Distinct Hippocampal Subfield Representational Shifts Underlie Category Exception Learning

When we encounter an exception to our prior category knowledge, such as a leaf-like butterfly that perceptually diverges from the butterflies we commonly see, we must update our knowledge to reconcile this exception item and generalize it to similar instances. Category knowledge updating has been suggested to rely on flexible representational modulation by the hippocampus, which involves two critical operations-pattern differentiation and integration. Yet, how separate hippocampal subfields carry out these operations has remained an outstanding puzzle in the exception learning domain. Here, we used functional magnetic resonance imaging to investigate the specific hippocampal subfield operations during which human participants (28 females and 25 males) learned exceptions that violated previously acquired regularities in two visual categories. We found that learning drove representations of exception and regular items toward differentiation in DG and integration in CA1, especially when the exception was confusable with members of the competing category instead of being a distinct oddball. Moreover, learning-related pattern differentiation and integration within DG, CA2/3, and CA1 were associated with participants' abilities to reconcile and generalize the confusable exception post-learning. Finally, the representational space in each subfield was distinctively reorganized in successful exception learners. Altogether, our findings provide fresh and nuanced insights into the hippocampal representational shifts over exception learning, contributing to a more complete picture of the hippocampal circuits in knowledge updating.