Restricted maximumlikelihood (REML) estimates of additive and residual variances and covariancesfor birth weight and adjusted weights at 120, 205, 240, 365, 420 e 550 daysof age were used to estimate growth covariance functions (CFs) of Tabapuãbeef calves. Data were observed on 41,415 animals born from 1975 to 1997 andraised under pasture conditions. Estimation of CFs is a very useful tool toanalyze beef cattle growth. It was possible to estimate covariance between anypair of ages and the analyses of eigenfunctions associated with the eigenvaluesof coefficients matrix of CFs showed that the growth curves of Tabapuãcalves could be easily changed by selection. Weaning stress, compensatory growthand selection of animals in the final period caused changes on (co)variancetrajectories. Therefore only CFs of more complex order were able to estimatevalues near to REML estimates. However, high order Legendre polynomials drewsharp waves on variances trajectories at the period edges, witch does not havea coherently biological reason.
