Neocortical differentiation and hippocampal integration of non-meaningful items and their spatial location

Resolving interference between similar inputs is a critical feature of adaptive memory systems. Computational theories of the medial temporal lobe posit that the dentate gyrus and CA3 (CA3DG) subfields of the hippocampus are ideally suited to reduce interference via a process called pattern separation. Whereas neocortical areas upstream of the hippocampus have been shown to play a role in content-specific (e.g., spatial or object-related) interference reduction, the CA3DG is traditionally viewed as a domain-general pattern separator. Recent work also drew attention to the role of frontal and parietal control areas in allocating resources according to task demands during mnemonic discrimination, exerting top-down control over the medial temporal lobe. However, extant evidence in humans is almost solely based on mnemonic discrimination tasks involving everyday items, potentially confounding retrieval processes with pattern separation. Here, we studied pattern separation in a “process-pure” manner, utilizing non-meaningful fractal images. We acquired full-brain high-resolution functional MRI data of 39 participants (mean age (SD) = 22.6 (2.4) years, 19 females) while they studied fractals with varying degrees of interference in either their spatial or object features. Building upon the idea that the repetition of a stimulus results in a diminished BOLD response in areas involved in the processing of that stimulus (repetition suppression), we expected that regions engaged in pattern separation of objects or locations would decrease their response to repetitions, but not to highly similar (interference-inducing) items. We found that the parahippocampal cortex contributes to interference reduction in the spatial domain, while the perirhinal cortex contributes to interference reduction in the object domain. The frontoparietal control network was recruited during the encoding of both object and location lures, and displayed strengthened within- and cross-network connectivity in response to lures. Contrary to our expectations, the CA3DG showed repetition suppression to both exact repetitions and highly similar lures, indicating a lack of sensitivity to small differences in interfering stimuli. Altogether, these results are in line with content-specific neocortical interference reduction in the medial temporal lobe, possibly orchestrated by the frontoparietal control network, but challenge the view of CA3DG as a universal pattern separator.