Marine protected areas (MPAs) are increasingly used as a management tool to preserve species and habitats. Testing hypotheses about the effectiveness of MPAs is important for their implementation and to identify informative criteria to support management decisions. This study tested the general proposition that MPAs affected assemblages of algae and invertebrates between 0.0 and 0.5 m above the mean low water level of rocky coasts on two islands in the Tuscan Archipelago (northwest Mediterranean). Protection was concentrated mainly on the west coasts of the islands, raising the possibility that neither the full range of assemblages nor the relevant scales of variation were properly represented within MPAs. This motivated the comparison of assemblages on opposite sides of islands (habitats). The effects of MPAs and habitat were assessed with a multifactorial sampling design; hypotheses were tested about differences in structure of assemblages, in mean abundance of common taxa and in univariate and multivariate measures of spatial variation. The design consisted of three replicate shores for each condition of protected and reference areas on the west side of each island and three unprotected shores on the eastern side. Assemblages were sampled independently four times on each island between June 1999 and January 2001. At each time of sampling two sites were selected randomly at each of two tidal heights to represent midshore and lowshore assemblages on each shore. Estimates of abundance were obtained using nondestructive sampling methods from five replicate 2020 cm quadrats at each site. Results indicated differences among habitats in structure of assemblages, in mean abundance of common taxa and in univariate and multivariate measures of spatial variation at the scale of shores. Most of these patterns were inconsistent with the predicted effect of management through MPAs. The data suggest that designation of MPAs in the Tuscan Archipelago should proceed through management of multiple shores and types of habitat selected to guarantee protection to a representative sample of assemblages and to the processes responsible for maintenance of spatial patchiness at different scales. This study also shows that considerations of spatial heterogeneity are important to underpin management decisions about the number, size and location of MPAs.