Reef sharks may influence the foraging behaviour of mesopredatory teleosts on coral reefs via both risk effects and competitive exclusion. We used a "natural experiment" to test the hypothesis that the loss of sharks on coral reefs can influence the diet and body condition of mesopredatory fishes by comparing two remote, atoll-like reef systems, the Rowley Shoals and the Scott Reefs, in northwestern Australia. The Rowley Shoals are a marine reserve where sharks are abundant, whereas at the Scott Reefs numbers of sharks have been reduced by centuries of targeted fishing. On reefs where sharks were rare, the gut contents of five species of mesopredatory teleosts largely contained fish while on reefs with abundant sharks, the same mesopredatory species consumed a larger proportion of benthic invertebrates. These measures of diet were correlated with changes in body condition, such that the condition of mesopredatory teleosts was significantly poorer on reefs with higher shark abundance. Condition was defined as body weight, height and width for a given length and also estimated via several indices of condition. Due to the nature of natural experiments, alternative explanations cannot be discounted. However, the results were consistent with the hypothesis that loss of sharks may influence the diet and condition of mesopredators and by association, their fecundity and trophic role. Regardless of the mechanism (risk effects, competitive release, or other), our findings suggest that overfishing of sharks has the potential to trigger trophic cascades on coral reefs and that further declines in shark populations globally should be prevented to protect ecosystem health.