288 
Fishery Bulletin 96(2), 1998 
outside waters, and at six locations in inside waters. 
All plankton tows were oblique from the surface to a 
depth of 50 m. Samples were preserved in a 10%- 
formalin-seawater solution after debris was removed. 
Plankton and neuston samples were sorted to re- 
move large organisms, such as gelatinous zooplank- 
ton. We then split the sample with a Folsom splitter 
until a subsample of about 500 organisms remained. 
Plankton organisms were identified to the lowest 
convenient taxon and counted. Detailed composition 
of many of the neuston samples has been presented 
elsewhere (Brodeur 2 ). 
Data analysis 
Stomach data were partitioned into subsets accord- 
ing to salmon species, geographic area, habitat, dis- 
tance offshore, month, and year. The index of rela- 
tive importance (IRI; Pinkas etal., 1971) was calcu- 
lated for each data subset. The modified IRI (dry 
weight rather than volume) was used to character- 
ize the diet of each species and to rank prey taxa: 
IRI = (N + W)F, 
where N is numerical percentage, W is weight per- 
centage, and F is frequency of occurrence (FO) per- 
centage. In all comparisons, the IRI is expressed as 
a percentage of total IRI for each data subset. 
Morisita’s index of overlap as modified by Horn 
(1966) was used to calculate overlap between spe- 
cies pairs; values range from 0 (no overlap) to 1 (com- 
plete overlap): 
s 
2 
i=l 
y, 
i= i 
i= 1 
where x ; and y ■ are proportions of the numbers of 
individuals of prey species i found in the predator 
species x andy, respectively. 
The percent similarity index (PSI; Whittaker, 1975) 
was used to compare stomach samples to plankton 
samples: 
PSI = 2 ^min(p g or p b ), 
where p a is percentage number for a given species in 
sample A, and p b is percentage number for the same 
species in sample B. A PSI value of 1.00 shows com- 
plete similarity; a value of 0 indicates no similarity. 
We considered values >0.60 to be significant for both 
overlap indices. Chinook salmon were not included 
in the analysis of overlap because of small sample 
sizes (Table 1). Following Brodeur and Pearcy ( 1990), 
we tested for differences in the occurrence of princi- 
pal prey between years, months, areas, and distance 
offshore using ^ 2 . 
We examined data for neuston and plankton prey 
samples collected at locations where stomachs of at 
least five specimens of one salmon species were avail- 
able and included a taxon in the prey collections. To 
measure prey selection, we used Strauss’s linear food 
selection index (Strauss, 1979): 
L = r i -P l , 
where r and p t are the proportional abundances of 
prey item i in the gut and habitat, respectively. Se- 
lection values range from -1, indicating avoidance 
or negative accessibility, to +1, indicating preference 
or positive selection; 0 indicates random feeding. 
Values are extreme only when the prey item is pro- 
portionately abundant but rarely consumed (-1), or 
is proportionately rare but consumed almost exclu- 
sively ( + 1). We tabulated selection values >0.10 or 
<-0.10 for an indication of the positive or negative 
selection of a particular taxon. 
To compare the number of stomachs required to 
characterize the breadth of diet for each species of 
salmon, we pooled the stomachs over all periods and 
habitats. Stomachs were selected randomly, and the 
cumulative number of taxa were plotted versus the 
number of stomachs until the asymptote was reached 
(Hurtubia, 1973). 
Results 
Description of diet 
All salmon species The prey spectrum for juveniles 
of five Pacific salmon species comprised at least 30 
taxa (Table 2). The six taxonomic groups of greatest 
importance (IRI) were calanoid copepods, hyperiid 
amphipods, euphausiids, decapods, tunicates, and 
fishes (Table 3). In pooled samples, fish were the most 
important prey for coho and chinook salmon 
(IRI=63.8% and 76.4%) but were only moderately 
important for the other species (IRI 28.3-40.3%). 
Hyperiid amphipods, most commonly Themisto spp., 
were also important prey for pink, chum, and sock- 
eye salmon (IRI 28.0-39.6%). However, the biomass 
of teleost prey made up more than 75% of the total 
biomass consumed by each of the juvenile salmon 
species in pooled samples. 
The full breadth of the prey spectrum for juvenile 
salmon species was obtained by randomly selecting 
