St John et al.: Diet of Plectropomus leopardus on the Great Barrier Reef 
183 
formal calcium acetate (FCA), a buffered formalin solu- 
tion (approximately 10 mL of FCA for every gram of 
stomach content) for a minimum of seven days and then 
stored in 70% ethanol. 
Stomach contents were classified broadly into natural 
prey or bait. All bait in our study were pilchards ( Sardinops 
neopilchardus), which does not occur naturally in the area. 
Every natural prey item was identified under low magni- 
fication to the lowest taxonomic group possible (St John, 
1995). Fish were identified following Randall et al. (1990) 
by using several keys (Allen, 1975; Masuda et al., 1984; 
Smith and Heemstra, 1986; Myers, 1991). 
Families of prey fishes were classified by the area where 
they were most commonly found (Randall et al., 1990; se- 
nior author, personal obs.). Prey fishes from the “demer- 
sal reef substrata” habitat swam around and above coral, 
but used it for shelter (e.g. Pomacentridae and Scaridae). 
Fishes categorized as using the “benthic reef substrata” 
habitat were benthic dwellers that remained very close to 
the substrata (e.g. Blenniidae). Fishes categorized as in 
the “adjacent sands” habitat dwelled over, on, or within 
the sandy areas adjacent to reefs (e.g. Mullidae). Pelagic 
fishes in the “midwater” habitat were found from the wa- 
ter surface to a depth of approximately 1 m over the reef 
(e.g. Clupeidae and Caesionidae). 
Wet weights of the stomach contents were measured 
after preservation in 70% ethanol. Before weighing, the 
stomach contents were emptied into a sieve (St John, 1995) 
and shaken or sponged, or both shaken and sponged, to 
remove excess surface liquid (Parker, 1963). Wet weights 
of stomach contents were excluded from weight analyses 
when there was evidence of regurgitation of food from the 
stomach, (e.g. digested prey in the mouth or gills or an 
empty stomach that was stretched or everted), but this 
condition was rare. 
Generally, weights of digested prey reflected prey size 
and therefore were a useful measure of the diet (St John, 
1995). Less than 14% of the prey was highly digested, and 
such fish could not be separated for weighing. In these 
cases, individual prey weights were estimated from total 
weights of the stomach sample, taking into account the di- 
gestion stage and the size (when known) of all individuals 
in the sample. 
The contents of the stomach were assumed to repre- 
sent daily feeding in P. leopardus because prey items were 
mostly digested after 24 hours (St John, 1995). 
Data analyses 
Kolmogorov-Smirnov (K-S) (Sokal and Rohlf, 1981) tests 
were used to compare size structures of leopard coral- 
grouper populations on reefs on both closed and open 
zones to fishing. 
Fisher’s exact tests were used to compare the frequen- 
cy of prey items in the diet because the categories (fish- 
ing zones, reefs, habitats, fishing methods and families of 
prey) were all nominal (Mehta and Patel, 1992). Because 
Fisher’s exact tests do not compare nested data sets, com- 
parisons within zones or locations used pooled data. Data 
from reefs were pooled when they did not differ signifi- 
cantly (i.e. P>0.05). Because the differences in number 
of prey per family in the diet was statistically marginal 
in reefs in the open zone (P=0.056), we used the P-0.01 
level of significance to compare open and closed zones 
when reefs were pooled. Also, the P=0.01 level of signifi- 
cance was used when sample sizes were small (e.g. in the 
comparison between the natural diet of line-caught and 
speared fish). 
An independent Ptest was used to compare the mean 
number of families consumed in each fishing zone. Prior 
to analysis, the variances were tested for homogeneity by 
using Cochran’s test, which was not significant (Cochran’s 
test statistic=0.84, P>0.05). Because the null hypothesis 
was not rejected, the power of the test to detect specified 
differences was calculated following Cohen (1988). In the 
calculations of power, the sample means were assumed to 
be representative of the parametric means for each treat- 
ment group. A nonsignificant result was considered to be 
inconclusive unless the power of the test (1-/3) was >0.80. 
The index of relative importance (IRI) determined for 
the diet of P. leopardus was 
IRI = 0.5 x (% prey number + % prey weight). 
This measure was used in Schoener’s a index of dietary 
overlap (Schoener, 1970) for pair-wise comparisons of the 
diet of P. leopardus between reefs within each of the two 
fishing zones. Dietary overlaps were classified by using 
Langton’s ( 1982) scale: low 0-0.29, medium 0.30-0.59, and 
high >0.60. 
Results 
Size structure of P. leopardus 
In total, 672 P. leopardus, ranging in size from 13 to 58.5 
cm FL, were collected by line and spear from the four reefs 
(Fig. 2) and fewer P. leopardus were caught by line (n=85) 
than by spear (n= 587) (Table 1). 
The speared catch was considered to represent popu- 
lations of P. leopardus on reefs because spear fishermen 
were deliberately not size selective in this study. The size- 
structure of the line catch was significantly larger than 
that of the speared catch (Dmax=0.243, K-S P<0.01) when 
samples from the four reefs were pooled (Fig. 2). Thus, line 
caught P. leopardus were excluded from further analyses 
of population size structure. 
When reef pairs within locations were compared, the size 
structure of the speared catch on the open reefs differed 
significantly from the closed reefs (north: Dmax=0.20, K-S, 
/2 = 143, P<0.01; south: Dmax=0.13, K-S, n=131 P<0.05; 
Fig. 2), but these zonal patterns differed between loca- 
tions. The size structure of the speared catch differed sig- 
nificantly between the two open reefs (Dmax=0.19, K-S, 
n = 155, PcO.Ol, Fig. 2); the proportion of larger fish ( >35 
cm FL) was 45% at Nathan Reef and 64% at Potter Reef 
compared with 57% at each of the closed reefs. The results 
of our study did not demonstrate an effect of fishing on 
size structure between open and closed reefs. 
