In Exp. I ( N = 32), 4 1/2-yr.-old children discriminated between two stimulus compounds, multiple and related stimulus components displayed together (3 pairs of parallel lines vs 3 pairs of nonparallel lines) or multiple but unrelated displayed components (3 unrelated objects vs 3 other unrelated objects). Although between-group acquisition was the same, significantly more children trained with multiple-related stimulus components responded appropriately to each of the relevant components of the compound when they were presented separately during test trials, i.e., less stimulus overselection. In Exp. II ( N = 15), multiple-related training of parallelness led to faster concept attainment, and proportionally more children got almost perfect transfer scores than when training consisted of presenting pairs of single examples of parallel and nonparallel lines during each trial.