Retrieval aids the creation of a generalised memory trace and strengthens episode-unique information

Representational formats of human memory traces

Neural representations are internal brain states that constitute the brain's model of the external world or some of its features. In the presence of sensory input, a representation may reflect various properties of this input. When perceptual information is no longer available, the brain can still activate representations of previously experienced episodes due to the formation of memory traces. In this review, we aim at characterizing the nature of neural memory representations and how they can be assessed with cognitive neuroscience methods, mainly focusing on neuroimaging. We discuss how multivariate analysis techniques such as representational similarity analysis (RSA) and deep neural networks (DNNs) can be leveraged to gain insights into the structure of neural representations and their different representational formats. We provide several examples of recent studies which demonstrate that we are able to not only measure memory representations using RSA but are also able to investigate their multiple formats using DNNs. We demonstrate that in addition to slow generalization during consolidation, memory representations are subject to semantization already during short-term memory, by revealing a shift from visual to semantic format. In addition to perceptual and conceptual formats, we describe the impact of affective evaluations as an additional dimension of episodic memories. Overall, these studies illustrate how the analysis of neural representations may help us gain a deeper understanding of the nature of human memory.

A complementary learning systems model of how sleep moderates retrieval practice effects

While many theories assume that sleep is critical in stabilizing and strengthening memories, our recent behavioral study (Liu & Ranganath, 2021, Psychonomic Bulletin & Review, 28[6], 2035-2044) suggests that sleep does not simply stabilize memories. Instead, it plays a more complex role, integrating information across two temporally distinct learning episodes. In the current study, we simulated the results of Liu and Ranganath (2021) using our biologically plausible computational model, TEACH, developed based on the complementary learning systems (CLS) framework. Our model suggests that when memories are activated during sleep, the reduced influence of temporal context establishes connections across temporally separated events through mutual training between the hippocampus and neocortex. In addition to providing a compelling mechanistic explanation for the selective effect of sleep, this model offers new examples of the diverse ways in which the cortex and hippocampus can interact during learning.

Effects of prior knowledge on brain activation and functional connectivity during memory retrieval

Previous studies have shown that the ventral medial prefrontal cortex (vmPFC) plays an important role in schema-related memory. However, there is an intensive debate to what extent the activation of subregions of the hippocampus is involved in retrieving schema-related memory. In addition, it is unclear how the functional connectivity (FC) between the vmPFC and the hippocampus, as well as the connectivity of the vmPFC with other regions, are modulated by prior knowledge (PK) during memory retrieval over time. To address these issues, participants learned paragraphs that described features of each unfamiliar word from familiar and unfamiliar categories (i.e., high and low PK conditions) 20 min, 1 day, and 1 week before the test. They then performed a recognition task to judge whether the sentences were old in the scanner. The results showed that the activation of the anterior-medial hippocampus (amHPC) cluster was stronger when the old sentences with high (vs. low) PK were correctly retrieved. The activation of the posterior hippocampus (pHPC) cluster, as well as the vmPFC, was stronger when the new sentences with high (vs. low) PK were correctly rejected (i.e., CR trials), whereas the cluster of anterior-lateral hippocampus (alHPC) showed the opposite. The FC of the vmPFC with the amHPC and perirhinal cortex/inferior temporal gyrus was stronger in the high (vs. low) PK condition, whereas the FC of the vmPFC with the alHPC, thalamus and frontal regions showed the opposite for the CR trials. This study highlighted that different brain networks, which were associated with the vmPFC, subregions of the hippocampus and cognitive control regions, were responsible for retrieving the information with high and low PK.

