From July 1988 to September 1992, Scuba divers used a non-destructive sampling method to study the fish communities of rocks and Posidonia oceanica seagrass beds, in Scandola Marine Reserve. Sites were established in the integral reserve (protection of the marine environment), in the non-integral reserve (partial protection) and outside the reserve (no protection). Average density and biomass of the sampled population in P. oceanica seagrass beds showed no significant difference between the sites in the integral reserve and the others. Only the integrally-protected shallow and deep water zones showed considerably reduced seasonal variations. For rocky substrata, the integral reserve site demonstrated higher density and biomass than the other sites. The impact of seasonal variations on the sampled population was not reduced in this biotope, as in the seagrass beds. Whatever the biotope, the demographic structures of the fish population sampled in the integral reserve and in the other sites were different: the number of large fishes (matured adults) was always the highest. Species diversity was slightly greater in the seagrass bed sites of the integral reserve, but site variations were not significant. This was also true for rocky substrata; in contrast, lesser seasonal differences were recorded in the integral reserve than outside. Three conclusions can be drawn from these pluriannual observations: 1) The reserve effect is real, not just theoretical, and affects the fish population in both rocks and seagrass beds. 2) The two sites studied, the Posidonia oceanica seagrass beds and rocky substrata, react very differently to protection which reduces anthropogenic disturbances. These differences may be attributed to the degree of sensitivity of the dominant families (Labridae in the seagrass, Sparidae in rocky substrata) to all disturbances. 3) The reserve effect comprises both a refuge and a buffer effect. The refuge effect is the most traditionally highlighted and is characterized by more large fishes and a greater variety of species and noble fishes (Dicentrarchus labrax, Sparus aurata, Epinephelus marginatus, Sciaena umbra). The buffer effect, never shown in previous studies, is characterized by the fact that the impact of seasonal variations on parameters such as density, biomass, and diversity is considerably lessened.