Amongst passive collecting gear types, light traps have been used extensively to sample the larvae and juveniles of fish in both marine and freshwater environments, and especially so in structurally complex habitats. Although a number of modifications have been proposed to increase efficiency, no account has hitherto been made of the possible loss of trapped larvae and juveniles as a result of within-trap predation by piscivorous fish. To address this issue, we evaluated the efficiency of modified light traps by the addition of mesh (3 mm knot-to-knot wrapped around the entrance chambers), by combining results from a long-term monitoring study with those from a designed field experiment. Abundance of larvae and juveniles collected during the monitoring program was higher in the years following mesh application (hence, beyond monthly variation), and this was supported by the experiment. Based on the latter, light/mesh traps consistently caught only larvae/juveniles, hence excluding adults; whereas the opposite was true of the light/no-mesh trap combination. Also, mesh/no-light traps did not capture any fish, unlike no-mesh/no-light traps, which sampled only adults. Our findings suggest that in situ predation on larvae and juveniles within light traps should not be overlooked. The occurrence of piscivorous and cannibalistic behaviour in Hypseleotris spp. is also discussed.