The ichtyofauna of Geyser and Zelee coral banks (north of Mozambique canal) was investigated in 1996, this work being the first quantitative study on this area. Quantitative data were collected on 11 stations (mean depth 18 m) by visual observations along 50 m transect, widened to 2 m for small sedentary species, and to 5 m for commercial species with a large action range. The main descriptors of the fish communities (species richness, abundance per species) and the trophic structure (omnivorous, herbivorous, carnivorous, browsers of sessile invertebrates) were described and analysed.A total of 294 fish species, belonging to 43 families were recorded. The species richness varied from 45 to 83 species / station, and the abundance, from 292 to 1322 individuals / 100 m(2), the lower values being recorded on the inner slope and the higher ones on the outer slope. In terms of number of individuals, diurnal planktivores were highly represented at all sites (mean 63.7%). Pomacentridae and Caesionidae were the most abundant families, damselfishes being mainly represented by Chromis ternatensis (mean 73 ind., max, 148 ind. 100 m(2)), C. dimidiata (mean 33 ind., max. 151 ind. / 100 m(2)) and Dascyllus carneus (mean 38 ind., max. 92 ind. 100 m(2)), and fusiliers by Pterocaesio chrysozona (mean 19 ind., max. 128 ind. / 100 m(2)), P. tile (mean 12 ind., max. 38 ind. / 100 m(2)) and Caesio teres (mean 6 ind., max. 32 ind. / 100 m(2)). The high values recorded showed that small and isolated coral banks can support relatively rich fish faunas and substantial populations of individuals. This result may be explained by different factors such as the oceanographic circulation favourable to a localised enrichment of nutrients from surface water, the low human impact, a good coral cover and a high degree of self-recruitment associated with regional recruitment. Nevertheless, the main commercial species such as Lutjanus bohar and Variola louti, observed at each station, were represented at a low density (mean 0.8 ind., max. 6 ind. / 100 m(2) and mean 0.2 ind., max. 0.8 ind. / 100 m(2) respectively). This result may be due to the exploitation of these coral banks between 1990 and 1992, which was interrupted following the fast decreasing output.Such an experience shows us the fragility of these small coral banks and the necessity to establish a rational management of the resource.