11.6 — The Limbic System and Memory

In 1953, a man known for fifty years only as H.M. had both his medial temporal lobes removed to control intractable epilepsy. The seizures improved. And he could never form a new conscious memory again, for the remaining 55 years of his life.

He could hold a conversation. His intelligence and personality were intact. He could learn new motor skills, improving at them day after day — while insisting each morning that he had never done them before. He read the same magazine repeatedly with fresh interest.

H.M. — Henry Molaison — taught neuroscience more about memory than any other single case, and almost everything on this page traces back to him.

The limbic structures

A ring of structures on the inner border of each hemisphere. Limbus is Latin for border.

Hippocampus — named for its seahorse shape. Forms new conscious memories and is central to spatial navigation.

Amygdala — almond-shaped, sitting just in front of the hippocampus. Emotional processing, especially fear, and attaching emotional significance to events.

Hypothalamus — the homeostatic and endocrine controller (Chapter 11.4).

Cingulate gyrus — arching over the corpus callosum. Emotion, attention, and the sense of conflict between competing responses.

Fornix and mammillary bodies — the output pathway from the hippocampus.

Nucleus accumbens — reward and motivation.

Kinds of memory

Memory is not one system, and the strongest evidence for that is that the systems can be damaged separately.

Sensory memory — under a second, holding raw sensory input.

Working (short-term) memory — seconds to a minute, and it has a famously limited capacity.

The traditional figure is "seven plus or minus two" items, from George Miller's 1956 paper. More recent work suggests the real capacity is closer to four chunks, with the higher figure achieved by grouping.

And chunking is why a phone number is broken into groups. 0-7-9-4-6-2-3-1-8-5 is ten items. 079 462 3185 is three. The capacity limit is in chunks, not in digits, which is why an expert can hold far more information in their own field — they chunk it more efficiently.

Working memory depends on the prefrontal cortex, not the hippocampus, which is why H.M.'s working memory was normal. He could hold a conversation; he simply could not convert any of it into anything lasting.

Long-term memory divides into two systems, and this is the division H.M. revealed.

Declarative (explicit) memory — facts and events, consciously recalled.

• Episodic — personal experiences, with a time and place. Your last birthday.
• Semantic — facts, without context. The capital of France.

Non-declarative (implicit) memory — expressed through performance rather than recall.

• Procedural — skills. Riding a bicycle, typing.
• Priming, classical conditioning, and simple learned associations.

H.M. lost the ability to form new declarative memories and retained the ability to form procedural ones. He learned mirror-drawing over successive days, improving steadily, while each day denying he had ever attempted it. His hands remembered and he did not.

And this is why procedural memory is retained so long in Alzheimer's disease. A person who cannot recognise their own children may still play the piano, because those two memories are stored in different systems and one is affected far later than the other.

How memory is formed

Encoding → consolidation → storage → retrieval.

The hippocampus is required for encoding and consolidation, not for storage.

Memories are ultimately stored in the cortex, distributed across the same regions that processed the original experience — visual aspects in visual cortex, sounds in auditory cortex. The hippocampus binds those fragments together and, over months to years, the cortical connections strengthen until the hippocampus is no longer needed.

This is systems consolidation, and it explains H.M.'s pattern exactly. He remembered his childhood clearly, because those memories had already been consolidated into cortex years before. He could not form new ones, because the binding mechanism was gone. And he had a graded loss of memories from the years immediately before his surgery — the ones still dependent on the hippocampus.

The cellular mechanism is long-term potentiation (Chapter 11.2). The hippocampus has an unusually high density of NMDA receptors, and blocking them in animals prevents new learning while leaving existing memories intact.

Sleep is where much of consolidation happens, and this is well established rather than speculative. During slow-wave sleep, hippocampal activity patterns from the day are replayed to the cortex — literally, the same sequences of neuron firing, at speed. Sleep deprivation impairs consolidation measurably, and studies consistently show that learning followed by sleep produces better retention than the same learning followed by the same interval awake.

This is the most practically useful finding in this chapter. Chapter 24.7.

How memory is retrieved, and why it is unreliable

Retrieval is reconstruction, not playback.

This is the single most important thing to understand about memory, and it contradicts almost everyone's intuition about their own.

Every time a memory is recalled it becomes temporarily unstable and must be re-stored — reconsolidation. And it can be altered during that window.

So a memory recalled many times is not more accurate; it is a memory of the last recall, with any distortions introduced along the way now incorporated.

Elizabeth Loftus's work demonstrated how easily this happens. Witnesses shown a video of a car accident and asked how fast the cars were going when they "smashed into" each other gave higher speed estimates than those asked about cars that "hit" each other — and a week later, those asked the "smashed" question were more likely to report seeing broken glass that was never in the film. A single word changed the memory.

Entirely false memories can be implanted. In a series of studies, a substantial minority of participants came to confidently remember childhood events that never happened, after being told by a relative that they had.