Lack of source memory as a potential marker of early assimilation of novel items into current knowledge

In daily life, humans process a plethora of new information that can be either consistent (familiar) or inconsistent (novel) with prior knowledge. Over time, both types of information can integrate into our accumulated knowledge base via distinct pathways. However, the mnemonic processes supporting the integration of information that is inconsistent with prior knowledge remain under-characterized. In the current study, we used functional magnetic resonance imaging (fMRI) to examine the initial assimilation of novel items into the semantic network. Participants saw three repetitions of adjective-noun word pairs that were either consistent or inconsistent with prior knowledge. Twenty-four hours later, they were presented with the same stimuli again while undergoing fMRI scans. Outside the scanner, participants completed a surprise recognition test. We found that when the episodic context associated with initially inconsistent items was irretrievable, the neural signature of these items was indistinguishable from that of consistent items. In contrast, initially inconsistent items with accessible episodic contexts showed neural signatures that differed from those associated with consistent items. We suggest that, at least one day post encoding, items inconsistent with prior knowledge can show early assimilation into the semantic network only when their episodic contexts become inaccessible during retrieval, thus evoking a sense of familiarity.

Learning from errors: Distinct neural networks for monitoring errors and maintaining corrects through repeated practice and feedback

How memory representations are eventually established and maintained in the brain is one of central issues in memory research. Although the hippocampus and various brain regions have been shown to be involved in learning and memory, how they coordinate to support successful memory through errors is unclear. In this study, a retrieval practice (RP) - feedback (FB) paradigm was adopted to address this issue. Fifty-six participants (27 in the behavioral group, and 29 in the fMRI group) learned 120 Swahili-Chinese words associations and underwent two RP-answer FB cycles (i.e., RP1, FB1, RP2, FB2). The responses of the fMRI group were recorded in the fMRI scanner. The trials were divided based on participant's performance (correct or incorrect, C or I) during the two RPs and the final test (i.e., trial type, CCC, ICC, IIC III). The results showed that the regions in the salience and executive control networks (S-ECN) during RP, but not during FB, was strongly predictive of final successful memory. Their activation was just before the errors were corrected (i.e., RP1 in ICC trials and RP2 in IIC trials). The anterior insula (AI) is a core region in monitoring repeated errors, and it had differential connectivity with the default mode network (DMN) regions and the hippocampus during the RP and FB phases to inhibit incorrect answers and update memory. In contrast, maintaining corrected memory representation requires repeated RP and FB, which was associated with the DMN activation. Our study clarified how different brain regions support error monitoring and memory maintenance through repeated RP and FB, and emphasized the role of the insula in learning from errors.

Continuously changing memories: a framework for proactive and non-linear consolidation

The traditional view of long-term memory is that memory traces mature in a predetermined 'linear' process: their neural substrate shifts from rapidly plastic medial temporal regions towards stable neocortical networks. We propose that memories remain malleable, not by repeated reinstantiations of this linear process but instead via dynamic routes of proactive and non-linear consolidation: memories change, their trajectory is flexible and reversible, and their physical basis develops continuously according to anticipated demands. Studies demonstrating memory updating, increasing hippocampal dependence to support adaptive use, and rapid neocortical plasticity provide evidence for continued non-linear consolidation. Although anticipated demand can affect all stages of memory formation, the extent to which it shapes the physical memory trace repeatedly and proactively will require further dedicated research.

Reactivation of schema representation in lateral occipital cortex supports successful memory encoding

Schemas provide a scaffold onto which we can integrate new memories. Previous research has investigated the brain activity and connectivity underlying schema-related memory formation. However, how schemas are represented and reactivated in the brain, in order to enhance memory, remains unclear. To address this issue, we used an object–location spatial schema that was learned over multiple sessions, combined with similarity analyses of neural representations, to investigate the reactivation of schema representations of object–location memories when a new object–scene association is learned. In addition, we investigated how this reactivation affects subsequent memory performance under different strengths of schemas. We found that reactivation of a schema representation in the lateral occipital cortex (LOC) during object–scene encoding affected subsequent associative memory performance only in the schema-consistent condition and increased the functional connectivity between the LOC and the parahippocampal place area. Taken together, our findings provide new insight into how schema acts as a scaffold to support the integration of novel information into existing cortical networks and suggest a neural basis for schema-induced rapid cortical learning.

