Disturbances over a range of intensities are common in intertidal communities, often causing partial removal of dominant species. However, most studies testing effects of disturbance on community structure use treatments where the dominant species is either present or absent, and usually record responses of only a few community components. Here, we test responses of the entire local community of benthic algae and invertebrates across a gradient of distur- bance to a habitat dominant in southern New Zealand, the fucoid alga Hormosira banksii. Replicate 0.25 m 2 plots of 100% cover of H. banksii were manipulated and maintained at 0, 25, 50 and 75% cover at 2 sites for 13 mo. Throughout the experiment, communities in the complete removal and control plots were the most different from each other, but intermediate plots had varied responses and were usually not significantly different from each other. Generally, as cover of H. banksii decreased, community variability increased through time. Taxa richness declined with reduced H. banksii cover, with up to 77% of its variation explained by variation in canopy cover. Other fucoid algae declined across the gradient, with complete canopy removals having 92% less fucoid cover than control plots. Complete canopy removal had positive effects on ephemeral algae, especially at one site which exhibited large seasonal blooms. Bare space was greatest where canopies were completely removed and understory algae died. Diversity was greatest in plots with complete canopies rather than at intermedi- ate levels, and there was no replacement of the domi- nant H. banksii by other species. There was no consis- tent effect on taxa across the gradient; some species responded linearly while others responded at thresh- old levels of canopy cover. This study shows the impor- tance of disturbance gradients interacting with key species, which may have considerable bearing on reef community structure.