Memory Consolidation

Abstract

Memory consolidation is a process within long-term memory by which a newly encoded, labile trace is transformed into a stable, durable form. It operates on two timescales: synaptic consolidation stabilizes the trace within individual neurons over minutes to hours, while systems consolidation reorganizes a memory across the hippocampus and neocortex over weeks to years. McGaugh established that a new memory passes through a time-dependent window during which it can be strengthened or disrupted, and Nader later showed that a reactivated memory returns to a labile state and must be reconsolidated to persist. Competing accounts, from standard consolidation through multiple-trace and contextual-binding theory, disagree over whether an episodic memory ever becomes independent of the hippocampus. This article develops synaptic and systems consolidation, multiple-trace theory, reconsolidation, sleep-dependent consolidation, and the engram.

Keywords: memory consolidation, synaptic consolidation, systems consolidation, reconsolidation, hippocampus

Memory consolidation is the set of time-dependent processes that convert a freshly encoded memory, initially fragile and easily disrupted, into a lasting representation resistant to interference (McGaugh, 2000). The idea is old: at the close of the nineteenth century it was already clear that a memory does not become permanent at the instant of learning but stabilizes gradually over time, so that a blow to the head or a competing experience can erase a recent memory while leaving older ones intact. What modern work added is the recognition that consolidation is not a single mechanism but a nested pair of them operating on radically different timescales. Within minutes to hours of learning, molecular and cellular changes fix the trace at the synapses that encode it, a fast process called synaptic or cellular consolidation. Over the following weeks, months, and years a slower reorganization takes place at the level of whole brain systems, as the hippocampus gradually trains lasting representations in the neocortex and the memory's dependence on the medial temporal lobe changes, the process called systems consolidation. The two are not rival descriptions of one event but distinct stages, and much of the field's history is the story of pulling them apart and, more recently, of discovering that consolidation is not the one-way street it was long assumed to be. The sections below develop the two timescales, the standard model of systems consolidation and its multiple-trace challenger, the reopening of the consolidated trace through reconsolidation, the active role of sleep, and the cellular engram that carries the memory.

What Memory Consolidation Is

Memory consolidation names the transition of a memory from a state in which it is vulnerable to disruption to one in which it is durable, and the defining property of that transition is that it takes time. The founding experimental demonstration is the phenomenon of retrograde amnesia gradients: an event that interferes with the brain shortly after learning, such as an electroconvulsive shock or a protein-synthesis inhibitor, abolishes a recent memory while sparing an older one, and the boundary between what is lost and what survives moves with the interval between learning and disruption (McGaugh, 1966). This time-dependence is the signature of consolidation: it shows that the memory was not fixed at encoding but continued to be processed afterward, passing through a window during which its fate could still be changed. McGaugh's synthesis a century after the phenomenon was first described drew the crucial distinction between the two consolidation processes that the gradient conflates (McGaugh, 2000). Synaptic consolidation is fast, local, and molecular, completing within hours at the synapses that hold the trace; systems consolidation is slow and distributed, reorganizing the memory across brain regions over years. The two are dissociable in principle and in evidence, since a manipulation that blocks the molecular machinery of synaptic consolidation acts only in the first hours after learning, whereas the systems-level gradient of retrograde amnesia extends over a far longer span. Dudai framed the enduring puzzle at the heart of the topic as the question of how stable the engram really is, given that a trace must be permanent enough to last a lifetime yet plastic enough to be updated by new experience (Dudai, 2004). The figure below places the two timescales, and the reconsolidation loop that reopens a consolidated trace, on a single schematic.

Synaptic Consolidation

Synaptic consolidation is the fast, local process that stabilizes a memory trace within the neurons and synapses that encode it, and its defining feature is a dependence on new protein synthesis during a narrow window after learning. A memory can be formed and briefly held without making new proteins, but converting that short-term trace into a lasting one requires a cascade of molecular events, gene transcription, and the synthesis of proteins that structurally remodel the activated synapses; blocking protein synthesis in the hours after learning leaves immediate memory intact but prevents the formation of a durable one. The cellular model for this stabilization is long-term potentiation, the persistent strengthening of synaptic transmission that follows brief high-frequency stimulation, first demonstrated in the hippocampus by Bliss and Lomo (Bliss & Lomo, 1973). Long-term potentiation supplied a concrete, measurable instance of the kind of lasting synaptic change that a physical memory trace would require, and its early phase, independent of protein synthesis, and its late phase, protein-synthesis-dependent and structurally expressed, map onto the distinction between a labile short-term trace and a consolidated long-term one. Work on the molecular biology of the trace has since detailed the signaling pathways, the role of the transcription factor CREB, and the synaptic tagging that lets a weak event be consolidated if it coincides with a strong one, building an account of how activity at a synapse is translated into a structural change that can persist (Asok et al., 2019). Synaptic consolidation is complete within hours and is largely finished before systems consolidation has meaningfully begun, which is why the two are best treated as sequential stages rather than as competing descriptions of one process.

