This study examined the effect of input modality in spatial transformation task. In a previous study (Sasaki, 1981), input modality had not affected the performance of perspective transformation in adult subjects; the present work was to analyse this modality effect developmentally. The two most important theoretical standpoints of the nature and the development of imagery have been advanced by Piaget & Inhelder (1966) and by Kosslyn (1978a, b). Piaget seemed to rule out the notion that images were derived from perception and developmentally it was operative aspect of cognition that produced changes in imagery. Kosslyn claimed imagery was a quasi-pictorial (perceptual) representation and suggested that young children tended to use imagery more than adults do. The former seemed to hypothesize that input modality did not affect spatial transformation in children while the latter considered this effect as probable. To clarify this theoretical problem was the second purpose of this experiment. Subjects were 119 children (57 boys and 62 girls) from 7 to 12 years old. The subjects at each age level were equally divided into two (only 7 years old) or three groups and assigned to one of the three input conditions: visual, haptic and verbal. In the first phase, subjects were presented with information about objects placed randomly on a board (FIG. 1). They identified those objects visually, haptically (touch and movement) and verbally. In haptic and verbal conditions, children were blindfolded. After a practice session with one object, the subjects were told to locate the position of the other three objects (simple reconstruction trials) and then, to imagine that he/she had moved to a position behind the other object and to make similar evaluation of the imagined position (perspective transformation trials). In simple reconstruction trials, a few differences were found. The difference of the three groups were no more than 10 degrees level (FIG. 2). But on perspective transformation trials, a highly significant interaction between age and input modality was found. The basic data were 12 angle setting produced by a subject in each of the mental manipulation tasks. These angle settings were converted into physical representations as FIG. 3. The number of triangles formed in the physical representation constructed from S's angle setting provided a global index of S's ability to manipulate mentally (FIG. 4). The second dependent measures were setting categories. There were 5 categories. 1) Correct: all settings were with in 10 degrees level. 2) Subcorrect: with in 30 degrees. 3) Degree error: degree error included over 30 degree error but positions of four objects were coordinated at last. 4) Position erro r: positions of four objects were uncoordinated and without any order. 5) Egocentric error. The proportions of each category with age in the input conditions shown in FIG. 5. These results suggested that there were developmental stage in the development of spatial transformation. This developmental stage was found in two dependent measures in the visual and haptic conditions, and response time (FIG. 6) in the verbal condition. And this developmental stage appeared at different ages in three input conditions. In visual condition, it was found at age 10 and at 11 in haptic condition. Hence, it may be concluded that: 1. The development of spatial transformation depended on the changes of underlying cognitive structure. 2. But being affected by the input modality, visual information had some superiority. 3. Thus its development was well understood by the interaction of underlying cognitive structure and the information having visual or spatial mode.