ABSTRACT: This study describes a method for modeling and predicting, from biological and physical variables, habitat use by a commercially harvested groundfish species. Models for eastern Bering Sea flathead sole Hippoglossoides elassodon were developed from 3 relationships describing the response of organism abundance along a resource continua. The model was parameterized for 1998 to 2000 trawl survey data and tested on 2001 and 2002 data. Catch per unit effort (CPUE) of flathead sole had a curvilinear relationship with depth, peaking at 140 m, a proportional relationship with bottom water temperature, a positive curvilinear relationship with potential cover (invertebrate sheltering organisms such as anemones, corals, sponges, etc.), a negative relationship with increasing mud:sand ratio in the sediment, and an asymptotic relationship with potential prey abundance. The predicted CPUE was highly correlated (r² = 0.63) to the observations (1998 to 2000) and the model accurately predicted CPUE (r² = 0.58) in the test data set (2001 and 2002). Because this method of developing habitat-based abundance models is founded on ecological relationships, it should be more robust for predicting fish distributions than statistically based models. Thus, the model can be used to examine the consequences of fishing activity (e.g. reduction in sheltering organisms), changes in temperature (e.g. climate effects) and interaction between variables, and can be modified to incorporate new variables as more information is collected about a species.