Change in motion discrimination was assessed after seven days training on a rapid serial visual presentation training task, which included exposure to below-threshold coherent motion that was irrelevant to the task the participant was involved in. Post-training, participants had improved sensitivity for supra-threshold motion discrimination, which was specific to the direction exposed during training. A computational decomposition of the effect shows that the improvement is a combination of (i) an increase in rate at which participants accumulate evidence for the direction to which they were exposed and (ii) a decrease in their criterion for a response. Together with these differences (consistent across participants) other cognitive processes vary non-systematically between the pre-test and the post-test session making an analysis only based on accuracy or reaction times potentially misleading. Our analysis shows the benefits of isolating the different processes that are involved in perceptual decision making and are affected by perceptual learning.