756 
Fishery Bulletin 96(4), 1998 
larities percentage (SIMPER) procedure (Clarke, 
1993; Clarke and Warwick, 1994). All multivariate 
analyses were performed with Plymouth Routines 
in Marine Environmental Research (PRIMER) pro- 
grams (copyright M. R. Carr and K. R. Clarke, Ma- 
rine Biological Laboratory, Plymouth, UK; Clarke 
and Warwick, 1994). 
Separate analyses were performed to compare 
stomach contents between various length groups of 
bonefish, between seasons, and between areas (Table 
1). To detect length-related differences in feeding, we 
pooled bonefish into 20-mm length intervals and used 
cluster and MDS analyses to compare stomach con- 
tents. Stomach contents of all 20-mm length groups 
within the 480- to 699-mm length range had a level 
of similarity >55%, so we restricted all other com- 
Table 1 
Sample sizes, and collection months and years for data 
included in ANOSIM 
comparisons. Numbers 
in parenthe- 
ses are the number of samples collected during a particu- 
lar month. Bonefish, Albula vulpes, fork lengths for all 
comparisons ranged from 480 to 699 mm. 
Comparison 
n 
Months 
Years 
Area 
Florida Bay 
50 
Jan (5), Feb (3), 
Mar (2), Apr (5), 
May (7), Jun (1), 
Jul (4), Aug (4), 
Sep (3), Oct (6), 
Nov (7), Dec (3) 
1991-1995 
Ocean side 
50 
same monthly 
sample sizes as 
Florida Bay 
1991-1995 
Season 
Ocean side 
Jan-Mar 
39 
1991-1995 
Apr-Jun 
43 
Jul-Sep 
6 
Oct-Dec 
Florida Bay 
33 
1991-1995 
Jan-Mar 
8 
Apr-Jun 
25 
Jul-Sep 
70 
Oct-Dec 
18 
Stomach throw trap 
Ocean side 
Stomachs 
39 
Jan. (7), Feb ( 13), 
Aug (4), Sept (3), 
Oct (12) 
1991-1995 
Throw traps 
Florida Bay 
54 
Jan (40), Sep ( 14) 
1996-1997 
Stomachs 
45 
Mar (2), May (9), 
Jun ( 10), Sep (24) 
1991-1995 
Throw traps 
30 
Mar (6), May (6), 
Jun (6), Sep (12) 
1994-1996 
parisons to this length group in order to minimize 
length-related dietary shifts that could have con- 
founded comparisons of areas and seasons. We com- 
pared stomach contents of bonefish from two areas, 
Florida Bay and the ocean (Florida Straits) side of 
the Keys, using the analysis of similarity ( ANOSIM) 
permutation test (Clarke, 1993; Clarke and Warwick, 
1994). We did not include the lower Keys or Biscayne 
Bay in our area comparisons because we examined 
relatively few bonefish stomachs from these two ar- 
eas. Bonefish move seasonally between Florida Bay 
and ocean-side areas; thus for any given month 
sample sizes were rarely the same for the two areas. 
To reduce confounding from seasonal effects that 
could result from unequal seasonal representation 
of the two areas, we eliminated some stomachs from 
our analysis to achieve equal monthly sample sizes 
for the two areas for each month. We pooled samples 
from all years, totaled the number of samples by 
month for each area, and then randomly eliminated 
stomachs for each month from the area with the 
greatest sample size so that, for any given month, 
both areas had equal sample sizes. The resulting 
sample of 50 stomachs from each area contained 
equal sample sizes from both areas for each month, 
but the total sample size varied from month to month 
(Table 1). To detect seasonal dietary shifts, collec- 
tions were pooled into four seasonal groupings: Janu- 
ary-March, April-June, July-September, and Octo- 
ber-December. We used ANOSIM to make seasonal 
comparisons of the diets of bonefish separately for 
the two principal sampling areas. Six pairwise com- 
parisons were made among seasons for each area. 
No adjustment of significance levels exists for 
ANOSIM to account for the increased possibility of 
type-1 error associated with multiple comparisons 
(Clarke and Warwick, 1994). 
To determine feeding selectivity, we used ANOSIM 
to compare the species of crustaceans and fishes 
found in the stomachs of bonefish 480-699 mm long 
to those found in samples collected in the potential- 
prey environment. We included only bonefish col- 
lected during the same seasons as those when the 
throw-trap collections were made. For Florida Bay 
comparisons, bonefish collected during March, May, 
June, and September were included; for ocean-side 
comparisons, bonefish collected during January, Feb- 
ruary, August, September, and October were included 
(Table 1). Significant differences in the suite of crus- 
taceans and fishes found in bonefish stomachs and 
the potential prey available in the environment would 
imply selective feeding. We used SIMPER to indi- 
cate the percentage of dissimilarity contributed by 
each prey species and thus show which prey were 
selected or not selected. Taxa that were not selected 
