Based on an analysis of 90 marine fish populations, collapses (the greatest proportional reduction in spawner biomass over 15 years) are predicated typically by dramatic increases in fishing mortality and recoveries are more likely to occur when exploitation is reduced. However, among populations for which fishing mortality declined after collapse, recovery was independent of exploitation rate, even when fishing mortality (F) post-collapse was expressed as a function of each population's maximum growth rate (r). After a period of 15 years, many populations that experienced 15 year declines >60% exhibited little or no recovery, despite considerable reductions in fishing mortality. This suggests that factors other than fishing may be considerably more important to recovery, and fishing less important, than previously thought. Furthermore, among populations for which fishing mortality decreased post-collapse, rate of population decline was a reliable predictor of recovery. With the possible exception of clupeids, variation in marine fish breeding population size was found to differ little from that of other vertebrates, and such variability appears to have no effect on rate of recovery. In addition to providing an empirical framework for the study of population collapse and recovery, the analyses presented here provide a means of assessing the precautionary nature of various population-decline thresholds used to assign extinction risks to marine fish.