Habitat complexity in temperate intertidal rockpools was manipulated to assess its effect on fish assemblage structure. We modified the complexity of the substratum within pools by adding or removing loose rocks. The complexity of the water column within pools was modified by the removal of macro-algae and the addition of artificial algae. We applied 5 treatments to 40 pools in 2 locations on the southeast Australian coast: addition of rocks and algae (+R+A), removal of rocks and algae (–R–A), removal of rocks but addition of algae (–R+A), addition of rocks and removal of algae (+R–A), and unmanipulated control pools (C). Changes in algal cover alone had no significant effect on fish assemblages. The total number of individual fishes, species richness, species composition and abundance of the 4 most abundant species only differed in the –R–A treatment. There was no change in the length-frequency distributions of the most abundant species after habitat manipulations. The effects of manipulating substratum habitat complexity were investigated further by reversing the treatments in the same pools, i.e. by adding rocks to pools where rock had previously been removed and vice versa. This resulted in a significant increase in species richness and the numbers of individuals in rockpools where substratum heterogeneity had been increased, relative to pools in which it had been decreased. This study revealed that rockpool fishes do not discriminate between the structural complexity within these 2 microhabitats (substratum and water column) so long as some shelter is available. However, substratum complexity may represent the most attractive shelter, since most species are benthic or at least demersal. We speculated that the lack of habitat specificity exhibited by fishes in this study may be due to many rockpool fishes only using rockpools as temporary refugia at low tide.