Production of low cost milk is the most relevant aspect in dairy goat systems. Therefore, studying the behavior of the lactation curve of dairy goats is important to implement animal breeding programs as well as for plan and introduce new management tools. The objective of this study was to select among mathematical models published in the literature the one that best describe the lactation curve of dairy goats. A dataset constituted of 23,502 observations of milk yield from 748 lactations of Saanen and Brown Alpine (Oberhaslis) goats was used. Estimation of the parameters was done by the non-linear regression procedure of SAS using the modified method of Gauss-Newton. Residual plot, absolute mean deviation, correlation coefficient and visual appraisal of predicted versus measured lactation curve all were used to determine the best model. The equality of parameters and the models identity of non-linear regression were used to test if the parameters of the lactation curve were similar for breed and order of lactation. Among the evaluated models, that proposed by Nelder (1966), , was the best compromise for adjustment of the lactation curve of goats. Even though differences among parameters of the chosen model for different lactation and breed were observed, only one equation was used in the analysis.