Predation can be affected by habitat structure, e.g., by reducing the foraging efficiency of predators by providing refuge space ). Based on differences in the use of habitats as a refuge against predators, habitat structure may provide axes for niche diversification among prey, even if there is no competition for resources .
In coral reefs, crustaceans have been identified as the largest component of the most abundant reef fishes’ diet . Hiatt and Strasburg  verified that crustaceans associated with corals are consumed by many species of reef fishes, especially when they move between coral colonies . However, when crustaceans are protected by a suitable refuge space within their coral host, fish are usually unable to predate them .
In the present study, the significant differences in richness and density patterns of crustaceans associated with Mussismilia corals between the treatments (total cage vs. partial cage, and total cage vs. no cage treatments) is evidence of the importance of predation in structuring the associated community. The absence of significant differences between partial cage and no cage treatments indicates there was no artefact influence of the structures used to construct exclusion cages. The exclusion of predators is known to impact communities by increasing their density and richness . Our results also show that once predators were excluded in caged M. braziliensis and M. hispida, the richness and abundance of crustaceans grew to values close to those found in M. harttii, which is a species structurally more complex and provides additional refuge against predators.
Of the three Mussismilia species, M. harttii has been previously identified to harbour the richest and most abundant fauna . This result was confirmed in the present study by comparison of Mussismilia species in uncaged and partial cage treatments: there was a significant difference in the richness and abundance of associated carcinofauna among species of Mussismilia corals.
Higher richness and densities associated with M. harttii in uncaged and partial cage treatments, compared with M. braziliensis and M. hispida, indicates that the partial cages allowed the predators to freely access the colonies. Nogueira et al.  identified the space among corallites, an exclusive feature of M. harttii, as the most important factor influencing the richness and abundance of associated fauna. The space among corallites acts as a refuge against predators, which is confirmed by the results obtained from caged corals, in which no statistical difference was observed for richness and density once the predators had been excluded.
Our results for the most abundant copepods and ostracods showed higher mean values of density in caged colonies of M. braziliensis. This indicates that copepods and ostracods associated with uncaged colonies and partial cages of M. braziliensis were affected by predators feeding preferentially on the most abundant species, reducing their population size. The habitat structure of M. braziliensis may provide other benefits beyond protection against predators, e.g., easier access to resources, allowing the growth of populations of small animals like copepods and ostracods. In normal conditions (uncaged), those benefits are suppressed by the harshness of predation.
For decapods, we found higher densities associated with caged M. harttii, compared with caged M. braziliensis and M. hispida. As found by Nogueira et al. , Mithraculus forceps was the most abundant decapod species, with higher values associated with M. harttii. The association of this crab with corals was described as mutualistic by Stachowicz and Hay . According to these authors, the crab obtains food and protection by living among the coral branches and benefits the coral host by preventing the overgrowth of algae.
The most abundant medium-sized species (2–10 mm), the peracarids Cheiriphotes megacheles and Carpias sp. , displayed the same pattern as the most abundant decapods, however, this could be due to cage artefacts. The mesh size used to construct cage artefacts are much smaller than most crabs species, thus, once cages were placed around coral species, crabs associated with the corals remained confined and others could not access the cage interiors. Consequently, the number of species and individuals associated with corals remained the same as at the beginning of the experiment. This is corroborated by the higher values recorded for small crustaceans, e.g., copepods (0.1–1.2 mm)  of caged M. braziliensis colonies.
Results of SIMPER showed high dissimilarities in the crustacean assemblage between exclusion cage treatments for all three Mussismilia species; even with differences in richness and density among cage treatments, the composition of crustaceans was very variable, which may have produced the high dissimilarities found.