Forgetting as an Adaptive Process: Interference, Inhibition and the Costs of Total Recall

Abstract

Forgetting is commonly treated as a failure of the memory system, a passive loss of information through decay or interference. The present review examines the alternative position that forgetting is, in substantial part, adaptive and often active. Four bodies of evidence are considered: the classical retention data of Ebbinghaus and their reinterpretation in the new theory of disuse; behavioural work on retrieval-induced forgetting and on the deliberate suppression of unwanted memories, together with its neural correlates; recent neurobiological accounts of intrinsic forgetting mechanisms, including synaptic weakening and hippocampal neurogenesis; and the study of individuals with exceptional retention, from Luria’s mnemonist to highly superior autobiographical memory. Across these literatures, forgetting emerges as a set of mechanisms that reduce interference, support generalisation and permit behavioural updating in changing environments, although the replication status and boundary conditions of some effects remain contested. The main conclusion is that a memory system without forgetting would be neither more accurate nor more useful. For any programme that seeks to preserve human emotional memory, this implies that fidelity should be defined by the functional role of a memory rather than by the completeness of its record.

Keywords: forgetting; retrieval-induced forgetting; memory suppression; inhibition; neurogenesis; hyperthymesia

1. Introduction

The everyday model of memory is a storage model: experiences are deposited, retained and later withdrawn, and any discrepancy between the two counts as loss. Within this model forgetting has no function, and the ideal system is one that never forgets.

This model has been challenged from several directions. Theoretical work reframed the loss of accessibility as a precondition for efficient learning (Bjork & Bjork, 1992). Experimental work showed that forgetting can be produced by the act of remembering itself (Anderson, Bjork, & Bjork, 1994) and by deliberate cognitive control (Anderson & Green, 2001), with identifiable prefrontal and hippocampal correlates (Anderson & Hanslmayr, 2014). Neurobiological reviews have catalogued molecular and cellular processes whose apparent purpose is to remove stored traces (Davis & Zhong, 2017; Richards & Frankland, 2017). Case studies of exceptional retainers reveal costs rather than unqualified benefits (LePort et al., 2012; Luria, 1968; Parker, Cahill, & McGaugh, 2006). Nørby (2015) has offered the most systematic functional taxonomy of these findings.

The present review has two aims: to assess how strong the evidence is that forgetting is adaptive and, where relevant, active rather than passive, distinguishing established from contested findings; and to derive constraints for any effort to preserve human emotional memory, a domain in which the equation of fidelity with completeness is particularly seductive and, we argue, misleading.

2. Classical retention data and the theory of disuse

2.1 Ebbinghaus and the savings method

The quantitative study of forgetting begins with Ebbinghaus (1885/1913), who learned lists of nonsense syllables to criterion and measured retention by the savings method: the reduction in trials needed to relearn a list relative to original learning, at intervals from about twenty minutes to thirty-one days. The resulting function is steep at first and then flattens, so that most measurable loss occurs within the first day. Two features are often neglected. First, savings remained positive at the longest intervals even when unaided recall was at floor: a trace can persist without being accessible. Second, repetitions distributed over days produced markedly better retention than the same number massed within a session. The data rest on a single participant, but the shape of the function has been replicated many times, and the dissociation between availability and accessibility became a cornerstone of later theory.

2.2 Storage strength and retrieval strength

Bjork and Bjork (1992) formalised this dissociation in the new theory of disuse. Each memory carries two independent indices: storage strength, which reflects how well an item is learned and interconnected and is assumed never to decrease, and retrieval strength, which reflects current accessibility and is what declines with disuse and competition. Retrieval strength is a limited resource, so that gains for some items entail losses for others; increments to storage strength are larger when retrieval strength is low; and retrieval is a more potent learning event than re-study.

Under these assumptions the loss of retrieval strength is not a defect but the condition that makes subsequent learning efficient while keeping older information available. The theory accounts for the spacing effect and for the advantage of testing over restudy, and it predicts that forgetting should be selective and use-dependent, because retrieval strength is redistributed according to what the organism actually retrieves.

3. Forgetting produced by remembering

3.1 Retrieval-induced forgetting

Anderson, Bjork, and Bjork (1994) introduced the retrieval-practice paradigm. Participants study category–exemplar pairs (Fruit–Orange, Fruit–Banana, Drink–Whisky), then practise retrieving half of the exemplars from half of the categories using stem cues (*Fruit–Or___). The critical comparison at final test is between unpractised items from practised categories (Banana) and items from unpractised categories (Whisky): the former were recalled less well, an effect termed retrieval-induced forgetting (RIF). The authors’ interpretation was inhibitory: when Fruit* cues retrieval of Orange, competing fruits are activated and must be suppressed, and this suppression persists as reduced accessibility.

Two findings were subsequently taken to favour inhibition over associative blocking, on which practised items simply become stronger competitors at test: RIF appears with independent cues absent during practice, and it depends on retrieval practice rather than on extra study of the same items. Both have been replicated, but neither is uncontested.