And confidence does not track accuracy. Flashbulb memories — vivid recollections of hearing shocking news — feel exceptionally accurate and are not. Studies collecting accounts immediately after major events and again years later find substantial changes, while confidence remains extremely high.

The consequences for justice are real. Eyewitness misidentification is the leading contributing factor in wrongful convictions later overturned by DNA evidence — present in around 70 percent of the first several hundred exonerations in the United States (Chapter 2.10).

Nothing here means memory is useless. It means memory is a reconstruction optimised for extracting meaning, not a recording optimised for fidelity — which is a good design for an animal that needs to learn from experience and a poor one for a witness.

The amygdala and emotional memory

The amygdala attaches emotional significance, and it has a fast route.

Sensory information reaches the amygdala directly from the thalamus, before it reaches the cortex. So there is a quick, crude pathway and a slower, accurate one.

This is why you jump back from a coiled shape on a path before recognising it is a rope. The fast route reacted; the slow route corrected it a fraction of a second later. The false alarm costs almost nothing; missing a real snake costs everything, and the asymmetry is why the system is built to be trigger-happy.

Emotional arousal strengthens memory, through stress hormones acting on the amygdala, which then modulates hippocampal consolidation. This is why you remember emotionally significant events far better than neutral ones, and it is generally adaptive.

Taken too far it becomes post-traumatic stress disorder, where the memory is consolidated so strongly that it intrudes involuntarily and is re-experienced rather than recalled. Reconsolidation offers a possible therapeutic route — reactivating the memory under conditions that allow it to be re-stored with less emotional charge — and this is the principle behind several trauma therapies. Chapter 20.8.

Bilateral amygdala damage produces a striking loss of fear. People with the rare condition Urbach–Wiethe disease, which calcifies both amygdalae, show almost no fear response to snakes, haunted houses or threatening films — and can still be frightened by carbon dioxide inhalation, which produces suffocation panic through a different route. Fear of external threat and internal alarm are separable.

Where memory fails

Anterograde amnesia — cannot form new memories. H.M.

Retrograde amnesia — cannot recall memories formed before the event. Usually graded, with recent memories lost more than remote ones, because recent ones are less consolidated.

Head injury typically causes both, with a brief retrograde gap and a longer anterograde one. The duration of post-traumatic amnesia is one of the better predictors of long-term outcome after head injury, and it is asked about specifically.

Transient global amnesia is a curious and benign event. A person, usually over 50, suddenly cannot form new memories for several hours, repeatedly asking the same question, while remaining fully alert, knowing who they are, and able to perform complex tasks like driving. It resolves completely within 24 hours, leaving a permanent gap for that period. The cause is not established; it is frightening for families and essentially harmless.

Korsakoff's syndrome — severe memory impairment from thiamine deficiency, damaging the mammillary bodies and thalamus. Usually follows untreated Wernicke's encephalopathy (Chapter 1.6).

Its most striking feature is confabulation — the person fills gaps with fluent, detailed, entirely fabricated accounts, delivered with complete conviction and no intent to deceive. They are not lying; the memory system is producing plausible content where there is none.

And it is largely preventable. Giving thiamine to at-risk patients — particularly anyone with alcohol dependence, and before any glucose — prevents the progression from a reversible Wernicke's to an irreversible Korsakoff's. The intervention costs almost nothing and the failure to give it is a recognised and continuing lapse.

Alzheimer's disease begins in the medial temporal lobes, which is why episodic memory fails first (Chapter 20.4).

Improving memory, from the evidence

This is one of the few areas of neuroscience with a genuinely useful practical output, and it is worth stating because most study advice ignores it.

Spaced repetition beats massed practice, decisively. The same total study time distributed over days produces far better long-term retention than the same time in one block. The effect is one of the largest and most reliable in cognitive psychology, and it is why flashcard systems that schedule reviews at increasing intervals work.

Retrieval practice beats re-reading, and it is not close. Testing yourself on material produces substantially better retention than reading it again, even though re-reading feels more productive. The feeling of fluency from re-reading is exactly the problem: it is mistaken for learning.

Elaboration — connecting new material to what you already know, and asking why it is true — outperforms rote repetition.

Interleaving — mixing different topics rather than blocking them — feels harder and produces better retention and transfer.

Sleep after learning, as above.

Physical exercise improves memory, with reasonable evidence, including increased hippocampal volume in trials of aerobic exercise in older adults.

And the classical technique still works. The method of loci — placing items to be remembered at specific locations along a familiar route — exploits the hippocampus's spatial machinery, and it is what every competitive memory champion uses. It is not a trick with limited application; it works because spatial memory is exceptionally good and it recruits it.

What does not have good evidence: commercial brain-training games improve performance at those games and show little transfer to general cognitive ability. Most supplements marketed for memory have little or no effect.

What the next page fixes

Twelve nerves leave the brain directly rather than through the spinal cord, and between them they carry vision, hearing, smell, taste, balance, facial sensation and movement, and the control of the heart and gut. Chapter 11.7 covers all twelve, what each does, and what its failure looks like.