The application of evolutionary principles to the management of fisheries has gained considerable attention recently. Harvesting of fish may apply directional or disruptive selection to key life history traits and evidence for fishery-induced evolution is growing. The traits that are directly selected upon are often correlated (genetically or phenotypically) with a suite of interrelated physiological, behavioral, and morphological characters. A question that has received comparatively little attention is whether or not, after cessation of fishery-induced selection, these correlated traits revert back to previous states. Here, we empirically examine this question. In experiments with the Atlantic silverside, Menidia menidia, we applied size-selective culling for 5 generations and then maintained the lines a further 5 generations under random harvesting. We found that some traits do return to pre-harvesting levels (e.g., larval viability), some partially recover (e.g., egg volume, size at hatch), and others show no sign of change (e.g., food consumption rate, vertebral number). Such correlations among characters could, in theory, greatly accelerate or decelerate the recovery of fish populations. These results may explain why some fish stocks fail to recover after fishing pressure is relaxed.