Did H.M. exhibit accelerated long-term forgetting? Measuring forgetting in amnesia.

The early investigations of patient H.M. inaugurated the modern era of memory research. During the 1970s and 1980s, a key debate over whether H.M. with bilateral medial temporal lobe lesions exhibited accelerated long-term forgetting attracted an increasing interest in forgetting research among amnestic patients. Huppert and Piercy (1979) examined H.M.'s performance in visual recognition at 10-minute, 1-day, and 7-day intervals and suggested that H.M. was subjected to rapid forgetting compared with Korsakoff patients and healthy participants reported in Huppert and Piercy (1978). In contrast, Freed et al. (1987) employed the same experimental paradigm and concluded that forgetting rates in H.M. did not differ from those in healthy controls. These incompatible findings highlighted a methodological challenge in measuring forgetting in the cross-group comparison design, where closely equalising the initial performance between patient and control groups is usually suggested. The re-analysis in this viewpoint, using both linear- and nonlinear-based modelling, reconciled the discrepancy between the aforementioned studies. Our results indicated that the rate of forgetting in H.M. did not differ from that in healthy controls, regardless of whether the initial performance was closely matched. Here, we suggest that the cross-group comparisons in forgetting studies do not necessarily seek a perfect match in initial performance unless the risks of confounding encoding and retrieval processes can be effectively controlled.

Detecting accelerated long-term forgetting remotely in a community sample of people with epilepsy: Evidence from the Crimes and Four Doors tests.

People with epilepsy often report experiencing memory problems though these are not always detectable using standard neuropsychological measures. One form of difficulty that may be relatively prevalent in epilepsy is termed accelerated long-term forgetting (ALF), typically described as relatively greater loss of memory over days or weeks following initial encoding. The current study used remote assessment to examine memory and forgetting over one week in a broad community sample of people with epilepsy and healthy control participants, using two recently developed tests, one verbal (the Crimes test) and one visual (the Four Doors test). These were administered as part of a short battery of cognitive measures, run remotely with participants over Zoom. Across this community-derived sample, people with epilepsy reported more memory complaints and demonstrated significantly faster forgetting on both the verbal and visual tests. This difference was not attributable to level of initial learning performance and was not detectable through delayed recall on a standard existing test. Our results suggests that ALF may be more common than suspected in people with epilepsy, leading to a potentially important source of memory problems that are currently undetected by standard memory tests.

Sparse and Expandable Network for Google's Pathways.

Introduction: Recently, Google introduced Pathways as its next-generation AI architecture. Pathways must address three critical challenges: learning one general model for several continuous tasks, ensuring tasks can leverage each other without forgetting old tasks, and learning from multi-modal data such as images and audio. Additionally, Pathways must maintain sparsity in both learning and deployment. Current lifelong multi-task learning approaches are inadequate in addressing these challenges. Methods: To address these challenges, we propose SEN, a Sparse and Expandable Network. SEN is designed to handle multiple tasks concurrently by maintaining sparsity and enabling expansion when new tasks are introduced. The network leverages multi-modal data, integrating information from different sources while preventing interference between tasks. Results: The proposed SEN model demonstrates significant improvements in multi-task learning, successfully managing task interference and forgetting. It effectively integrates data from various modalities and maintains efficiency through sparsity during both the learning and deployment phases. Discussion: SEN offers a straightforward yet effective solution to the limitations of current lifelong multi-task learning methods. By addressing the challenges identified in the Pathways architecture, SEN provides a promising approach for developing AI systems capable of learning and adapting over time without sacrificing performance or efficiency.

Durability of retrieval-induced forgetting: Effects of different practice schedules.

If retrieval-induced forgetting (RIF) is to play a role in the formation of collective memories, it should be long lasting. Although several studies have found that RIF is short-lived, there is other evidence to suggest that repeated selective practice schedules with a temporal gap between each practice trial may increase the durability of RIF. We tested this possibility in three experiments, focusing on socially shared retrieval-induced forgetting (SSRIF). In two experiments, participants studied scientific or story materials, then listened to someone selectively recall the material repeatedly, either in rapid succession or over an extended time period, and finally recalled the original materials either immediately, after a 1-week delay, or after a 3-week delay. A third experiment examined selective practice in free-flowing conversations. In each instance, RIF was found with repeated selective practice with a temporal gap between trials. The results are discussed in terms of the role RIF might play in the formation of collective memory.

Preventing fixation: Evidence of item-method directed forgetting protecting against mental impasses in creative problem-solving.

