Processes occurring at the end of the larval stage are of major importance in shaping spatial structure of fish assemblages in coral reefs. However, because of the difficulty in identifying larvae to species, many studies dealing with these stages are limited to the family level. It remains unknown if variation in the spatial structure of coral-reef fish assemblages across life stages can be detected at such a coarse taxonomic level. Two different surveys conducted in a similar area of New Caledonia, Southwest Pacific, provided the opportunity to compare the structure of coral-reef fish assemblages collected as pre-settlement larvae, juveniles and adults along a coast to barrier reef gradient. Adult and juvenile fish were sampled using underwater visual counts (UVC) during the warm seasons of 2004 and 2005. Pre-settlement larvae were sampled with light-traps during the same seasons. In order to standardize data between sampling methods, analyses were conducted on the relative abundance (for larvae) and density (for juveniles and adults) of 21 families commonly collected with both methods. Relative abundances/densities of families were analysed as a function of life stage (larvae, juveniles or adults), large-scale spatial location (coast, lagoon or barrier) and years (2004, 2005) using non-parametric multidimensional scaling (nMDS) and permutational multivariate analysis of variance (permanova). Kruskal–Wallis tests were then used to examine differences among life stages and locations for individual families. Different levels of spatial and temporal variability characterized fish assemblages from different life stages, and differences among life stages were detected at all locations and years. Differences among life stages were also significant at the level of individual families. Overall results indicate that studies conducted at the family level may efficiently reveal changes in coral-reef fish spatial structure among successive life stages when large spatial scales are considered.