Hippocampal and medial prefrontal cortices encode structural task representations following progressive and interleaved training schedules

Memory generalisations may be underpinned by either encoding- or retrieval-based generalisation mechanisms and different training schedules may bias some learners to favour one of these mechanisms over the other. We used a transitive inference task to investigate whether generalisation is influenced by progressive vs randomly interleaved training, and overnight consolidation. On consecutive days, participants learnt pairwise discriminations from two transitive hierarchies before being tested during fMRI. Inference performance was consistently better following progressive training, and for pairs further apart in the transitive hierarchy. BOLD pattern similarity correlated with hierarchical distances in the left hippocampus (HIP) and medial prefrontal cortex (MPFC) following both training schedules. These results are consistent with the use of structural representations that directly encode hierarchical relationships between task features. However, such effects were only observed in the MPFC for recently learnt relationships. Furthermore, the MPFC appeared to maintain structural representations in participants who performed at chance on the inference task. We conclude that humans preferentially employ encoding-based mechanisms to store map-like relational codes that can be used for memory generalisation. These codes are expressed in the HIP and MPFC following both progressive and interleaved training but are not sufficient for accurate inference.

Suppression weakens unwanted memories via a sustained reduction of neural reactivation

Aversive events sometimes turn into intrusive memories. However, prior evidence indicates that such memories can be controlled via a mechanism of retrieval suppression. Here, we test the hypothesis that suppression exerts a sustained influence on memories by deteriorating their neural representations. This deterioration, in turn, would hinder their subsequent reactivation and thus impoverish the vividness with which they can be recalled. In an fMRI study, participants repeatedly suppressed memories of aversive scenes. As predicted, this process rendered the memories less vivid. Using a pattern classifier, we observed that suppression diminished the neural reactivation of scene information both globally across the brain and locally in the parahippocampal cortices. Moreover, the decline in vividness was associated with reduced reinstatement of unique memory representations in right parahippocampal cortex. These results support the hypothesis that suppression weakens memories by causing a sustained reduction in the potential to reactivate their neural representations.

OUP accepted manuscript

Retrieval practice improves retention of information in long-term memory more than restudy, but the underlying neural mechanisms of this “retrieval practice effect” (RPE) remain poorly understood. Therefore, we investigated the behavioral and neural differences between previously retrieved versus restudied items at final retrieval. Thirty younger (20–30 years old) and twenty-five older (50+ years old) adults learned familiar and new picture stimuli either through retrieval or restudy. At final recognition, hemodynamic activity was measured using functional magnetic resonance imaging (fMRI). Behaviorally, younger and older adults showed similar benefits of retrieval practice, with higher recollection, but unchanged familiarity rates. In a univariate analysis of the fMRI data, activation in medial prefrontal cortex and left temporal regions correlated with an individual’s amount of behavioral benefit from retrieval practice, irrespective of age. Compatible with this observation, in a multivariate representational similarity analysis (RSA), retrieval practice led to an increase in pattern similarity for retested items in a priori defined regions of interest, including the medial temporal lobe, as well as prefrontal and parietal cortex. Our findings demonstrate that retrieval practice leads to enhanced long-term memories in younger and older adults alike, and this effect may be driven by fast consolidation processes.

Rapid neural reorganization during retrieval practice predicts subsequent long-term retention and false memory

Active retrieval can alter the strength and content of a memory, yielding either enhanced or distorted subsequent recall. However, how consolidation influences these retrieval-induced seemingly contradictory outcomes remains unknown. Here we show that rapid neural reorganization over an eight-run retrieval practice predicted subsequent recall. Retrieval practice boosted memory retention following a 24-hour (long-term) but not 30-minute delay, and increased false memory at both delays. Long-term retention gains were predicted by multi-voxel representation distinctiveness in the posterior parietal cortex (PPC) that increased progressively over retrieval practice. False memory was predicted by unstable representation distinctiveness in the medial temporal lobe (MTL). Retrieval practice enhanced the efficiency of memory-related brain networks, through building up PPC and MTL connections with the ventrolateral and dorsolateral prefrontal cortex that predicted long-term retention gains and false memory, respectively. Our findings indicate that retrieval-induced rapid neural reorganization together with consecutive consolidation fosters long-term retention and false memories via distinct pathways.

