FISHERY BULLETIN: VOL. 73, NO. 4 



tion of amounts among the 5 items or classes. 

 Values are scaled as B/S between zero, the most 

 uneven distribution possible among 5 items, and 

 unity, the most even possible. 



Proportionate overlap of resource use may be 

 thought of as a measure of mutual use by two 

 species of the same items. The coefficient used here 

 measures overlap between species ./ and k (Horn 

 1966): 



Cx = 



(?i^ 



2 P,/+ 2 P,f 



i = 1 



i = 1 



Among the three possible species pairs, activity 

 overlap calculated from field observations of feed- 

 ing rate varied in the same manner as activity 

 overlap calculated from laboratory observations of 

 swimming rate. For the subsequent calculation of 

 total overlap in food, activity, and habitat, 

 therefore, we used the arithmetic mean of these 

 two independent estimates to express a single ac- 

 tivity component. 



Two estimates were made of total overlap. If, 

 e.g., food and habitat are mutually independent, 

 i.e., the same foods are available in the same 

 proportions in different habitats, overlap between 

 any species pair is the product of the separate 

 measures for food and habitat. But if food is 

 completely dependent on habitat, i.e., different 

 habitats contain totally different foods, then 

 overlap is the arithmetic mean of the two 

 measures (Pianka 1974). (In the latter case, food 

 and habitat measure the same resource axis, and 

 the separate measures can be thought of as 

 independent estimates of the same overlap.) In 

 general, the actual relation between resource dis- 

 tributions is somewhere between complete 

 independence and dependence. Thus, total overlap 

 as expressed by the mean of the separate overlaps 

 in food, activity, and habitat is a maximum value, 

 and true overlap is somewhere between this and 

 the product estimate (see Pianka 1974). 



Kolmogorov-Smirnov tests of goodness-of-fit 

 showed that the distributions of the variates of 

 food breadth (.s, 6, b/.s) differed significantly from 

 normal (were skewed) for the white seaperch and 

 senorita (P = O.O.S-0.01), though not for the kelp 

 perch (/^0.05). Therefore, we computed medians, 

 ranges, and the Kruskal-Wallis measure H of 



differences in location of the ranked variates to 

 test for differences among the three species (see 

 Sokal and Rohlf 1969). 



RESULTS 



Food 



Seasonal Effect 



Food arrays were generally correlated among 

 seasons. The rank concordance of diets was sig- 

 nificant for all species (Table 1). The same items, 

 which often made up more than 70% of the total 

 food volume, usually occupied one or another of the 

 first three ranks from one season to the next; sel- 

 dom did the first three positions for any one season 

 contain items not among the first three for the 

 others. Consequently, all samples were pooled by 

 species, regardless of season, for the remaining 

 analyses. 



Individual Variability 



Food items were generally uncorrelated among 

 fish of the same species collected at the same time 

 and place. Rank concordance was significant for 

 but one of five collections of kelp perch {W = 0.55, 

 P<0.01), and for no collection of white seaperch or 

 senoritas. This might indicate that members of the 

 same species that co-occur were choosing different 

 items from the same forage base. But such a 

 conclusion may be misleading for white seaperch 

 and senoritas. A single item (plant-encrusting 

 bryozoans) usually dominated their foregut con- 

 tents, and although the second and third ranked 

 items varied considerably among fish, they made 

 up but a minor part of the total. 



Kelp Perch 



Tiny plankters, mainly copepods, made up more 

 than half of the food consumed by the kelp perch 

 (Table 2). Of 18 categories of items eaten by kelp 

 perch, small calanoid and cyclopoid copepods led in 

 both abundance and frequency; foreguts of six fish 

 contained copepods only, and one was packed with 

 more than 300 individuals. The distinctive genus 

 Corycaeiii^ occasionally dominated the contents. 

 Large calanoids longer than 4 mm, which were 

 found in fish collected during late winter only, 

 were relatively rare. Tiny ostracods were 

 frequently consumed, but only in small numbers. 



818 



