Amnesia

Abstract

Amnesia is a disproportionate loss of long-term memory caused by brain injury or disease, in which the capacity to acquire or retrieve declarative facts and events is impaired while perception, language, and reasoning remain largely intact. The defining case is bilateral damage to the medial temporal lobe, which abolishes the ability to form new conscious memories (anterograde amnesia) and erases a graded band of memories from before the injury (retrograde amnesia), yet spares the acquisition of perceptual-motor and other nondeclarative skills. That selective profile, first documented in patient H.M., established that memory is not one faculty but a set of dissociable systems. Modern work reframes the disorder around the consolidation of episodic memory and the accessibility of memory engrams, sharpening a century-old question: does amnesia destroy stored memories, or only the means of reaching them?

Keywords: amnesia, anterograde amnesia, retrograde amnesia, medial temporal lobe, memory consolidation

Amnesia is a pathological impairment of memory that is out of proportion to any other cognitive deficit. The amnesic patient can hold a conversation, solve a puzzle, and score normally on tests of intelligence and perception, yet cannot remember the conversation minutes later or recall events from a defined period of life. The disorder is diagnostically important precisely because it is selective: it strikes the declarative memory for facts and events while leaving working memory, language, and skilled performance comparatively untouched. MeSH classifies amnesia as a memory disorder and, in its disease facet, as a neurocognitive disorder — that is, amnesia is itself a disorder of memory rather than a variety of normal forgetting (Kopelman, 2002). The scientific interest of amnesia lies in this dissociation: by removing one kind of memory and sparing others, brain injury reveals the seams along which the memory system is built (Scoville & Milner, 1957; Squire, 1992).

Types of Amnesia

MeSH indexes amnesia under the broader heading of memory disorders, and, within its disease facet, under neurocognitive disorders; the descriptor in turn subsumes four narrower kinds. These subtypes are cross-cut by two independent distinctions, and it helps to keep them apart. One axis is temporal direction — whether the deficit blocks new learning (anterograde) or reaches back to erase the past (retrograde); most organic amnesias combine both in differing proportion. A second axis is aetiology — whether the cause is a focal lesion, a metabolic insult such as thiamine deficiency, a transient vascular or ictal event, or a psychological trauma with no detectable lesion. The MeSH children below are organised partly by direction and partly by cause, so they overlap rather than partition cleanly; a Korsakoff patient, for instance, shows both anterograde and retrograde loss. As always, the MeSH tree is an indexing classification built for retrieval, not a mechanistic taxonomy of the disorder, and the categories should be read in that spirit.

Anterograde and Retrograde Amnesia

The single most important distinction divides amnesia by its temporal relation to the injury. Anterograde amnesia is the failure to lay down new declarative memories from the moment of injury forward: the patient's experience is continuous and comprehensible in the present, but leaves no trace, so each encounter can feel like the first. Retrograde amnesia is the loss of memories established before the injury, and it characteristically follows a temporal gradient — memories from the months or years just before onset are the most vulnerable, while childhood and remote memories are relatively spared (Squire, 1992; Bayley et al., 2005). This gradient, first described by Théodule Ribot in the nineteenth century, is one of the most robust facts about the disorder and a central constraint on any theory of it: a mechanism that explained anterograde loss but predicted uniform retrograde loss would be wrong.

The two forms dissociate. A patient may have dense anterograde amnesia with only a brief retrograde gap, as in the classic medial temporal case, or a disproportionately extensive retrograde amnesia reaching back decades (Cipolotti et al., 2001). Their partial independence tells us that acquiring a new memory and holding an old one are not the same operation, and that the structures damaged in amnesia serve the first more obviously than the second.

The Medial Temporal Lobe

The founding observation of the modern field was surgical. In an attempt to control intractable epilepsy, the patient known as H.M. underwent bilateral resection of the medial temporal lobes, including the hippocampus and adjacent cortex; the seizures abated, but he was left with a profound and permanent anterograde amnesia and a graded retrograde loss, against an otherwise preserved intellect (Scoville & Milner, 1957). The precision of the deficit — memory devastated, everything else intact — made H.M. the most studied patient in the history of psychology and localised the human capacity to form new declarative memories to the medial temporal lobe (Corkin, 2002).

