Two animal models were compared with respect to potential bias in genetic trend estimates for female fertility and for their predictive ability. In addition to either a fixed effect for month of first insemination or for month-year of first insemination, the models had fixed effects of age and double insemination and random effects of herd-year and animal. The model with a fixed effect of month of first insemination had a larger positive genetic trend for 56-d nonreturn rate in virgin heifers (0.16% yr), smaller downward bias, and somewhat higher predictive ability. These results demonstrate the importance of verifying models to be used in the calculation of breeding values.