This longitudinal intervention assessed children's memory at 2-1/2 years (short-term posttest; N = 115) and their memory and narrative at 3-1/2 years (long-term posttest; N = 100) as a function of maternal training in elaborative reminiscing when children were 1-1/2 to 2-1/2 years. At both posttests, trained mothers were more elaborative in their reminiscing than untrained mothers. At the long-term posttest, trained mothers were also more repetitive than untrained mothers. Children of trained mothers provided richer memories than children of untrained mothers at both posttests. Children of trained mothers also produced more accurate memories with researchers but only if they had high initial levels of self-awareness. Results are discussed with respect to theory and practice in autobiographical memory development.