Learning new words: Memory reactivation as a mechanism for strengthening and updating a novel word's meaning

In the present study we explored the postlearning changes in a novel word's definition using a cue-induced memory reactivation. Native speakers of Spanish (N = 373) learned low-frequency words with their corresponding definitions. The following day, reactivated groups were exposed to a reminder and provided a subjective assessment of reactivation for each word, while control groups did not receive a reactivation. Study A demonstrated that memory reactivation enhances both explicit recall and semantic integration of new meanings. Study B investigated the effect of memory reactivation in the modification of the new meanings, through three different experiments. Results show an improvement of the updated definitions according to each word's reactivation strength. In addition, congruence with previous knowledge was suggested to be a boundary condition, while consolidation time had a positive modulatory effect. Our findings call attention to reactivation as a factor allowing for malleability as well as persistence of long-term memories for words.

Stability of neural encoding moderates the contribution of sleep and repeated testing to memory consolidation

There is evidence suggesting that online consolidation during retrieval-mediated learning interacts with offline consolidation during subsequent sleep to transform memory. Here we investigate whether this interaction persists when retrieval-mediated learning follows post-training sleep and whether the direction of this interaction is conditioned by the quality of encoding resulting from manipulation of the amount of sleep on the previous night. The quality of encoding was determined by computing the degree of similarity between EEG-activity patterns across restudy of face pairs in two groups of young participants, one who slept the last 4 h of the pre-training night, and another who slept 8 h. The offline consolidation was assessed by computing the degree of coupling between slow oscillations (SOs) and spindles (SPs) during post-training sleep, while the online consolidation was evaluated by determining the degree of similarity between EEG-activity patterns recorded during the study phase and during repeated recognition of either the same face pair (i.e., specific similarity) or face pairs sharing sex and profession (i.e., categorical similarity) to evaluate differentiation and generalization, respectively. The study and recognition phases were separated by a night of normal sleep duration. Mixed-effects models revealed that the stability of neural encoding moderated the relationship between sleep- and retrieval-mediated consolidation processes over left frontal regions. For memories showing lower encoding stability, the enhanced SO-SP coupling was associated with increased reinstatement of category-specific encoding-related activity at the expense of content-specific activity, whilst the opposite occurred for memories showing greater encoding stability. Overall, these results suggest that offline consolidation during post-training sleep interacts with online consolidation during retrieval the next day to favor the reorganization of memory contents, by increasing specificity of stronger memories and generalization of the weaker ones.

Feature-specific reaction times reveal a semanticisation of memories over time and with repeated remembering

Memories are thought to undergo an episodic-to-semantic transformation in the course of their consolidation. We here test if repeated recall induces a similar semanticisation, and if the resulting qualitative changes in memories can be measured using simple feature-specific reaction time probes. Participants studied associations between verbs and object images, and then repeatedly recalled the objects when cued with the verb, immediately and after a two-day delay. Reaction times during immediate recall demonstrate that conceptual features are accessed faster than perceptual features. Consistent with a semanticisation process, this perceptual-conceptual gap significantly increases across the delay. A significantly smaller perceptual-conceptual gap is found in the delayed recall data of a control group who repeatedly studied the verb-object pairings on the first day, instead of actively recalling them. Our findings suggest that wake recall and offline consolidation interact to transform memories over time, strengthening meaningful semantic information over perceptual detail.

Consolidation and generalisation across sleep depend on individual EEG factors and sleep spindle density

Sleep is involved in both the consolidation of discrete episodes, as well as the generalisation of acquired memories into schemata. Here, we have isolated early versus late periods of sleep in order to replicate previous behavioural findings and to demonstrate: i) that distinct sleep and sleep electroencephalography (EEG) factors influence the generalisation of learned information, and; ii) that the consolidation and generalisation of memory across sleep depends on individual alpha frequency (IAF) and strength of initial encoding. Subjects underwent a night-half protocol with polysomnography (PSG), and completed a Chinese character-English paired associates learning task. Recognition accuracy of learned word-pairs, the extent to which the subject was able to generalise this knowledge, and the extent of explicit transfer of knowledge were measured. Results demonstrate that quality of initial learning determined the relationship between sleep neurophysiology and outcome, with IAF modulating this effect. We also note an effect of IAF in modulating the effect of sleep spindles in determining generalisation of learned materials. Finally, we note a complex relationship between initial learning, IAF and sleep spindle density in determining when information will reach explicit awareness across sleep. Together, these data implicate encoding factors in subsequent offline processing, demonstrate a potential role for individual differences in the EEG and subsequently add to our understanding of the the conditions in which sleep may benefit both memory and learning.

