A dual-subsystem model of the brain's default network: self-referential processing, memory retrieval processes, and autobiographical memory retrieval

Most internally oriented mental activities are known to strongly activate the default network, which includes remembering the past, future thinking and social cognition, and are heavily self-referential, and demanding of memory retrieval processes. Based on these observations and building on related findings from the literature, the present article proposed a simple, dual-subsystem model of the default network. The ability of the model to estimate brain activity during autobiographical memory (AM) retrieval and related reference conditions was then tested by performing a quantitative meta-analysis of relevant literature. The model divided the default network into two subsystems. The first, called the 'cortical midline subsystem (CMS)', was comprised of the anteromedial prefrontal cortex and posterior cingulate cortex, and primarily mediates self-referential processing. The other, termed the 'parieto-temporal subsystem (PTS)', included the inferior parietal lobule, medial temporal lobe and lateral temporal cortex, and mainly supports memory retrieval processes. The meta-analysis of AM retrieval contrasts yielded a double dissociation that was consistent with this model. First, CMS regions associated more with an AM>laboratory-based memory (LM) contrast than with an AM>rest contrast, confirming that these regions play more critical roles in self-referential processing than memory retrieval processes. Second, all three PTS regions showed a greater association with an AM>rest contrast than with an AM>LM contrast, confirming that their role in memory retrieval processes is greater than in self-referential processing. Although the present model is limited in scope, both in terms of anatomical and functional specifications, it integrates diverse processes such as self-referential processing, episodic and semantic memory and subsystem interface, and provides useful heuristics that can guide further research on fractionation of the default network.     

Functional network in the prefrontal cortex during episodic memory retrieval

A recent consistent finding in neuroimaging studies of human memory is that the prefrontal cortex (PFC) is activated during episodic memory retrieval. To date, however, there has been no direct evidence to explain how activity in the right and left PFC and in the anterior and posterior PFC are functionally interconnected. The goal of the present study was to obtain such evidence by event-related functional magnetic resonance imaging (MRI) and the functional connectivity method. Subjects were first asked to try to remember a series of associate-word lists outside the MRI scanner in preparation for a later recognition test. In the MRI scanning phase, they were asked to make recognition judgments in regard to old words, semantically related lure words, and unrelated new words. The analysis of functional connectivity revealed that the posterior PFC in each hemisphere had strong functional interconnections with the contralateral posterior PFC, whereas the anterior PFC in each hemisphere had only weak functional interconnections with the contralateral anterior PFC. No strong functional interconnections were found between the anterior and posterior PFC in either hemisphere. These findings support the hypothesis of an associative contribution of the bilateral posterior PFC to episodic memory retrieval and a dissociative contribution of the bilateral anterior PFC.     

Left mediotemporal structures mediate the retrieval of episodic autobiographical mental images

The aim of this functional magnetic resonance imaging (fMRI) study was to investigate the neuroanatomical substrates associated with the process of mental generation of specific (i.e., exemplar) and episodic autobiographical (i.e., an image of a unique life episode connected with an object) images. The fMRI paradigm in this experiment included a non-image generation baseline and two activation conditions requiring the generation of either specific or episodic autobiographical images. Image generation times and brain activation were recorded. Behavioral results showed that generating specific mental images took significantly less than generating episodic autobiographical images. Individuals generated specific images that were well distinct from the episodic autobiographical ones, semantic in nature without an episodic reference. Episodic autobiographical images did not show a significant bias towards preferential retrieval from any particular life period but were retrieved from across the entire life span. Conjunction analysis of the fMRI data showed that the two image generation conditions significantly activated a common set of neural structures, including mediofrontal areas. This shared pattern of activation might be the result of an underlying similar format and characteristics (e.g., richness in details) between the two types of images and might reflect the involvement of similar cognitive processes. Distinct patterns of significant activation were also present. Activation in the right parietal regions, cuneus, precuneus and left temporal regions was associated solely with the generation of specific images. Regions more specifically devoted to episodic memory retrieval and imagery, such as the left parahippocampal gyrus and precuneus, and the posterior cingulate cortex bilaterally, were significantly activated exclusively by the generation of episodic autobiographical images.     

