We examined the effects of process and observation errors on the selection of the stock-recruitment relationship (SRR) curve using simulations. When the process and observation errors were added to both spawning stock biomass and recruitment, the results were as follows:(1) When the proportional model was set as the true SRR model, there was a high probability that the Ricker or Beverton and Holt model was selected in response to the errors; (2) When the Ricker or Beverton and Holt model was set as the true SRR model, the proportional model was seldom selected in response to the errors; (3) The proportional SRR model should be accepted as the optimum SRR model for the Pacific stock of Japanese sardine; (4) We should use an SRR model that is constructed from more than two independent variables (spawning stock biomass, environmental factors, etc.) when we discuss management of fisheries resources.