RESUMEN
Nonlinear mixed models were used to describe longitudinal scrotal circumference (SC) measurements of Nellore bulls. Models comparisons were based on Akaike's information criterion, Bayesian information criterion, error sum of squares, adjusted R2 and percentage of convergence. Sequentially, the best model was used to compare the SC growth curve in bulls divergently classified according to SC at 18-21 months of age. For this, bulls were classified into five groups: SC < 28cm; 28cm ≤ SC < 30cm, 30cm ≤ SC < 32cm, 32cm ≤ SC < 34cm and SC ≥ 34cm. Michaelis-Menten model showed the best fit according to the mentioned criteria. In this model, ß1 is the asymptotic SC value and ß2 represents the time to half-final growth and may be related to sexual precocity. Parameters of the individual estimated growth curves were used to create a new dataset to evaluate the effect of the classification, farms, and year of birth on ß1 and ß2 parameters. Bulls of the largest SC group presented a larger predicted SC along all analyzed periods; nevertheless, smaller SC group showed predicted SC similar to intermediate SC groups (28cm ≤ SC < 32cm), around 1200 days of age. In this context, bulls classified as improper for reproduction at 18-21 months old can reach a similar condition to those considered as good condition. In terms of classification at 18-21 months, asymptotic SC was similar among groups, farms and years; however, ß2 differed among groups indicating that differences in growth curves are related to sexual precocity. In summary, it seems that selection based on SC at too early ages may lead to discard bulls with suitable reproductive potential.