Transient Global Amnesia

Abstract

Transient global amnesia is a striking, self-limiting form of amnesia in which an otherwise alert adult abruptly loses the ability to lay down new memories for several hours, repeatedly asking the same orienting questions before recovering almost completely. The deficit is unusually pure: dense anterograde amnesia with a variable retrograde gap, yet fully preserved identity, language, and procedural skill, and a working-memory span long enough to sustain a conversation that will not be remembered. That profile localises the fault to the hippocampus, and diffusion-weighted MRI confirms it, revealing transient punctate lesions confined to the CA1 subfield 24–72 hours after onset that vanish within a week. Why CA1 fails so selectively and reversibly remains debated, but the syndrome offers a natural experiment in which episodic-memory encoding is switched off and on.

Keywords: transient global amnesia, anterograde amnesia, hippocampus

Transient global amnesia (TGA) was named and defined in the 1960s, but its modern clinical identity rests on Hodges and Warlow's careful separation of it from the other syndromes of transient amnesia — transient epileptic amnesia and the amnesia of cerebral ischaemia — on the basis of a witnessed attack of pure anterograde memory loss, without clouding of consciousness, focal neurological signs, or epileptic features, resolving within 24 hours (Hodges & Warlow, 1990a). Prospective study of 142 patients confirmed that during the attack the failure is specifically one of *encoding* new episodic information, with the rest of cognition — attention, language, praxis, and old semantic knowledge — remarkably spared (Quinette et al., 2006). That dissociation, and the discovery two decades ago that it corresponds to a visible, reversible lesion in one hippocampal subfield, is what makes TGA a rare window on how a single structure builds the memory of an event (Bartsch & Butler, 2013).

What Transient Global Amnesia Is

Transient global amnesia is a syndrome in which a previously well adult, most often in late middle age, suddenly becomes unable to form new memories. The onset is abrupt and usually witnessed: the patient becomes bewildered, repeatedly asking the same orienting questions — where they are, what day it is, how they came to be there — and, told the answer, forgets it within a minute or two and asks again. The word *global* is a historical misnomer; the amnesia is not global at all. Personal identity is intact (the patient always knows who they are), language and comprehension are normal, complex learned skills such as driving are retained, and the patient remains fully alert throughout (Hodges & Warlow, 1990a). What fails is one specific capacity: the encoding of new episodic events into a form that will survive more than the passing moment.

Because the picture can be alarming and superficially resembles a stroke, TGA is defined by strict clinical criteria that exclude its mimics. The attack must be witnessed, there must be clear anterograde amnesia during it, there must be no clouding of consciousness or loss of personal identity, no focal neurological signs and no epileptic features, the attack must resolve within 24 hours, and the patient must not have had recent head injury or active epilepsy (Hodges & Warlow, 1990a). Applied prospectively to 142 patients, these criteria pick out a strikingly homogeneous group whose only deficit during the episode is a failure of new episodic learning, with attention, working memory, semantic knowledge, and executive function preserved (Quinette et al., 2006). This purity is what distinguishes TGA from the broader family of transient amnesic syndromes, and it is why the condition has become a model case for the cognitive neuroscience of memory (Bartsch & Butler, 2013).

Table 1 places TGA among the transient amnesias it must be separated from, across the features clinicians use to tell them apart.

Table 1

Transient Global Amnesia Distinguished From Its Main Mimics

The Amnesia Profile

The defining feature of TGA is the *shape* of the memory loss, not its depth alone. During an attack there are two amnesias at once. The anterograde deficit is near-total: the patient cannot retain new information beyond the span of immediate attention, so anything told to them slips away within a minute, which is what drives the relentless repetitive questioning. The retrograde deficit is partial and graded: memories from the recent past — days, weeks, sometimes months or years before the attack — are also inaccessible, but this retrograde gap shrinks as the episode resolves, with the most remote memories returning first and the events of the attack itself never being recovered (Quinette et al., 2006).

