730 



Fishery Bulletin 103<4> 



by using a test of independence (Pearson chi-square. 

 Sokal and Rohlf, 1981; McCleave et al. 1987). Depth 

 distributions were averaged over each station. Compari- 

 sons were then made between all pairs of taxa within 

 a station, and a Bonferoni correction was applied to 

 assess the significance of the tests of independence. 

 Comparisions were not made between stations, because 

 sampling methods varied and depth distributions were 

 not directly comparable. The following null hypothesis 

 was evaluated: during each station occupation, average 

 larval depth distributions were independent of species. 

 Column and row variables were species and depth strata; 

 cell values were the average proportion of the larvae 

 captured in a depth stratum at a station. Comparisons 

 of center of mass were also made and the results were 

 very similar to the results of the test of independence 

 reported in the present study. 



The calculation of average proportion was made in 

 two steps. First, the proportion of larvae (P) collected in 

 each depth stratum id) at each sampling time (i) during 

 each station occupation (J) was calculated: 



the individual species comparisons were pooled across 

 station by the a priori assigned outcome of transport. 

 The number of significant differences found between 

 species were then compared to the number of significant 

 differences expected with a 5% error rate by using the 

 G-statistic (Sokal and Rohlf 1981). For example, in 

 a comparison of B. tyrannus to exported species, five 

 pairwise comparisons of larval depth distributions were 

 found to be significantly different and 12 were not sig- 

 nificantly different. At «=0.05, one significant and 16 

 nonsignificant differences are expected. The G-statistic 

 demonstrates that more significant differences were 

 found between B. tyrannus and exported species than 

 expected by chance. The classifications of significant 

 depth differences (shallower, deeper, different) were 

 then examined to determine the relation between larval 

 vertical distributions and the general outcome of larval 

 transport. 



Results 



dij 



'dij 





where C = larval concentration 100/m 3 . 



Then the average proportion of larvae (P) for each depth 

 stratum (d) was calculated for each station (J): 



IP* 



where n n = the number of sampling times (;') during 

 station occupation (J). 



Because the significance of a test of independence 

 depends, in part, on the magnitude of the cell values 

 (i.e., sample sizes), average larval concentration of each 

 species during each station occupation (number of lar- 

 vae/100 m 3 ) was used as a weighting factor. The av- 

 erage proportion of larvae at depth during a station 

 occupation (P d .) was multiplied by the weighting factor 

 to derive the cell values for use in the test of indepen- 

 dence. The weighting factor approximated the number 

 of fish larvae collected, and incorporated the effect of 

 variability in sampling volume. 



Values of the standardized residuals, which are a 

 result of the test of independence, were used to classify 

 significant differences in depth distribution as follows: 

 species A shallower (<) than species B, species A deeper 

 (>) than species B, and species A distributed differently 

 (< or >) than species B. This last category was assigned 

 when one species was not clearly deeper or shallower 

 than the other species, yet its depth distributions were 

 significantly different. 



To evaluate whether larval fish vertical distributions 

 were associated with larval transport, the results of 



Comparison of larval vertical distributions indicated 

 that B. tyrannus often had the shallowest larval verti- 

 cal distribution. There were more significant differences 

 than expected by chance between the vertical distribu- 

 tions of B. tyrannus and exported, estuarine, and shelf- 

 resident taxa (Table 1). For all significant differences, 

 the standard deviates from the test of independence 

 indicated that B. tyrannus were found in shallower water 

 than were other taxa (Appendix 1). 



Exported taxa generally were higher in the water col- 

 umn than estuarine and shelf-resident taxa. There were 

 more significant differences than expected by chance 

 between the vertical distributions of exported taxa and 

 estuarine and shelf-resident taxa (Table 1). Further, 9 

 of 12 significant differences between exported and es- 

 tuarine taxa indicated that exported taxa were found in 

 shallower water; 8 of 11 significant differences between 

 exported and shelf resident taxa indicated that exported 

 taxa were found in shallower water (Appendix 1). 



The vertical distributions of estuarine and shelf-resi- 

 dent taxa were different more often than expected by 

 chance, but taxa of neither group were consistently 

 found in shallower water (Table 1). Significant differ- 

 ences in larval vertical distributions were distributed 

 evenly among the three classifications of the direction 

 of difference (»=4 shallower; n=2 deeper; ;; = 5 different) 

 (Appendix 1). 



Discussion 



The results indicate an overall hierarchy of larval ver- 

 tical distributions; B. tyrannus was found in shallower 

 water than were exported taxa, and exported taxa 

 were shallower than estuarine and shelf-resident taxa. 

 Although this general pattern emerged, considerable 

 variability in larval vertical distributions was observed, 

 which is a common result of many studies (e.g., Boehlert 



