8238 - Laserfiche WebLink

Home Browse Search

(7fq ,., /i. r, /'v1.tll('/iltf"C.,! L f I ? (:. ;ju:lj/ t( 7 ) V! [ Z r) e - a f Evolution, 53(1), 1999, pp. 238-246 0'-'1-.,/ D :f. '/ U DID VICARIANCE MOLD PHENOTYPES OF WESTERN NORTH AMERICAN FISHES? EVIDENCE FROM GILA RIVER CYPRINIDS MICHAEL E. DOUGLAS,! W. L. MINCKLEY, AND BRUCE D. DEMARAIS 2 Department of Biology and Museum, Arizona State University, Tempe, Arizona 85287-1501 IE-mail: m.douglas@asu.edu r Abstract.-Pairwise, two- and three-way Mantel tests were used to evaluate a null hypothesis of no significant co- variation when morphological features of three cyprinid fish taxa of the genus Gila were compared. Tests involved ecological conditions and past and present hydrography in the Gila River Basin of western North America. A vicariance hypothesis was the only model statistically proficient in explaining diversity of fish phenotypes. Of paleohydrographic reconstructions compared, those of the mid-Miocene and Pliocene epochs were significantly associated with present- day distributions of phenotypes. Of these, the Pliocene was paramount. , Key words.-Fishes, genus Gila, geologic history, hydrography, Mantel test, vicariance, western North America. Received December 30, 1997. Accepted September 10, 1998. , Patterns of variation in terrestrial organisms often reflect fragmentation of gene flow between and among formerly sympatricpopulations. These underlying causes also generate patterns of variability in freshwater fishes, particularly those in geologically stable zones with abundant groundwater, where connectedness among aquatic systems is essentially complete. In the water-poor American West, however, con- nectedness in aquatic systems is at best transitory and at worst illusory. As a result, some western American fishes, like their habitats, display little system-to-system congruence in pat- terns of morphology. Species' geographic ranges cut across modern watershed divides and distinct forms may exhibit baffling variational mosaics with few apparent geographic components (Hubbs and Miller 1948; Hubbs et al. 1974). Historically, such patterns were dealt with by arguing for local selection, genetic drift, or other evolutionary phenom- ena known to occur within stocks long isolated from one another (Hedrick 1986). Indeed, even today these concepts often provide safe hooks from which to hang discordant data. Western American fishes exhibiting such variation were first "split" as a multitude of taxa (Jordan et al. 1930) and more recently "lumped" with other distinct forms, regarded as ecologic or geographic variants within polytypic species (e.g., Smith 1966; for a similar perspective regarding anurans see Hillis 1988). To us, neither approach is intuitively sat- isfactory or realistic. We posit instead an approach that sorts through the jumble of discordant data and evaluates multiple rather than indi- vidual hypotheses of differentiation. This is because single causation and its accompanying tests can be misleading, par- ticularly when phenotypic traits are involved (Sokal 1979; Douglas and Endler 1982). Alternative models need to be invoked simultaneously; none should be discarded a priori. Similarly, when testing multiple hypotheses, care must be taken to evaluate independence of patterns. Derivation of simple matrices to represent complex situations may generate patterns highly correlated among themselves (Thorpe and 2 Present address: Oklahoma Biological Survey, University of Oklahoma, Norman, Oklahoma 73019. @ 1999 The Society for the Study of Evolution. All rights reserved. Brown 1989). In a sense, all variance is accommodated, but none is explained. A traditional method of approaching seemingly intractable evolutionary questions is to reconstruct the process using theoretical models. This procedure can be accomplished read- ily enough, but practical representation of models, either mathematically or in matrix format, becomes difficult (Smou- se and Long 1992, p. 193). Yet, ifreasonable limitations can be accepted, appropriate tests can be and have been per- formed. These frequently employ statistical techniques of a geographic nature, which evaluate partial correlations and estimates of congruence between matrices (Smouse et al. 1986; Douglas and Matthews 1992; Smouse and Long 1992). We use this approach to derive, evaluate, then test multiple hypotheses that explain long-confusing patterns of morpho- logical variation in a group of cyprinid fishes endemic to western North America. Specifically, we deal with those taxa often referred to a "Gila robusta complex" (e.g., Robins et al. 1980) in the Gila River Basin, Arizona and New Mexico and Sonora, Mexico, excluding Gila elegans Baird and Gi- rard, which was formerly present but is now extirpated in the basin and not involved in the present situation. Included are Gila robusta Baird and Girard and two nominal taxa re- ferred by Minckley (unpubl.) to G. intermedia (Girard) and G. nigra Cope (the last formerly called Gila grahami Baird and Girard, a synonym of robusta), with nigra including some confusingly intermediate populations (DeMarais 1986, 1992) that cast doubt on its distinctiveness. For the present study, our approach was to delineate the geographic ranges of each phenotype (Fig. 1), then statisti- cally examine the resulting mosaic of variation against three separate hypotheses: (1) ecophenotypy or ecotypy (individual or populational response to local conditions); (2) hybridiza- tion or intergradation (gene flow between differentiated spe- cies or races); or (3) vicariance (a historical explanation re- lated to regional hydrographic evolution). One or more ma- trices representing these hypotheses were tested against a matrix of body-shape variation using two- and three-way Mantel tests. 238