What is spared is as diagnostic as what is lost. Working memory — the ability to hold and manipulate a small amount of information over seconds — is intact, so the patient can repeat a phone number, follow a sentence, and sustain a conversation they will not later remember (Hodges & Warlow, 1990a). Procedural memory is untouched: patients have driven home, cooked, and played instruments during attacks. Semantic memory for facts about the world and, above all, personal identity are preserved. This selective sparing tells us that the lesion cannot be affecting memory in general; it must be striking the specific system that converts a moment's experience into a durable episodic trace — the hippocampal encoding machinery — while leaving the short-term buffer, the stores of old knowledge, and the skill systems alone (Bartsch & Deuschl, 2010). TGA thus behaves like a temporary, reversible version of the permanent amnesia produced by bilateral hippocampal destruction, which is precisely why it is so informative.

The CA1 Lesion Timeline

For decades TGA was a syndrome without a visible lesion, and its cause was inferred entirely from the behaviour. That changed when diffusion-weighted MRI was applied at the right moment and revealed the culprit: tiny, focal, high-signal lesions in the hippocampus, almost always confined to the CA1 subfield — the same population of pyramidal neurons that is selectively vulnerable to metabolic stress elsewhere in neurology (Bartsch et al., 2006). The lesions are typically 1–3 mm, often single, and lie in the lateral part of the hippocampus where CA1 curves along the ventricle.

The lesions have a characteristic and clinically crucial time course. Imaging done in the first hours after onset is frequently *normal*: the diffusion change takes time to develop. Sensitivity climbs to its peak around 24–72 hours after the attack, then the lesions fade, and by a week or two they are usually gone, leaving no permanent scar (Bartsch et al., 2007). This evolution — invisible at onset, maximal on day 2–3, resolved within a week — explains why an early scan can be falsely reassuring and why the optimal time to confirm the diagnosis on MRI is two to three days later (Sander & Sander, 2005). The transience of the lesion mirrors the transience of the amnesia, and the confinement to CA1 provides the anatomical bridge between the two: the deficit is severe because CA1 is essential to episodic encoding, and it recovers because the neurons are stunned rather than destroyed (Szabo, 2014).

The Trigger Pathway

If the lesion is a reversible insult to CA1, what delivers it? TGA is notable for its precipitants: attacks are often preceded by a sudden physical exertion, immersion in cold or hot water, a Valsalva-like strain, sexual intercourse, or an episode of acute emotional stress (Quinette et al., 2006). This clustering of triggers has generated three overlapping mechanistic hypotheses, none yet decisive.

The venous congestion hypothesis notes that many triggers raise intrathoracic or intra-abdominal pressure, which, in people with incompetent jugular venous valves, can transmit a pressure wave back into the cerebral venous system and congest the medial temporal lobe (Lewis, 1998). The spreading-depression hypothesis proposes that a slow wave of neuronal and glial depolarisation, of the kind that underlies the migraine aura, sweeps across the hippocampus and transiently disables CA1 — an idea supported by the well-documented association between TGA and migraine, quantified in a meta-analysis showing migraineurs are substantially over-represented among TGA patients (Liampas et al., 2022). The metabolic-stress hypothesis emphasises the selective vulnerability of CA1 neurons to a mismatch between energy demand and supply, which an emotional or physical stressor could precipitate (Bartsch & Deuschl, 2010). Hodges and Warlow's early case-control study had already shown that TGA patients, unlike patients with transient ischaemic attacks, do not carry the classic vascular risk-factor profile, arguing against a simple ischaemic-stroke mechanism and pointing toward these more idiosyncratic pathways (Hodges & Warlow, 1990b). The three hypotheses are not mutually exclusive: each could be a route by which a stereotyped final event — transient CA1 dysfunction — is reached (Sander & Sander, 2005).

Worked Example

One of the most reassuring facts about TGA is how rarely it recurs — a sharp contrast with transient epileptic amnesia, which recurs by definition, and with transient ischaemic attack, which heralds a high stroke risk (Morris et al., 2020). We can make the reassurance quantitative with a simple constant-hazard model.

