Understanding the influence of geographic factors in structuring fish assem
blages is crucial to developing a comprehensive assessment of stream condit
ions. We compared the classification strengths (CS) of geographic groups (e
coregions and catchments), stream order, and groups based on cluster analys
is of fish assemblage data from 200 wadeable streams in the Mid-Atlantic Hi
ghlands. We 1st calculated intersite similarity indices (Bray-Curtis on rel
ative abundance; Dice-Sorensen on presence/absence), then compared overall
mean within-group similarities ((W) over bar) with among-group similarities
((B) over bar). We used subsets of the data to test CS from: 1) 31 referen
ce sites defined on chemistry and habitat criteria, and 2) 21 samples from
8 sites that had been resampled within and between years to estimate the ma
ximum similarity expected for any of the classifications. We assessed the s
trength of each classification by determining the degree to which (W) over
bar was greater than (B) over bar. Sites classified by taxonomic clusters h
ad higher CS than did sites grouped by stream order, US Geological Survey 4
-digit Hydrologic Unit Code (HUC) catchments, and ecoregions. Except for ta
xonomic clusters, the CS values were greater when all sites were used in th
e analysis than when only reference sites were used. The mean similarities
for the revisits were 2 to 3 times greater than for all other classificatio
ns. We used nonmetric multidimensional scaling as an alternative approach t
o detecting geographic structure in the data. We found little separation of
ecoregion or catchment groups except at very broad spatial scales. The rel
atively weak CS of any of the geographic groups suggests that the interacti
on of complex zoogeographic patterns and a long history of human disturbanc
e has masked any fine-scale structure of regional fish assemblages.