The capacity to use past experience to infer connections between things not yet directly experienced — inferential reasoning — relies on the brain’s ability to detect regularities in complex environments and use them to generate abstract representations of environmental features. However, it is unclear how abstract representations are learned, where they are encoded and how important they are for adaptive human behaviour. Now, Courellis et al. reveal that the population activity of human hippocampal neurons encodes abstract representations of task variables that support subsequent inferential reasoning.
The authors recorded population activity from different brain regions of individuals as they performed an inferential reasoning task in which they learned, via trial and error, stimulus–response–outcome associations (four images; push left or right; low or high monetary reward). Participants learned two related stimulus–response–outcome mappings (contexts), which they did not know were orthogonalized (that is, once they knew all stimulus–response mappings in one context then, by inverting them, they could also know all stimulus–response mappings in the other context). During the task, the two learned contexts could randomly switch every so often without participants receiving an overt signal about this change. Thus, after a context switch, participants needed to infer from the trial feedback error that the context had changed and then update the stimulus–response–outcome mapping they used during the subsequent trial. The authors used the response accuracy on that subsequent trial to determine the presence or absence of such inferential reasoning in their analysis of the neural population activity.
留言 (0)