Suppose the risk of a recurrence is constant from year to year at an annual hazard $p$. Then the probability of remaining recurrence-free for $n$ years is $(1-p)^n$, and the cumulative probability of *at least one* recurrence within $n$ years is:

$$P(\text{recurrence within } n \text{ years}) = 1 - (1-p)^n$$

Take an annual recurrence hazard of $p = 0.025$ (about 2.5% per year, consistent with the low rates reported in large follow-up series):

$$P(1 \text{ year}) = 1 - (0.975)^1 = 0.025 = 2.5\%$$

$$P(5 \text{ years}) = 1 - (0.975)^5 = 1 - 0.881 = 0.119 \approx 11.9\%$$

$$P(10 \text{ years}) = 1 - (0.975)^{10} = 1 - 0.776 = 0.224 \approx 22.4\%$$

Even after a decade, then, roughly three-quarters of patients would be expected never to have a second attack. The model also clarifies what risk-factor research is really doing: when a study identifies a subgroup — say, those with a personal history of migraine or a shorter attack — whose annual hazard is elevated, it is estimating a larger $p$ for that subgroup, which the same formula turns into a steeper cumulative curve (Morris et al., 2020). Because the underlying event is benign and self-limiting, even the elevated subgroups remain far from the recurrence rates seen in the epileptic and ischaemic mimics, which is the clinical bottom line: a confirmed TGA is a frightening event with an excellent prognosis (Arena & Rabinstein, 2015).

Discussion

The scientific value of transient global amnesia lies in the coincidence of three things that rarely occur together: a pure and complete deficit, a visible and localised lesion, and full reversibility. Permanent amnesia from hippocampal damage tells us that the hippocampus is necessary for episodic memory, but the lesions are usually large, static, and confounded by surrounding injury. TGA offers the same lesion in miniature — a stunned CA1 field — and then removes it, letting researchers watch the memory system fail and recover on a timescale of hours to days (Bartsch & Deuschl, 2010). The three demonstrations on this page track the three faces of the syndrome: the amnesia profile shows *what* is lost and spared, the CA1 timeline shows *where and when* the lesion appears, and the trigger pathway shows the candidate routes *by which* it is delivered.

The convergence across these levels is the strongest part of the account. A cognitive psychologist charting a patient's spared working memory against their abolished new learning, a neuroradiologist finding a 2 mm CA1 dot on a day-2 diffusion scan, and a physiologist modelling a spreading-depression wave through the hippocampus are describing one event at three resolutions (Bartsch & Butler, 2013). Where the account is still incomplete is at the trigger end: we can localise the lesion far more confidently than we can explain what sets it off, and the venous, spreading-depression, and metabolic hypotheses remain in productive tension rather than resolved (Spiegel et al., 2017). That gap is itself instructive — it marks the boundary between what a reversible-lesion model can settle (the anatomy of episodic encoding) and what it cannot yet settle (the vascular and metabolic physiology upstream of the neurons).

Cognitive Implications

For cognitive psychology, TGA is a clean demonstration of the *modularity of memory*. The classic distinction between short-term (working) memory and long-term memory, and within long-term memory between episodic, semantic, and procedural systems, was built largely from permanent-lesion cases. TGA reproduces the crucial dissociation acutely and reversibly: the anterograde failure to lay down new episodes coexists with intact working memory, intact semantic knowledge, and intact procedural skill, so the systems must be separable and must depend on different neural substrates (Bartsch & Butler, 2013). The fact that a lesion in a single hippocampal subfield can produce this exact pattern narrows the anatomy still further — it identifies CA1 as a critical node in the specific circuit that binds the elements of an experience into a storable episode (Bartsch et al., 2006).

The retrograde gradient adds a second lesson, this one about memory consolidation. That the retrograde amnesia is temporally graded — recent memories more vulnerable than remote ones — and that it shrinks from the remote end inward as the attack resolves fits the view that memories become progressively independent of the hippocampus over time. A transient hippocampal lesion should, on that view, impair exactly the memories that still depend on the hippocampus (the recent ones) while sparing those that have consolidated elsewhere (the remote ones), which is what the TGA retrograde gradient shows (Quinette et al., 2006). The syndrome therefore does double duty: it isolates the encoding function of the hippocampus and, through the shape and recovery of its retrograde gap, illustrates the time-course of systems-level consolidation.

