Detection patterns of coral reef fish were assessed from the meta-analysis of distance sampling surveys performed by visual census in New Caledonia and French Polynesia, from 1986 to 1999. From approximately 100,000 observations relating to 593 species, the frequency distributions of fish detection distances perpendicular to the transect line were compared according to species characteristics and sampling conditions. The shape and extension of these detection profiles varied markedly with fish size, shyness, and crypticity, indicating strong differences of detectability across species. Detection of very small and cryptic fish decreased strongly 1 m away from the line. Conversely, sightings of shy and large species were excessively low in the first meters due to diver avoidance prior to detection. The larger the fish, the greater the fleeing distance. Distance data underscore how inconsistent detectability biases across species and sites can affect the accuracy of visual censuses when assessing coral reef fish populations.
