Two cuing, free-recall studies were conducted to test Bach and Underwood's (1970) hypothesis that acoustic encoding is dominant among second graders and semantic encoding is dominant among sixth graders. When retrieval cues were presented with to-be-remembered items at both input and output (Experiment 1), and when cues were presented only at output (Experiment 2), semantic cues were more efficient in elevating recall than were acoustic cues for both second and sixth graders. When these and other results generally found using recognition, sorting, incidental learning, and free-recall experimental designs are compared, it seems plausible that item presentation and memory-testing formats interact with age, and that these factors account for the different patterns of attribute dominance found in the literature. The knowledge base cannot be understood by focusing on either subject or task analyses, but only by focusing on interactions between subject and task variables as they change over time. The educational implications for young grade-school children are discussed.