We have circumvented this dilemma by adapting
<br />an instrument described by Eschmeyer and poss
<br />(1977). It provides a direct, repeatable measure-
<br />ment, accurate to 0.1 mm, of the development of
<br />the nuchal 'hump in association with the depressed
<br />(often concave) dorsal surface of the skull,
<br />features most conspicuous in Gila cypha (Figure 6).
<br />A ratio derived by measuring the depth of the
<br />frontal depression (the maximum distance between
<br />a straight line from highest part of nuchal hump
<br />and dorsal tip of snout, and dorsal surface of
<br />skull) and dividing this figure into the distance
<br />between the insertion of the pectoral and pelvic
<br />fins, provides an effective means for distinguish-
<br />ing adults of the three chubs of the middle and
<br />upper Colorado River basin (Table 1).
<br />
<br />Number of precaudal vertebrae (Table 3) proves
<br />to be useful, along with gill-raker number
<br />(Table 2), in discriminating Gila robusta seminuda
<br />Cope (Figure 7), from the typical subspecies,
<br />G. r. robusta (Figure 8). Counts of these verte-
<br />brae provide a better separation than the total
<br />number which (excluding Weberian vertebrae)
<br />varies from' 40 to 45, modally 42, in seminuda,
<br />and 39 to 44, modally 42, in robusta. Gila r.
<br />seminuda is closest to G. elegans (Figure 9) in
<br />number of precaudal vertebrae, as well as in gill-
<br />raker number, but separates well from that species
<br />on the basis of nuchal hump development (Table 1)
<br />although more and (especially larger) individuals
<br />need to be examined to verify this. Modally,
<br />G. robusta has nine dorsal and anal rays; G. cypha
<br />usually has nine dorsal and 10 anal rays; and
<br />G. elegans has 10 dorsal and 10 or 11 anal rays.
<br />G. r. seminuda tends to be intermediate.
<br />
<br />DISCUSSION
<br />
<br />621
<br />
<br />Our analyses indicate that Gila robusta, cyph~,
<br />and elegans coexist as three separate, reproduc-
<br />tively isolated species in the main channels of
<br />
|