DISTRIBUTED NEURAL DECISION VARIABLES IN HUMAN CORTEX SHAPE DECISION CONFIDENCE

Recent efforts in neuroscience aim to understand the distributed nature of decision-making in the brain. Theoretical models hold that decisions emerge from competitive interactions between neural populations encoding alternative options. Choice-selective signals, often referred to as neural decision variables (DV), have been observed in multiple brain regions. Here, we asked how these distributed dynamics of neural DVs relate to the sense of confidence.
We used source-level magnetoencephalography (MEG) decoding to simultaneously track neural DVs across multiple cortical areas while human participants performed a visual evidence accumulation task. On each trial, they viewed circular drifting gratings presented in both, left and right visual hemifields, the contrasts of which fluctuated independently over time. After stimulus viewing, participants reported which side had the higher average contrast along with their confidence about that decision.
Within premotor/motor cortex, neural DVs encoding alternative choices were largely divided between left and right hemispheres, respectively, and intrinsically anti-correlated, consistent with competitive interactions. Further, both the “winning” and “losing” DVs in this competition made unique contributions to confidence. Critically, neural DVs from several areas of fronto-parietal association cortex also predicted choice and contributed to confidence, over and above the premotor/motor DVs: confidence scaled with (i) the number of DVs favoring the option chosen on that trial and (ii) the average strength of both winning and losing DVs across this distributed network of decision-related areas.
Our findings support the idea that confidence emerges from the consistency of neural DVs across many loosely connected local decision circuits in the cortex.