Abstract Trap fishing is widespread on coral reefs but the sustainability of this practice is causing concern because it is efficient and unselective. The effects of trap fishing were investigated by comparing fish assemblages among six Caribbean islands subject to different trapping pressures. These ranged from none in Bonaire and Saba increasing through Puerto Rico, St Lucia, Dominica and Jamaica respectively. Fish were censused at depths of 5 m and 15 m on fore‐reef slopes by counting the numbers within replicate 10 m diameter areas for 15 min. Between 64 and 1375 counts were made in each country. In St Lucia and Jamaica abundance of fish censused on the reef was compared to representation in traps which were visually sampled underwater in the area of fish counts. Twenty‐three traps were sampled in Jamaica and 75 in St Lucia. For some comparisons between these islands, St Lucian sampling effort was reduced to that of Jamaica (23 traps and 112 counts) by randomly sub‐sampling 10 times. Traps contained 54 different species in St Lucia and 22 in Jamaica, while there were 90 and 57 respectively in counts. After reducing St Lucian sampling effort to Jamaican levels, an average of 35 species were found in traps and 70 seen in counts. Of these, 76% in St Lucia and 73% in Jamaica were relatively more abundant in traps than they were on the reef. Species were considered to be highly susceptible to trapping if the ratio of their abundance in traps compared to that on the reef exceeded 3:1. Trapping pressure was approximately three and a half times greater in Jamaica than St Lucia. After equalizing sampling effort, there was an average of 16 highly trappable species in St Lucia compared to 13 in Jamaica. Species did not always appear highly trappable in both countries. Eleven of St Lucia's highly trappable species were absent from Jamaica (falling to 8.5 on average after equalizing sampling effort), but none vice versa, suggesting that trapping may have contributed to their absence or rarity on Jamaican reefs. The Tetraodontiformes, which include many non‐target species, were particularly susceptible to trapping in both countries. Their abundance in the six islands censused was inversely related to trap fishing pressure, as was that of two other non‐target families, butterflyfish (Chaetodontidae) and angelfish (Pomacanthidae). To determine whether fish that are common in traps in St Lucia are reaching sexual maturity before capture, size frequency data for 23 species from a sample of trap catches were gathered and examined for their state of maturity. In seven species, more than a third of 705 trapped fish were immature, indicating that trap fishing causes growth over‐fishing (premature removal of fish), and calling into question the sustainability of yields for these species. In conclusion, at the intensities seen in this study, trap fisheries cause serious over‐fishing, reduce biodiversity, and alter ecosystem structure. While commonly perceived as low impact, coral reef trap fisheries in the Caribbean and further afield, need tighter regulation and control. Copyright © 2006 John Wiley & Sons, Ltd.