The extent to which a landscape is fragmented affects persistence of predator-prey dynamics. Increasing fragmentation concomitantly imposes conditions that stabilise and destabilise metapopulations. For the first time, we explicitly assessed the hypothesis that intermediate levels provide optimal conditions for stability. We examine four structural changes arising from increased fragmentation: increased fragment number; decreased fragment size; increased connectedness (corridors scaled to fragment); increased fragment heterogeneity (based on connectedness). Using the model predator-prey system (Didinium-Paramecium) we support our hypothesis, by examining replicated metapopulations dynamics at five fragmentation levels. Although both species became extinct without fragmentation, prey survived at low and high levels, and both survived at intermediate levels. By examining time to extinction, maximum abundances, and population asynchrony we conclude that fragmentation produces structural heterogeneity (independent of environmental heterogeneity), which influences stability. Our analysis suggests why some theoretical, field and microcosm studies present conflicting views of fragmentation effects on population persistence.