A growth curve model used for analyzing growth is characterized by a mathematical function with respect to time, called a growth function. As the results of analysis from a growth curve model strongly depend on the growth function used for the analysis, the selection of growth functions is important. A choice of growth function based on the minimization of a model selection criterion is one of the major selection methods. In this paper, we compare the performances of growth-function selection methods using these criteria (e.g., Mallows' Cp criterion) through Monte Carlo simulations. As a result, we recommend the use of a method employing the Bayesian information criterion for the selection of growth functions.