Spatially complex habitats provide refuge for prey and mediate many predator-prey interactions. Increasing anthropogenic pressures are eroding such habitats, reducing their complexity and potentially altering ecosystem stability on a global scale. Yet, we have only a rudimentary understanding of how structurally complex habitats create ecological refuges for most ecosystems. Better informed management decisions require an understanding of the mechanisms underpinning the provision of physical refuge and this may be linked to prey size, predator size and predator identity in priority habitats. We tested each of these factors empirically in a model biogenic reef system. Specifically, we tested whether mortality rates of blue mussels (Mytilus edulis) of different sizes differed among: (i) different forms of reef structural distribution (represented as 'clumped', 'patchy' and 'sparse'); (ii) predator species identity (shore crab, Carcinus maenas and starfish, Asterias rubens); and (iii) predator size. The survival rate of small mussels was greatest in the clumped experimental habitat and larger predators generally consumed more prey regardless of the structural organisation of treatment. Small mussels were protected from larger A. rubens but not from larger C. maenas in the clumped habitats. The distribution pattern of structural objects, therefore, may be considered a useful proxy for reef complexity when assessing predator-prey interactions, and optimal organisations should be considered based on both prey and predator sizes. These findings are essential to understand ecological processes underpinning predation rates in structurally complex habitats and to inform future restoration and ecological engineering practices.