Previous research on infant swimming has reported contradictory findings. Cross-sectional observations revealed a disorganized phase between about 3 and 12 months, which was attributed to "cortical inhibition" and implied slow learning (McGraw, 1939). However, training with a single infant during this period revealed rapid acquisition (McGraw, 1935). This study reconciles these findings and examines evidence for a subsequent deliberate phase. Five swimming behaviors were measured during 5 initial trials cross-sectionally at 4, 8, 12, and 16 months; effects of 4 months of training were observed at 12, 16, and 20 months. Results for initial trials were comparable for kicks, arm flexions, a righting response, swimming, and a combinatorial task--turning 180[degrees] and reaching for the pool wall. Disorganized behavior organized quickly with training during initial and subsequent trials. The more rapid acquisition of the combinatorial task for 16- and 20-month-olds under reduced gravitational demands is consistent with a cognitive-maturational transition at about 12 months. (Contains 5 figures and 1 table.)