Probing the neural dynamics of mnemonic representations after the initial consolidation

Memories are not stored as static engrams, but as dynamic representations affected by processes occurring after initial encoding. Previous studies revealed changes in activity and mnemonic representations in visual processing areas, parietal lobe, and hippocampus underlying repeated retrieval and suppression. However, these neural changes are usually induced by memory modulation immediately after memory formation. Here, we investigated 27 healthy participants with a two-day functional Magnetic Resonance Imaging study design to probe how established memories are dynamically modulated by retrieval and suppression 24 h after learning. Behaviorally, we demonstrated that established memories can still be strengthened by repeated retrieval. By contrast, repeated suppression had a modest negative effect, and suppression-induced forgetting was associated with individual suppression efficacy. Neurally, we demonstrated item-specific pattern reinstatements in visual processing areas, parietal lobe, and hippocampus. Then, we showed that repeated retrieval reduced activity amplitude in the ventral visual cortex and hippocampus, but enhanced the distinctiveness of activity patterns in the ventral visual cortex and parietal lobe. Critically, reduced activity was associated with enhanced representation of idiosyncratic memory traces in the ventral visual cortex and precuneus. In contrast, repeated memory suppression was associated with reduced lateral prefrontal activity, but relative intact mnemonic representations. Our results replicated most of the neural changes induced by memory retrieval and suppression immediately after learning and extended those findings to established memories after initial consolidation. Active retrieval seems to promote episode-unique mnemonic representations in the neocortex after initial encoding but also consolidation.

Retrieval practice facilitation of family psychoeducation in people with early psychosis

BACKGROUND

Providing early psychosis (EP) individuals with family psychoeducation (FPE) can reduce symptoms and improve clinical outcomes. However, relational memory problems may limit prospective utilization of FPE information. This study examines whether memory for FPE can be improved by testing participants during the initial FPE workshop presentation.

METHOD

Data were obtained from 20 people with EP and 20 demographically matched healthy comparison subjects (HC). During session one, FPE was presented in small group workshops, with half of the information re-studied twice (re-study condition) and the remaining information tested twice using cued recall tasks (retrieval practice condition). One week later (session two), delayed cued recall was tested for all FPE information. "Testing effects" (i.e., better memory following retrieval practice versus re-study) were examined across all items (standard analysis) and also limited to items successfully retrieved during session one (conditionalized analysis).

RESULTS

HC had better initial recall and learned more over the two retrieval practice trials than EP. However, HC also lost more information than EP over the one-week delay. Both groups produced a significant testing effect. This effect was smaller in EP versus HC across all test items, but did not differ for the conditionalized analysis. Negative symptoms were inversely correlated with delayed cued recall in EP.

CONCLUSIONS

EP participants benefit from retrieval practice, with participants with less severe negative symptoms showing the greatest benefit. These results encourage use of memory tests during group psychoeducation to improve subsequent long-term recall of clinically relevant information.

Memory Engrams in the Neocortex

While in the past much of our knowledge about memory representations in the brain has relied on loss-of-function studies in which whole brain regions were temporarily inactivated or permanently lesioned, the recent development of new methods has ushered in a new era of downright "engram excitement." Animal research is now able to specifically label, track, and manipulate engram cells in the brain. While early studies have mostly focused on single brain regions like the hippocampus, recently more and more evidence for brain-wide distributed engram networks is emerging. Memory research in humans has also picked up pace, fueled by promising magnetic resonance imaging (MRI)-based methods like diffusion-weighted MRI (DW-MRI) and brain decoding. In this review, we will outline recent advancements in engram research, with a focus on human data and neocortical representations. We will illustrate the available noninvasive methods for the detection of engrams in different neocortical regions like the medial prefrontal cortex and the posterior parietal cortex and discuss evidence for systems consolidation and parallel memory encoding. Finally, we will explore how reactivation and prior knowledge can lead to and enhance engram formation in the neocortex.