Current Directions

Contemporary TGA research has moved beyond localising the lesion to asking how the lesion disturbs the wider memory *network*. Resting-state functional MRI during and after attacks shows that the transient CA1 insult is accompanied by measurable changes in the functional connectivity of the hippocampus with the rest of the memory network, and that these network-level disturbances track and then resolve alongside the clinical deficit — evidence that TGA is a transient network event, not merely an isolated focal one (Zidda et al., 2019).

A second front is risk stratification. Large follow-up cohorts have begun to identify the factors associated with recurrence, so that the small subgroup at higher risk can be distinguished from the large majority for whom a single attack is a one-off (Morris et al., 2020). A third front is the migraine link: the meta-analytic confirmation that migraine is over-represented in TGA has sharpened interest in the spreading-depression hypothesis and in whether the two conditions share a common cortical-hyperexcitability substrate (Liampas et al., 2022). Across all three fronts the open question is the same one the Discussion flagged — the mechanism of the trigger — and progress there will decide whether the venous, spreading-depression, and metabolic accounts are competitors or complementary descriptions of one pathway (Spiegel et al., 2017).

Common Misconceptions

The patient loses their identity, as in the movies.

No. The dramatic amnesia of fiction, in which a person forgets their own name and history, essentially never happens in TGA. Personal identity is always preserved; the patient knows their name and recognises family. What is lost is the ability to form new memories and to recall the recent past, not the sense of self (Hodges & Warlow, 1990a).

It is a kind of mini-stroke.

No. Although it can look like one, TGA patients do not carry the vascular risk-factor profile of transient ischaemic attack, the lesion is confined to CA1 rather than following an arterial territory, and — critically — TGA does not carry the elevated stroke risk that follows a TIA (Hodges & Warlow, 1990b).

The amnesia is global, affecting all memory.

No, despite the name. Working memory, semantic knowledge, and procedural skills are all intact during an attack; the failure is specific to forming new episodic memories, with a partial, recovering loss of recent old ones (Quinette et al., 2006).

A normal early MRI rules it out.

No. The CA1 diffusion lesion is often invisible in the first hours and only becomes detectable at 24–72 hours, so an early normal scan is expected and does not exclude the diagnosis (Bartsch et al., 2007).

Glossary

Anterograde amnesia.

The inability to form new memories after the onset of a condition; in transient global amnesia this deficit is near-total and drives the repetitive questioning.

CA1 field.

A subfield of the hippocampus whose pyramidal neurons are selectively vulnerable to metabolic stress; the site of the transient lesion in transient global amnesia.

Diffusion-weighted imaging (DWI).

An MRI sequence sensitive to the restricted movement of water molecules in stressed tissue; it reveals the punctate CA1 lesion, best at 24–72 hours after onset.

Episodic memory.

Memory for specific personally experienced events, bound in time and place; the memory system whose encoding fails in transient global amnesia.

Hippocampus.

A medial temporal-lobe structure essential for forming new episodic memories; the CA1 subfield within it is the focus of the transient global amnesia lesion.

Memory consolidation.

The process by which new memories become progressively stable and independent of the hippocampus; the temporally graded retrograde gap in transient global amnesia reflects it.

Procedural memory.

Memory for skills and habits, expressed through performance rather than recollection; fully preserved during a transient global amnesia attack.

Retrograde amnesia.

Loss of memories formed before the onset of a condition; in transient global amnesia it is partial, temporally graded, and shrinks as the attack resolves.

Spreading depression.

A slow, self-propagating wave of neuronal and glial depolarisation, implicated in migraine aura and proposed as a trigger that transiently disables CA1.

Transient epileptic amnesia.

A recurrent, brief amnesic syndrome caused by focal seizure activity; a key differential diagnosis distinguished by its short duration and frequent recurrence.

Transient global amnesia.

A sudden, self-limiting syndrome of dense anterograde and partial retrograde amnesia in an alert adult, lasting up to 24 hours, with a transient CA1 lesion.

Valsalva manoeuvre.

