644
<br />COPEIA, 1995, NO. 3
<br />have supported the view that G. robusta and G.
<br />cypha represent distinct species, though "inter-
<br />grades" have been detected in varying frequen-
<br />cies. Our results similarly argue for treating G.
<br />robusta and G. cypha as separate species; despite
<br />the presence of intermediates, the two forms
<br />were clearly distinct both in sympatry and in
<br />allopatry. In fact, the intermediate specimens
<br />in our sample could be assigned to species with
<br />confidence, suggesting that morphological dis-
<br />tinctiveness of G. robusta and G. cypha is suffi-
<br />cient to categorize individuals despite high lev-
<br />els of variability. This would not necessarily be
<br />expected if the bases for a priori assignment of
<br />intermediates were more random than real
<br />(McElroy and Kornfield, 1993). Further, this
<br />variability may not to be a ubiquitous feature
<br />of populations of Gila (Smith et al., 1979; Doug-
<br />las et al., 1989; Kaeding et al., 1990): the ma-
<br />jority of intermediates in our data set were col-
<br />lected from Cataract and Desolation canyons,
<br />and these populations may be subject to evo-
<br />lutionary forces not operating, or less impor-
<br />tant, at other localities (see below).
<br />Only one study to date has addressed the issue
<br />of intraspecific population divergence in upper
<br />basin Gila. Holden and Stalnaker (1970) argued
<br />that populations of both G. robusta and G. cypha
<br />were homogeneous morphologically. However,
<br />their analysis relied on ordinal data and, thus,
<br />provided limited resolving power (Smith et al.,
<br />1979; Douglas et al., 1989). By contrast, we
<br />found subtle but significant differences in mor-
<br />phology among all populations in both species.
<br />This structure is overlain on patterns of varia-
<br />tion at the species level, as a posteriori assign-
<br />ments of all individuals to species agreed with
<br />a priori classications despite the existence of
<br />among-population differences. Our results also
<br />indicate that an isolation by distance model of
<br />gene flow among populations cannot account
<br />for these differences (Fig. 4): the lack of matrix
<br />correlation between canonical and river mile
<br />distances among populations of both species
<br />suggests that local processes must contribute to
<br />distinctiveness of groups. The nature of these
<br />additional biological forces (e.g., hybridization,
<br />selection, gene flow) remains to be elucidated.
<br />The most intriguing aspect of our data con-
<br />cerns the relationships of populations at the ge-
<br />neric level. In particular, although G. robusta
<br />and G. cypha are in general distinct morpholog-
<br />ically both in sympatry and allopatry, the two
<br />species converge morphologically at Cataract
<br />and Desolation canyons. At these sites, a locality
<br />effect appears to contribute significantly to re-
<br />lationships among populations; by contrast, a
<br />species effect appears to dominate relationships
<br />among other populations (Fig. 7). The Cataract
<br />and Desolation population pairs are clearly dis-
<br />tinct from all other groups (and from one an-
<br />other) in all analyses. This dichotomy between
<br />species and locality effects suggests that popu-
<br />lation dynamics of G. robusta and G. cypha in
<br />Cataract and Desolation canyons may differ
<br />qualitatively from those of other populations of
<br />Gila in the upper basin. The structure of these
<br />populations may be driven by biological forces
<br />(see below) whose effects are less pronounced
<br />elsewhere. The uniqueness of the G. cypha pop-
<br />ulation at Cataract Canyon has been noted pre-
<br />viously: specimens from this area tend to be
<br />smaller, deeper bodied, and less extreme mor-
<br />phologically than are G. cypha from other lo-
<br />calities (R. Valdez, Bio/West, Inc., pers. comm.).
<br />It has been suggested that these fish represent
<br />remnants of a much larger population that in-
<br />habited the region prior to closure of Glen Can-
<br />yon Dam (Holden and Stalnaker, 1975). Time
<br />series data that might provide insight into this
<br />hypothesis are currently being analyzed to ex-
<br />amine changes in morphology over the past cen-
<br />tury U. Lynch, pers. comm.).
<br />Potential bases for the locality effect. -We postulate
<br />three explanations for the morphological con-
<br />vergence of heterospecific populations at Des-
<br />olation and Cataract canyons. First, the simi-
<br />larity of G. robusta and G. cypha at these localities
<br />may reflect extensive introgressive hybridiza-
<br />tion, either occurring naturally or as a result of
<br />human intervention. Second, sympatric popu-
<br />lations may be subject to similar selection pres-
<br />sures within each of these localities, i.e., the
<br />observed locality effect may reflect parallel local
<br />adaptation of the two species in Desolation and
<br />Cataract canyons. Third, populations of both
<br />species at these sites may retain a high propor-
<br />tion of ancestral traits, and their apparent sim-
<br />ilarity may be a consequence of shared primitive
<br />characteristics. These explanations are not mu-
<br />tually exclusive. It is unlikely, however, that our
<br />results reflect an inability to consistently char-
<br />acterize species, because this would lead to poor
<br />classification of individual specimens. Although
<br />complementary genetic studies might be useful
<br />in evaluating the relative contribution of each
<br />alternative, the likelihood of each based on ex-
<br />isting data is considered below.
<br />Hybridization appears to have played a po-
<br />tentially significant role in the evolutionary dy-
<br />namics of the G. robusta group (DeMarais et al.,
<br />1992; Dowling and DeMarais, 1993). A hybrid
<br />origin for G. seminuda, long suspected (Ellis,
<br />1914; Miller, 1946; Smith et al., 1979), has been
<br />substantiated by comparison of morphological,
|