Investigating the functional interaction between semantic and episodic memory: convergent behavioral and electrophysiological evidence for the role of familiarity

Throughout our lives we acquire general knowledge about the world (semantic memory) while also retaining memories of specific events (episodic memory). Although these two forms of memory have been dissociated on the basis of neuropsychological data, it is clear that they typically function together during normal cognition. The goal of the present study was to investigate this interaction. One influence of semantic memory on episodic retrieval is 'Levels Of Processing'; recognition is enhanced when stimuli are processed in a semantically meaningful way. Studies examining this semantic processing advantage have largely concluded that semantic memory augments episodic retrieval primarily by enhancing recollection. The present study provides strong evidence for an alternative relationship between semantic and episodic memory. We employed a manipulation of the semantic coherence of to-be-remembered information (semantically related vs. unrelated word pairs) during an associative recognition memory test. Results revealed that associative recognition is significantly enhanced for semantically coherent material, and behavioral estimates (using the process dissociation procedure) demonstrated concomitant changes in the contribution of familiarity to retrieval. Neuroimaging data (event-related potentials recorded at test) also revealed a significant increase in familiarity based retrieval. The electrophysiological correlate of familiarity (the mid-frontal ERP old/new effect) was larger for semantically related compared to unrelated word pairs, but no difference was present in the electrophysiological correlate of recollection (the left parietal old/new effect). We conclude that semantic memory and episodic memory do indeed interact in normal functioning, and not only by modulating recollection, but also by enhancing familiarity.     

Similarities and differences in the neural correlates of episodic memory retrieval and working memory

Functional neuroimaging studies have shown that different cognitive functions activate overlapping brain regions. An activation overlap may occur because a region is involved in operations tapped by different cognitive functions or because the activated area comprises subregions differentially involved in each of the functions. To investigate these issues, we directly compared brain activity during episodic retrieval (ER) and working memory (WM) using event-related functional MRI (fMRI). ER was investigated with a word recognition test, and WM was investigated with a word delayed-response test. Two-phase trials distinguished between retrieval mode and cue-specific aspects of ER, as well as between encoding/maintenance and retrieval aspects of WM. The results revealed a common fronto-parieto-cerebellar network for ER and WM, as well as subregions differentially involved in each function. Specifically, there were two main findings. First, the results differentiated common and specific subregions within the prefrontal cortex: (i) left dorsolateral areas were recruited by both functions, possibly reflecting monitoring operations; (ii) bilateral anterior and ventrolateral areas were more activated during ER than during WM, possibly reflecting retrieval mode and cue-specific ER operations, respectively; and (iii) left posterior/ventral (Broca's area) and bilateral posterior/dorsal areas were more activated during WM than during ER, possibly reflecting phonological and generic WM operations, respectively. Second, hippocampal and parahippocampal regions were activated not only for ER but also for WM. This result suggests that indexing operations mediated by the medial temporal lobes apply to both long-term and short-term memory traces. Overall, our results show that direct cross-function comparisons are critical to understand the role of different brain regions in various cognitive functions.     

Abnormal insula functional network is associated with episodic memory decline in amnestic mild cognitive impairment

Abnormalities of functional connectivity in the default mode network (DMN) recently have been reported in patients with amnestic mild cognitive impairment (aMCI), Alzheimer's disease (AD) or other psychiatric diseases. As such, these abnormalities may be epiphenomena instead of playing a causal role in AD progression. To date, few studies have investigated specific brain networks, which extend beyond the DMN involved in the early AD stages, especially in aMCI. The insula is one site affected by early pathological changes in AD and is a crucial hub of the human brain networks. Currently, we explored the contribution of the insula networks to cognitive performance in aMCI patients. Thirty aMCI and 26 cognitively normal (CN) subjects participated in this study. Intrinsic connectivity of the insula networks was measured, using the resting-state functional connectivity fMRI approach. We examined the differential connectivity of insula networks between groups, and the neural correlation between the altered insula networks connectivity and the cognitive performance in aMCI patients and CN subjects, respectively. Insula subregional volumes were also investigated. AMCI subjects, when compared to CN subjects, showed significantly reduced right posterior insula volumes, cognitive deficits and disrupted intrinsic connectivity of the insula networks. Specifically, decreased intrinsic connectivity was primarily located in the frontal-parietal network and the cingulo-opercular network, including the anterior prefrontal cortex (aPFC), anterior cingulate cortex, operculum, inferior parietal cortex and precuneus. Increased intrinsic connectivity was primarily situated in the visual-auditory pathway, which included the posterior superior temporal gyrus and middle occipital gyrus. Conjunction analysis was performed; and significantly decreased intrinsic connectivity in the overlapping regions of the anterior and posterior insula networks, including the bilateral aPFC, left dorsolateral prefrontal cortex, dorsomedial prefrontal cortex, and anterior temporal pole was found. Furthermore, the disrupted intrinsic connectivity was associated with episodic memory (EM) deficits in the aMCI patients and not in the CN subjects. These findings demonstrated that the functional integration of the insula networks plays an important role in the EM process. They provided new insight into the neural mechanism underlying the memory deficits in aMCI patients.     

