There is growing awareness that patterns of habitat use by animals cannot be isolated from issues of scale. Recently, techniques have been devised which allow empirical testing of hypotheses related to the effects of spatial scale on habitat use. We used spatially explicit statistical procedures to examine the roles of scale and habitat arrangement in determining fish distribution patterns in a reach of an Australian lowland river. Native golden perch and introduced common carp were tracked day and night using radiotelemetry over a 10 day period and their distributions compared with mapped habitat variables. Golden perch were significantly associated with depth, current velocity, substratum and cover at larger scales of analysis both day and night, and at smaller scales during the day. At night, however, associations between golden perch and habitat variables were generally much weaker at small scales. Common carp were generally not significantly associated with habitat variables at larger scales of analysis, whereas at smaller scales they were associated with depth, current velocity, substrate and cover. Associations were generally stronger during the day than at night. Our study emphasises the need to consider scale-dependence in studies of fish-habitat associations. Management and restoration of fish habitat in lowland rivers should be based on an understanding of habitat use at scales relevant for the fish rather than at scales arbitrarily defined by humans.