Patterns of distribution of a Mediterranean fish assemblage, with special attention to juvenile labrids, were studied at different depths on the basis of their relationships with environmental and benthic habitat features on a rocky coast at Marseille, south-eastern France. Fish species were distributed according to depth, with juveniles and small individuals being found in shallow waters, whereas larger individuals were found in deeper waters. However, depth by itself was not a significant factor in explaining fish species pattern, whereas sandy bottom habitat was significant. The most significant factors explaining fish species distribution were small and large blocks, and rock cover. Juvenile labrids, such as the most abundant ones Coris julis and Symphodus ocellatus, were only rarely observed deeper than 10 m, and the distribution of these juveniles was mainly correlated with a cover of macroalgae, rubble and small blocks. Fish density and fish biomass were higher in shallow than in deeper waters, although some species displayed an opposite pattern, such as the adult of C. julis and Diplodus sargus. The most significant factors explaining the fish density pattern found were the presence of macroalgae, Posidonia oceanica sea grass and sand cover. Among environmental factors assessed in this study, exposure to dominant winds was never significant, either for the whole fish assemblage or for juvenile labrids, or for the communities at any of the depth strata investigated. Depth was a significant factor influencing the distribution of juvenile labrids, whereas its significance in explaining fish density was lower. Bottom slope was generally more significant than depth, and thus constituted the most significant environmental factor. Benthic habitat characteristics were all significant, except for rubble cover, in explaining both fish species and fish density distributions. However, the amplitude of the relationships between substrate variables and fish assemblages was dependent on the descriptor studied (i.e. species distribution, including considerations of fish sizes, and fish density), suggesting that extrapolation of a particular relationship to other components of a fish assemblage would most likely be unfounded. Our results as a whole indicate that benthic habitat use by fish on Mediterranean rocky infralittoral coast is clearly non-random, and suggest that differences in habitat and/or environmental characteristics have significant effects at a small spatial scale on these fish assemblages.