1. Understanding how animals interact with their environment is a fundamental ecological question with important implications for conservation and management. The relationships between animals and their habitat, however, can be scale dependent. If ecologists work at suboptimal spatial scales, they will gain an incomplete picture of how animals respond to the landscape. Identifying the scale at which animal-landscape relationships are strongest (the ‘scale of effect’) will improve our ability to better plan management and conservation activities. 2. Several recent studies have greatly enhanced our knowledge about the scale of effect, and the potential drivers of inter-specific variability, in particular life-history traits. However, while many marine systems are inherently multi-scalar, research into the scale of effect has been mainly focussed on terrestrial taxa. As the scales of observation in fish-habitat association studies are often selected based on convention rather than biological reasoning, they may provide an incomplete picture of the scales where these associations are strongest. 3. We examined fish-habitat associations across four nested spatial scales in a temperate reef system to ask: (1.) at what scale are fish-habitat associations the strongest, (2.) are habitat elements consistently important across scales, and (3.) do scale-dependent fish-habitat associations vary in relation to either body size, geographic range size or trophic level? 4. We found that: (1.) the strongest fish-habitat associations were observed when these relationships were examined at considerably larger spatial scales than usually investigated; (2.) the importance of environmental predictors varied across spatial scales, indicating that conclusions about the importance of habitat elements will depend on the scales at which studies are undertaken; and (3.) scale-dependent fish-habitat associations were consistent across all life-history traits. 5. Our results highlight the importance of considering how animals relate to their environment and suggest the small scales often chosen to examine fish-habitat associations are likely to be suboptimal. Developing a more mechanistic understanding of animal-habitat associations will greatly aid in predicting and managing responses to future anthropogenic disturbances.