Abstract–Habitat selection theory assumes that individuals either know the quality of habitatsa priorior can learn it through experience. This work tests the effects of total food supply and of water velocity on the ability of drift‐feeding stream fish (Rhinicthys atratulus) to discriminate between two patches in a laboratory flow‐tank that differ only in the amount of food in the patch. If the fish can discriminate between the patches, their distributions should be predicted by ideal free distribution models; otherwise, their distributions should be random. The results show that at high total food supply levels, fish distributions differ from random in general agreement with ideal free distribution models; total food supply decreases, fish distributions become random, indicating that total food supply levels affect the ability of fish to discriminate between habitats that differ only in food supply. Increased water velocity was expected to affect fish distributions similarly, but its effect was not significant. The results suggest that habitat selection theory should consider both differences between habitats and the ability of fish to perceive those differ