Converging evidence from many patients and from experimental lesions in animals defined a medial temporal lobe memory system: the hippocampus together with the surrounding entorhinal, perirhinal, and parahippocampal cortices, working as a functional unit whose damage produces the amnesic syndrome in proportion to its extent (Squire, Stark, & Clark, 2004). Reviews of hippocampal amnesia across large case series confirm the core profile and its variations (Spiers, Maguire, & Burgess, 2001). A developmental variant sharpened the picture further: patients who sustained hippocampal damage early in life develop severely impaired episodic memory for events yet acquire near-normal semantic knowledge of facts and vocabulary, showing that the hippocampus is more critical for remembering experiences than for accumulating context-free knowledge (Vargha-Khadem et al., 1997).

Spared Learning: Declarative and Nondeclarative Memory

What amnesia spares is as informative as what it destroys. H.M. and patients like him, unable to recall having practised a task, nonetheless improved at it across days: their performance on mirror-drawing and other perceptual-motor skills climbed steadily even as they denied any memory of the training (Milner, Corkin, & Teuber, 1968). This dissociation — skill acquired, episode forgotten — forced a fundamental revision. Memory is not one store but at least two broad kinds: declarative memory for facts and events, which is conscious, flexible, and lost in amnesia, and nondeclarative memory for skills, habits, priming, and simple conditioning, which is expressed through performance and largely preserved (Squire, 1992). Because the amnesic patient can learn without knowing that they have learned, the two systems must be neurally distinct, the declarative one depending on the medial temporal lobe that amnesia destroys and the nondeclarative ones on the basal ganglia, cerebellum, and neocortex that it spares.

Figure 1

The Taxonomy of Long-Term Memory Revealed by Amnesia

Consolidation and the Temporal Gradient

Why should recent memories be more vulnerable than remote ones? The standard model of systems consolidation answers that the medial temporal lobe is only a temporary player. On this view the hippocampus rapidly binds the distributed cortical traces of a new experience and, over subsequent months to years, a slow reorganisation transfers the memory to a hippocampus-independent form in the neocortex; damage therefore erases recent, still-hippocampus-dependent memories while sparing older, already-consolidated ones, generating Ribot's gradient (Squire, 1992; Bayley et al., 2005).

This account is contested for one important class of memory. Multiple trace theory argues that detailed, re-experienced episodic memories never become fully independent of the hippocampus: each retrieval lays down a new hippocampal trace, so richly episodic memories remain permanently hippocampus-dependent, and only their semanticised gist migrates to cortex (Nadel & Moscovitch, 1997). The prediction that separates the theories concerns very old autobiographical memories: the standard model expects them spared after hippocampal damage, whereas multiple trace theory expects vivid episodic recollection to be impaired at every age. Evidence of extensive, ungraded retrograde amnesia for episodic detail after hippocampal lesions supports the second view (Cipolotti et al., 2001), and a transformation account now reconciles much of the debate: with time and retrieval a memory changes in kind, from a detailed hippocampal episode toward a gist-like cortical representation, so the gradient observed depends on which quality of memory is tested (Winocur & Moscovitch, 2011).

Functional and Diencephalic Amnesias

Not all amnesia follows from medial temporal damage. Diencephalic amnesia results from lesions of the midline structures — the mammillary bodies and the anterior and mediodorsal thalamic nuclei — most often in alcoholic Korsakoff syndrome, where chronic thiamine deficiency produces a dense anterograde and retrograde amnesia frequently accompanied by confabulation, the unwitting production of false memories offered with conviction (Kopelman, 2002; Markowitsch & Staniloiu, 2012). That two distant sites — medial temporal and diencephalic — produce a similar syndrome indicates that they belong to a single extended memory circuit.