A robust deep neural network for denoising task-based fMRI data: An application to working memory and episodic memory

In this study, a deep neural network (DNN) is proposed to reduce the noise in task-based fMRI data without explicitly modeling noise. The DNN artificial neural network consists of one temporal convolutional layer, one long short-term memory (LSTM) layer, one time-distributed fully-connected layer, and one unconventional selection layer in sequential order. The LSTM layer takes not only the current time point but also what was perceived in a previous time point as its input to characterize the temporal autocorrelation of fMRI data. The fully-connected layer weights the output of the LSTM layer, and the output denoised fMRI time series is selected by the selection layer. Assuming that task-related neural response is limited to gray matter, the model parameters in the DNN network are optimized by maximizing the correlation difference between gray matter voxels and white matter or ventricular cerebrospinal fluid voxels. Instead of targeting a particular noise source, the proposed neural network takes advantage of the task design matrix to better extract task-related signal in fMRI data. The DNN network, along with other traditional denoising techniques, has been applied on simulated data, working memory task fMRI data acquired from a cohort of healthy subjects and episodic memory task fMRI data acquired from a small set of healthy elderly subjects. Qualitative and quantitative measurements were used to evaluate the performance of different denoising techniques. In the simulation, DNN improves fMRI activation detection and also adapts to varying hemodynamic response functions across different brain regions. DNN efficiently reduces physiological noise and generates more homogeneous task-response correlation maps in real data.     

Routes to the past: neural substrates of direct and generative autobiographical memory retrieval

Models of autobiographical memory propose two routes to retrieval depending on cue specificity. When available cues are specific and personally-relevant, a memory can be directly accessed. However, when available cues are generic, one must engage a generative retrieval process to produce more specific cues to successfully access a relevant memory. The current study sought to characterize the neural bases of these retrieval processes. During functional magnetic resonance imaging (fMRI), participants were shown personally-relevant cues to elicit direct retrieval, or generic cues (nouns) to elicit generative retrieval. We used spatiotemporal partial least squares to characterize the spatial and temporal characteristics of the networks associated with direct and generative retrieval. Both retrieval tasks engaged regions comprising the autobiographical retrieval network, including hippocampus, and medial prefrontal and parietal cortices. However, some key neural differences emerged. Generative retrieval differentially recruited lateral prefrontal and temporal regions early on during the retrieval process, likely supporting the strategic search operations and initial recovery of generic autobiographical information. However, many regions were activated more strongly during direct versus generative retrieval, even when we time-locked the analysis to the successful recovery of events in both conditions. This result suggests that there may be fundamental differences between memories that are accessed directly and those that are recovered via the iterative search and retrieval process that characterizes generative retrieval.     

Ecphory of autobiographical memories: an fMRI study of recent and remote memory retrieval

Ecphory occurs when one recollects a past event cued by a trigger, such as a picture, odor, or name. It is a central component of autobiographical memory, which allows us to "travel mentally back in time" and re-experience specific events from our personal past. Using fMRI and focusing on the role of medial temporal lobe (MTL) structures, we investigated the brain bases of autobiographical memory and whether they change with the age of memories. Importantly, we used an ecphory task in which the remote character of the memories was ensured. The results showed that a large bilateral network supports autobiographical memory: temporal lobe, temporo-parieto-occipital junction, dorsal prefrontal cortex, medial frontal cortex, retrosplenial cortex and surrounding areas, and MTL structures. This network, including MTL structures, changed little with the age of the memories.     

