We investigated the community-wide effects of no-take marine reserves at the regional scale in the Mediterranean Sea. Reserves (Cabo de Gata, Cabo de Palos, Cabrera , Medes, Banyuls-sur-Mer, Portofino, Cinque Terre , Scandola, Ustica Island , Torre Guaceto, Tremiti Islands, Miramare) had general positive effects of protection on fish targeted by fishing, and variable effects on non-target species. Differences in the structure of fish assemblages (in terms of abundance of species and trophic groups) were greater between geographic locations than when compared among all protected and adjacent fished areas at the regional scale. These results suggest that the assemblage-wide response to protection at the species level may be spatially idiosyncratic, as a function of local factors. However, the functional (trophic group) response of the fish assemblages to protection appears to be more general. Response of fish assemblages to protection was significantly related to reserve age (i.e. duration of protection) only when evaluated at functional level, whereas reserve size did not appear to influence fish assemblages at either species or functional level. Reserves can enhance trophic cascades and ecological shifts once the density of fish predators of sea urchins reaches a density threshold of about 15 adult sea breams (Diplodus sargus and D. vulgaris, the most important sea urchin predators) per 100 m2. These non linearities in the community-wide effect of reserves suggest that caution is needed in simplistically treating reserves and unprotected areas as 2 experimental treatments in ecological studies.