A forced exhalation against a closed airway that raises intrathoracic pressure; a Valsalva-like strain is a recognised precipitant, central to the venous-congestion hypothesis.

Venous congestion.

Backflow of raised pressure into the cerebral venous system, proposed to congest the medial temporal lobe in people with incompetent jugular venous valves.

Working memory.

The system that holds and manipulates a small amount of information over seconds; intact in transient global amnesia, which is why conversation is possible during an attack.

Key Researchers

Thorsten Bartsch

Professor for Memory Disorders and Plasticity in the Department of Neurology at Christian-Albrechts-University Kiel, whose diffusion-MRI studies localised the transient lesion of TGA to the hippocampal CA1 field and traced its time course. Faculty Page

Günther Deuschl

Professor of Neurology at Christian-Albrechts-University Kiel and University Hospital Schleswig-Holstein, co-author of the CA1 imaging work and the *Lancet Neurology* synthesis of the functional anatomy of TGA. ORCID - Faculty Page

Francis Eustache

Director of the INSERM neuropsychology and memory unit at the Université de Caen Normandie, whose large prospective series characterised the precise cognitive profile of TGA. ORCID - Wikidata

C. Miller Fisher

(1913–2012). Canadian-American neurologist at Massachusetts General Hospital and Harvard Medical School who, with Raymond Adams, named and defined transient global amnesia in the 1960s. Wikipedia - Wikidata

John R. Hodges

Behavioural neurologist at the University of Sydney and Neuroscience Research Australia, formerly at Cambridge, whose case-control and classification studies established the modern clinical criteria for TGA. Google Scholar - Faculty Page

Alejandro A. Rabinstein

Professor of Neurology at the Mayo Clinic, Rochester, co-author of contemporary reviews and follow-up cohorts defining the prognosis and recurrence risk of TGA. ORCID - Google Scholar - Faculty Page

Kristina Szabo

Neurologist at the University Medical Centre Mannheim, Heidelberg University, whose neuroimaging work includes resting-state connectivity studies of the memory network during and after TGA. Faculty Page

Frequently Asked Questions

Is transient global amnesia dangerous?

Rarely. It is frightening to witness and to experience the after-report of, but the attack itself is benign and self-limiting, memory recovers almost completely within 24 hours, and, unlike a transient ischaemic attack, it does not carry an elevated risk of stroke. The main clinical task is to confirm the diagnosis and exclude the mimics.

How long does an attack last?

Most attacks last four to eight hours, and by definition no more than 24. During that window the patient cannot retain new information; once it lifts, the ability to form memories returns, though the events of the attack itself are never recovered.

Will the patient remember the episode afterward?

No. Because memories were never encoded during the attack, there is a permanent blank for the period of the amnesia and often for a short time just before it. Everything from before that window, and everything after recovery, is remembered normally.

What causes transient global amnesia?

The immediate cause is a transient, reversible disturbance of the hippocampal CA1 field, visible as a tiny lesion on diffusion MRI at 24 to 72 hours. What triggers that disturbance is debated: leading hypotheses are venous congestion from a Valsalva-like strain, a spreading-depression-like wave related to migraine, and acute metabolic stress on vulnerable CA1 neurons.

Why do patients keep asking the same questions?

Because the anterograde deficit is near-total, the answer they are given is forgotten within a minute, so the same disorienting uncertainty returns and the same question is asked again. This repetitive questioning is the most characteristic feature of an attack.

Is it the same as amnesia in films, where people forget who they are?

No. Loss of personal identity essentially never occurs in TGA. Patients always know who they are and recognise their families; what fails is the formation of new memories and the recall of the recent past, not the sense of self.

Does a normal MRI mean it was not TGA?

Not necessarily. The CA1 lesion is often invisible in the first hours and only becomes detectable a day or two later, so an early normal scan is expected. When imaging confirmation is needed, the scan is best repeated at 24 to 72 hours.

Will it happen again?

Usually not. Recurrence is low, far lower than in the recurrent transient epileptic amnesia it must be distinguished from, and most people who have one attack never have another, although a small subgroup with particular risk factors is more prone to recurrence.

Support Organizations

References