SELECTIVE MODULATION OF CORTICAL VARIABILITY FOLLOWING CHANGE POINTS DURING EVIDENCE ACCUMULATION

When making decisions in volatile environments, humans update their belief about the state of the environment in a manner that balances stable evidence accumulation with sensitivity to change. In such settings, brainstem arousal systems respond to change points, which may modulate cortical belief updating computations. Here, we studied magnetoencephalography (MEG) dynamics during decision-making in a volatile environment and characterized trial-to-trial variability in decision-related regions, as a putative signature of neuromodulation.
During MEG recordings, human participants judged whether sequentially presented visual discs with variable position were drawn from a source distribution centered in the right or left hemifield. The source could switch within each trial (probability = 0.08) in a hidden fashion. By fitting a normative model to behavior, we tracked participants’ belief about the source. We analyzed the dynamics of spectral MEG power in source space within 22 atlas-defined regions covering the cerebral cortex.
The belief at trial end was decodable from alpha-band activity of multiple areas, specifically in posterior parietal (inferior parietal and dorsal visual stream) and motor/premotor cortices. The belief was encoded in the lateralization of alpha-power in these regions. After change points, alpha-power lateralization flipped sign in these regions, tracking the model belief state. Further, in parietal cortex, trial-by-trial variability of alpha-power decreased contralateral to the upcoming choice. Such a selective modulation of trial-to-trial variability was neither observed in (pre-)motor cortical regions, nor for the model belief state.
We conclude that the intrinsic variability of neural decision variables in parietal cortex is actively modulated following change points.