7022 - Laserfiche WebLink

Home Browse Search

828 COMA, 1991, NO. 3 V IATE ANALYSIS OF STANDA RDIZED PCA SC ORES OF MEASUREMENTS AND STANDARD- TABLE 4. CANONICAL AR s are the results of an f CVA loadin alyses with dorsal-fin rays included and IZED COUNTS. The two omitted. g sets o The same set of standardized PCA scores w as used for each analysis. Canonical vector loadings First analysis Second analysis Character 1 2 1 2 PCA axis 1 0.13175 -0.02654 0.29673 0.04392 43603 -0 PCA axis 2 0.80062 -1.02306 1.46241 23940 1 . 0.73585 PCA axis 3 0.62336 52598 0 -1.27442 -0.39820 . 0.92915 -0.97829 PCA axis 4 PCA axis 5 . 0.08375 -0.38383 0.24055 0.49310 PCA axis 6 -0.09881 0.49592 -0.38436 35367 0 -0.41492 0.39593 Lateral-line scales 0.21902 -0.41333 . omitted Dorsal-fin rays -2.45691 -0.91952 08175 0 0.00905 -0.16599 Anal-fin rays 0.03886 03276 -0 . 0.21546 -0.18225 -0.12537 Pectoral-fin rays . 01176 -0 0.08675 -0.07120 -0.05523 Pelvic-fin rays . 00342 -0 0.12206 -0.05297 -0.19980 Gill rakers . species, being 32.11 ± 2.40 in P. oregonensis and 37.27 ± 2.89 in P. umpquae, indicating slight overlap. In general, if only the differences in proportions had been used to distinguish these species, such differences could be viewed as spu- rious, resulting from allometric differences in the individuals sampled. However, the methods used to choose the interorbital width character should ensure that the difference is real. To ascertain the usefulness of this character over the range of size, a linear regression of inter- orbital width and standard length for P. oregon- ensis and P. umpquae was performed, the sample prediction intervals were distinct (Fig. 4), with little overlap. For this regression, five very large individuals of P. oregonensis were eliminated to provide a more readable plot. If the large in- dividuals had been included, the entire plot for P. umpquae would have been compressed into the left quarter of the graph. In all, the inter- orbital width/prepectoral length proportion represented the best single proportion available for confident identification of these two species. Apparent from the regression scatterplot is a polymorphism in morphology between the two samples of P. umpquae from the Umpqua River. The two samples were taken from different streams in the Umpqua drainage, and they were also taken five years apart. Both populations had a nearly perfect linear relationship between the two measurements, clearly separable from P. oregonensis, but the sample of large individ- uals (UCLA W86-4) had a different linear re- lationship than the sample of small individuals (UCLA W81-15). Although there was no rec- ognizable difference in counts or other mor- phological features, these populations deserve further study to evaluate the origin of this vari- ation. Phylogenetic relationships.-Gap-coding of six meristics and 45 proportional measures (Table 5) yielded 38 multistate characters. Statistical evaluation of variation between ingroup and outgroup taxa supported the recognition of all six counts, 17 of 18 head proportions, and 15 of 27 body proportions as multistate characters. These characters produced a single Wagner tree (Fig. 5) with a consistency index (CI) of 0.868. General consistency of the different groups of characters demonstrated that counts had a CI of 0.778, head proportions a CI of 0.857, and body proportions a CI of 0.920. We note that those features most easily quantified (meristics) had the lowest CI. Although only a subset of the body proportions contributed useful mul- tistate characters, those that did yielded char- acters of high consistency. A homoplasy matrix for this analysis (Table 6) indicated that most (20 of 28 homoplastic steps) were associated with P. Lucius with nearly half of these between P. Lucius and M. conocephalus. This tree supported the monophyly of the genus Ptychocheilus, and lineages within the genus were supported by multiple synapomorphies (Table 7). These re- sults confirmed the sister-group relationship of