Many sport skills require equal performance with limbs from both sides of the body, however, considerable debate remains as to the optimal direction or asymmetry of bilateral transfer with regard to limb preference and initial preference. This study compared the traditional view that initial practice should be with one limb before beginning practice with the opposing limb to a more current hypothesis that initial practice might be most effective when both limbs are practiced in a varied order. Asymmetry of transfer comparisons were also investigated within a learning context to determine whether initial practice should start with the preferred or nonpreferred hand. Subjects practiced a novel lacrosse throwing task for accuracy under one of three practice sequences: (a) preferred group (practiced initially with the preferred hand before switching to the opposing hand); (b) nonpreferred group (practiced initially with the nonpreferred hand before switching); and (c) variable group (switched practice hands in blocks of 5 trials throughout the entire acquisition period). Results were compared across time (pretest, short-term retention and delayed retention) and for the directional effects of preference (preferred and nonpreferred hand performance). Transfer test data indicated clear beneficial effects of practicing with a variable sequence. Subjects in the variable condition demonstrated less radial error than subjects in a constant practice condition on retention, p <.05 and transfer, p<.05. Retention differences were observed between initial preferred and nonpreferred practice groups, with the preferred sequence demonstrating less radial error at retention, p<.05. Practice condition showed an interaction with gender on transfer, p<.05, such that males demonstrated a particular advantage for radial error (greater accuracy) in the variable practice condition. However, females in the variable condition showed a significant benefit in terms of decreased variance (increased consistency) in transfer performance, p<.05 (average variance), p<.05 (target variance). These results lend considerable support to Schmidt’s (1975) variability of practice hypothesis with respect to the learning of bilateral skills. Partial support is provided for the utility of the preferred to nonpreferred practice sequence. It is suggested that the traditional perspective for the practice of bilateral skills be reconsidered from a variability of practice standpoint, particularly if retention and transfer are the performance goals. The nature of the gender interactions were discussed with respect to the amount of original learning and task difficulty.