Ecosystems are shaped by processes occurring and interacting over multiple temporal and spatial scales. Theory suggests such complexity can be simplified by focusing on processes sharing the same scale as the pattern of interest. This scale‐dependent approach to studying communities has been challenged by multiscale meta‐ecosystem theory, which recognizes that systems are interconnected by the movement of “ecological subsidies” and suggests that cross‐scale feedbacks between local and regional processes can be equally important for understanding community structure. We reconcile these two perspectives by developing and testing a hierarchical meta‐ecosystem model. The model predicts local community responses to connectivity over multiple oceanographic spatial scales, defined as macro‐ (100s of km), meso‐ (10s of km), and local scale (100s of m). It assumes that local communities occur in distinct regions and that connectivity effects are strongest among local sites. Predictions are that if macroscale processes dominate, then regardless of mesoscale differences, (1) local communities will be similar, and (2) will be even more so with increased connectivity. With dominance of mesoscale (i.e., regional) processes, (3) local structure will be similar within but distinct between regions, and (4) with increased connectivity similar both within and among regions. With dominance of local‐scale processes, (5) local communities will differ both within and among regions, and (6) with increased connectivity be similar within but not between regions. We tested the model by evaluating rocky intertidal community structure patterns with variation in ecological subsidies and environmental conditions at 13 sites spanning 725 km of the northern California Current system. External factors operating at meso‐ and local scales had strong effects, explaining 52% and 27% of the variance, respectively, in community structure. Sessile invertebrate and predator dominance was associated with weaker upwelling, higher phytoplankton abundance, and higher recruitment, and the opposite was true for macrophyte dominance. Overall, our results support the theory that meta‐ecosystems are organized hierarchically, with environmental processes dominating at meso‐ to macroscales and ecological processes playing a more important role at local scales, but with important bidirectional cross‐scale interactions.