Overexploitation and subsequent collapse of marine fishes has focused attention on the ability of affected populations to recover to former abundance levels and on the degree to which their persistence is threatened by extinction. Although potential for recovery has been assessed indirectly, actual changes in population size following long-term declines have not been examined empirically. Here I show that there is very little evidence for rapid recovery from prolonged declines, in contrast to the perception that marine fishes are highly resilient to large population reductions. With the possible exception of herring and related species that mature early in life and are fished with highly selective equipment, my analysis of 90 stocks reveals that many gadids (for example, cod, haddock) and other non-clupeids (for example, flatfishes) have experienced little, if any, recovery as much as 15 years after 45-99% reductions in reproductive biomass. Although the effects of overfishing on single species may generally be reversible, the actual time required for recovery appears to be considerable. To exempt marine fishes from existing criteria used to assign extinction risk would be inconsistent with precautionary approaches to fisheries management and the conservation of marine biodiversity.