Creative problem-solving can be impeded by mental impasses which are caused by fixation on associated but misguiding ideas. Recently, research has focused on the question of which processes might be involved in overcoming fixation and how it can be prevented from occurring altogether. In the present study, we investigated whether initially encoded fixation words can be forgotten by means of item-method directed forgetting, a procedure in which participants are presented with a series of items that are each followed by a remember cue or forget cue. Here, the series of items partly consisted of words that were misleadingly associated with compound remote associate problems that had to be solved later on. Two groups were compared: For one group, fixation words were followed by forget cues (fixation-F), whereas for the other group, they were followed by a remember cue (fixation-R). Results revealed three major findings: (1) The fixation-R group showed typical detrimental effects of fixation (i.e., impaired problem-solving performance), whereas there was no significant fixation effect in the fixation-F group, and (2) the magnitude of fixation was positively related to memory for fixation words. The present findings add to existing literature about how fixation negatively affects creative problem-solving and how it can be prevented by reducing the activation level of misleading thoughts.

A novel study: fragmented and holistic forgetting.

According to recent theoretical work, certain event memories are more likely to be remembered or forgotten in their entirety. This prior work focused on collections of concepts, such as person-location-object triples. To explore this idea with complex materials, we created triples of people, locations, objects, or activities from events in real-world novels. People who had read one of the included novels were provided with one element from these triples (the cue) and asked to identify which of six alternatives best went with it. The results revealed that memory for the narrative events remained stable across many years. Moreover, people recalled events in a more holistic manner than would be expected by chance. This was more likely the more causally important an event was. This pattern of performance also remained stable over time. Our results are consistent with the idea that event models involve integrating separate elements into a single coherent representation, and this is likely to stay integrated over long periods of time. However, the degree to which this is so appears to be related to how well-integrated the information is within a larger set of events.

A class-incremental learning approach for learning feature-compatible embeddings.

Humans have the ability to constantly learn new knowledge. However, for artificial intelligence, trying to continuously learn new knowledge usually results in catastrophic forgetting, the existing regularization-based and dynamic structure-based approaches have shown great potential for alleviating. Nevertheless, these approaches have certain limitations. They usually do not fully consider the problem of incompatible feature embeddings. Instead, they tend to focus only on the features of new or previous classes and fail to comprehensively consider the entire model. Therefore, we propose a two-stage learning paradigm to solve feature embedding incompatibility problems. Specifically, we retain the previous model and freeze all its parameters in the first stage while dynamically expanding a new module to alleviate feature embedding incompatibility questions. In the second stage, a fusion knowledge distillation approach is used to compress the redundant feature dimensions. Moreover, we propose weight pruning and consolidation approaches to improve the efficiency of the model. Our experimental results obtained on the CIFAR-100, ImageNet-100 and ImageNet-1000 benchmark datasets show that the proposed approaches achieve the best performance among all the compared approaches. For example, on the ImageNet-100 dataset, the maximal accuracy improvement is 5.08%. Code is available at https://github.com/ybyangjing/CIL-FCE.

Does epilepsy differentially affect different types of memory?

Despite the recognition that epilepsy can substantially disrupt memory, there are few published accounts of whether and how this disruption varies across different types of memory and/or different types of epilepsy. This review explores four main questions: (1) Are working, episodic and semantic memory differentially affected by epilepsy? (2) Do various types of epilepsy, and their treatment, have different, specifiable effects on memory? (3) Are the usual forms of neuropsychological assessments of memory - many or most designed for other conditions - appropriate for patients with epilepsy? (4) How can research on epilepsy contribute to our understanding of the neuroscience of memory? We conclude that widespread and multifactorial problems are seen in working memory in all patient groups, while patients with temporal lobe epilepsy seem particularly prone to episodic memory deficit, and those with frontal lobe epilepsy to executive function deficits that may in turn impair semantic control. Currently, it is difficult to make individual patient predictions about likely memory deficits based on seizure aetiology and type, but it is possible to guide and tailor neuropsychological assessments in an individualised way. We make recommendations for future directions in validating and optimising neuropsychological assessments, and consider how to approach effective shared decision making about the pros and cons of seizure treatment strategies, especially at crucial educational stages such as adolescence.

Medial Prefrontal Cortex Stimulation Reduces Retrieval-Induced Forgetting via Fronto-parietal Beta Desynchronization.

