Towards domain-general predictive coding: Expected TMS excites the motor system less effectively than unexpected stimulation

Abstract

The brains response to sensory input is modulated by prediction. For example, sounds that are produced by ones own actions, or those that are strongly predicted by environmental cues, are perceived as less salient and elicit an attenuated N1 component in the auditory evoked potential. Here we examined whether the neural response to direct stimulation of the brain is attenuated by prediction in a similar manner. Transcranial magnetic stimulation (TMS) applied over primary motor cortex can be used to gauge the excitability of the motor system. Motor-evoked potentials (MEPs), elicited by TMS and measured in peripheral muscles, are larger when actions are being prepared and smaller when actions are voluntarily suppressed. We tested whether the amplitude of MEPs was attenuated under circumstances where the TMS pulse can be reliably predicted, even though control of the relevant motor effector was never required. Self-initiation of the TMS pulse and reliable cuing of the TMS pulse both attenuated MEP amplitudes, compared to MEPs generated programmatically in an unpredictable manner. These results suggest that predictive coding may be governed by domain-general mechanisms responsible for all forms predictive learning.