Six sigmoidal growth curves and two growth curves derived from a two-phase relative growth rate model were evaluated, using an experimental body-weight data from male and female turkeys of two genetic lines; a fast-growing (F) line and a randombred control (RBC) line from which the F line was developed. When their root mean square error was compared to the root mean square error of the local regression smoother, all sigmoidal growth curves: the logistic, Gompertz, von Bertalanffy, Richards, Weibull, and Morgan-Mercer-Flodin growth curves demonstrated a lack of fit. The primary source of the systematic lack of fit was identified with nonparametric estimates of the relative growth rate (the growth rate as a fraction of the body weight) of 20 turkeys. When the relative growth rate was estimated from the above sigmoidal growth curves, none could accommodate features of the nonparametric estimates of the relative growth rate. Based on the feature of the relative growth rate, two new growth curves were derived from a segmented two-phase model. Both models, in which the relative growth rate decreases in two linear phases with slopes of beta1 and beta2 joined together at time=kappa, gave growth curves that fit the experimental data acceptably. The linear-linear model with the smooth transition rendered better fit over the model with the abrupt transition. When the growth curves of male and female turkeys were compared, beta1, beta2, and kappa were smaller in males. When the F line was compared to the RBC line, beta1 and kappa were smaller and beta2 was closer to zero, indicating that the relative growth rate declined rapidly until about 61 days of age in the F line, while it declined less rapidly until about 71 days of age in the RBC line.