Taylor and Gervasi: Feeding habits and dietary overlap of age-0 flounder 
171 
The percent index of relative importance (%IRI) was 
used to estimate the overall contribution of a prey tax¬ 
on to the diet of winter and summer flounder, such that 
INI 
%IRR = —— X100, (1) 
n^^IRh 
where %IRI^ = a compound index calculated for prey tax¬ 
on k and equal to and 
np = the total number of prey taxa identi¬ 
fied in stomachs of winter and summer 
flounder (np: 33 and 32, respectively). 
The IRI index was selected as a descriptor of the diet 
of winter and summer flounder in this study for 2 prin¬ 
cipal reasons: 1) to minimize biases associated with in¬ 
dividual component indices (Hyslop, 1980; Cortes, 1997; 
Liao et ai., 2001; Hart et aL, 2002), although others 
have noted that compound indices may exacerbate the 
error term and are affected by the taxonomic resolution 
of prey (Hyslop 1980; Hansson, 1998), and 2) to facili¬ 
tate comparisons with other studies of juvenile floun¬ 
der diet that have a similar approach (Burke, 1995; 
Carlson et ah, 1997; Grover, 1998; Zloch and Sapota, 
2010; Sagarese et ah, 2011). 
Lastly, each seine haul yielded a cluster of winter 
and summer flounder, and these individuals likely 
have increased similarities in diet relative to conspe- 
cifics sampled at different sites or dates (Bogstad et 
aL, 1995). As such, the aforementioned component and 
compound diet indices (by percentages) were recalcu¬ 
lated by using a cluster sampling estimator (Buckel et 
aL, 1999; Latour et aL, 2008), and these data were used 
in all subsequent analyses (e.g., hierarchical cluster 
and permutational multivariate analyses; see “Intra- 
specific dietary analysis” section). The cluster sampling 
estimator is represented as 
where qik 
%Zk 
M, 
Xj 
^ik 
'M' 
X100, (2) 
X; 
one of several diet indices {%F, %N, %V, or 
%IRI) for prey taxon k; 
the number of clusters (e.g., number of 
seine hauls containing winter or summer 
flounder); 
the number of winter or summer flounder 
collected from site i on a specific date; 
the total frequency, number, volume, or in¬ 
dex of relative importance of all prey in 
the stomachs of winter or summer floun¬ 
der collected from site i; and 
the total frequency, number, volume, or in¬ 
dex of relative importance of prey taxon 
k in flounder stomachs from site i. 
The variance estimate for %X]^ was calculated as 
var(%Xk) = • 
(3) 
where M = ■ —- and represents the average num¬ 
ber of winter oPcsummer flounder collected at site i. 
Intraspedfic dietary analysis 
Hierarchical cluster analyses of %IRI data were used 
to examine the diet composition of winter flounder and 
summer flounder as a function of body size (TL in mil¬ 
limeters). Winter and summer flounder used in the diet 
study ranged from 20 to 90 mm TL and from 19 to 
172 mm TL, respectively (preserved lengths). To assess 
the effect of fish size on diet, before cluster analyses, 
winter and summer flounder were grouped into 5-mm- 
TL and 10-mm-TL size-class intervals, respectively, and 
%IRI was recalculated with Equation 2 (i.e., 1 seine 
haul produced more than 1 cluster when multiple size 
classes were present). The statistical software pack¬ 
age PRIMER 7.0 (PRIMER-E Ltd., Plymouth, UK) was 
used to create resemblance matrices of the diet data. 
For each flounder species, diet data were first log- 
transformed (log[x-i-l]) to account for log-normal distri¬ 
butions (Latour et aL, 2008), and the Bray-Curtis index 
was then used to construct a similarity matrix. 
Cluster analyses were conducted on the resulting 
resemblance matrices by using a similarity profiling 
routine (SIMPROF), which defines statistically distinct 
groupings among samples (Clarke et aL, 2014). Den¬ 
drograms derived from cluster analyses were used to 
visually represent the dietary similarities among size 
classes of winter and summer flounder (group average 
method), and similarity percentage analyses (SIMPER) 
were used to identify the prey taxa that accounted for 
the dietary similarities or differences within or among 
groupings. Accordingly, the hierarchical cluster analy¬ 
ses yielded 4 distinct groups of winter and summer 
flounder (Fig. 3), corresponding to 4 broad size catego¬ 
ries. For winter flounder, the size categories were <39 
mm TL (small), 40-59 mm TL (small-medium), 60-79 
mm TL (medium-large), and >80 mm TL (large), and, 
for summer flounder, the categories were <59 mm TL 
(small), 60-79 mm TL (small-medium), 80-119 mm TL 
(medium-large), and >120 mm TL (large). 
Spatial (site) and temporal (monthly) variations in 
diet of winter and summer flounder within each riv¬ 
erine system were examined by using 2-way permuta¬ 
tional multivariate analysis of variance (PERMANOVA) 
models, as provided in the PRIMER 7.0 software pack¬ 
age (Anderson et aL, 2008). Bray-Curtis resemblance 
matrices of log(x+l)-transformed data were created by 
using the previously described methods; however, %IRI 
was recalculated from Equation 2 by grouping winter 
and summer flounder into their respective broad size 
categories (4 size categories for each species; Fig. 3). 
Therefore, each element in a resemblance matrix rep¬ 
resented the mean %IRI for winter or summer flounder 
as a function of its size category (small, small-medi¬ 
um, medium-large, and large), site (Seekonk River site 
1-4 [SR1-SR4] or Taunton River site 1-4 [TR1-TR4]; 
Fig. 1), and month (May, June, July, and August- 
-1 
