During the day, we used settlement-stage reef-fish larvae from light-traps to study in situ orientation, 100 to 1000 m from coral reefs in water 10 to 40 m deep, at Lizard Island, Great Barrier Reef. Seven species were observed off leeward Lizard Island, and 4 species off the windward side. All but 1 species swam faster than average ambient currents. Depending on area, time, and species, 80 to 100% of larvae swam directionally. Two species of butterflyfishes Chaetodon plebeius and Chaetodon aureofasciatus swam away from the island, indicating that they could detect the island's reefs. Swimming of 4 species of damselfishes Chromis atripectoralis, Chrysiptera rollandi, Neopomacentrus cyanomos and Pomacentrus lepidogenys ranged from highly directional to nondirectional. Only in N. cyanomos did swimming direction differ between windward and leeward areas. Three species (C. atripectoralis, N. cyanomos and P. lepidogenys) were observed in morning and late afternoon at the leeward area, and all swam in a more westerly direction in the late afternoon. In the afternoon, C. atripectoralis larvae were highly directional in sunny conditions, but nondirectional and individually more variable in cloudy conditions. All these observations imply that damselfish larvae utilized a solar compass. Caesio cuning and P. lepidogenys were non-directional overall, but their swimming direction differed with distance from the reef, implying the reef was detected by these species. Larvae of different species of reef fishes have differing orientations and apparently use different cues for orientation while in open, pelagic waters. Current direction did not influence swimming direction. Net movement by larvae of 6 of the 7 species differed from that of currents in either direction or speed, demonstrating that larval behaviour can result in non-passive dispersal, at least near the end of the pelagic phase.