Visual motor imagery recruits forward models to predict the sensory consequences of imagined movement

Kinesthetic motor imagery (KMI) likely involves forward model-based prediction of the somatosensory consequences of imagined actions, as suggested by sensory attenuation observed during KMI. As visual motor imagery (VMI) and KMI have been suggested to share partially overlapping neural mechanisms, we hypothesized that VMI may also engage forward model-based prediction. Here, we tested this hypothesis by examining sensory attenuation during VMI using a force-matching task. In this task, a reference force was applied to the relaxed left index finger during rest, motor execution, KMI, or VMI involving right index finger pressing, and participants then reproduced its magnitude. In Experiment 1, the right index finger was positioned directly above the left index finger, allowing prediction of somatosensory input to the left index finger via forward models. Sensory attenuation, defined as a reduction in the reproduced force relative to the rest condition, was observed during KMI and VMI. In Experiment 2, which served as a control condition to rule out the possibility that sensory attenuation was driven by differences in cognitive demands, the right index finger was positioned 250 mm to the right of the left index finger. In this configuration, forward models would not predict finger contact, whereas cognitive demands were comparable to those in Experiment 1. No sensory attenuation was observed during KMI or VMI. Therefore, the sensory attenuation observed in Experiment 1 was not attributable to differences in cognitive demands across conditions. These findings suggest that VMI may recruit forward models to predict the sensory consequences of imagined movements.