Results from the field assay identified larger Thalassoma quinquevittatum as strongly competing with focal T. hardwicke. A factorial experiment was used to evaluate the effect of competition with larger T. quinquevittatum and the provisioning of structural refuge on the survival of transplanted T. hardwicke. Structural refuge was increased by adding Turbinaria ornata (a structurally complex, locally common, macroalga) to reefs. Twenty reefs were haphazardly selected, with an average surface area of 5.46 m2 (SE = 0.36), a mean height of 0.79 m (SE = 0.03) and isolated from their nearest neighbour by 3.02 m (SE = 0.19). All reefs initially had three to five (mean = 3.38; SE = 0.16) resident Thalassoma quinquevittatum (22.1 mm SL, SE = 0.56) and lacked Turbinaria ornata. The presence/absence of T. ornata was cross-factored with the presence/absence of T. quinquevittatum and treatments were randomly assigned to reefs. On reefs assigned to the presence of T. quinquevittatum, selective removals were conducted so that each reef had three, similarly sized, T. quinquevittatum individuals that were larger than focal T. hardwicke individuals. For T. ornata present treatments, whole T. ornata plants were attached to recipient reefs with cable ties, so that ~15% of habitat cover was T. ornata. Before the experiment, all resident T. hardwicke were removed from the reefs. Three Thalassoma hardwicke individuals (13.69 mm SL, SE = 0.51: collected and tagged as per the field assay) were transplanted onto each reef and their survival surveyed daily (~ 9 a.m.) for five days. Each morning, immigrant T. quinquevittatum (three immigrants were observed) and new T. hardwicke that had settled the previous night (a total of eight settlers) were removed. Neighboring non-experimental reefs were also searched for tagged T. hardwicke immigrants. No tagged immigrants were found; therefore, the disappearance of a tagged fish was assumed due to mortality rather than migration (as for the field assay). Using the same reefs, the experiment was run in two temporal blocks (beginning12 June 2005 and 12 May 2007), yielding ten replicates (five in each temporal block) for each of the four treatments. The mean proportion of T. hardwicke remaining in each treatment on the last day of the experiment was used as the response variable because by the end of each experimental run survival trajectories had stabilized.