3.2 Replication and moderators

In the meta-analysis by Murayama, Miyatsu, Buchli, and Storm (2014), RIF was reliable and of moderate size, and its magnitude was moderated by the integration of materials, the strength of competitors and the delay between practice and test. Some moderator patterns are consistent with inhibition and some equally with interference-based accounts. RIF is therefore a robust behavioural phenomenon whose mechanism remains debated; claims that hinge specifically on inhibition should be presented as provisional.

3.3 Deliberate suppression: the think/no-think paradigm

Anderson and Green (2001) asked whether forgetting could be produced intentionally. In the think/no-think (TNT) paradigm, participants learn weakly related word pairs (Ordeal–Roach). Cues are then presented repeatedly; for Think items participants recall the response, whereas for No-Think items they prevent the response from entering awareness. At the final test, recall of No-Think items fell below that of baseline items learned but not presented in the second phase, and the deficit increased with the number of suppression trials (0, 1, 8 and 16). The deficit was also observed with independent probes, which argues against participants merely associating the cue with something else.

The TNT effect has been replicated in many laboratories, although it is modest, a proportion of participants show no effect, and null results have been reported, particularly when compliance is not enforced. Individual differences correlate with executive control, and reduced effects have been reported in post-traumatic stress and depression, although these associations rest on small samples.

3.4 Neural mechanisms of motivated forgetting

Anderson and Hanslmayr (2014) reviewed the neuroimaging and electrophysiological evidence and distinguished two mechanisms. Direct suppression engages right dorsolateral prefrontal cortex during No-Think trials and reduces hippocampal activity, and the degree of both predicts later forgetting. Thought substitution recruits left ventrolateral prefrontal regions associated with selective retrieval. The direct-suppression network overlaps with that implicated in stopping motor responses, suggesting that memory suppression is one application of a domain-general inhibitory control system. The causal role of hippocampal downregulation, however, has been established mainly by correlation with subsequent recall.

Table 1 summarises the principal paradigms.

4. The biology of active forgetting

The behavioural literature establishes that forgetting can be selective and controllable, not that the brain is built to forget. Richards and Frankland (2017) distinguish persistence, the mechanisms that stabilise engrams, from transience, the mechanisms that degrade them, and argue that both are regulated. Evidence for active transience includes activity-dependent weakening of synaptic connections, loss of dendritic spines, and adult hippocampal neurogenesis: newly integrated neurons remodel existing circuits, and in rodents increasing neurogenesis after learning accelerates forgetting whereas suppressing it slows forgetting. Their functional argument is drawn from machine learning: a learner that stores every training example overfits, whereas transience discards idiosyncratic detail, promotes generalisation and, in non-stationary environments, removes information that has become misleading. On this view the goal of memory is adaptive decision-making in the present rather than accurate transmission of the past.

Davis and Zhong (2017) reach a parallel conclusion from molecular data, largely from Drosophila. They argue that the default state of newly formed memories is to be forgotten, through an intrinsic pathway in which the small GTPase Rac1 and cytoskeletal remodelling erode labile traces and dopaminergic signalling after learning drives ongoing forgetting. Blocking these pathways prolongs memory; activating them accelerates its loss. Generalisation from invertebrate conditioning to human autobiographical memory must be cautious. What the two reviews jointly establish is that some forgetting is the product of dedicated biological machinery rather than entropy: a process the organism pays for presumably yields a return.

5. The costs of exceptional retention

If forgetting is functional, individuals who forget less should exhibit the costs of its absence. The evidence is case-based and cannot support causal inference, but it is consistent in direction.

Luria (1968) followed the mnemonist S. for nearly thirty years. S. reproduced lists of seventy or more items in any order and recalled lists learned fifteen years earlier, relying on pronounced synaesthesia and an involuntary conversion of words into images. Luria’s clinical observations concern the costs. S. had difficulty extracting the gist of a text because each word generated an image that competed with the next; he found metaphor hard to process; he had difficulty recognising faces, because each instance had been stored separately; and material from previous performances intruded on later ones. Its central observation, that indiscriminate retention interferes with abstraction, anticipates the overfitting argument above.

Parker, Cahill, and McGaugh (2006) reported the case of AJ, who could recall the events of nearly any day of her adult life when given a date, verified against diaries; her performance on many standard laboratory tests was unremarkable, and her recall was described as automatic, uncontrollable and at times burdensome. LePort et al. (2012) studied eleven individuals with highly superior autobiographical memory (HSAM) against matched controls. HSAM participants were dramatically superior on tests of public and personal event memory, yet performed at control levels on most standard laboratory measures. Structural imaging revealed differences in temporal and parietal regions, and HSAM participants scored higher on measures of obsessional tendencies. Subsequent work has found HSAM individuals as susceptible as controls to laboratory-induced false memories. The literature is small, but it demonstrates that superior retention does not generalise to superior cognition and carries a distinctive profile of costs.

