Abstract
Introduction:
Declarative memory is supposed to rely mainly on the hippocampal and the neocortical systems. Recent evidence has shown that not only the hippocampus, but also the neocortex can acquire genuine memory engrams rapidly (Brodt, Gais et al. 2018). These findings challenge the traditional assumption of systems consolidation that time is the main factor differentiating the contribution of the two systems. This study investigates the alternative idea of concurrent memory encoding in both systems, but coding for different aspects of the same experience (Gilboa and Moscovitch 2021).
Methods:
80 participants (44 f/36 m) divided into two groups completed a visual learning paradigm consisting of abstract stimuli in different contexts during fMRI scanning. Groups only differed in the instructions on how to encode the stimuli. One group (SIM) was instructed to identify categories based on similar features while the other (DIF) had to learn to differentiate the single items based on individual features including context. 24h later a categorization and an item-context recognition task were conducted. For the encoding task, functional activity in response to different stimulus types and repetitions was compared between the two groups.
Results:
Both groups performed equally well during encoding (t78=1.65; p=0.103). SIM performed significantly better in the categorization task (t78=-6.908; p<0.001), DIF performed better in the context recognition (t78=6.307; p<0.001). Both groups showed robust encoding-related activity along the dorsal and ventral visual streams extending into the medial posterior parietal cortex (mPPC) and posterior hippocampus ([-26 -92 -10], k=18457, tpeak=20.6, pFWE<0.05). Stimuli that belonged to a category elicited higher bilateral activation of the mPPC ([16 -60 28], k=1760, tpeak=9.20, pFWE<0.05) than items without category membership. There was a greater increase of activity over category repetitions in SIM compared to DIF in higher order visual areas along the dorsal and ventral stream ([22 -50 -6], k=4118, tpeak=6.83, pFWE<0.05). In DIF compared to SIM, the second vs. first item presentation elicited higher bilateral subcortical activation of the caudate and thalamus ([-8 0 10], k=455, tpeak=5.88; pFWE<0.05).
Conclusions:
In both groups, the hippocampal and neocortical memory systems were involved in mnemonic processing, and the mPPC activated in response to item as well as category repetitions, which is in line with findings of the parietal memory network activating in response to familiar stimuli (Gilmore, Nelson et al. 2015). The manipulation of encoding strategy successfully prioritized the extraction of different mnemonic features for both groups: 24 h later, SIM showed better generalization performance, whereas DIF better remembered contextual features. These findings are complemented by neural differences during encoding. In line with their known role in visual categorization (Seger and Miller 2010), we find increased encoding-related activation in higher-order visual areas in response to category repetition in SIM. Conversely, higher activation in caudate and thalamus in response to exact item repetition in DIF might indicate that these areas are involved in the recall of specific item features, which matches their proposed roles in recognition memory (Scimeca and Badre 2012, Pergola and Suchan 2013). The thalamic activation represents a particular interesting finding, as the thalamus has recently gained more attention in memory research (Pardi, Vogenstahl et al. 2020, Ferraris, Cassel et al. 2021). Together, while the hippocampus shows similar activation between both groups, our data indicate a functional specialization of neocortical vs. subcortical areas during prioritized encoding of similar vs. differential features.
