<br />VICARIANCE AND WESTERN NORTH AMERICAN FISHES
<br />
<br />243
<br />
<br />TABLE 1. Matrix comparisons between 10 taxonomic-distance matrices (each matrix represents a different characteristic measured or
<br />estimated for the 33 study sites in the Gila River Basin). H, a phenotypic-distance matrix based on sheared principal component scores
<br />for fish samples; SIZ, a matrix of stream size (discharge) at each site. ELE, sample-site elevations above mean sea level; BAR, sample-
<br />site isolation. aLl, EMI, MMI, PLI, PLE, and REC, respectively, paleodrainage matrices for: Oligocene, Early Miocene, mid-Miocene,
<br />Pliocene, Pleistocene, and Recent. The upper triangle of the table presents values for two-way Mantel tests and the lower triangle contains
<br />matrix correlations. Significance (3.06, P < 0.001) is represented by an asterisk.
<br />
<br />Matrix H BAR SIZ ELE OLl EMI MMI PLI PLE REC
<br />H -1.27 -0.06 0.55 0.33 -0.08 3.69* 4.36* -0.18 0.84
<br />BAR -0.10 4.01* -1.28 2.63 2.37 2.41 -0.22 3.47* 0.16
<br />SIZ -0.01 0.30 0.07 1.97 1.50 4.46* 1.73 1.39 2.01
<br />ELE 0.05 -0.10 0.01 0.10 0.10 -0.82 -0.37 0.63 0.32
<br />aLl 0.02 0.17 0.14 0.01 15.74* 13.31 * 7.45* 10.86* 3.96*
<br />EMI -0.01 0.14 0.10 0.01 0.87 15.45* 6.98* 10.50* 5.90*
<br />MMI 0.15 0.10 0.19 0.03 0.56 0.66 14.89* 9.57* 16.15*
<br />PLI 0.29 -0.01 0.12 -0.03 0.43 0.37 0.63 4.42* 7.91*
<br />PLE -0.02 0.26 0.12 0.05 0.76 0.66 0.40 0.27 1.30
<br />REC 0.06 0.01 0.14 0.02 0.23 0.31 0.68 0.44 0.09
<br />
<br />rived from paleohydrographic maps (i.e., OLG, EMI, MMI,
<br />PLI, PLE, REC).
<br />Models I and II offered little explanatory power (Table 1).
<br />Instead, historical distributions of physiographic features (re-
<br />lated to Model III), particularly those of mid-Miocene and
<br />Pliocene, appear paramount in explaining the existing mosaic
<br />of variation. A three-way Mantel analysis was performed to
<br />unravel interrelations between mid-Miocene and Pliocene hy-
<br />drography and phenotype. The resulting correlations and par-
<br />tial correlations (Table 2) indicate that although fish body
<br />shape has both mid-Miocene and Pliocene components (e.g.,
<br />both X-matrices are useful predictors of the Y-matrix), shape
<br />is more highly correlated with Pliocene (rY 2 = 0.29, P <
<br />0.0005) than with mid-Miocene hydrography (rY 1 = 0.15, P
<br />< 0.004). The partial correlation of shape and Pliocene hy-
<br />drography is significant (rY 2 = 0.26, P < 0.008), whereas
<br />that of shape and mid-Miocene hydrography is not (rY2 =
<br />-0.04, P > 0.76). The partial correlation analysis (Table 2)
<br />can be more succinctly understood through an evaluation of
<br />coefficients of multiple determination (R2). Once having fit
<br />Pliocene hydrography to the model (top, Table 2). there was
<br />virtually nothing gained by adding mid-Miocene, because R2
<br />only increased from 0.085 to 0.087. However, once having
<br />fit mid-Miocene hydrography to the model (R2 = 0.023),
<br />considerable information was derivable by adding the Plio-
<br />cene component (R2 = 0.087).
<br />Other significant correlations in Table 1 are intuitively ex-
<br />
<br />TABLE 2. Results of three-way Mantel tests between body shape,
<br />mid-Miocene, and Pliocene distance matrices compiled for 33 study
<br />sites in the Gila River Basin; formulations per Smouse et al. (1986).
<br />
<br /> Body shape = Mid-Miocene occurrence,
<br /> Pliocene occurrence
<br />Model (Y) (XI) (X2)
<br />Comparison Formulation Correlation Probability
<br />Shape, mid-Miocene rYI 0.154 < 0.004
<br />Shape, Pliocene r Y2 0.293 < 0.0005
<br />Mid-Miocene, Pliocene r YI.Z 0.636 < 0.00001
<br />Shape, mid-Miocene
<br />(Pliocene) r YI.2 -0.044 > 0.76
<br />Shape, Pliocene
<br />(mid-Miocene) r YZ,1 0.255 < 0.0008
<br />
<br />plicable and do not detract from these interpretations. Those
<br />between epochs demonstrate regional continuity of major
<br />structural development through later Tertiary apparent in Fig-
<br />ure 2. That between SIZ and MMI may relate to MMI basins
<br />being dominantly large geographic features that, for that rea-
<br />son alone, are associated with watersheds of higher water
<br />yield. The same explanation likely exists for the significant
<br />relationship between BAR and SIZ because a greater prob-
<br />ability exists that larger (higher-discharge) streams will cross
<br />geologic formations of differing erodability, thus promoting
<br />barrier (e.g., waterfall) formation. That between BAR and
<br />PLI, more speculatively, may result from younger basin de-
<br />posits still experiencing the process of rapid erosion. Places
<br />where downcutting channels encounter parent material at the
<br />margin of an incising basin fill could result in barriers, either
<br />due to differential durability of substrate (thus creating rapids
<br />or waterfalls, as above) or through water flowing over im-
<br />pervious parent rock passing onto and percolating into porous
<br />alluvium to create dry reaches.
<br />
<br />CONCLUSIONS
<br />
<br />It is not surprising that vicariant events are correlated with
<br />patterns of morphological variation in minnows of the genus
<br />Gila. Vicariance has been implicated as an important factor
<br />in generating orderly patterns of diversity and variability in
<br />a variety of taxa (Signor 1990, p. 529), and hypotheses of a
<br />vicariant nature have already been generated for both aquatic
<br />and terrestrial components of the western American aquatic
<br />and semiaquatic fauna (Hendrickson 1986; Minckley et al.
<br />1986).
<br />Our study is nonetheless novel in several ways. First, we
<br />are able to statistically assess existing phenotypic diversity
<br />in relation to patterns of ancient hydrography controlled by
<br />tectonism. To our knowledge, this has not before been done.
<br />Second, the resolution of our analysis is noteworthy. Riddle
<br />(1995) performed a relatively fine-grained analysis by using
<br />molecular data to demonstrate arid-land rodent divergence
<br />and distribution were attributable to Late Miocene devel-
<br />opment of western North American Cordillera. His study
<br />would be more appropriately described as "regional," where-
<br />as ours would be "local." Third, our results argue strongly
<br />for the importance of deep rather than shallow history in the
<br />
|