Simple maternal effect animal models provide biased estimates of additive genetic and maternal variation

Maternal effects (the consistent effect of a mother on her offspring) can inflate estimates of additive genetic variation (V_A) if not properly accounted for. As they are typically assumed to cause similarities only among maternal siblings, they are often accounted for by modelling maternal identity effects. However, if maternal effects have a genetic basis, they create additional similarities among relatives with related mothers that are not captured with maternal identity effects. Unmodelled maternal genetic variance (V_{Mg}) may therefore still inflate V_A in common quantitative genetic models, which is under-appreciated in the literature. Using published data and simulations, we explore the extent of this problem. Estimates from 14 studies of eight species suggest that a large proportion of maternal variation is genetic. Both this data and simulations confirmed that unmodelled V_{Mg} can inflate V_A and underestimate total maternal variation (V_M), the bias increasing with the amount of non-sibling maternal relatives in a pedigree. Simulations show these biases are further influenced by the size and direction of any direct-maternal genetic covariance. The estimation of total V_A (i.e., the weighted sum of V_A and V_{Mg}) is additionally affected, limiting inferences about evolutionary potential from simple maternal effect models. Unbiased estimates require modelling V_{Mg} explicitly, but these models are often avoided due to perceived data limitations. We demonstrate that estimating V_{Mg} is possible even with small pedigrees, reducing bias in V_A and maintaining accuracy in estimates of V_A, V_M, and total V_A. We therefore advocate for the broader use of these models.