The densities of recruits on caged and uncaged areas were compared in an experiment done to show the extent of predation on recently metamorphosed coral reef fishes. The design was unlike typical caging experiments, however, in that areas were caged only for short periods of 20–30 days and several independent trials, testing the same null hypothesis, were run. This was done to avoid confounding the effects of excluding herbivorous fishes with the effects of excluding piscivorous fishes. A third treatment, partially-meshed cages, revealed that the experiment was complicated by several other factors. Some prey species were attracted to the high relief offered by the experimental structures. Others responded to the differences in shelter from predators by redispersing themselves among the treatments shortly after settlement. There was also at least one significant “edge” effect caused by fishes preferring to settle near the boundaries of all treatments. In spite of these difficulties, observations on known individuals revealed that rates of mortality were age-dependent and decreased rapidly after metamorphosis. More than 25% of such fishes disappeared during their first five days in the benthic habitat compared with >10% of fishes aged 6–10 days and no losses of fishes aged 11–15 days. These early losses are the greatest instantaneous rates of mortality yet documented for recruited reef fishes. The experiment also suggested different rates of early mortality for various groupings of species: individuals of solitary, sedentary species disappeared approximately half as fast as individuals of the more mobile, and the more gregarious, species. This is probably a true reflection of the different vulnerability of these groups to predation and it may be caused by the different ways in which these fishes use the coral substratum. Our experience suggests that caging artifacts can have major impacts on the results obtained from this type of experiment and they must be controlled for adequately. We conclude that studies of predation on reef fishes may be done more easily using other methods.