Motor learning has traditionally been associated with the concept of automaticity. Automaticity refers to the reduction of the cognitive effort required to perform a motor task, as learning progresses. However, there is little detailed consensus in the literature on what the process of automatization actually involves. We measured tracking performance in two groups of participants while either the target or the manual cursor was suppressed for a brief period during each trial. Subjects learned to maintain accurate tracking through periods of target or cursor suppression. We have used this approach to investigate the internal models used during tracking, and their updating during motor learning. We have simultaneously measured tracking performance and pupil dilation as a measure of cognitive load. The results showed that pupil diameter decreases with learning process of tracking performance. Decrease of pupil diameter suggests that automatization is linked specifically to the learning of internal models.