Familiarity modulates the functional relationship between theory of mind and autobiographical memory

Qualitative and quantitative reviews of the neuroimaging literature show that overlapping brain regions support theory of mind (ToM) and autobiographical memory (AM). This overlap has been taken to suggest that individuals draw on past personal experiences to infer others' mental states, but work with amnesic people shows that ToM does not always depend on AM. One variable that may determine the extent to which one relies on AM when inferring another's thoughts and feelings during ToM is whether that individual is personally known. To test this possibility, participants were scanned with fMRI as they remembered past experiences in response to personal photos ('AM' condition) and imagined others' experiences in response to photos of personally familiar ('pToM' condition) and unfamiliar ('ToM' condition) others. Spatiotemporal Partial Least Squares was used to identify the spatial and temporal characteristics of neural activation patterns associated with AM, pToM, and ToM. We found that the brain regions supporting pToM more closely resembled those supporting AM relative to ToM involving unfamiliar others, with the greatest degree of overlap within midline regions. A complementary finding was the observation of striking differences between pToM and ToM such that midline regions associated with AM predominated during pToM, whereas more lateral regions associated with social semantic memory predominated during ToM. Overall, this study demonstrates that ToM involves a dynamic interplay between AM and social semantic memory that is biased towards AM when a personally familiar other is the subject of the mental state inference.     

The role of precuneus and left inferior frontal cortex during source memory episodic retrieval

The posterior medial parietal cortex and left prefrontal cortex (PFC) have both been implicated in the recollection of past episodes. In a previous study, we found the posterior precuneus and left lateral inferior frontal cortex to be activated during episodic source memory retrieval. This study further examines the role of posterior precuneal and left prefrontal activation during episodic source memory retrieval using a similar source memory paradigm but with longer latency between encoding and retrieval. Our results suggest that both the precuneus and the left inferior PFC are important for regeneration of rich episodic contextual associations and that the precuneus activates in tandem with the left inferior PFC during correct source retrieval. Further, results suggest that the left ventro-lateral frontal region/frontal operculum is involved in searching for task-relevant information (BA 47) and subsequent monitoring or scrutiny (BA 44/45) while regions in the dorsal inferior frontal cortex are important for information selection (BA 45/46).     

Characterizing spatial and temporal features of autobiographical memory retrieval networks: a partial least squares approach

Conway (Conway, M.A., 1992. A structural model of autobiographical memory. In: Conway, M.A., Spinnler, H., Wagenaar, W.A. (Eds.), Theoretical Perspectives on Autobiological Memory. Kluwer Academic Publishers, Dordrecht, The Netherlands, pp. 167-194) proposed that two types of autobiographical memories (AMs) exist within a hierarchical AM system: unique, specific events and repeated, general memories. There is little research on whether retrieval of these AMs relies on different neural substrates. To investigate this issue, we used a multivariate image analysis technique, spatiotemporal partial least squares (PLS), to identify distributed patterns of activity most related to AM tasks that we have found to be associated with a medial and left-lateralized network. Using PLS, specific and general memories were more strongly associated with different parts of this retrieval network. Specific AM retrieval was associated more with activation of regions involved in imagery in episodic memory, including the left precuneus, left superior parietal lobule and right cuneus, whereas general AM retrieval was associated with activation of the right inferior temporal gyrus, right medial frontal cortex, and left thalamus. These two patterns emerged at different lags after stimulus onset, with the general AM pattern peaking between 2 and 6 s, and the specific AM pattern between 6 and 8 s. These lag differences are consistent with Conway's theory which posits that general AMs are the preferred level of entry to the AM system. A seed PLS analysis revealed that the regions functionally connected to the hippocampus during retrieval did not differentiate specific from general AM retrieval, which confirms our earlier univariate analysis indicating that some aspects of the memory retrieval network are shared by these memories.     

