FV 3 Large-scale informational connectivity tracks flexible sensory-motor mapping rules in the human brain

Background: Many choice tasks entail arbitrary rules that govern the mapping from states of the sensory environment (e.g. orientation of a grating stimulus) to motor response (e.g., left- or right-hand movement). Animals and humans can flexibly change this sensory-motor mapping following a change of rule. Yet, it has remained unknown how the brain implements such flexible sensory-motor mapping.

Objective: We aimed to probe the flexible configuration of large-scale networks of neural populations encoding stimulus features and motor actions during sensory-motor mapping behavior.

Methods: Human participants switched between two rules for reporting visual orientation judgments during fMRI recordings. Rule switches were either instructed explicitly or inferred by the participants from ambiguous cues. We used behavioral modeling to reconstruct the time course of participants’ hidden belief about the active rule. We trained and applied decoders on the activity of visual and motor cortical regions to extract spontaneous (ongoing) fluctuations in the population codes for stimulus features and action, and examined rule-specific co-variation between these population codes. Rather than correlating region-averaged fMRI signals as commonly done in conventional ‘functional connectivity’ analyses, this so-called ‘informational connectivity’ quantified co-variation of specific task-related features contained within activity fluctuations.

Results: In both task contexts (instructed and inferred rule), the patterns of correlations between ongoing fluctuations in stimulus- and action-selective activity across a network of visual and action-related brain regions tracked participants’ belief about the active rule. The instantaneous, model-inferred belief state could be robustly predicted (with cross-validation) from local patterns of informational connectivity between stimulus- and action-codes. Furthermore, the rule-specific correlation patterns were strongly linked to behavior: they broke down around the time of behavioral errors, and the strength of these patterns co-varied across participants with their individual model-estimated internal noise corrupting the implementation of the rule based on belief state. Locally encoded representations of the active rule in parietal and frontal areas also co-varied with stimulus-action decoder coupling. As expected, conventional functional connectivity between the average signals from the same regions did not reflect the rule or rule belief.

Conclusion: Our results indicate that sensory-motor mapping rules manifest in specific patterns of informational connectivity within the sensory-motor network, which can be reconfigured adaptively to support task demands. We conclude that internal beliefs about task state are instantiated in brain-wide patterns of correlated variability of stimulus and action codes. More generally, informational connectivity provides a more powerful approach to unraveling distributed cognitive computations than afforded by conventional functional connectivity analyses.