How to optimize knowledge construction in the brain

Well-structured knowledge allows us to quickly understand the world around us and make informed decisions to adequately control behavior. Knowledge structures, or schemas, are presumed to aid memory encoding and consolidation of new experiences so we cannot only remember the past, but also guide behavior in the present and predict the future. However, very strong schemas can also lead to unwanted side effects such as false memories and misconceptions. To overcome this overreliance on a schema, we should aim to create robust schemas that are on the one hand strong enough to help to remember and predict, but also malleable enough to avoid such undesirable side effects. This raises the question as to whether there are ways to deliberately influence knowledge construction processes, with the goal to reach such optimally balanced schemas. Here, we will discuss how the mnemonic processes in our brains build long-term knowledge and, more specifically, how different phases of memory formation (encoding, consolidation, retrieval, and reconsolidation) contribute to this schema build-up. We finally provide ways how to best keep a balance between generalized semantic and detailed episodic memories, which can prove very useful in, e.g., educational settings.

Role of the hippocampus in the spacing effect during memory retrieval

It is well known that distributed learning (DL) leads to improved memory performance compared with massed learning (ML) (i.e., spacing effect). However, the extent to which the hippocampus is involved in the spacing effect at shorter and longer retention intervals remains unclear. To address this issue, two groups of participants were asked to encode face-scene pairs at 20-min, 1-day, and 1-month intervals before they were scanned using fMRI during an associative recognition task. The pairs were repeated six times in either a massed (i.e., six times in 1 day) or a distributed (i.e., six times over 3 days, twice per day) manner. The results showed that compared with that in the ML group, the activation of the left hippocampus was stronger in the DL group when the participants retrieved old pairs correctly and rejected new pairs correctly at different retention intervals. In addition, the posterior hippocampus was more strongly activated when the new associations were rejected correctly after DL than ML, especially at the 1-month interval. Hence, our results provide evidence that the hippocampus is involved in better memory performance after DL compared to ML at both shorter and longer retention intervals.

Semantic Memory and the Hippocampus: Revisiting, Reaffirming, and Extending the Reach of Their Critical Relationship

Since Tulving proposed a distinction in memory between semantic and episodic memory, considerable effort has been directed towards understanding their similar and unique features. Of particular interest has been the extent to which semantic and episodic memory have a shared dependence on the hippocampus. In contrast to the definitive evidence for the link between hippocampus and episodic memory, the role of the hippocampus in semantic memory has been a topic of considerable debate. This debate stems, in part, from highly variable reports of new semantic memory learning in amnesia ranging from profound impairment to full preservation, and various degrees of deficit and ability in between. More recently, a number of significant advances in experimental methods have occurred, alongside new provocative data on the role of the hippocampus in semantic memory, making this an ideal moment to revisit this debate, to re-evaluate data, methods, and theories, and to synthesize new findings. In line with these advances, this review has two primary goals. First, we provide a historical lens with which to reevaluate and contextualize the literature on semantic memory and the hippocampus. The second goal of this review is to provide a synthesis of new findings on the role of the hippocampus and semantic memory. With the perspective of time and this critical review, we arrive at the interpretation that the hippocampus does indeed make necessary contributions to semantic memory. We argue that semantic memory, like episodic memory, is a highly flexible, (re)constructive, relational and multimodal system, and that there is value in developing methods and materials that fully capture this depth and richness to facilitate comparisons to episodic memory. Such efforts will be critical in addressing questions regarding the cognitive and neural (inter)dependencies among forms of memory, and the role that these forms of memory play in support of cognition more broadly. Such efforts also promise to advance our understanding of how words, concepts, and meaning, as well as episodes and events, are instantiated and maintained in memory and will yield new insights into our two most quintessentially human abilities: memory and language.