724 
Fishery Bulletin 95(4), 1997 
represent the limiting dimension for ingestion, were 
measured to 0.01 mm with an ocular micrometer in a 
compound microscope. Mouth width of larvae at the 
widest point of the upper jaw was similarly measured 
on a subsample of at least 50 larvae of each species. 
Widths of prey and mouth widths of larvae of each spe- 
cies were then plotted against body length of larvae. 
Proportional utilization (p.) of each prey type was 
calculated for length classes within each species with 
data pooled across regions and months. Proportional 
utilization was also calculated across length classes 
for the larvae of each species within a sample. Guts 
that were empty or contained only unidentifiable 
material were not included in these calculations. 
Relative feeding prevalence of all larvae within 
samples were correlated against the corresponding 
estimates of zooplankton abundance to determine if 
there was any evidence that zooplankton abundance 
was limiting feeding success. The average DNO ex- 
hibited between all species within samples was like- 
wise compared with zooplankton abundance. 
Calculation of dietary niche overlap 
Interspecific DNO was calculated for species-pairs 
within individual samples, i.e. for each site within a 
given month. Because this part of our study focused 
on examining niche relations between species, pooled 
diets for each species within a sample were consid- 
ered to represent the average diet of each species. 
Furthermore, comparisons were limited to those 
samples in which >10 larvae of each of two or more 
species contained food. By pooling data across size 
classes we were able to compare more DNO data. 
Although the use of average diets would reduce the 
robustness of a parametric test of significance, we 
used nonparametric techniques for assessing the sig- 
nificance of DNO. 
Sufficient numbers of larvae were obtained for analy- 
sis in 13 of the 28 ichthyoplankton samples (7 months 
x 4 regions) to allow 43 pairwise comparisons to be made 
between the diets of co-occurring species. 
Besides the DNO values that were calculated and 
that incorporated prey abundance data, the results 
of this technique were also evaluated against calcu- 
lations of DNO that did not incorporate such data. 
Because prey abundance data are incorporated into 
prey utilization data prior to calculating DNO (see 
below), the same formula was used to calculate DNO 
both with and without consideration of prey concen- 
trations. DNO was measured with the symmetric 
niche overlap coefficient (Pianka, 1973) 
= (X PijPik ) / Pi 5X)’ 
where p tj and p ik = the proportional utilization of prey 
type i by species j and k, respectively. 
Using the p t data directly, we were able to provide 
a basis for calculating DNO without prey abundance 
data. To incorporate prey abundance data, the geo- 
metric mean (g ; ) of p ; and electivity (e ; ) was used 
(Winemiller and Pianka, 1990), instead of p ; as in 
the original formula. The geometric mean gives a 
better indication of ecological similarity by reducing 
those biases within both p ; and e ; that can result from 
the presence of very abundant or very rare prey types 
(Winemiller and Pianka, 1990). Electivity is the 
value that has been weighted by resource availabil- 
ity (R } ) as e ; = p/R r These values were calculated 
within the DNO algorithm and are not presented. 
Note that g : can be used with other overlap indices 
because it is calculated prior to the calculation of the 
overlap value. 
Bootstrapping of the resource matrix of a pair of 
species was used to obtain a null distribution of 1,000 
pseudo-DNO values against which the significance 
of observed DNO could be assessed (Winemiller and 
Pianka, 1990). These calculations were performed for 
species pairs at sites within months. In each one of 
the 1,000 runs, the algorithm randomly reassigned 
theg ; values for each prey type (e.g. resource states 
i...p) within each larval species j and k, but among 
the resource types used by both j and k (e.g. amongst 
Siy-Snj and Sik-Spk^- A Si value for one of the species 
pair may thus be reassigned to a resource state which 
was used only by the other species. 
The null hypothesis ( H 0 ) for each test was that the 
dietary compositions of the larvae of the two species 
were not the same. The null hypothesis was rejected 
if more than 95% of the 1,000 pseudo-DNO values 
were less that the observed DNO. Such cases indi- 
cated that the observed value was larger than would 
be randomly expected at P<0.05. 
The prevalence of significant DNO calculated with 
g t and bootstrapping was then compared with that 
obtained with p t , i.e. when R : was not taken into ac- 
count. These results were also compared with those 
obtained when the significance of DNO was arbi- 
trarily set at values >0.6. 
Results 
Zooplankton 
Zooplankton were very abundant in Wilson Inlet be- 
tween October 1988 and April 1989; there was a to- 
tal mean monthly concentration of 342,746 organ- 
isms/m 3 (range = 48,641-2,951,209/m 3 ). Concentra- 
tions exceeded 100,000/m 3 in 26 of the 28 zooplank- 