The act of recalling memories can paradoxically lead to the forgetting of other associated memories, a phenomenon known as retrieval-induced forgetting (RIF). Inhibitory control mechanisms, primarily mediated by the prefrontal cortex, are thought to contribute to RIF. In this study, we examined whether stimulating the medial prefrontal cortex (mPFC) with transcranial direct current stimulation modulates RIF and investigated the associated electrophysiological correlates. In a randomized study, 50 participants (27 males and 23 females) received either real or sham stimulation before performing retrieval practice on target memories. After retrieval practice, a final memory test to assess RIF was administered. We found that stimulation selectively increased the retrieval accuracy of competing memories, thereby decreasing RIF, while the retrieval accuracy of target memories remained unchanged. The reduction in RIF was associated with a more pronounced beta desynchronization within the left dorsolateral prefrontal cortex (left-DLPFC), in an early time window (<500 ms) after cue onset during retrieval practice. This led to a stronger beta desynchronization within the parietal cortex in a later time window, an established marker for successful memory retrieval. Together, our results establish the causal involvement of the mPFC in actively suppressing competing memories and demonstrate that while forgetting arises as a consequence of retrieving specific memories, these two processes are functionally independent. Our findings suggest that stimulation potentially disrupted inhibitory control processes, as evidenced by reduced RIF and stronger beta desynchronization in fronto-parietal brain regions during memory retrieval, although further research is needed to elucidate the specific mechanisms underlying this effect.

Removing information from working memory with a delay: Effective but not beneficial.

Ideally, removing outdated information from working memory (WM) should have two consequences: The removed content should be less accessible (removal costs), and other WM content should benefit from the freeing up of WM capacity (removal benefits). Robust removal benefits and removal costs have been demonstrated when people are told to forget items shortly after they were encoded (immediate removal). However, other studies suggest that people might be unable to selectively remove items from an already encoded set of items (delayed removal). In two experiments (n = 219; n = 241), we investigated the effectiveness and consequences of delayed removal by combining a modified version of Ecker's et al. (Journal of Memory and Language, 74, 77-90, 2014) letter updating task with a directed-forgetting in WM paradigm. We found that while delayed removal resulted in reduced memory for the to-be-forgotten item-location relations (removal costs), it failed to enhance performance for existing WM content. This contrasts sharply with immediate removal, where removal benefits can be observed. A fine-grained analysis of removal benefits shows that removal from WM proactively facilitates the subsequent encoding of new information but does not retroactively aid stored WM content.

Responsible remembering: The role of metacognition, forgetting, attention, and retrieval in adaptive memory.

In our everyday lives, we must remember important information, especially if there are consequences for forgetting. In this review, I discuss recent work on responsible remembering: the strategic and effortful prioritization of important information with consequences for forgetting. Thus far, research regarding responsible remembering has revealed several key factors and mechanisms that work together to enhance memory for important information that will continue to be refined: the identification and selection of what to remember (metacognitive reflectivity), the forgetting of less important information to facilitate memory for items that do need to be remembered (responsible forgetting), the functional prioritization of attention at the expense of competing factors (responsible attention), and the selective recall of important information via efficient retrieval strategies (responsible retrieval). Together, these functions form a cohesive system that aims to selectively prioritize, encode, and recall information that is deemed important based on its anticipated utility or the consequences of forgetting, and considering the importance of information may be a critical memory adaptation as we age. Specifically, if younger and older adults learn to self-assess and prioritize important information that has negative consequences if forgotten, engage in strategic forgetting, efficiently allocate their attentional resources, and utilize effective retrieval operations, memory for said important information can be enhanced.

Effortful retrieval of semantic memories induces forgetting of related negative and neutral episodic memories.

Retrieval-induced forgetting (RIF) experiments show that the act of retrieving some recently encoded items from a given conceptual category leads to greater forgetting of competing items from that same category. However, RIF studies using emotional stimuli have produced mixed results, perhaps due to the reinstatement of arousal or negative affect during retrieval practice. To induce forgetting of negative episodic memories more indirectly, we examined if retrieving neutral semantic memories leads to RIF of related negative memories. In two experiments, participants studied eight categorized lists comprised of an equal number of negative and neutral words (Experiment 1) or neutral words preceded by neutral or negative images (Experiment 2). To avoid re-exposing individuals to negative material during retrieval practice, participants then performed a semantic memory retrieval task in which they generated (i.e., completed word-stems for) new neutral words from half of the studied categories. We found that semantic retrieval, or word generation, induced forgetting of recently studied words irrespective of their emotional valence or original emotional context. Additionally, across both experiments, less successful word generation was associated with stronger RIF effects. In Experiment 2, the magnitude of RIF was also correlated with higher subjective ratings of retrieval effort during word generation. Together, these results suggest that even when retrieving neutral semantic memories, effortful retrieval may enhance inhibitory processes that lead to forgetting of both neutral and negative episodic memories.