The brain network associated with acquiring semantic knowledge

There is ongoing debate about how semantic information is acquired, whether this occurs independently of episodic memory, and what role, if any, brain areas such as hippocampus are required to play. We used auditory stimuli and functional MRI (fMRI) to assess brain activations associated with the incidental acquisition of new and true facts about the world of the sort we are exposed to day to day. A control task was included where subjects heard sentences that described novel scenarios involving unfamiliar people, but these did not convey general knowledge. The incidental encoding task was identical for two stimulus types; both shared the same episodic experience (lying in the brain scanner) and conveyed complex information. Despite this, and considering only those stimuli successfully encoded, compared to a baseline task, a more extensive network of brain regions was found to be associated with exposure to new facts including the hippocampus. Direct comparison between the two stimulus types revealed greater activity in dorsal, ventrolateral and dorsomedial prefrontal cortex, medial dorsal nucleus of the thalamus, and temporal cortex for fact stimuli. The findings suggest that successful encoding is not invariably associated with activation of one particular brain network. Rather, activation patterns may depend on the type of materials being acquired, and the different processes they engender when subjects encode. Qualitatively, from postscan debriefing sessions, it emerged that the factual information was found to be potentially more useful. We suggest that current or prospective utility of incoming information may be one factor that influences the processes engaged during encoding and the concomitant neuronal responses.     

Structural brain network characteristics in patients with episodic and chronic migraine

BackgroundMigraine is a primary headache disorder that can be classified into an episodic (EM) and a chronic form (CM). Network analysis within the graph-theoretical framework based on connectivity patterns provides an approach to observe large-scale structural integrity. We test the hypothesis that migraineurs are characterized by a segregated network.Methods19 healthy controls (HC), 17 EM patients and 12 CM patients were included. Cortical thickness and subcortical volumes were computed, and topology was analyzed using a graph theory analytical framework and network-based statistics. We further used support vector machines regression (SVR) to identify whether these network measures were able to predict clinical parameters.ResultsNetwork based statistics revealed significantly lower interregional connectivity strength between anatomical compartments including the fronto-temporal, parietal and visual areas in EM and CM when compared to HC. Higher assortativity was seen in both patients’ group, with higher modularity for CM and higher transitivity for EM compared to HC. For subcortical networks, higher assortativity and transitivity were observed for both patients’ group with higher modularity for CM. SVR revealed that network measures could robustly predict clinical parameters for migraineurs.ConclusionWe found global network disruption for EM and CM indicated by highly segregated network in migraine patients compared to HC. Higher modularity but lower clustering coefficient in CM is suggestive of more segregation in this group compared to EM. The presence of a segregated network could be a sign of maladaptive reorganization of headache related brain circuits, leading to migraine attacks or secondary alterations to pain.Supplementary InformationThe online version contains supplementary material available at 10.1186/s10194-021-01216-8.     

SD-CNN: A static-dynamic convolutional neural network for functional brain networks

Static functional connections (sFCs) and dynamic functional connections (dFCs) have been widely used in the resting-state functional MRI (rs-fMRI) analysis. sFCs, calculated based on entire rs-fMRI scans, can accurately describe the static topology of the brain network. dFCs, estimated by dividing rs-fMRI scans into a series of short sliding windows, are used to reveal time-varying changes in FC patterns. Currently, how to jointly use sFCs and dFCs to identify brain diseases under the framework of deep learning is still a hot issue. To this end, we propose a static-dynamic convolutional neural network for functional brain networks, which involves a static pathway and a dynamic pathway for taking full advantages of sFCs and dFCs. Specifically, the static pathway, using high-resolution convolution filters (i.e., convolution filters with a high number of channels) at a single adjacency matrix of sFCs, is performed to capture static FC patterns. The dynamic pathway, using low-resolution convolution filters at each adjacency matrix of dFCs, is performed to capture time-varying FC patterns. Two types of diffusion connections are used in this model for encouraging the transfer of information between the static pathway and the dynamic pathway, which can make the learned features more discriminative. Furthermore, a static and dynamic combination classifier is introduced to combine features from two pathways for identifying brain diseases. Experiments on two real datasets demonstrate the effectiveness and advantages of our proposed method.     

A study of semantic memory after brain injury: Learning newly coined French words

ObjectiveTo investigate semantic memory in brain-injured patients.MethodsWe used the new word questionnaire (QMN) to assess the ability of 12 brain-injured patients and 12 healthy controls to define French words, which had been admitted to the dictionary in 1996 to 1997 or in 2006 to 2007.ResultsDespite amnesia or severe executive disorders, the brain-injured patients were able to learn new words and remember those that they already learnt. They successfully selected the relevant phrase in which the new word was placed and were reasonably good at recognizing the right definition from among decoys. In contrast, they had trouble defining the words and compensated for this by giving examples. These problems were correlated with their vocabulary and executive function scores in a battery of neuropsychological tests.ConclusionOur results suggest that frontal injury leads to an impairment in accurate word selection and the scheduling abilities required to generate word definitions.     

