Quickly 'learning' to move optimally

People take account of the variability in their movements in a near-optimal manner in various visuo-motor tasks. Is knowledge of one's variability needed for such near-optimal performance, or could it arise from responding to one's success in previous attempts in some simple manner? We asked subjects to move a pen back and forth across a tablet to make a cursor move as quickly as possible between two targets. The cursor had to stop within the targets. Task difficulty was varied between blocks. Part of the variation in difficulty was explicit (three target sizes) whereas the rest had to be discovered during the movements (two mappings between the movements of pen and cursor). In all cases, subjects sped up after stopping within a target and slowed down after failing to do so. We interpret this as evidence that explicit knowledge of one's variability is not necessary for performing close to optimally.