Transient global amnesia is a striking, benign syndrome in which an otherwise healthy person abruptly loses the ability to form new memories for several hours, repeats the same questions, and then recovers fully, leaving only a permanent gap for the episode itself; its functional anatomy centres on transient hippocampal dysfunction (Bartsch & Deuschl, 2010). Different again is dissociative (psychogenic) amnesia, in which autobiographical memory — sometimes including personal identity — is lost after psychological trauma with no identifiable brain lesion, and in which the profile is often the reverse of the organic case, with a disproportionate retrograde loss of personal history and relatively intact new learning (Staniloiu & Markowitsch, 2014). The existence of a functional amnesia that mimics some features of the organic syndrome while inverting others underlines that amnesia names a family of dissociable failures, not one lesion.

Worked Example

Consider a densely amnesic patient tested over three consecutive days on a mirror-drawing task, in which a shape must be traced while seeing only its mirror image — a perceptual-motor skill that is nondeclarative. Suppose the mean number of errors per trial falls from 30 on Day 1 to 20 on Day 2 to 12 on Day 3. The saving from first to last session is (30 − 12) / 30 = 18 / 30 = 0.60, a 60% reduction in errors; the day-to-day improvements are (30 − 20) / 30 = 33.3% and (20 − 12) / 20 = 40.0%. The learning curve is not only positive but accelerating slightly, exactly as for a neurologically healthy learner.

Now suppose that at the start of each session the same patient is given a yes/no recognition test asking whether they have ever performed this task, and that across 20 such probes they respond affirmatively on 10 and negatively on 10 with no relation to the truth — 10 hits out of 20 targets and 10 false alarms out of 20 lures. The sensitivity index is d′ = z(hit rate) − z(false-alarm rate) = z(0.50) − z(0.50) = 0 − 0 = 0: recognition is exactly at chance. The patient therefore shows robust skill learning (60% error reduction) with zero declarative memory of the learning (d′ = 0) — the numerical signature of the declarative/nondeclarative dissociation that defines the amnesic syndrome (Milner, Corkin, & Teuber, 1968; Squire, 1992).

Discussion

Amnesia has been the single most productive lesion in the study of memory because its deficits are so cleanly bounded. From the profile of a handful of patients came the medial temporal lobe memory system, the declarative/nondeclarative distinction, the episodic/semantic distinction, and the framework of systems consolidation (Scoville & Milner, 1957; Squire, 1992; Tulving, 1985). Each was discovered not by adding a faculty but by subtracting one and watching what remained. The disorder thereby turned an abstract question — is memory unitary? — into an anatomical one, and answered it: memory is a confederation of systems that fail independently. Tulving's analysis added the subjective dimension the anatomy alone could not supply: episodic retrieval is accompanied by autonoetic consciousness, the self-knowing awareness that lets a rememberer mentally travel back and re-experience an event as their own, whereas semantic knowing carries only a noetic awareness of facts stripped of that first-person re-living. What medial temporal amnesia most deeply removes, on this view, is autonoetic re-experience — the patient may relearn a fact yet can no longer relive the episode in which it was met (Tulving, 1985).

The theoretical fault line that remains is old and sharp. When an amnesic patient cannot retrieve an event, is the memory gone — its cortical trace never consolidated, or its representation destroyed — or is it present but unreachable, intact in storage yet cut off from the machinery of recall? The graded temporal profile of retrograde amnesia, the recovery seen in transient global amnesia, and the preservation of memory expressed through nondeclarative performance all point toward retrieval failure playing a larger role than a pure storage-loss account allows (Bartsch & Deuschl, 2010; Tulving, 1985). It is this question that contemporary cellular neuroscience has taken up, and it is why amnesia remains an active problem rather than a solved one.

