Data from 26.314 animals of Guzerá cattle, weighted every 90 days from birth to 630 days, between 1975 and 2001, were used to estimate covariance components by REML methodology using four different models. The first model was the complete one and included four random components: direct genetic (GA), maternal genetic (GM), maternal permanent environmental (AM) and the residual; model 2 did not include the GM effect; model 3 did not include the AM effect and model 4 did not include both GM and AM effects. Models 1 and 2 did not differ by likelihood ratio test in almost all age classes. Direct heritability estimates obtained from models 1 and 2 were quite similar, and increased from first to second age, remained the same until weaning and then increased. Direct heritability estimates for weight in age classes close to 205, 365 and 550 days, by models 1 and 2, were respectively 0.15, 0.12 and 0.14. Estimates of the same parameter obtained by models 3 and 4 were respectively 0.15, 0.14, 0.15 and 0.26, 0.19, 0.17. Maternal heritability estimates by model 3 were higher because it did not include the AM effect. The additive genetic variance was probably superestimated by model 4 because it did not include GM and AM effects. The comparison between the models indicated that the model without GM was equivalent to the complete model 1.
beef cattle; estimates of variances; maternal effects; heritability