The aim of this study was to model individual lactation curves for milk production in Alpine goats by comparing four models based on their fit to typical or atypical lactation curves. Three thousand five hundred and eighty-one test-day records were analyzed for 535 first lactations of Alpine goats controlled monthly between 1994 and 2008 in the Genetics Enterprise “Los Naranjos”, in the province of La Havana, Cuba. The mathematical models used to describe the lactation curve were: Wood, Wilmink, Ali-Schaeffer and Legendre’s orthogonal polynomials. Parameters were estimated utilizing nonlinear regression and iterative procedures. The criterion used to verify the goodness of fit was the adjusted coefficient of determination (R 2 ADJ ). Goodness of fit based on R 2 ADJ >0.75 was accomplished for 46.3, 47.4, 60.5 and 69.9% of lactations for Wood, Wilmink, Ali-Schaeffer and Legendre´s polynomials, respectively. The three parameter models detected four types of curves, meanwhile Ali-Schaeffer and Legendre’s polynomials models were able to recognize 9 and 14 types of curves, respectively, from the 32 possible theoretical groups that were determined by combinations of the signs of the parameters of each model. Correlations between parameter values estimated for Ali-Schaeffer were higher than those estimated for Legendre´s polynomials. The different curve shapes found in the 5-parameter models can be considered as modifications of the two main groups of curve shapes known in the literature: typical and atypical. Results of this study indicate that Ali-Schaeffer and Legendre’s orthogonal polynomials models were the best functions to model adequately individual lactation curves of Alpine goats. Furthermore, the appearance of several groups of individual curves suggests the need of more research on comparison of individual models within group.