Marine protected areas (MPAs) are aimed at managing and protecting marine environments.Their design, however, often disregards both a thorough knowledge of the distribution of habitats and assemblages and the use of proper experimental evaluations of the efficacy of MPAs by comparing protected vs. unprotected zones. About 200 MPAs have been recently instituted in the Mediterranean area, but the evidence of their efficacy is scant. The MPA of Torre Guaceto (Southern Adriatic Sea, Italy) is one of the rare cases of effective protection enforcement. The reserve was instituted more than 10 years ago, a period currently considered as sufficient to show responses by organisms to protection. The MPA is divided into a C zone, the general reserve, where many activities are permitted, a B zone, the partial reserve where restrictions increase and two A zones, the integral reserve where access is prohibited. The goals of the paper were to map the distribution of benthic assemblages to assess if they were properly represented in the differently protected zones, and to test the efficacy of protection by quantifying possible differences between the assemblages in two control areas and in the two A zones, where human impact is completely excluded. The analysis of habitat and assemblage distribution within the MPA showed that the zones with total protection do not include most valuable environmental types. Most of the considered variables (i.e. cover of substratum, number of taxa, and average abundance of the most common taxa) were not significantly different in and out of the A zones, at each time of sampling. Results, however, suggested a possible effect of protection in modifying patterns of abundance of sponges under Cystoseira canopy (more abundant in the fully protected zone). In the subtidal habitat, differences were found in the structure of the whole assemblage and in the abundance of encrusting coralline red algae (more abundant outside the fully protected area). Notwithstanding the correct general methodology employed in the study, a lack of statistical power could have a role in preventing the detection of ecologically relevant effects of protection. In some instances, data pooling allowed a discrimination between cases where there was clearly no effect of protection and cases where there might be. On this basis, the optimization of this experimental design should be considered in further studies. In any case, if the goals of MPAs have not been clearly stated, efficacy of protection might prove very difficult to test even with the use of sound experimental designs.