Current Directions

The storage-versus-retrieval question has been reopened at the level of the engram — the physical ensemble of neurons whose coordinated activity stores a specific memory. Optogenetic experiments in rodents can label the cells activated during learning and later reactivate them artificially; when animals rendered amnesic by an interference protocol fail to recall a memory, direct stimulation of the tagged engram cells restores the behaviour, implying that the memory was stored but inaccessible — a retrieval failure rather than a storage failure (Josselyn & Tonegawa, 2020). This work has begun to supply a mechanistic foundation for the psychology of retrieval, framing recall as the reinstatement of an engram and forgetting as a change in its accessibility rather than its destruction (Frankland, Josselyn, & Köhler, 2019). A complementary proposal recasts some forgetting as adaptive: engram cells may be actively switched to a silent, inaccessible state by circuit plasticity, so that forgetting is a regulated process the brain performs, not merely decay or damage (Ryan & Frankland, 2022). In the human domain, high-resolution study of hippocampal amnesia has extended the hippocampus's role beyond memory of the past to the construction of imagined and future scenes, suggesting that the amnesic deficit is partly a failure to build coherent mental scenes at all (Clark & Maguire, 2016). Whether the engram findings in animals scale to the graded, decades-spanning retrograde amnesias of human patients is the open question these lines are converging on.

Common Misconceptions

Amnesia erases the person's identity and their whole past.

The organic amnesic syndrome overwhelmingly impairs the formation of new memories (anterograde) and a graded band of recent premorbid ones; identity and remote autobiographical memory are typically preserved. Total loss of personal identity is characteristic of dissociative amnesia, a different, non-lesional condition (Staniloiu & Markowitsch, 2014).

Amnesic patients cannot learn anything new.

They cannot form new declarative memories, but they acquire perceptual-motor skills, habits, priming, and conditioned responses normally, often with no awareness of the training — the dissociation that revealed nondeclarative memory (Milner, Corkin, & Teuber, 1968).

Amnesia is a failure of general intelligence.

By definition the deficit is disproportionate: working memory, language, perception, and reasoning are intact, and IQ can be normal. It is this selectivity that makes amnesia a probe of memory rather than of cognition in general (Scoville & Milner, 1957).

Glossary

Amnesia.

A pathological loss of memory, disproportionate to any other cognitive impairment, caused by brain injury or disease.

Anterograde amnesia.

The inability to form new declarative memories after the onset of injury, so that ongoing experience leaves no lasting trace.

Autonoetic consciousness.

The self-knowing awareness that accompanies episodic retrieval, enabling a rememberer to mentally re-experience a past event as their own; its loss is central to the episodic deficit in amnesia.

Confabulation.

The unwitting production of false or distorted memories offered with conviction, characteristic of Korsakoff syndrome.

Declarative memory.

Conscious memory for facts and events that can be verbally reported; the system selectively lost in amnesia.

Diencephalic amnesia.

Amnesia caused by damage to midline diencephalic structures such as the mammillary bodies and thalamic nuclei, as in Korsakoff syndrome.

Dissociative amnesia.

A functional loss of autobiographical memory, sometimes including identity, following psychological trauma without a detectable brain lesion.

Engram.

The physical substrate of a stored memory, conceived as an ensemble of neurons whose coordinated reactivation expresses recall.

Episodic memory.

Memory for specific personally experienced events, bound to their time and place, and re-experienced on retrieval.

Korsakoff syndrome.

A chronic diencephalic amnesia caused by thiamine deficiency, typically in chronic alcoholism, with dense anterograde and retrograde loss.

Medial temporal lobe.

The hippocampus and adjacent entorhinal, perirhinal, and parahippocampal cortices, whose bilateral damage produces the amnesic syndrome.

Multiple trace theory.

The proposal that detailed episodic memories remain permanently dependent on the hippocampus, each retrieval creating a new trace.

Nondeclarative memory.

Memory expressed through performance rather than recollection — skills, habits, priming, and conditioning — largely spared in amnesia.

Retrograde amnesia.

Loss of memories acquired before the injury, usually graded so that recent memories are more affected than remote ones.

Ribot's law.

The generalisation that retrograde amnesia follows a temporal gradient, with the most recent memories the most vulnerable.

Systems consolidation.

The slow, post-encoding reorganisation by which a memory becomes progressively independent of the hippocampus and stored in neocortex.

Transient global amnesia.

A sudden, self-limiting episode of dense anterograde amnesia lasting hours, with full recovery except for the episode itself.