How varying cue duration influences item-method directed forgetting: A novel selective retrieval interpretation.

A series of four experiments tested the assumptions of the most prominent and longstanding account of item-method directed forgetting: the selective rehearsal account. In the item-method directed forgetting paradigm, each presented item is followed by its own instructional cue during the study phase - either to-be-forgotten (F) or to-be-remembered (R). On a subsequent test, memory is poorer for F items than for R items. To clarify the mechanism underlying memory performance, we manipulated the time available for rehearsal, examining instructional cue durations of 1 s, 5 s, and 10 s. Experiments 1a and 1b, where the order of cue durations was randomized, showed no effect of cue duration on item recognition of unrelated single words, for either R or F items. Experiment 2, using unrelated word pairs, again showed no effect of randomized cue duration, this time on associative recognition. Experiments 3 and 4 blocked cue duration and showed equivalent increases in recognition of both R and F single words and word pairs with increasing cue duration. We suggest that any post-cue rehearsal is carried out only when cue duration is predictable, and that such limited rehearsal is equally likely for F items and R items. The consistently better memory for R items than for F items across cue duration depends on selective retrieval involving (1) a rapid retrieval check engaged for R items only and (2) a rapid removal process implemented for F items only.

Continual learning for seizure prediction via memory projection strategy.

Despite extensive algorithms for epilepsy prediction via machine learning, most models are tailored for offline scenarios and cannot handle actual scenarios where data changes over time. Catastrophic forgetting(CF) for learned electroencephalogram(EEG) data occurs when EEG changes dynamically in the clinical setting. This paper implements a continual learning(CL) strategy Memory Projection(MP) for epilepsy prediction, which can be combined with other algorithms to avoid CF. Such a strategy enables the model to learn EEG data from each patient in dynamic subspaces with weak correlation layer by layer to minimize interference and promote knowledge transfer. Regularization Loss Reconstruction Algorithm and Matrix Dimensionality Reduction Algorithm are introduced into the core of MP. Experimental results show that MP exhibits excellent performance and low forgetting rates in sequential learning of seizure prediction. The forgetting rate of accuracy and sensitivity under multiple experiments are below 5%. When learning from multi-center datasets, the forgetting rates for accuracy and sensitivity decrease to 0.65% and 1.86%, making it comparable to state-of-the-art CL strategies. Through ablation experiments, we have analyzed that MP can operate with minimal storage and computational cost, which demonstrates practical potential for seizure prediction in clinical scenarios.

Forgetting unwanted memories in sleep.

Memories are sometimes best forgotten, but how do our brains weaken unwanted details of the past? We propose a theoretical framework in which memory reactivation during sleep supports adaptive forgetting. This mnemonic rebalancing underpins the affective benefits of sleep by ensuring that our memories remain aligned with our emotional goals.

Towards Robust Decision-Making for Autonomous Highway Driving Based on Safe Reinforcement Learning.

Reinforcement Learning (RL) methods are regarded as effective for designing autonomous driving policies. However, even when RL policies are trained to convergence, ensuring their robust safety remains a challenge, particularly in long-tail data. Therefore, decision-making based on RL must adequately consider potential variations in data distribution. This paper presents a framework for highway autonomous driving decisions that prioritizes both safety and robustness. Utilizing the proposed Replay Buffer Constrained Policy Optimization (RECPO) method, this framework updates RL strategies to maximize rewards while ensuring that the policies always remain within safety constraints. We incorporate importance sampling techniques to collect and store data in a Replay buffer during agent operation, allowing the reutilization of data from old policies for training new policy models, thus mitigating potential catastrophic forgetting. Additionally, we transform the highway autonomous driving decision problem into a Constrained Markov Decision Process (CMDP) and apply our proposed RECPO for training, optimizing highway driving policies. Finally, we deploy our method in the CARLA simulation environment and compare its performance in typical highway scenarios against traditional CPO, current advanced strategies based on Deep Deterministic Policy Gradient (DDPG), and IDM + MOBIL (Intelligent Driver Model and the model for minimizing overall braking induced by lane changes). The results show that our framework significantly enhances model convergence speed, safety, and decision-making stability, achieving a zero-collision rate in highway autonomous driving.

Effect of levels-of-processing on rates of forgetting.

