Fig 1. Cart-pole extrapolation task.

(A) Sequence of the blocks and tasks. Subjects in the group VF followed the upper (orange) path, while subjects in the group MF followed the lower (blue) path. In block E, the extrapolation task was explained. In block B, the baseline performance was measured. The subsequent four blocks were preceded by a visual familiarization block (orange) or motor control block (blue). Feedback about the correct response was only provided during the last five blocks (dark gray, F1-F5, see below). (B) Phases of each trial during the cart-pole extrapolation task. During the observation phase, a red arrow indicated the direction and magnitude of the force, which was applied during balancing. During the no input phase, the force applied to the cart was set to zero (see main text) while the pole was still visible. In the prediction phase, the pole was occluded but the dynamical system was simulated further with the force set to zero. Thus, the cart and (invisible) pole kept moving. Subsequently the time bar turned red and the subject gave a response (keypress) representing the expected orientation of the pole. The feedback phase indicated the correct pole angle and was only provided during the last five blocks (F1-F5, see A). RT: response time. (C) Schematic illustration of the prediction error regarding the pole angle. Notice that an underestimation of the pole movement (blue) corresponds to a positive error irrespective of the side.