Key Researchers

Sheena A. Josselyn (contemporary). Senior scientist at the Hospital for Sick Children and professor at the University of Toronto; her engram research reconceives amnesia at the cellular level as a failure to retrieve memories from surviving engram cells rather than a loss of storage. Faculty Page - ORCID - Google Scholar - Wikipedia

Eleanor A. Maguire (1970-2025). Professor at the Wellcome Centre for Human Neuroimaging, University College London; she characterised hippocampal amnesia and showed that hippocampal damage impairs not only memory of the past but the ability to imagine and construct novel scenes. Tribute - ORCID - Wikipedia

Brenda Milner (b. 1918). Emerita professor at the Montreal Neurological Institute, McGill University; her studies of patient H.M. founded the modern science of amnesia, showing that medial temporal removal abolishes new declarative memory while sparing perceptual-motor skill learning. Faculty Page - Wikipedia

Morris Moscovitch (b. 1945). Professor at the University of Toronto and the Rotman Research Institute, Baycrest; co-author of multiple trace theory, the leading challenge to the standard consolidation model, holding that the hippocampus remains permanently necessary for detailed episodic retrieval. Faculty Page - ORCID - Google Scholar - Wikipedia

Larry R. Squire (b. 1941). Distinguished professor at the University of California, San Diego and the VA San Diego Healthcare System; he defined the declarative/nondeclarative taxonomy and the standard model of systems consolidation, and mapped the human medial temporal lobe memory system. Faculty Page - ORCID - Google Scholar - Wikipedia

Endel Tulving (1927-2023). University Professor Emeritus at the University of Toronto; he drew the episodic/semantic distinction and the concept of autonoetic consciousness, framing amnesia as a selective loss of the ability to re-experience the personal past. In Memoriam - Google Scholar - Wikipedia

Frequently Asked Questions

What is the difference between anterograde and retrograde amnesia?
Anterograde amnesia is the inability to form new memories after an injury, so ongoing experience leaves no lasting trace; retrograde amnesia is the loss of memories laid down before the injury, usually graded so recent memories are more affected than remote ones (Squire, 1992).

What part of the brain is damaged in amnesia?
The classic cause is bilateral damage to the medial temporal lobe (the hippocampus and adjacent cortices), but midline diencephalic structures produce a similar syndrome, indicating a single extended memory circuit (Scoville & Milner, 1957; Squire, Stark, & Clark, 2004).

Can someone with amnesia still learn new skills?
Yes. Amnesic patients acquire perceptual-motor skills, habits, and priming normally, often without any memory of the practice, showing the dissociation between spared nondeclarative and impaired declarative memory (Milner, Corkin, & Teuber, 1968).

Why are old memories often spared while recent ones are lost?
The standard consolidation model holds that memories gradually become independent of the hippocampus and are stored in neocortex, so hippocampal damage erases recent, still-dependent memories while sparing older, consolidated ones, producing Ribot's temporal gradient (Bayley et al., 2005).

What is transient global amnesia?
It is a sudden, benign episode in which an otherwise healthy person cannot form new memories for several hours and repeats the same questions, then recovers fully apart from a permanent gap for the episode; it reflects temporary hippocampal dysfunction (Bartsch & Deuschl, 2010).

Is dissociative (psychogenic) amnesia the same as organic amnesia?
No. Dissociative amnesia follows psychological trauma with no detectable lesion and often reverses the organic profile, with a disproportionate loss of personal history and identity but relatively intact new learning (Staniloiu & Markowitsch, 2014).

Does amnesia destroy memories or just block access to them?
This is the central open question. Graded retrograde loss, recovery in transient global amnesia, and animal engram experiments in which stimulating tagged cells restores a seemingly lost memory all suggest that retrieval failure plays a large role alongside any true storage loss (Josselyn & Tonegawa, 2020).

Does amnesia lower intelligence?
No. The amnesic deficit is disproportionate to any other impairment: working memory, language, perception, and reasoning are intact, and general intelligence can be entirely normal (Scoville & Milner, 1957).

References