Nørby (2015) organises these findings into three adaptive functions: emotional regulation, as the affective charge of negative experiences attenuates; cognitive economy, as competing information that retrieval must search is reduced; and knowledge acquisition, as episodes are abstracted into schemas. These functions can conflict with one another and with accuracy, so the value of forgetting is contingent rather than absolute. The taxonomy clarifies where forgetting becomes pathological: not when much is lost, but when what is lost is required for the functions that remain.

6. Implications for emotion preservation

The findings reviewed impose constraints on any programme for preserving human emotional memory, stated here as propositions open to empirical test.

First, the fidelity of an emotional memory cannot be defined by the completeness of its record. An affective trace that persists across a lifetime has been shaped by transience: peripheral detail has been discarded and the residual tone, evaluation and relational meaning have been abstracted (Nørby, 2015; Richards & Frankland, 2017). The testable proposition is that judgements of authenticity by the rememberer will track the abstracted form of an emotional memory more closely than its detailed original.

Second, accessibility and availability are distinct (Bjork & Bjork, 1992; Ebbinghaus, 1885/1913). A programme may aim to preserve availability while leaving accessibility to the person’s own retrieval history. The proposition is that re-exposed material will be experienced as more intrusive, and less as one’s own, when it restores accessibility the individual has allowed to lapse.

Third, the capacity to suppress is a competence with a neural basis and clinical significance (Anderson & Green, 2001; Anderson & Hanslmayr, 2014). Material the individual has deliberately kept from awareness should not be reinstated without consent, and any protocol should be evaluated for its effects on intrusion frequency in vulnerable groups.

Fourth, retrieval strengthens what is retrieved and impairs what is not (Anderson et al., 1994; Murayama et al., 2014). Repeated cued recall of selected emotional episodes will strengthen them at the expense of related unrehearsed ones; selection criteria are therefore an intervention on the memory system, and their effects on the unselected remainder should be measured.

Fifth, the HSAM and mnemonist data (LePort et al., 2012; Luria, 1968; Parker et al., 2006) caution that an exhaustive emotional archive is not a model of psychological health. Which features of an emotional memory carry its long-term function is an open question that should precede decisions about what to retain.

7. Conclusion

The evidence reviewed converges on a position stronger than the claim that forgetting is tolerable: at least some forgetting is produced by mechanisms the organism maintains at a cost, is regulated by the same control systems that govern action, and yields benefits in generalisation, updating and affective regulation that a non-forgetting system would lack. The position has limits: the inhibitory interpretation of retrieval-induced forgetting is contested, suppression effects are modest and variable, and the molecular evidence comes largely from invertebrates. What is not in doubt is that the exceptional retainers studied to date do not think better for retaining more. Programmes concerned with emotional memory should begin from the functions that forgetting serves, not from the assumption that the best memory loses nothing.

References

Anderson, M. C., Bjork, R. A., & Bjork, E. L. (1994). Remembering can cause forgetting: Retrieval dynamics in long-term memory. Journal of Experimental Psychology: Learning, Memory, and Cognition, 20(5), 1063–1087.

Anderson, M. C., & Green, C. (2001). Suppressing unwanted memories by executive control. Nature, 410(6826), 366–369.

Anderson, M. C., & Hanslmayr, S. (2014). Neural mechanisms of motivated forgetting. Trends in Cognitive Sciences, 18(6), 279–292.

Bjork, R. A., & Bjork, E. L. (1992). A new theory of disuse and an old theory of stimulus fluctuation. In A. F. Healy, S. M. Kosslyn, & R. M. Shiffrin (Eds.), From learning processes to cognitive processes: Essays in honor of William K. Estes (Vol. 2, pp. 35–67). Erlbaum.

Davis, R. L., & Zhong, Y. (2017). The biology of forgetting: A perspective. Neuron, 95(3), 490–503.

Ebbinghaus, H. (1913). Memory: A contribution to experimental psychology (H. A. Ruger & C. E. Bussenius, Trans.). Teachers College, Columbia University. (Original work published 1885)

LePort, A. K. R., Mattfeld, A. T., Dickinson-Anson, H., Fallon, J. H., Stark, C. E. L., Kruggel, F., Cahill, L., & McGaugh, J. L. (2012). Behavioral and neuroanatomical investigation of Highly Superior Autobiographical Memory (HSAM). Neurobiology of Learning and Memory, 98(1), 78–92.

Luria, A. R. (1968). The mind of a mnemonist: A little book about a vast memory (L. Solotaroff, Trans.). Basic Books.

Murayama, K., Miyatsu, T., Buchli, D., & Storm, B. C. (2014). Forgetting as a consequence of retrieval: A meta-analysis of retrieval-induced forgetting. Psychological Bulletin, 140(5), 1383–1409.

Nørby, S. (2015). Why forget? On the adaptive value of memory loss. Perspectives on Psychological Science, 10(5), 551–578.

Parker, E. S., Cahill, L., & McGaugh, J. L. (2006). A case of unusual autobiographical remembering. Neurocase, 12(1), 35–49.

Richards, B. A., & Frankland, P. W. (2017). The persistence and transience of memory. Neuron, 94(6), 1071–1084.