The levels-of-processing (LOP) framework, proposing that deep processing yields superior retention, has provided an important paradigm for memory research and a practical means of improving learning. However, the available levels-of-processing literature focuses on immediate memory performance. It is assumed within the LOP framework that deep processing will lead to slower forgetting than will shallow processing. However, it is unclear whether, or how, the initial level of processing affects the forgetting slopes over longer retention intervals. The present three experiments were designed to explore whether items encoded at qualitatively different LOP are forgotten at different rates. In the first two experiments, depth of processing was manipulated within-participants at encoding under deep and shallow conditions (semantic vs. rhyme judgement in Experiment 1; semantic vs. consonant-vowel pattern decision in Experiment 2). Recognition accuracy (d prime) was measured between-participants immediately after learning and at 30-min, 2-h, and 24-h delays. The third experiment employed a between-participants design, contrasting the rates of forgetting following semantic and phonological (rhyme) processing at immediate, 30-min, 2-h, and 6-h delays. Results from the three experiments consistently demonstrated a large effect size of levels of processing on immediate performance and a medium-to-large level effect size on delayed recognition, but crucially no LOP × delay group interaction. Analysis of the retention curves revealed no significant differences between the slopes of forgetting for deep and shallow processing. These results suggest that the rates of forgetting are independent of the qualitatively distinct encoding operations manipulated by levels of processing.

Influence of degree of learning on rate of forgetting of tonal sequences.

Initial performance is frequently equated in studies that compare forgetting rates across groups. However, since the encoding capacity of different groups can be different, some procedures to match initial degree of learning need to be implemented, adding confounding variables such as longer exposures to the material, which would create memories of a different age. Slamecka and McElree Journal of Experimental Psychology: Learning, Memory, and Cognition, 9, 384-397, (1983) and our previous work found that the rate of forgetting was independent from initial degree of learning using verbal material. The present study seeks to determine whether this pattern holds true when undertaken with nonverbal material. In two experiments, we manipulate initial degree of learning by varying the number of presentations of the material and studying the effect on the forgetting rates. A set of 30 tonal sequences were presented to young, healthy participants either once or three times. Forgetting was evaluated in a yes/no recognition paradigm immediately and 1 hour or 24 hours after the study phase. A different subset of 10 sequences was tested along with 10 nontargets at each retention interval. The results of these experiments showed that initial acquisition was modulated by the number of repetitions. However, the forgetting rates were independent of initial degree of learning. These results are in keeping with the pattern found by Slamecka and McElree, and in our own previous studies. They suggest that the pattern of parallel forgetting after different levels of initial learning is not limited to verbal material.

Estimating the human bottleneck for contact tracing.

The SARS-CoV-2 pandemic has highlighted the importance of contact tracing for epidemiological mitigation. Contact tracing interviews (CTIs) typically rely on episodic memory, which is prone to decline over time. Here, we provide a quantitative estimate of reporting decline for age- and gender-representative samples from the United Kingdom and Germany, emulating >15,000 CTIs. We find that the number of reported contacts declines as a power function of recall delay and is significantly higher for younger subjects and for those who used memory aids, such as a scheduler. We further find that these factors interact with delay: Older subjects and those who made no use of memory aids have steeper decline functions. These findings can inform epidemiological modeling and policies in the context of infectious diseases.

Could Accelerated Long-Term Forgetting Be a Feature of the Higher Rate of Memory Complaints Associated with Subjective Cognitive Decline? An Exploratory Study.

Background: Recently, subjective cognitive decline (SCD) was proposed as an early risk factor for future Alzheimer's disease (AD). Objective: In this study, we investigated whether accelerated long-term forgetting (ALF), assessed with extended testing intervals than those adopted in clinical practice, might be a cognitive feature of SCD. Using an explorative MRI analysis of the SCD sample, we attempted to investigate the areas most likely involved in the ALF pattern. Methods: We recruited 31 individuals with SCD from our memory clinic and subdivided them based on their rate of memory complaints into mild SCDs (n = 18) and severe SCDs (n = 13). A long-term forgetting procedure, involving the recall of verbal and visuo-spatial material at four testing delays (i.e., immediate, 30 min, 24 h, and 7 days post-encoding) was used to compare the two sub-groups of SCDs with a healthy control group (HC; n = 16). Results: No significant between-group difference was found on the standard neuropsychological tests, nor in the immediate and 30 min recall of the experimental procedure. By contrast, on the verbal test severe SCDs forgot significantly more than HCs in the prolonged intervals (i.e., 24 h and 7 days), with the greatest decline between 30 min and 24 h. Finally, in the whole SCD sample, we found significant associations between functional connectivity values within some cortical networks involved in memory (default mode network, salience network, and fronto-parietal network) and verbal long-term measures. Conclusions: Our preliminary findings suggest that long-term forgetting procedures could be a sensitive neuropsychological tool for detecting memory concerns in SCDs, contributing to early AD detection.