In the training sessions, participants completed 3000 trials in which they were presented with only dIOVD random-dot stereograms at 100% coherence distributed over three sessions on different days. In those trials, they had to decide whether the stimuli moved towards or away from them. Additionally during the training, they were given the option to press a third key to indicate that they were unsure about the direction. We introduced the third response option to get a more nuanced measure of how the participants’ confidence would change during the training. However, participants used this response option very sparingly. No feedback was given. The rationale for the exclusive use of dIOVD stimuli in the training was as follows: the null hypothesis of our study was that performance for the three types of stimuli should be similar. The stimulus properties were chosen to be similar to those used in a study that had found similar performances for aIOVD and FULL cue stimuli [35]. As described above, stimulus properties were optimised for aIOVD but if both aIOVD and dIOVD stimuli isolate the same mechanism their optimal stimulation conditions should be similar. However, it could be that certain aspects in which the stimuli differ result in differences in the optimal stimulus properties. Thus, to mitigate the potential disadvantage of the dIOVD stimulus, we decided to train the participants on dIOVD. So, the training favoured the null hypothesis, i.e., similar performances for all stimuli against which we tested.