466 
Fishery Bulletin 107(4) 
the two island samples and the combined overall sample, 
the vacuity rate (i.e., the proportion of empty stomachs) 
was determined. Differences in vacuity between islands 
were assessed with a chi-square test. For full stomachs, 
the number of prey items was counted, and the SL and 
TL of fish prey and the carapace length of crustacean 
prey were determined to the nearest mm, where diges- 
tion state allowed reliable measurements. For fishes, SL 
could be determined more often than TL because skel- 
etons are slow to disintegrate during digestion. In these 
cases, TL was calculated from SL according to published 
SL-TL equations for the respective taxon (Froese and 
Pauly, 2009). The M of all prey items was recorded to 
the nearest mg. Mean relative stomach content M was 
then calculated as total M of prey items divided by the 
total M of C. argus specimens. 
We identified each prey item to the lowest possible 
taxonomic level, using Randall’s (1996; 2007) key for 
fish prey, and Hoover’s (1998) key for crustacean prey. 
Cumulative prey curves (Ferry and Cailliet, 1996) de- 
rived from plotting the cumulative number of unique 
prey taxa against the cumulative number of analyzed 
stomachs allowed us to assess whether sample sizes 
were large enough to accurately characterize dietary 
breadth. These curves reach an asymptote if sample 
size is sufficient. To determine dietary importance of 
prey, for all identified prey types and families, the nu- 
merical importance (%N), frequency of occurrence (%0) 
(calculated from full stomachs), and gravimetric impor- 
tance ( %M ), as well as the index of relative importance 
(IRI) and the %IRI (proportion of the IRI of a taxon to 
the sum of IRIs of all taxa) were calculated. The IRI 
incorporates the individual indices in the formula 
IRI = (%N + %M) x %0 (1) 
and may provide a more accurate description of dietary 
importance than its components by canceling out their 
individuals biases, such as the overestimation of the 
importance of an abundant but small prey item by the 
%N (Cortes, 1996). Because %N, %M, %0, and %IRI 
indicated that fishes dominated the diet of C. argus, for 
the comparison of diets between islands and the calcula- 
tion of electivity, the indices were recalculated for the 
fish component of the diet alone. 
Composition of the reef fish assemblage 
Underwater visual surveys with scuba were used to 
determine reef fish abundances and sizes (in 5-cm bins, 
i.e., 0-5 cm, 5-10 cm, etc.) at 23 sites (depth range 8.2 
m-18.2 m, mean depth 11.9 m). All sites were located in 
the dominant reef habitat of the Kona coast, reef shelves 
with moderate to high finger coral (Porites compressa) 
cover. Each survey of a site involved four divers (two 
pairs), who between them surveyed four 25x4 m (100 
m 2 ) belt transects that were permanently installed at 
each site. Each transect count consisted of one rapid 
swim to count mobile and midwater species, and a slow 
return swim closer to the bottom to record fishes in and 
around the benthos. Sites were surveyed 4 to 6 times 
during the year 2003, generally between 0840 and 1600 
hours. The detailed sampling regime was described 
by Tissot et al. (2004). Surveys were conducted under 
the direction of the West Hawaii Aquarium Project 
( WHAP, a collaboration of the Hawaii Division of Aquatic 
Resources (HDAR), the University of Hawaii at Hilo, and 
Washington State University), and are therefore referred 
to as “WHAP surveys” here. 
WHAP survey counts were used to calculate mean 
fish densities (individuals/100 m 2 ) and relative numeri- 
cal importance (%N) of reef fishes in Kona in 2003 
(i.e., grand mean of densities for the year 2003 at the 
23 sites). In addition, relative importance in terms of 
biomass (%M) was determined for large (mean body M 
>50 g) piscivores. For this purpose, the M of individuals 
was estimated from their TL by using conversion equa- 
tions for the respective taxa (Froese and Pauly, 2009). 
Finally, size-frequency distributions of reef fishes in 
Kona were calculated from the combined WHAP survey 
counts for 2003. 
Abundances of nocturnally active taxa tend to be 
underestimated by data collected during daytime sur- 
veys (Ackerman and Belwood, 2000). For the calcula- 
tion of electivity (see next section), abundances of the 
nocturnal apogonids, holocentrids, and priacanthids 
were therefore estimated by nighttime surveys (“Night 
WHAP”) that took place in 2003 at the same sites sur- 
veyed during daytime. The ratios of nighttime to day- 
time abundances for these families were 90.6, 2.6, and 
1.5, respectively. 
Prey selection 
To determine the taxonomic focus of predation, we used 
Ivlev’s electivity index (Ivlev, 1961): 
E i = (r i -p i )Ur i +p i ), (2) 
where r ; = numeric importance (%N) of fish family i in 
the diet of C. argus ; and 
p t - %N of the same family in the reef environment. 
E l can take values between -1 and 1. Positive values 
indicate “preference” (a taxon overrepresented in the di- 
et in relation to its availability in the environment), and 
negative values “avoidance” (a taxon underrepresented 
in the diet in relation to its availability) (Lechowicz, 
1982). Because of the scarcity of reef fish abundance 
data for Oahu for the year 2003, when stomach con- 
tents for this study were obtained, we based electivity 
calculations on diet composition data obtained from the 
Kona sample, and on reef fish abundance data for the 
Kona coast from WHAP surveys in 2003. 
Prey-size selection 
To assess the size focus of C. argus predation, length- 
frequency distributions of important fish families in the 
diet of C. argus were compared with length-frequency 
