Non-additive genetic effects are usually ignored in animal breeding programs due to data structure (e.g., incomplete pedigree), computational limitations, and over parameterization of the model. However, non-additive genetic effects may play an important role in the expression of complex traits in livestock species, such as reproduction and fertility traits. We assessed the use of pedigree and SNP-marker-based models to estimate additive and non-additive genetic variances for reproduction and fertility traits in Canadian Holstein heifers (n = 5,825) and cows (n = 6,090). Four traits were analyzed including age at first service for heifers (AFS), calving to first service interval for cows (CTFS), and 56-d non-return rate for heifers and cows (NRR). Four linear models were used (1) additive genetic model (MA); (2) a model including both additive and epistatic (additive by additive) genetic effects (MAE); (3) a model including both additive and dominance effects (MAD); and (4) a full model including additive, epistatic, and dominance genetic effects (MAED). The models which included non-additive genetic effects for AFS and CTFS indicated that epistasis, dominance, or a combination thereof, are as important as additive effects, and sometimes contribute a larger proportion to the total phenotypic variance than the contribution of the additive effects. The partitioning of variance components resulted in a re-ranking of animals in the top 10% based exclusively on the additive genetic effects between models. The change in rank indicates that adjusting for non-additive genetic effects could change selection decisions made in dairy cattle breeding programs. These results suggest that non-additive genetic effects play an important role in some reproduction and fertility traits in Canadian Holsteins and their inclusion in genetic evaluations may improve accuracy of selection based on additive genetic effects.