Systems Consolidation

Systems consolidation is the slow reorganization of a memory across brain regions, and its classic statement is the standard model, in which the hippocampus rapidly binds the elements of an experience at encoding and then, over an extended period, gradually trains stable representations in the neocortex until the memory can be retrieved without the hippocampus at all (Squire & Alvarez, 1995). The model was built to explain the temporally graded retrograde amnesia that follows hippocampal damage: because recent memories still depend on the hippocampus while remote ones have been consolidated into the cortex, a lesion abolishes the recent and spares the remote, and the gradient of loss traces the progress of consolidation. The founding human evidence is the patient H.M., whose bilateral resection of the medial temporal lobes to relieve epilepsy left him unable to form new lasting declarative memories while sparing his remote memories and general intellect, first localizing the consolidation of new memories to the hippocampus and adjacent structures (Scoville & Milner, 1957). Frankland and Bontempi reframed the standard model in terms of the changing neural circuits that support a memory as it ages, marshalling evidence that the retrieval of remote memories comes to depend on distributed neocortical networks, particularly in the prefrontal cortex, as the hippocampal contribution wanes (Frankland & Bontempi, 2005). On this view the function of the slow transfer is to extract what is stable and generalizable across many episodes and to store it in cortical networks suited to integrated, schematic knowledge, leaving the hippocampus free to encode new experiences without interference. That rationale was given its computational statement by McClelland, McNaughton, and O'Reilly, whose complementary learning systems framework held that a single network cannot both learn new information rapidly and integrate it into structured knowledge without catastrophic interference, so the brain needs two systems: a fast-learning hippocampus that stores each experience quickly and keeps it separate, and a slow-learning neocortex to which those experiences are gradually replayed and interleaved with existing knowledge, which is precisely what systems consolidation carries out (McClelland et al., 1995). Squire's later synthesis treated systems consolidation as a well-established framework spanning the molecular, cellular, and systems levels, unifying the retrograde-amnesia evidence from humans, monkeys, and rodents into a single account of how a memory changes its neural home over time (Squire et al., 2015). The demonstration below plots the retrograde-amnesia gradient predicted by the standard model against the flat gradient that its principal challenger predicts for detailed episodic memories, so the reader can vary a memory's age at the time of hippocampal damage and read off how much of it each theory expects to survive.

Multiple-Trace and Transformation Theories

The standard model's claim that a memory eventually leaves the hippocampus altogether is precisely what multiple-trace theory denies. Nadel and Moscovitch argued that every time an episodic memory is retrieved, the hippocampus binds a new trace of the retrieval event, so that older memories are represented by more numerous and more widely distributed hippocampal traces rather than by a cortical representation that has become hippocampus-independent (Nadel & Moscovitch, 1997). The theory was motivated by evidence that retrograde amnesia for genuinely episodic, detail-rich autobiographical memories is often not temporally graded at all but extends across the whole lifespan when hippocampal damage is extensive, which the standard model does not predict. On this account the graded amnesia that does appear reflects the semanticization of memories over time, not their physical migration out of the hippocampus: a memory that has been retrieved and re-encoded many times develops a schematic, semantic version that the cortex can support alone, while its vivid episodic form continues to depend on the hippocampus for as long as it retains its detail. This was later generalized into trace-transformation theory, which holds that consolidation changes the character of a memory rather than merely its location, so that a detailed episodic representation and a gist-like semantic one coexist and are supported by different structures, the hippocampus for the former and the neocortex for the latter. The dispute is not merely terminological, because the two accounts make opposite predictions about whether a decades-old but still vivid memory should engage the hippocampus, a prediction that high-resolution imaging and careful lesion mapping continue to test. Winocur and Moscovitch developed this into an explicit systems-consolidation framework in which the standard and transformation accounts are not rivals but describe two parallel fates of a memory: a hippocampally dependent detailed trace that can transform into a gist-like neocortical version, with the two coexisting and competing, so that whether a remote memory looks hippocampus-independent depends on which version the retrieval test happens to probe (Winocur & Moscovitch, 2011).

