Three theories of habitat use proposed for marine fishes -the constant density model, the proportional density model, and the basin model -make contrasting predictions of how the geographical range, local density, and fitness change as population size changes. We tested model predictions with survey data on yellowtail flounder Limanda ferruginea from the Georges Bank region, where abundance changed by a factor of 4 over a decade. Surveys took place in spring and fall, and data on individual length, mass, sex, and reproductive status were available. Analysis of spatial pattern revealed that the overall area occupied by flounder increased by a factor of 2 when abundance was high, and local density increased predominantly in high quality habitat that had been closed to commercial fishing. Condition, which served as a proxy for fitness, was lower in females when abundance was high. Geospatial analysis revealed mesoscale variability in condition, over 10s to >100 km, except in the spring season during low abundance periods. Spatial autocorrelation explained as much as 25% of the variability in condition, indicating that site dependence was a factor in explaining the spatial distribution that we observed. These results are most supportive of both the constant density model and the basin model. This approach detected an important population center for yellowtail flounder and determined its extent using only measures of abundance, location, and condition of individual fish, data commonly collected during routine fishery assessment surveys. Here we demonstrate that analyses linking population responses to variation in such measures at local spatial scales can have significant implications for identifying areas of important fish habitat and suggest greater use of geospatial approaches in conservation and management of exploited species.