Factors and processes affecting community structures operate at various spatial and temporal scales. We analyzed how similarities of rocky intertidal assemblages vary at different spatial scales using a nested, hierarchical sampling design. Intertidal assemblages consisting of algae, sessile animals, and mobile animals were censused on five rock walls at each of five shores chosen within each of six regions along the Pacific coast of Japan, encompassing 1,800 km of coastlines. Based on this sampling design, similarities in assemblages were calculated using both qualitative (presence/absence) and quantitative (percent cover and density) data, and compared at three spatial levels: (1) rock level (the finest spatial scale, encompassing several to hundreds of meters), (2) shore level (the intermediate spatial scale, encompassing several to tens of kilometers), and (3) region level (the broadest spatial scale, encompassing hundreds to thousands of kilometers). Cluster analysis showed that assemblages were categorized into distinct regional groups except for the two southern regions, but they did not separate clearly from each shore. A nested analysis of similarities revealed significant variation in similarities among regions and among shores within each region, with the former showing greater variation. Similarity was negatively correlated with geographic distance at the regional level but not at the shore or the rock levels. At the regional level, similarity decreased more rapidly with distance for mobile animals than sessile organisms. The analyses highlighted the importance of broad-scale abiotic/biotic factors such as oceanic current conditions and biogeographic factors in determining observed patterns in similarity of rocky intertidal assemblages.