These measures were computed on data from pooled replicates of suction 
samples at each level since pooling the replicates provided a larger 
sample size and a more representative estimate of community diversity 
at a site. Diversity of the sessile biota which generally could not be 
counted was limited to comparisons of the number of species (s) observed 
in photographs and along line transects. 
Cluster analysis was used to determine patterns of similarity among 
stations. The quantitative measure used in all analyses was the Bray- 
Curtis coefficient (Boesch, 1977): 
estat ey lige 
Ex + xy) 
L 
where X44 and X,4 are the number of individuals of the gen species in 
two collections under comparison. A normal analysis was completed on the 
site groups using modified data sets and a flexible sorting strategy with 
a standard 8 value of -0.25. Data sets represented pooled collections 
from the different levels at a site (station), separated by seasons. 
Additional modifications to the data sets included log transformation 
and deletion of taxa which occurred in only one collection, as well as 
deletion of those taxa of uncertain identity. These deletions were made 
to simplify the data sets and because "rare" species usually do not have 
definable distribution patterns, and can confuse interpretation of cluster 
analysis. 
Quantification techniques for food habits of fish are biased, depend- 
ing on the method (Hynes, 1950; Pinkas et al., 1971; Windell, 1971). 
Therefore, the relative contribution of different food items to the 
total diet was determined using three methods: percent frequency occurrence 
(F), percent numerical abundance (N), and percent volume displacement 
(V). From these, an index of relative importance (IRI) (Pinkas et al., 
1971) was calculated for each prey species and higher taxon as follows: 
IRI = (N+ V) F 
where N, V and F are the numerical, volumetric, and frequency percentages 
as defined above. This index has proven useful in evaluating the 
relative importance of different food items found in fish stomachs 
(Pinkas et al., 1971; McEachran et al., 1976; Sedberry, 1983) and was 
used in the present study to describe the food habits of each species. 
15 
