Fish stock-recruitment dynamics may be difficult to elucidate because of nonstationary relationships resulting from shifting environmental conditions and fluctuations in important vital rates such as individual growth or maturation. The Great Lakes have experienced environmental stressors that may have changed population demographics and stock-recruitment relationships while causing the declines of several prey fish species, including rainbow smelt (Osmerus mordax). We investigated changes in the size and maturation of rainbow smelt in Lake Michigan and Lake Huron and recruitment dynamics of the Lake Michigan stock over the past four decades. Mean lengths and length-at-maturation of rainbow smelt generally declined over time in both lakes. To evaluate recruitment, we used both a Ricker model and a Kalman filter-random walk (KF-RW) model which incorporated nonstationarity in stock productivity by allowing the productivity term to vary over time. The KF-RW model explained nearly four times more variation in recruitment than the Ricker model, indicating the productivity of the Lake Michigan stock has increased. By accounting for this nonstationarity, we were able identify significant variations in stock productivity, evaluate its importance to rainbow smelt recruitment, and speculate on potential environmental causes for the shift. Our results suggest that investigating mechanisms driving nonstationary shifts in stock-recruit relationships can provide valuable insights into temporal variation in fish population dynamics.