The light–dark preference test has been extensively used to screen anxiolytic drugs and investigate mechanisms of anxiety with rodents. Recently, this task has been adapted to zebrafish, but a number of inconsistent findings have emerged. For example, some found zebrafish to avoid and others to prefer dark. Given the translational relevance of the zebrafish, its utility in high throughput drug screens, and that anxiety still represents a large unmet medical need, there is an urgent need to resolve these inconsistencies. We propose these inconsistencies are due to lack of distinction between two separate factors: background shade and level of illumination. Here, we systematically manipulated background shade (black vs. white) while keeping the illumination level constant (uniformly illuminated). We also manipulated the level of illumination (illuminated vs. not illuminated) while keeping the background constant (either uniformly black or white). We examined the time-course of numerous behavioural responses under these conditions, and found zebrafish to exhibit a significant preference for the black side of the tank within the first 3 min of the test when the illumination level was constant. We found this response, along with other anxiety-like behaviours, to diminish over time. In contrast, we found zebrafish did not exhibit a preference for the unilluminated (dark) side of the tank when the background shade was kept constant. Our results demonstrate a dissociation between illumination level and background shade, shown by a preference for black to white, but not for dark to light, confirming the importance of differentiating these two distinct factors.