Predation is an important and ubiquitous selective force that can shape habitat preferences of prey species, but tests of alternative mechanistic hypotheses of habitat influences on predation risk are lacking. 2. We studied predation risk at nest sites of a passerine bird and tested two hypotheses based on theories of predator foraging behaviour. The total-foliage hypothesis predicts that predation will decline in areas of greater overall vegetation density by impeding cues for detection by predators. The potential-prey-site hypothesis predicts that predation decreases where predators must search more unoccupied potential nest sites. 3. Both observational data and results from a habitat manipulation provided clear support for the potential-prey-site hypothesis and rejection of the total-foliage hypothesis. Birds chose nest patches containing both greater total foliage and potential nest site density (which were correlated in their abundance) than at random sites, yet only potential nest site density significantly influenced nest predation risk. 4. Our results therefore provided a clear and rare example of adaptive nest site selection that would have been missed had structural complexity or total vegetation density been considered alone. 5. Our results also demonstrated that interactions between predator foraging success and habitat structure can be more complex than simple impedance or occlusion by vegetation. ?? 2008 British Ecological Society.