The ability to make goal-directed movements relies on estimates of limb position, based on vision, proprioception, as well as efference-based predictions. We measure the plasticity of proprioceptive and efferent-based estimates of the hand using a series of visuomotor adaptation experiments, which involve reaching with a rotated cursor. This rotated cursor training leads to changes in movements, proprioception and prediction. We measure the speed of these changes as well as fit them to the multi-rate model (Smith et al., 2006) which consists of a fast and slow process. We used a multiphase experiment which alternated between one rotated training trial and then one of 3 intervening tasks. To measure changes in hand localization, participants estimated the location of the unseen hand when it is moved by the robot (passive localization) or when they generated their own movement (active localization). By comparing the differences between these hand estimates after passive (only proprioception) or active (both proprioception and efferent-based prediction) movements, we are able to measure predicted sensory consequences of movement. The 3rd intervening task type was a no-cursor reach where no visual feedback of hand position or trial success was given. The trial-by-trial data suggest that proprioception recalibrates extremely fast, and is simply proportional to the visual-proprioceptive discrepancy, and does not reflect either model process. Prediction and and no-cursor reaches do not reflect either process fitted to training, but appear to be best explained by their own set of dual processes. Thus, these changes seem to reflect separate adaptation mechanisms.