Most studies of coral reef fish communities have been restricted to site-attached species on small, isolated patches of habitat. Few have investigated spatial variation in fish species richness in relation to predictions based on stochastic or deterministic processes of community organisation. Our aims were to: (1) compare species richness on contiguous and fragmented reef habitats, and (2) investigate the mechanisms underlying spatial variation in species richness. Quantitative comparison of species-area curves for contiguous and patchy coral reef indicated that patch reefs support more species than equivalent areas of contiguous reef. However, Monte-Carlo simulated rarefaction curves indicated little difference in the species-individuals relationship for both habitats. Rarefaction was employed to eliminate variation in species richness among sites due to differences in sample size (number of fish present). After removal of sample-size effects, multiple regression models explained 30% and 25% of total variability in species richness on contiguous and patchy coral reef based on variation in habitat structure (e.g., depth, shelter availability, substratum characteristics). To investigate the likely importance of stochastic processes in determining spatial variation in species richness, we compared the species-individuals relationship from contiguous reef sites with the relationships derived from null models involving the random reallocation of fish among sites. Comparisons of the observed data with the outcomes of the null models indicated that spatial variation in species richness was not wholly attributable to stochastic processes. We suggest that the observed patterns of species richness may reflect species interactions (e.g., competition and predation) within fish communities.