Predictive models such as the River InVertebrate Prediction And Classification System (RIVPACS) and AUStralian RIver Assessment System (AUSRIVAS) model the natural variation in macroinvertebrate communities across geographic regions by comparing observed (O) taxa to those expected (E) to occur in the absence of anthropogenic stress.  Benthic macroinvertebrate data collected from 85 reference sites of the Western Allegheny Plateau (WAP) by Ohio EPA between 1982 and 2005 was used to develop and validate a RIVPACS-type model to assess the biological condition of wadeable streams in this region.   Cluster analysis grouped the reference sites into groups based on biological similarity.  Multiple discriminant function analysis (DFA) determined which environmental variables most accurately predicted group membership.   The model discriminated groups based on hydrologic, geographic, and in-stream habitat variables.  DFA correctly classified 92% of model reference sites and O/E ratios were normally distributed for both model (standard deviation = 0.20) and validation (standard deviation = 0.12) sites.  The model was subsequently applied to Ohio EPA non-reference (impaired) site data within the WAP for possible use as a regional bioassessment tool.