Bollens et al.: Feeding ecology of |uvenile Oncorhynchus spp. 
403 
and teleost prey. Coho and pink salmon both strongly 
selected for insects, whereas decapod larvae were im- 
portant to the former and gammarid amphipods to the 
latter. Our study supports the importance of insect prey 
to young juvenile salmon in transitional environments 
(Moulton, 1997; Romanuk and Levings, 2005; Weitkamp 
and Sturdevant, 2008). 
The diet of juvenile chum salmon further differed 
from the other salmon species by the abundance of lar- 
vaceans (primarily Oikopleura sp.) in their gut contents, 
and their consumption of ctenophores. These observa- 
tions are consistent with other reports (Simenstad and 
Salo, 1980; Black and Low, 1983; Landingham et ah, 
1998) and may be related to anatomical gut specializa- 
tion, which enables chum salmon to assimilate prey 
items that other salmon cannot digest (Welch, 1997; 
Arai et al., 2003). 
Calanoid copepods have been described as a major 
diet item for juvenile salmon generally (Pearcy, 1992), 
and for chum and pink salmon specifically (Godin, 1981; 
Sturdevant et al., 2004). However, despite a diverse and 
abundant assemblage of copepod species in Dabob Bay, 
copepods only represented a modest component of our 
salmon diets and were particularly limited to smaller 
predators. Instead, juvenile salmon in Dabob Bay were 
found more often feeding on numerically less abundant 
macrozooplankton such as euphausiids, hyperiid am- 
phipods, and decapod larvae. Similarly, Peterson et al. 
(1982) showed that juvenile coho and Chinook salmon 
off Oregon fed more on hyperiid amphipods than on 
the numerically dominant copepods. These results sup- 
port the results from other studies that indicate that 
salmon are more likely to feed on larger, more visible 
prey items (Healey, 1980; Schabetsberger et al., 2003), 
in which case abiotic factors (e.g., light intensity and 
turbidity) and biotic processes (e.g., vertical migration 
and predator evasion behavior) will be important vari- 
ables that will help determine stomach fullness and 
feeding success. 
Ontogenetic diet thresholds for juvenile salmon at 
approximately 80 mm, before which teleost prey are 
less important, have been indicated by other studies 
(Brodeur, 1991; Keeley and Grant, 2001). In contrast, 
our electivity results provide evidence that teleost prey 
were strongly selected for by small (<75 mm) chum and 
pink salmon during spring, when fish larvae may have 
been particularly small (Bollens et al., 1992a; Fulmer 
and Bollens 2005). Simenstad and Salo (1980) found 
that juvenile chum salmon transitioned from nearshore 
habitats with epibenthic food sources to neritic habi- 
tats with pelagic and nektonic food sources when they 
reached approximately 45-55 mm FL. In other studies, 
seasonal variability in salmon gut contents has been at- 
tributed to ontogenetic shifts in feeding preferences or 
feeding behavior (Beacham, 1986; Brodeur, 1991; Daly 
et al., 2009). Similarly, our results indicate that small 
chum, Chinook, and coho salmon select small prey, 
then larger prey as the fish develop. Small Chinook 
and coho salmon selected Calanus pacificus roughly in 
proportion to its abundance, but other copepods were 
avoided. Similarly, only small coho salmon selected 
gammarid and hyperiid amphipods. In contrast, only 
large Chinook salmon selected for gammarids. Thus, 
both species-specific and ontogenetic shifts in prey pref- 
erence were observed. 
Our diet and electivity results should be interpret- 
ed cautiously because our samples were pooled across 
broad size, temporal, and spatial scales, and because 
of limitations associated with sample sizes, net sam- 
pling biases, and pooling of prey species and life history 
stages. For example, the range of euphausiid electivity 
values observed may be due to the pooling of euphausiid 
species and life-history stages, potentially obscuring 
euphausiid prey selection patterns observed in other 
studies (Schabetsberger et al., 2003). 
Another major caution concerns our ability to deter- 
mine “available prey” with plankton nets. More mo- 
bile and larger nektonic prey, such as cephalopods and 
young fish, are able to avoid conventional plankton nets, 
with the consequence that electivity indices for these 
prey types would be biased upward. Conversely, small 
prey types that are unable to avoid the plankton net 
(e.g., small copepods) would be proportionately over- 
represented in the net samples, with the consequence 
that electivity indices for these forms would be biased 
downward. We recommend that further research be 
undertaken into adequately sampling macrozooplank- 
ton and micronekton (e.g., Gewant and Bollens, 2005), 
such that a broader and potentially more appropriate 
range of potential prey for fishes can be quantitatively 
sampled. 
Several additional complicating factors should be 
considered when interpreting electivity indices. First, 
strongly positive electivity (e.g., E ; = 1.00) often results 
from a rare presence of a species in the gut contents 
and a corresponding absence of that same species in 
the plankton. In some cases zero abundance in the 
plankton may be due to low-volume plankton hauls 
which under-sample the available prey field. Conversely, 
an E x of —1.00 could result from a rare (but nonzero) 
occurrence of a species in the environment, combined 
with its absence from the gut contents, perhaps simply 
because of a low probability of encounter between preda- 
tor and prey. 
A final caution concerns the vertical resolution of 
sampling. Landingham et al. (1998) used both neuston 
and oblique plankton tows and showed that salmon 
diet most closely resembled that of the neuston assem- 
blage. The upper 50 m were sampled with our sampling 
methods and therefore electivity values may have been 
biased. For example, if juvenile salmon are primarily 
feeding near the surface, abundant zooplankton (i.e., co- 
pepods) that are more deeply distributed may not fully 
be part of the “available” prey community. We recom- 
mend finer-scale, vertically resolved sampling of juve- 
nile salmon and their potential prey in future studies. 
Dabob Bay has been the site of numerous studies for 
which the interactions between planktivorous fishes and 
