After a century of research into the drivers of early life (EL) growth and mortality, fisheries science has acquired limited capacity to predict future recruitment. A meta-analysis of stock assessment time series revealed that it may be difficult to identify stock– or environmental–recruitment drivers given limited variability in spawner biomass, recruitment, and survivorship in most populations. In nearly 50% of the stocks, there was limited information at low spawner biomass, limiting the reliability of fits to stock–recruitment models. Furthermore, variations in survivorship in 50% of year-classes resulted in less than a 2.5-fold change in recruitment. Simulations of three scenarios of change in EL growth and mortality rates demonstrated that they must covary positively to reproduce variations in survivorship consistent with observations. The potentially limited reliability of stock–recruitment relationships to predict year-class strength in many stocks and the low variability in survivorship in a large proportion of year-classes has important implications for the development of projections of stock productivity used in scientific advice. Furthermore, if a positive growth–mortality relationship underlies variations in survivorship, new research approaches are required to understand the trophic relationships that govern the dynamics of early life stages of fish and patterns of recruitment variability.