July 13, 2018
15:00 ~ 16:00
CiNet 1F Conference Room
“Multisensory and Unisensory Integration with Divisive Normalization for Human Motor Adaptation”
Tokyo University of Agriculture and Technology / JSPS Postdoctoral Fellow
Host : Atsushi Yokoi (Nishimoto Group)
Sensory information is important for human motor control and learning. Theoretically, motor adaptation is driven by a discrepancy between the actual and predicted sensory feedbacks (i.e., sensory prediction error). Both vision and proprioception could contribute to the computation of the sensory prediction error, but the manner via which the sensory information of these different modalities are integrated and utilized for motor adaptation is still under debate. We aimed to clarify the way of integration by examining motor adaptation when visual and haptic (i.e., proprioceptive) perturbations were independently applied. We used a virtual reality system in which participants held a manipulandum and moved a visual cursor representing their hand position towards a target on a display. This motor learning paradigm could perturb the cursor and handle (or hand) movements to produce the sensory prediction errors. We arbitrarily determined sizes of both errors and measured how largely they compensated these errors (aftereffect). The results demonstrated that either visual (i.e., cursor) or proprioceptive (hand) errors induced the aftereffect, but the magnitude was not linearly increased with the size of them. Most notably, we observed a complicated pattern of the aftereffects when both visual and proprioceptive errors were simultaneously imposed, which could not be explained by a prevailing idea in perceptual literatures, Bayesian integration. Rather, we found that divisive normalized integration, which is thought be a canonical neural computation (Carandini & Heeger, Nat Rev Neurosci, 2011), can successfully reproduce the complicated pattern. An additional experiment illustrated that the sensorimotor system stores and integrates visual and proprioceptive motor memories, suggesting that the integration occurs not in sensory (i.e., computing a sensory prediction error) but in motor domain (i.e., updating motor memories). Furthermore, preliminary results showed that divisively normalized integration may account for motor adaptation when multiple sensory feedbacks were applied in one modality (i.e., unisensory integration, Kasuga & Nozaki, PLoS ONE, 2013). Thus, these results illustrated that the multisensory and unisensory integration for human motor adaptation are operated with divisive normalization in motor domain.
In line with previous studies, our results extend the idea, divisive normalization, to explain the sensorimotor system in human.