Résumé
While previous research has established that errors are committed during motor imagery (MI), a remaining question is whether the commission of errors drives learning in MI in a manner comparable to physical execution. Here we investigated whether error detection and correction mechanisms exist during MI and if the presence of errors predicts learning. Participants completed a complex upper limb motor task via MI across 5 sessions, re-producing randomly generated or a repeated trajectory. After each trial, participants self-reported their accuracy from 1 – 10, (1 = completely inaccurate, 10 = completely accurate). Immediately after the 5th (final) MI session, participants physically executed the task to permit assessment of performance. Participants’ accuracy during physical execution was determined by calculating the point-by-point distance between the stimulus trajectory and participants’ generated trajectory for each trial. Participants’ self-reported accuracy increased significantly between session 1 and 5 for the repeated but not the random shapes, providing evidence that an error detection and correction mechanism exists during MI. Furthermore, the change in error over the 5 sessions predicted performance on subsequent physical execution in that participants reporting lower accuracy of the repeated shape during session 1 and subsequently reported improved accuracy on session 5 had better physical performance than participants who reported higher accuracy (made less errors) earlier on. These results highlight that in addition to errors being committed during MI, the commission and correction of errors during MI drives learning. These findings provide indirect support for the existence of internal models in MI, and their updating.
