FISHERY BULLETIN: VOL. 85, NO. 1 



Table 3.— Chi-square test of homogeneity of ranking of areas 

 by decreasing mean D values and correlation of ranking of 

 areas in each species with overall ranking of all other species. 

 Statistics computed for 8 species collected at all four areas and 

 for the remaining 7 species after data for Ca. princeps were 

 omitted. L = La Jolla; C = Channel Islands; E = Punta 

 Eugenia; G = Isia de Guadalupe. 



value is not quite significant (0.1 > P > 0.05; critical 

 value 7.81). Although the pattern of differentiation 

 over all eight species cannot be shown to depart 

 significantly from randomness by this nonpara- 

 metric test, it is instructive to continue the analysis 

 to see whether anomalous results in one or two 

 species may be obscuring an underlying pattern in 

 the others. Aberrant species can be identified by 

 measuring the correlation (rj of rankings for each 

 species with the overall rankings for all other species 

 combined. To do this, rankings for the localities were 

 computed as each species in turn was deleted from 

 the analysis. These rankings were then compared 

 with those for the species deleted. The r^ values for 



this analysis clearly indicate a core group of five 

 species (Chromis punctipinnis, Clinocottus analis, 

 E. jacksoni, M. califomiensis, P. ciathratus), rank- 

 ings for each of which are highly correlated with 

 those of all other species (Table 3). At the other ex- 

 treme, rankings of Caulolatilu^ princeps are essen- 

 tially the opposite of those of the other species (r^ 

 = - 0.80). Thus C. princeps is the only species for 

 which La Jolla was ranked the most divergent local- 

 ity, as it is the only species for which Guadalupe is 

 the locality with the highest overall genetic similar- 

 ity to the other populations. 



In order to evaluate the influence of Caulolatilus 

 princeps on the overall analysis, Friedman's test 

 was repeated after data for C. princeps had been 

 deleted. The resulting chi-square value for seven 

 species (10.59; 3 df) is significant at the 0.05 level. 

 After omitting C. princeps, the correlation (r^) for 

 each species with all other species was again com- 

 puted (Table 3). It is apparent that the remaining 

 species form a more coherent group with C. prin- 

 ceps omitted, values for each species being positively 

 correlated with those from all other species. 



More detail regarding possible pathways of lar- 

 val drift can be obtained by considering the relative 

 degree of divergence of each pair of populations. D 

 values for the six possible pairwise comparisons of 

 the four study areas have been ranked within each 

 species in Table 4. An analysis similar to the pre- 

 ceding indicates that the two northern populations 

 (La Jolla and the Channel Islands) are consistently 

 the most similar genetically, and the two southern 

 populations (Punta Eugenia and Guadalupe) are the 

 most divergent. There are no consistent differences 

 in rankings of the four other comparisons, each of 



Table 4.— Chi-square test of homogeneity of ranking of pairs of localities 

 by decreasing mean D values and correlation of ranking in each species 

 with overall ranking of all other species. Statistics computed for 8 species 

 collected at all four localities and for the remaining 7 species after data 

 for Ca. princeps were omitted. Abbreviations as in Table 3. 



