Direct brain recordings identify hippocampal and cortical networks that distinguish successful versus failed episodic memory retrieval

Abstract

Human data collected using noninvasive imaging techniques have established the importance of parietal regions towards episodic memory retrieval, including the angular gyrus and posterior cingulate cortex. Such regions comprise part of a putative core episodic retrieval network. In free recall, comparisons between contextually appropriate and inappropriate recall events (i.e. prior list intrusions) provide the opportunity to study memory retrieval networks supporting veridical recall, and existing findings predict that differences in electrical activity in these brain regions should be identified according to the accuracy of recall. However, prior iEEG studies, utilizing principally subdural grid electrodes, have not fully characterized brain activity in parietal regions during memory retrieval and have not examined connectivity between core recollection areas and the hippocampus or prefrontal cortex. Here, we employed a data set obtained from 100 human patients implanted with stereo EEG electrodes for seizure mapping purposes as they performed a free recall task. This data set allowed us to separately analyze activity in midline versus lateral parietal brain regions, and in anterior versus posterior hippocampus, to identify areas in which retrieval-related activity predicted the recollection of a correct versus an incorrect memory. With the wide coverage afforded by the stereo EEG approach, we were also able to examine interregional connectivity. Our key findings were that differences in gamma band activity in the angular gyrus, precuneus, posterior temporal cortex, and posterior (more than anterior) hippocampus discriminated accurate versus inaccurate recall as well as active retrieval versus memory search. The left angular gyrus exhibited a significant power decrease preceding list intrusions as well as unique phase-amplitude coupling properties, whereas the prefrontal cortex was unique in exhibiting a power increase during list intrusions. Analysis of connectivity revealed significant hemispheric asymmetry, with relatively sparse leftsided functional connections compared to the right hemisphere. One exception to this finding was elevated connectivity between the prefrontal cortex and left angular gyrus. This finding is interpreted as evidence for the engagement of prefrontal cortex in memory monitoring and mnemonic decision-making.

Neurons in the human medial temporal lobe track multiple temporal contexts during episodic memory processing

“…For analyzing retrieval-related activity, we excluded any list intrusions and only included items isolated from previous and subsequent recall events to avoid noise from adjacent vocalizations. As with previous studies, however ( Tan et al, 2020 ), this requirement did not significantly alter the response probability relationships on the subject level (see Fig. 2d for an example, Fig.…”

“…In our free recall paradigm, memory retrieval occurred within a 30 or 45 s period when no stimuli appeared on the screen and subjects vocalized items remembered from the immediately preceding list. We analyzed 1000-ms time windows preceding the vocalization of retrieved memory items, limiting our analysis to retrieval events that were sufficiently temporally isolated from others to avoid effects of vocalization ( Manning et al, 2011 ; Burke et al, 2014a ; Tan et al, 2020 ). As a result, between 5 and 112 (median 27) paired items were used for reinstatement analyses among SME and NonSME neurons.…”

“…Future investigations employing alternative paradigms, such as categorized free recall or repeated free recall (in which items are repeated across lists), can provide data capable of dissociating item-level (semantic) information versus contextual information. Likewise, the establishment of larger datasets, with subjects that provide a larger number of list intrusions, would allow us to address important questions such as how the representation of temporal context in SME neurons changes when the retrieval of appropriate temporal context fails but a retrieval attempt is made (for a discussion of the interpretation of list intrusions, please see Tan et al (2020) ). We also note that a limitation of free recall is the lack of control over the precise timing of retrieval-related activity, and therefore identifying common patterns of SME neurons across both associative recognition ( Kota et al, 2020 ) and free recall paradigms would provide important additional information regarding the behavior of SME neurons.…”

“…For the analysis of behavior-related data, we examined 1.5 s window following item presentation during encoding, and 1 s preceding a recall event during the retrieval period. For valid recall events for further analyses, we selected those were temporally separated from adjacent vocalizations by 1 s to avoid including vocalization-related activity ( Tan et al, 2020 ). This method of filtering the “temporally isolated” recall events was adopted from previous studies ( Burke et al, 2014b ; Lin et al, 2017 ).…”

Intracranial EEG in Episodic Memory Research: Possibilities, Limitations, Results

“…Как правило, эти исследования используют данные, накопленные в нескольких крупных хирургических центрах в течение нескольких лет. За счет сотрудничества нескольких клиник эти выборки получаются достаточно крупными и насчитывают десятки и сотни пациентов Tan et al, 2020). Как правило, используется методика свободного припоминания вербального материала: пациенты запоминают списки слов и после выполнения задачи-дистрактора в свободном порядке вспоминают так много слов из списка, как только смогут.…”