The relative importance of extrinsic and intrinsic causes is among the oldest unresolved problems in ecology. An important source of intrinsic biological variability has been discovered in the form of large-scale spatial patterns emerging from local biotic processes. However, the importance of such intrinsic variability in heterogeneous environments is still unknown. Here I use a simple metapopulation and metacommunity models to resolve the interaction between scales of environmental heterogeneity, biotic processes and of large-scale intrinsic variability. I adopt a general disturbance-recovery system previously applied to marine intertidal and tropical forest dynamics to predict large-scale spatial patterns in homogeneous environments. I report how population density increases with environmental variability only when its scale matches that of intrinsic patterns of abundance. This result is explained by the ability of populations to develop such intrinsic patterns when exposed to heterogeneous environments. The strength of local species interactions, through its control of intrinsic variability, can in turn control metacommunity response to environmental scales, and is used to clarify the role of intrinsic variability. Our results generate testable predictions, showing how the environment and species distribution might fail to show any correlation despite their strong causal association. Results more generally suggest that the spatial scale of ecological processes might not be sufficient to build a predictive framework for spatially heterogeneous habitats. This study constitutes a first step toward theories of large-scale variability in coastal environments with strong heterogeneities, including those resulting from marine reserve networks.