Reconsolidation

For most of the twentieth century consolidation was assumed to happen once: a memory stabilized and thereafter stayed stable. Reconsolidation overturned that assumption. Nader, Schafe, and LeDoux showed that a fully consolidated fear memory, reactivated by a reminder cue, becomes labile again and requires a fresh round of protein synthesis to persist, so that infusing a protein-synthesis inhibitor into the amygdala after reactivation abolished a memory that the same inhibitor left untouched if no reminder was given (Nader et al., 2000). The finding was startling because it meant that retrieval is not a neutral readout but an event that can destabilize the very trace it accesses, returning a consolidated memory to a state resembling the one it occupied just after learning, from which it must be reconsolidated. This reopened a question the field had considered closed and supplied a mechanism by which established memories can be updated with new information, since the labile post-retrieval window is exactly when a memory can incorporate what has changed since it was formed. Lee, Nader, and Schiller synthesized the subsequent decade of work, recasting reconsolidation as a memory-updating process rather than a simple restabilization and detailing the boundary conditions, the age and strength of the memory, the presence of new information at retrieval, that determine whether a reactivated memory becomes labile at all (Lee et al., 2017). The clinical promise is direct: if a maladaptive memory, such as the fear trace underlying a phobia or the intrusive memory of trauma, can be reactivated and then blocked from reconsolidating, it might be weakened at its source rather than merely suppressed. The demonstration below reproduces the reactivation-by-inhibitor design, letting the reader set how much of a memory is rendered labile by reactivation and see why only the condition that combines reactivation with an amnestic agent impairs later retention.

Sleep and Consolidation

Sleep is not a passive interval during which memory merely fails to decay; it is an active phase of consolidation with a distinctive contribution. Diekelmann and Born set out the active systems consolidation account, in which the slow oscillations of deep non-REM sleep orchestrate a dialogue between hippocampus and neocortex, repeatedly reactivating newly encoded traces and driving their gradual redistribution to long-term cortical stores (Diekelmann & Born, 2010). On this view the systems consolidation that unfolds over weeks in waking life is compressed and accelerated during sleep, when the hippocampus replays the day's experiences in the temporal frame of slow oscillations, sleep spindles, and sharp-wave ripples, and this coordinated activity is what transfers the labile memory into a stable neocortical form. The behavioral signature is robust: memory tested after a period containing sleep is retained better than memory tested after an equal period of waking, and the benefit is larger for the hippocampus-dependent declarative memories that systems consolidation reorganizes than for procedural skills. Klinzing, Niethard, and Born extended the account to the mechanisms of systems memory consolidation during sleep, specifying how the precise temporal coupling of slow oscillations, spindles, and ripples times the reactivation of memory traces to the windows of cortical plasticity that can receive them (Klinzing et al., 2019). Sleep-dependent consolidation ties the whole topic together, since it is the physiological process in which the fast synaptic mechanisms and the slow systems-level reorganization meet: the reactivations of sleep re-engage the molecular machinery of plasticity in the service of the long-range transfer that defines systems consolidation. The demonstration below contrasts the retention of a memory across an interval of sleep with retention across an equal interval of waking, for a declarative memory and for a procedural skill.

The Engram

Underlying every account of consolidation is the physical memory trace itself, the engram, and the central modern development is that the engram has moved from a theoretical postulate to an object that can be identified and manipulated. Josselyn and Tonegawa summarized a body of work in which the specific population of neurons activated during learning, tagged by the expression of activity-dependent genes, can be labeled, and then artificially reactivated to evoke the memory or silenced to block its recall, so that the sparse ensemble carrying a given memory can be found and controlled (Josselyn & Tonegawa, 2020). This work recasts consolidation in cellular terms: allocation of a memory to particular neurons is governed by their relative excitability at encoding, and systems consolidation can be read as a gradual shift in which ensembles, hippocampal or cortical, are necessary and sufficient to support recall as a memory ages. The engram framework gives the abstract debates of consolidation a concrete substrate on which to be tested, turning questions about whether a remote memory still depends on the hippocampus into experiments that silence a tagged hippocampal ensemble and ask whether the memory survives. The molecular account of the trace and the systems account of its reorganization thus converge on the engram cell as the unit that both stores a memory and relocates it. Table 1 summarizes the major theoretical positions developed across this article and the signature evidence for each.

