Presenting two or more stimulus types randomly interleaved, so-called roving stimuli, disrupts perceptual learning in many paradigms. It was recently reported that learning with disrupting stimuli types is possible when stimuli are presented in an alternating sequence, ie stimulus from type A, then type B, then type A, etc. In our experiment we used used bisection stimuli, but found the opposite pattern of results. Presentation of bisection stimuli in a sequence disrupted perceptual learning. We tried to explain these seemingly contradictory results by conducting a meta-analysis. Participants who initially performed the task at a low level, ie 'bad performers', were able to learn, whereas the 'good performers' did not (good performers were still outside the ceiling range). Therefore, interleaving stimuli may not abolish perceptual learning, they may just make it more difficult and more prone to interact with other factors, such as the initial performance level.