We used eight microhabitat variables to examine the assumption that the variables normally included in instream flow studies are adequate to discriminate among species' microhabitats. When eight variables were available in stepwise-discriminant analysis models to distinguish among the microhabitats of four fish species in a northern California stream, the variance explained ranged from 52 to 77%, and 59 to 86% of the observational records were correctly classified to species. The variables measured were temperature, total depth, focal point elevation (distance of fish from the bottom), focal point velocity (water velocity at fish's snout), mean water column velocity, surface velocity, substrate, and cover. When the number of variables available was reduced to the three normally used in instream flow studies (i.e., total depth, mean water column velocity, and substrate), the variance explained ranged from 0 to 20%, and 46 to 55% of the observational records were correctly classified to species. When all eight variables were available, two variables not normally measured in instream flow studies, temperature and focal point elevation, were important in discriminating among species. Focal point elevation explained between 32 and 43% of the variance in the five models in which it was available. Temperature was included in 14 of 15 models in which it was available and made significant contributions in 12. Total depth was included and significant in 12 of the 20 models in which it was available. When total depth was included in a model, it was always more important than temperature; however, temperature and focal point elevation were the only significant variables on two sampling dates. Mean water column velocity and substrate only made minor contributions in a few of the 20 models in which they were available.