Table 1

Major Theoretical Positions on Memory Consolidation

Worked Example

The reconsolidation experiment yields a clean prediction that can be worked out arithmetically, and doing so shows why only one of its four conditions impairs memory. Suppose a fear memory is measured by a retention score that is 1.00 when the memory is fully intact. Reactivation by a reminder cue renders a fraction d of the memory labile, and a protein-synthesis inhibitor prevents any labile portion from being reconsolidated, so that the labile fraction is lost while the rest, never destabilized, survives untouched. Take d = 0.65. In the control condition, no reactivation and no inhibitor, nothing is destabilized and retention is 1.00. With the inhibitor but no reactivation, the memory is never made labile, so the inhibitor has nothing to act on and retention is again 1.00, which is the critical control that rules out a nonspecific drug effect. With reactivation but no inhibitor, the labile fraction reconsolidates normally and retention returns to 1.00, showing that reactivation alone does no lasting harm. Only when reactivation and inhibitor are combined is the labile fraction blocked from reconsolidating: retention falls to 1 − d = 1 − 0.65 = 0.35. The lesson is that neither manipulation impairs memory on its own and that the deficit is the interaction of the two, which is exactly the pattern that makes reconsolidation a distinct process rather than an artifact, because a memory must first be reopened by retrieval before an amnestic agent can disrupt it (Nader et al., 2000; Lee et al., 2017).

Discussion

Memory consolidation has moved from a single graded phenomenon to a layered account spanning molecules, synapses, systems, and identified cells, and the through-line is the recognition that a memory's stability is achieved over time rather than at the moment of learning (McGaugh, 2000). The two-timescale framework separated the fast, protein-synthesis-dependent fixing of a trace within neurons from the slow, systems-level reorganization that reshapes where in the brain a memory lives, and long-term potentiation gave the fast process a concrete cellular model (Bliss & Lomo, 1973; Asok et al., 2019). The standard model of systems consolidation organized the retrograde-amnesia evidence into an account of hippocampal-to-cortical transfer, and it remains the default framework even as multiple-trace and transformation theories contest its central claim that detailed episodic memories ever become truly independent of the hippocampus (Squire et al., 2015; Nadel & Moscovitch, 1997). The most consequential disturbance to the classical picture is reconsolidation, which showed that a consolidated memory is not permanently sealed but can be reopened by retrieval and must be restabilized each time, turning consolidation from a one-time event into a recurring process and opening a route to updating or weakening established memories (Nader et al., 2000). Sleep research meanwhile established that consolidation is not something that merely happens to pass while an organism rests but is actively driven by the coordinated physiology of non-REM sleep (Diekelmann & Born, 2010). What remains genuinely open is the fate of the detailed episodic trace over long intervals, and the engram work now supplies the tools to settle it, by identifying and silencing the specific ensembles a theory says should or should not still carry a remote memory (Josselyn & Tonegawa, 2020). The topic that began with the observation that a recent memory is easier to erase than an old one has become a mechanistic account of how a brain builds a trace, moves it, and reopens it.

Current Directions

Contemporary work on consolidation is concentrated where its systems-level questions meet the cellular tools that can now answer them. The engram program has turned the abstract dispute over hippocampal dependence into a manipulable one, tagging the neurons active during learning and testing directly whether silencing a hippocampal or a cortical ensemble abolishes a memory of a given age, which promises to adjudicate between standard and transformation accounts on evidence rather than inference (Josselyn & Tonegawa, 2020). Reconsolidation research continues to map the boundary conditions that decide when a reactivated memory becomes labile and to translate the post-retrieval window into behavioral interventions for fear and addiction memories, where the aim is to weaken a maladaptive trace at its source, though the reliability of reconsolidation-based clinical effects remains under active test (Lee et al., 2017). Sleep research is refining the causal role of the specific oscillatory events of non-REM sleep, using closed-loop stimulation timed to slow oscillations to strengthen the spindle-ripple coupling that the active systems account holds responsible for the overnight transfer, moving from correlation toward intervention (Klinzing et al., 2019). The molecular biology of the trace is being pushed toward the question of persistence, asking how a synaptic change can outlast the turnover of the very proteins that express it, which is the form Dudai's question about the stability of the engram now takes (Dudai, 2004; Asok et al., 2019). Running through all of it is the effort to reconcile the timescales, to show how the reactivations of a single night of sleep and the reorganization of a memory over years are expressions of one consolidation process operating at the level of the same engram cells.

