McELROY AND DOUGLAS-MORPHOLOGICAL VARIATION IN GILA 645
<br />allozyme, and mtDNA data sets (DeMarais et
<br />al., 1992). Similarly, a discordance between well-
<br />resolved allozyme and mtDNA phylogenies, and
<br />sharing of derived molecular characters across
<br />lineages have been used to argue that "Colo-
<br />rado River Gila represent a complex of self-
<br />maintaining, genetically-distinctive species that
<br />are capable of exchanging genetic material"
<br />(Dowling and DeMarais, 1993). Our morpho-
<br />logical data are consistent with this hypothesis.
<br />If introgressive hybridization underlies the ob-
<br />served locality effect, we would expect genetic
<br />studies of populations of G. robusta and G. cypha
<br />to reveal incongruous patterns of allozyme and
<br />mtDNA variation (particularly in Desolation and
<br />Cataract canyons) similar to those seen at higher
<br />taxonomic levels (Dowling and DeMarais, 1993).
<br />Research to evaluate this hypothesis is in prog-
<br />ress (T. Dowling, pers. comm.).
<br />Introgressive hybridization may also be a con-
<br />sequence of human impact. Holden and Stal-
<br />naker (1970) attributed the unique morphology
<br />of G. cypha in Cataract Canyon to extensive hy-
<br />bridization resulting from habitat changes
<br />brought about by construction of Glen Canyon
<br />Dam in 1963. Although the potential effects of
<br />such water management projects cannot and
<br />should not be discounted, it appears that hy-
<br />bridization among some species of Gila predate
<br />known human modifications of the upper Col-
<br />orado basin (Dowling and DeMarais, 1993).
<br />Thus, hybridization in Gila is to some degree a
<br />natural phenomenon, and comparison of extant
<br />populations with museum collections made pri-
<br />or to known habitat alteration may help quan-
<br />tify the degree of human impact in Cataract
<br />Canyon and elsewhere.
<br />Members of the genus Gila have long been
<br />assumed to show adaptations to local habitat
<br />conditions, both within (Smith et al., 1979) and
<br />among (Miller, 1946; Holden and Stalnaker,
<br />1970; Rinne, 1976) species. Miller (1946) sug-
<br />gested that a number of putative taxa of the G.
<br />robusta group in fact constituted ecological sub-
<br />species. Subsequent studies failed to support this
<br />hypothesis (Holden and Stalnaker, 1970; Rinne,
<br />1976; Smith et al., 1979). Nevertheless, the de-
<br />gree of spatial heterogeneity may influence both
<br />morphological and species diversity of Gila at a
<br />given locality (Smith et al., 1979; Grant and
<br />Grant, 1989). Desolation and Cataract canyons
<br />are both extremely energetic habitats (R. Val-
<br />dez, pers. comm.), characterized by rapid and
<br />turbulent flow regimes. Such conditions may
<br />impose strong selection for a common mor-
<br />phology in both species, whereas less severe
<br />habitats may allow several distinct morpholo-
<br />gies to persist. Even so, our data indicate that
<br />fish of both species from Desolation and Cata-
<br />ract canyons are relatively more robusta-like in
<br />morphology; this would not be expected if the
<br />morphology of G. cypha indeed reflects adap-
<br />tation to high current regimes (Miller, 1946;
<br />Minckley, 1973; Kaeding et al., 1990) and con-
<br />ditions at Desolation and Cataract canyons fa-
<br />vor such adaptations. A local adaptation sce-
<br />nario for the locality effect observed here would
<br />be supported if genetic studies provided no ev-
<br />idence of introgression between G. robusta and
<br />G. cypha at the population level; however, evi-
<br />dence for hybridization would not in itself pre-
<br />clude the concomitant existence of local adap-
<br />tation.
<br />Symplesiomorphy seems an unlikely expla-
<br />nation for the similarity of G. robusta and G.
<br />cypha at Desolation and Cataract canyons. If
<br />these areas support relatively ancestral mor-
<br />photypes, we would expect that, in addition to
<br />similarity of heterospecifics within localities, fish
<br />from the two localities would be similar to one
<br />another. This is not the case; although both
<br />Cataract and Desolation canyons contain robus-
<br />ta-like forms, the node separating conspecifics
<br />from these localities is deep. This implies that
<br />the two sites support morphologically distinct
<br />forms of both species. Nevertheless, rejecting
<br />this hypothesis with any confidence is problem-
<br />atic given the paucity of cladistically informa-
<br />tive characters capable of differentiating species
<br />of Gila (Suttkus and Clemmer, 1977; Smith et
<br />al., 1979).
<br />Implications for conservation.-The existence of
<br />significant population divergence in both G. ro-
<br />busta and G. cypha suggests that all populations
<br />represent (to some degree) independent evo-
<br />lutionary, and thus conservation, units. Al-
<br />though this is most apparent with respect to
<br />Desolation and Cataract canyon populations of
<br />both species, all populations are distinct.
<br />Whether this divergence is a consequence of
<br />local adaptation, hybridization, or genetic drift
<br />(or any combination) cannot be ascertained from
<br />our data. The lack of a consistent geographic
<br />component to relationships among populations,
<br />however, suggests that an isolation by distance
<br />scenario is insufficient to explain the results.
<br />Based on these findings, it is clear that caution
<br />and biological foresight must be employed when
<br />making management decisions.
<br />We also emphasize that the existence of hy-
<br />bridization, if confirmed by genetic studies, does
<br />not imply that introgressed populations are less
<br />valuable from a conservation standpoint, for two
<br />reasons. First, it appears that hybridization in
<br />Gila represents to some degree an evolutionary
|