It has been suggested that learning is specific to the source of information available during practice (Proteau, Marteniuk, & Levesque, 1992). This hypothesis is quite robust for rapid aiming tasks that have defined spatial and temporal goals, but it is unclear whether it extends to tasks that are more sensory driven and with no clear spatio-temporal goal, such as ball balancing. In this experiment, 24 young adults practiced balancing a ball on their thumb and forefinger either with or without vision. Performance was measured early in practice (after 40 min.) and late in practice (after 180 min.) in both conditions. Both groups improved their total balancing time from the early to late testing sessions. Transfer data from the late testing session revealed that all participants performed better with vision regardless of their practice condition. This suggests that vision is the dominant source of afferent information for this task and learning was not specific to the source of information available during practice. Thus, the specificity of practice hypothesis does not apply to this type of task.