A total of 100 3- and 5-year-old children were trained and assessed individually on the concept of equilateral triangle at three levels of attainment: concrete, identity, and classificatory. Five training conditions existed: (1) visual inspection; (2) visual inspection and verbal orienting instruction; (3) visual inspection, free haptic activity, and tactile-kinesthetic training; (4) visual inspection, free haptic activity, tactile-kinesthetic training, and verbal orienting instruction; and (5) unrelated play activity (control). Subjects were trained with 36 geometric blocks varying in shape, color, size, and thickness. After training, each subject's knowledge of the concept was tested at the three levels. Two sets of tasks were administered for testing: transfer of training using the training material as test stimuli and transfer of training using two-dimensional representations of geometric forms. The results indicated that: (1) there were transfer effects for both sets of tasks; (2) training conditions 2, 3, and 4 were more facilitating than 1 and 5; (3) condition 4 was most effective; (4) 5-year-old subjects performed better than 3-year-olds at each level; and (5) in some tasks, 3-year-old subjects did better when aided by manual activity and 5-year-old subjects performed better with verbal orienting instruction. Concluding discussion focuses on some theoretical and educational implications of the results. (Author/SDH)