Winter ice cover was measured as the per cent of lake surface area covered by ice46 during the winter (February-March) before spawning in (a) western and (b) central Lake Erie in year t (year in which each cohort was hatched). Two-dimensional Kolmogorov–Smirnov tests53 determined whether changes in the variance of juvenile abundance were related to ice cover from the previous winter. This distribution-free test identifies a threshold value of ice cover (indicated by vertical dashed lines) that maximizes the difference in variance of juvenile abundance between the two sides of the threshold. P values indicate if the variance of juvenile abundance significantly differs between the two sides of the threshold. Above the threshold, juvenile abundance may be high or low, but below the threshold, only low abundances occur. Thus, winter duration appears to ‘set the stage’ for future high recruitment to the fishery. Juvenile (age-0) abundance was determined from annual, fisheries-independent Ohio Department of Natural Resources-Division of Wildlife bottom trawling surveys conducted during October of year t, and is presented as catch-per-unit-effort (CPUE; # of individuals per trawl minute) for both (a) western and (b) central Lake Erie.