The role of the basal forebrain in episodic memory retrieval: a positron emission tomography study

Human lesion data indicate that the basal forebrain or orbitofrontal cortex, or both, as well as medial temporal and diencephalic structures, is important for normal memory and that its disruption causes the pure amnesic syndrome, in which episodic memory is grossly impaired while other kinds of memory remain preserved. Among these critical areas, functional imaging studies have so far failed to detect activation of the basal forebrain, although activation in the nearby orbitofrontal cortex has been reported during episodic memory retrieval. We employed positron emission tomography to elucidate the neural basis of episodic memory recall utilizing two types of time cues and successfully detected activity in the basal forebrain for the first time. Specifically, recall of previously memorized words from temporal cues was associated with activity in the basal forebrain, right middle frontal gyrus, right superior temporal gyrus, and posterior cingulate gyrus, whereas their recall from person cues was associated with activity in the left insula, right middle frontal gyrus, and posterior cingulate gyrus. Furthermore, percentage increases of regional blood flow in the basal forebrain were correlated with behavioral data of successful recall. Our results provide clear evidence that the human basal forebrain has a specific role in episodic memory recall, especially that from time-contextual information.     

Acquisition versus retrieval deficits in traumatic brain injury: implications for memory rehabilitation

OBJECTIVE: To examine whether impaired memory in persons with traumatic brain injury (TBI) is caused by impaired initial acquisition or compromised retrieval from long-term storage. DESIGN: Prospective matched controlled trial. SETTING: Postacute rehabilitation institute. PARTICIPANTS: Patients with moderate to severe TBI (n = 28) and 21 matched, healthy controls (HCs). MAIN OUTCOME MEASURES: Patients with TBI and HC subjects were equated on initial acquisition on a verbal list-learning task. Recall and recognition performance was then evaluated at 30- and 90-minutes after learning. RESULTS: All HC subjects and 20 subjects with TBI (TBI-MET) were able to meet the learning criterion, but the TBI-MET group took significantly more trials than HC subjects to do so. However, after equating groups on acquisition, the TBI-MET group did not differ from controls on recall and recognition at both the 30- and 90-minute delays. Eight TBI subjects showed severe learning deficits (TBI-NOT MET) because they never learned the task, and showed significantly impaired recall and recognition performance. The 2 TBI groups did not differ on measures of severity of injury, but the TBI-NOT MET group performed significantly below the TBI-MET group on executive functioning. Rate of forgetting did not differ across the 3 groups. CONCLUSIONS: Results suggest that memory impairment after TBI is caused primarily by deficiencies in initial acquisition of verbal information rather than in compromised retrieval. The findings have significant implications for the rehabilitation and treatment of individuals with TBI.     

The representation of social interaction in episodic memory: a functional MRI study

The representation of social interaction in episodic memory is a critical factor for the successful navigation of social relationships. In general, it is important to separate episodic memory during social interaction from episodic memory during the self-generation of action events. Different cortical representations have been associated with social interaction vs. self-generated episodic memory. Here we clarified the cortical representation of the effect of context (social vs. solitary) on episodic memory by comparing it with the generation effect (self vs. other) on episodic memory. Each participant learned words while engaged in a sentence generation and a reading task, and subsequently each participant was scanned with functional magnetic resonance imaging (fMRI) while they performed a recognition task using the words that had been learned. The experiment was comprised of four conditions and we looked at two situations that involved a social context and non-social (solitary) context task. In the learning session before entering the MRI, two participants collaborated in a social context either generating (social-contextual self-generation condition: SS) or reading (social-contextual other-generation condition: SO) a sequence of sentences alternately to construct a meaningful story narrative. In the non-social context, the participants generated (non-social-contextual self-generation condition: NS) or read (non-social-contextual other-generation condition: NO) a sequence of sentences individually. The stimuli for the recognition session consisted of learned words and novel words. Activation for social context retrieval was identified in the right medial prefrontal cortex (mPFC), and activation for self-generated retrieval was identified in the left mPFC and the left middle cingulate cortex. These results indicate that dissociable regions within the medial prefrontal cortices contribute to the processes involved in the representation of social interaction, including social context and self-generation in the retrieval of episodic memory.