Common Misconceptions

A memory is permanent as soon as it is formed.

A newly encoded memory is labile and can be disrupted for a period after learning, becoming durable only through time-dependent consolidation. A protein-synthesis inhibitor or a competing experience shortly after learning can prevent a lasting memory while sparing an older one, which is the temporally graded retrograde amnesia that first revealed consolidation (McGaugh, 1966).

Consolidation happens only once.

Retrieval can return a consolidated memory to a labile state from which it must be reconsolidated to persist. A reactivated fear memory is abolished by a protein-synthesis inhibitor that leaves the same memory untouched when no reminder is given, so a consolidated trace can be reopened, updated, or disrupted rather than being permanently sealed (Nader et al., 2000).

Sleep helps memory only by preventing new interference.

Sleep is an active phase of consolidation, not a passive gap in which decay is merely paused. During non-REM sleep the hippocampus replays newly encoded traces, and the coupling of slow oscillations, spindles, and ripples drives their redistribution to neocortical stores, so the benefit of sleep is a positive contribution to systems consolidation (Diekelmann & Born, 2010).

Glossary

Active systems consolidation.

The account on which non-REM sleep actively reactivates newly encoded traces and drives their redistribution from hippocampus to neocortex, rather than sleep merely protecting memory from interference.

Complementary learning systems.

The framework on which a fast-learning hippocampus that rapidly stores separated experiences and a slow-learning neocortex that gradually interleaves them into structured knowledge together avoid the catastrophic interference a single network would suffer; it gives systems consolidation its computational rationale.

Consolidation.

The set of time-dependent processes that convert a labile, newly encoded memory into a stable, durable one, operating at both the synaptic and the systems level.

Engram.

The physical memory trace, understood as the sparse ensemble of neurons whose activity encodes a specific event and whose reactivation constitutes its recall.

Hippocampus.

A medial-temporal structure that binds the elements of an experience at encoding and, on the standard model, gradually trains cortical representations before relinquishing the memory.

Labile.

The unstable, disruptible state of a memory trace before consolidation, or again after reactivation returns a consolidated memory to a fragile condition.

Long-term potentiation.

The persistent strengthening of synaptic transmission following brief high-frequency stimulation; the leading cellular model of the synaptic change that stores a memory.

Multiple-trace theory.

The account on which each retrieval lays down a new hippocampal trace, so detailed episodic memories remain hippocampus-dependent at any age while only a semanticized gist becomes cortical.

Protein synthesis.

The making of new proteins required to stabilize a trace during synaptic consolidation; its inhibition in the window after learning blocks lasting memory while sparing immediate memory.

Reconsolidation.

The restabilization a memory must undergo after retrieval has returned it to a labile state; the window in which an established memory can be updated or disrupted.

Retrograde amnesia.

Loss of memories formed before an injury or disruption; its temporal gradient, sparing remote and abolishing recent memories, is the founding evidence for systems consolidation.

Semanticization.

The transformation over time of a detailed episodic memory into a schematic, fact-like form that cortical networks can support without the hippocampus.

Slow oscillations.

The large, low-frequency electrical rhythms of deep non-REM sleep that coordinate the reactivation of memory traces and their transfer between hippocampus and neocortex.

Standard model of consolidation.

The account on which the hippocampus gradually trains stable neocortical representations until a memory can be retrieved without it, explaining temporally graded retrograde amnesia.

Synaptic consolidation.

The fast, local, protein-synthesis-dependent fixing of a memory trace within the neurons and synapses that encode it, completed within hours of learning.

Systems consolidation.

The slow reorganization of a memory across brain regions over weeks to years, changing its dependence on the hippocampus relative to the neocortex.

Trace-transformation theory.

The generalization of multiple-trace theory on which consolidation changes the character of a memory, so detailed and gist versions coexist and are supported by different structures.

