7022 - Laserfiche WebLink

Home Browse Search

MAYDEN ET AL.-SYSTEMATICS OF WESTERN CYPRINIDS 823 tical analyses and graphical display of results were written by W. J. Rainboth in A Program- ming Language (APL). To study the phylogenetic relationships of these species, measurements were transformed to reduce differences because of size variation in the samples. The transformations were ac- complished through division by standard length (SL sense Hubbs and Lagler, 1958; for most longer body and fin measurements) or prepec- toral length (PPL = tip of snout to pectoral insertion; for most shorter measures associated with cranial features). Prepectoral length was chosen to standardize the cranial measures in- stead of the traditional "head length" (HL; sen- su Hubbs and Lagler, 1958). The former can be more accurately measured between two points, whereas the latter requires an arc mea- surement to a more subjective point. The char- acters were gap-coded using the simple-gap of Archie (1985) of one average standard devia- tion. We employed a standard procedure of or- dering the states of each character in numerical sequence (the smallest value assigned zero) giv- en their single-dimension, magnitude-only re- lationship. The evolutionary polarity of the or- der was determined from the outgroup perspective of H. symmetricus. A Wagner tree was constructed by Phylogenetic Analysis Using Parsimony (PAUP; Swofford, 1985), ordering all characters as above with the outgroup, and employing the branch and bound algorithm of Hendy and Penny (1982) to find all most par- simonious trees. For comparison, an alternative analysis examined the effect of such ordering by using all characters unordered. RESULTS Discrimination of operational taxonomic units (OTUs).-For the principal component analysis (PCA) of log-transformed measurements (Ta- ble 3), 98.3% of the total variance was account- ed for the first (size) axis because of great dis- parity in size of the specimens. Five additional roots were required to account for the next 1% of the variance. These six largest roots, along with the six counts were analyzed using canon- ical variate analysis (CVA). The CVA result for the entire group (Fig. 2) illustrates the distinc- tiveness of Hesperoleucus, Mylopharodon, Ptycho- chedus lucius, and Siuslaw River population of Ptychocheilus umpquae. The remaining three taxa of Ptychocheilus demonstrated significant spatial overlap among axes I and 11. The two groups of specimens, both currently recognized as P. umpquae, revealed in this anal- ysis correspond to specimens from the Umpqua drainage and those from the adjacent but more northern Siuslaw drainage. The distinctive re- sult for the Siuslaw population of P. umpquae was particularly interesting and may alter our present view of the number of species in the genus. However, given the problematic status of the Siuslaw population (C. Bond, pers. comm., see below), the few specimens available, and their small size„ and we defer a definitive treatment until more material becomes available. For our purposes, the specimens of P. umpquae from the Umpqua drainage, the drainage that includes the type locality (Bond, 1980), were used to rep- resent the taxon. Ptychocheilusgrandis, P. oregon- ensis, and P. umpquae (Umpqua River) had con- siderable overlap in the large-scale analysis but were reanalyzed with the easily separable spe- cies removed to evaluate diagnostic features of each (Fig. 3A, B). In this analysis, all three spe- cies could be distinguished easily. The best sin- gle identifying feature was the number of dorsal rays, eight in P. grandis and nine in both P. umpquae and P. oregonensis. Of the 96 specimens examined, a single P. oregonensis had the count typical of P. grandis. The near-perfect nature of this character for group identification caused a loading many times the value of the next high- est character loading (Table 4) and resulted in the single aberrant specimen plotting near spec- imens of P. grandis. A single specimen of P. umpquae had a sin$ilarly aberrant, but noncon- flicting, count of 10 dorsal rays and is plotted far to the left of the other specimens (Fig. 3A). These extreme outliers resulted in bivariate prediction ellipses assuming a seemingly nonpredictive alignment. A third CVA (Fig. 3B) conducted on these three species, omitting the dorsal ray character, resulted in little overlap of prediction ellipses. Margins of ellipses for P. oregonensis and P. umpquae overlapped slightly, but only one of the 46 P. oregonensis fell within the predicted values for the other species. Likewise, P. grandis was distinguished from the other two species with little overlap. Thus, even using all characters, except dorsal rays, P. oregonensis and P. umpquae were as distinguishable from one another as ei- ther was from P. grandis, a species already known to be distinct. Most of the separation of P. ore- gonensis and P. umpquae (Umpqua River) oc- curred along the horizontal first axis. This axis had high loadings by principal components 2