Key Researchers

Jan Born (b. 1958). Professor at the University of Tuebingen; he established the active role of sleep in systems consolidation, showing that non-REM sleep reactivates and redistributes newly encoded memories. ORCID - Google Scholar - Faculty Page - Wikipedia

Sheena A. Josselyn. Senior Scientist at the Hospital for Sick Children and Professor at the University of Toronto; she showed that memory engrams are allocated to neurons by relative excitability and that manipulating an engram controls recall. ORCID - Google Scholar - Faculty Page - Wikipedia

Joseph E. LeDoux (b. 1949). Professor at New York University; he mapped the amygdala fear circuit and, with Nader, established that a reactivated fear memory must be reconsolidated to persist. Google Scholar - Faculty Page - Wikipedia

James L. McGaugh (b. 1931). Research Professor at the University of California, Irvine; he founded the modern study of memory consolidation and of the hormonal and amygdala modulation of memory strength. Google Scholar - Faculty Page - Wikipedia

Morris Moscovitch (b. 1945). Professor at the University of Toronto and the Rotman Research Institute; he co-developed multiple-trace and trace-transformation theory, holding that detailed recollection stays hippocampus-dependent. ORCID - Google Scholar - Faculty Page - Wikipedia

Lynn Nadel (b. 1942). Regents' Professor Emeritus at the University of Arizona; he co-authored multiple-trace theory and the account of the hippocampus as a cognitive map underlying memory and space. ORCID - Faculty Page - Wikipedia

Karim Nader. James McGill Professor at McGill University; his 2000 amygdala study reopened the reconsolidation debate by showing that a reactivated memory requires new protein synthesis to persist. Google Scholar - Faculty Page

Larry R. Squire (b. 1941). Professor at the University of California, San Diego and the VA San Diego; he formalized the standard model of systems consolidation and the declarative-nondeclarative taxonomy. Google Scholar - Faculty Page - Wikipedia

Susumu Tonegawa (1939-2026). Nobel laureate and Professor at MIT's Picower Institute; he pioneered the optogenetic identification and manipulation of memory engram cells. ORCID - Google Scholar - Faculty Page - Wikipedia

Frequently Asked Questions

What is memory consolidation?
Memory consolidation is the set of time-dependent processes that convert a newly encoded, easily disrupted memory into a stable, durable one, operating both within neurons over hours and across brain systems over weeks to years (McGaugh, 2000).

What is the difference between synaptic and systems consolidation?
Synaptic consolidation is the fast, local, protein-synthesis-dependent stabilization of a trace within the neurons that encode it, completed within hours, whereas systems consolidation is the slow reorganization of the memory across the hippocampus and neocortex over weeks to years (McGaugh, 2000; Squire et al., 2015).

Does a memory ever become independent of the hippocampus?
This is contested: the standard model holds that memories are gradually transferred to the neocortex and become hippocampus-independent, while multiple-trace and transformation theories hold that detailed episodic recollection stays hippocampus-dependent no matter how old the memory (Squire & Alvarez, 1995; Nadel & Moscovitch, 1997).

What is reconsolidation?
Reconsolidation is the finding that retrieving a consolidated memory can return it to a labile state, so that it must be restabilized through new protein synthesis to persist; this opens a window in which an established memory can be updated or disrupted (Nader et al., 2000; Lee et al., 2017).

How does sleep help memory?
Sleep is an active phase of consolidation: during non-REM sleep the hippocampus replays newly encoded traces, and the coupling of slow oscillations, spindles, and ripples drives their transfer to neocortical stores, so declarative memory is retained better across sleep than across an equal period of waking (Diekelmann & Born, 2010; Klinzing et al., 2019).

What is long-term potentiation and how does it relate to consolidation?
Long-term potentiation is the lasting strengthening of synaptic transmission after brief high-frequency stimulation; its protein-synthesis-dependent late phase is the leading cellular model of the synaptic change that synaptic consolidation fixes in place (Bliss & Lomo, 1973; Asok et al., 2019).

What is an engram?
An engram is the physical memory trace, now identified with the sparse ensemble of neurons activated during learning; such ensembles can be tagged and artificially reactivated to evoke a memory or silenced to block its recall (Josselyn & Tonegawa, 2020).

Why does blocking protein synthesis erase some memories but not others?
Protein synthesis is required only during the consolidation window, so an inhibitor given in the hours after learning, or after a reminder reactivates a memory, blocks the durable trace, while the same inhibitor has no effect on a memory that is neither newly formed nor reactivated (McGaugh, 1966; Nader et al., 2000).

References