Memory: Amnesia and Memory Systems

Imagine improving at tracing a shape in a mirror while sincerely believing that you have never attempted the task before. The skill has changed, but your recollection of practising has not. Amnesia reveals that remembering an event, knowing a fact, and acquiring a skill are partly separable achievements.

This lecture extends the earlier Memory notes by examining what patient studies establish about brain organization. The central question is not simply “Where is memory?” but “Which contribution to which kind of remembering depends on which network?”

Learning goals and Feynman practice

Distinguish anterograde from retrograde amnesia, explain H.M.’s preserved and impaired abilities, compare memory systems, and evaluate evidence for consolidation without treating one theory as settled fact.

Read a section, close it, and teach it to a beginner. Explain the task, the finding, the inference, and one limitation. Use a fresh example. When your explanation breaks down, return to the missing step, then retry before opening the hidden answer.

1. Localization and distributed processing can both be true

An engram is a physical substrate or trace of a memory. Lashley’s lesion studies, particularly of maze learning, helped motivate mass action: performance could depend on the amount of cortex damaged rather than one uniquely critical spot. His related idea of equipotentiality concerned the capacity of remaining tissue to support some functions under particular conditions.

This historical position should not become a claim that every brain region stores every memory equally. Maze performance combines sensation, movement, motivation, and learning; distributed contributions can make a single storage location difficult to identify.

Modern accounts accommodate both specialization and distribution. A network can represent information across many regions while relying on a particular structure to bind or retrieve its components. Think of a library catalogue: information is distributed across books, but losing a system that links entries can still cause a specific access problem. The hippocampus is not literally a card catalogue, and memory is not a collection of unchanged files.

Teach it back: Why does finding a structure important for new learning fail to prove that every detail is permanently stored inside it?

2. H.M.: a profound deficit with important spared abilities

Henry Molaison, known in research as H.M., underwent bilateral medial temporal surgery in 1953 to treat severe epilepsy. It involved hippocampal regions and neighbouring tissue, rather than removal of the entirety of both temporal lobes or a clean intervention limited to the hippocampi. Seizures were reduced; memory impairment became a lasting consequence. Scoville and Milner’s report helped transform the study of memory. [1]

H.M. showed profound anterograde amnesia, especially for new declarative learning. Yet language, general intellectual abilities, some immediate memory tasks, and some skill learning were relatively preserved. “Global” describes the breadth of his new declarative learning difficulty across materials, not the disappearance of every memory process.

This case provided strong evidence for specialized medial temporal contributions. It did not, alone, isolate every necessary structure or prove a complete model. Converging patient, animal, and experimental evidence matters because the surgery affected several structures and connections.

Question	Illustrative ability	What it helps distinguish
Can information be maintained briefly?	Repeating a short digit sequence.	Immediate maintenance versus durable new learning.
Can a new event later be recalled?	Remembering meeting an examiner.	Episodic learning.
Can new knowledge be acquired?	Learning a new person’s name and significance.	Semantic learning, assessed over time.
Can performance improve with practice?	Mirror-tracing improvement.	Skill learning versus recollection of the practice episode.

3. Anterograde amnesia is not a stopwatch failure

Anterograde means difficulty learning information after the onset of injury or illness. It does not mean that a person loses consciousness every few seconds or cannot hold any thought.

H.M. could maintain limited information when task demands allowed sustained rehearsal. This does not establish that every aspect of working memory was completely intact or that temporary memory lasts exactly 18–30 seconds. Interruptions, relational complexity, amount of information, and interference can expose different difficulties.

A classic pattern in some amnesic patients is relatively preserved immediate recency with reduced primacy in list recall. This is compatible with a distinction between recently accessible information and contributions from longer-term learning. It is not a universal signature, and serial-position curves depend on task design and retrieval conditions.

Episodic versus semantic new learning

Episodic memory concerns events situated in a particular time and place. Semantic memory concerns facts, concepts, and meanings. Medial temporal amnesia can seriously impair acquisition of both, even while older knowledge remains useful.

H.M. had substantial difficulty learning vocabulary and information introduced after surgery. [2] But later research found some evidence of new semantic acquisition, so “he learned no new facts after 1953” is too absolute. [3] Opportunities for repeated exposure, residual tissue, and the sensitivity of the test affect what can be detected.

Teach-back check: how could someone discuss a familiar subject fluently yet fail to remember today's conversation?

Previously acquired semantic knowledge and language can support the discussion. Forming a durable new episodic record of this particular conversation may depend on processes that are impaired. Fluency is not a test of remembering a recent event.

4. Material-specific memory: relative lateralization

Left medial temporal systems often contribute more strongly to verbal learning, while right medial temporal systems often contribute more strongly to aspects of spatial and nonverbal learning. These are relative specializations within connected networks.

“Left removal affects only words; right removal affects only space” is too strong. Material, strategy, language dominance, lesion extent, and compensation matter. A visual task can be solved with verbal labels, and a verbal task may require spatial context. Unilateral injury also does not guarantee normal performance on every immediate memory task.

Teach it back: Why could naming the objects in a visual array change which processes a memory test recruits?

5. Retrograde amnesia and systems consolidation

Retrograde amnesia affects information acquired before onset. A temporal gradient means that more recent memories are disproportionately affected relative to remote ones. Some patients show this pattern, while others lose detail across much longer periods or have relatively ungraded impairments.

After acute injury, the apparent gap can sometimes shrink as recovery proceeds. This does not guarantee orderly reverse-chronological restoration, and some losses persist. Recovering access, resolving physiological disruption, and changes in assessment can all affect the observed pattern.

Systems consolidation concerns changes over time in how brain systems support memory. Distinguish this from cellular processes that stabilize synaptic changes on other timescales. Neither process implies that a fixed recording moves intact from a temporary box into an invulnerable permanent one.

Evidence	Why it is informative	What it does not establish alone
Temporally graded retrograde amnesia	Older and newer learning can differ in vulnerability.	Every remote memory is hippocampus-independent.
Lesions after different animal training intervals	Tests causal contributions after different retention periods.	A universal human consolidation timetable or equivalence across tasks.
ECT-associated memory changes	Perturbation can affect recent and autobiographical remembering.	That all effects are brief, identical, or limited to one or two years.
Imaging during recent versus remote retrieval	Measures how activity varies with memory age and task.	That an inactive-looking area is unnecessary or that age is the only difference.

ECT findings need particular care: memory effects vary with treatment and assessment, and autobiographical losses can persist. A clinical study examining public and personal events illustrates that these outcomes are not interchangeable. [4] These notes use the findings to discuss evidence, not to guide treatment decisions.

An influential famous-face fMRI study found temporally graded medial temporal activity, but its interpretation emphasized entorhinal involvement rather than a simple hippocampal switch between particular decades. [5] Familiarity, rehearsal, emotional importance, and retrieval detail can also differ between old and new material.

Competing interpretations

Standard consolidation accounts propose that some memories become less dependent on the hippocampus as cortical networks reorganize. Multiple-trace and transformation accounts emphasize continuing hippocampal involvement in richly detailed episodic recollection, with more schematic or semantic representations potentially supported differently.

Patient findings of impaired detail even for remote events helped motivate these alternatives. [6] Memory type and the quality of recollection matter alongside memory age. There is no single proven rule that all explicit memories become permanently cortical after a fixed number of months or years.

Teach-back check: what is the difference between knowing a childhood school's name and vividly re-experiencing one lesson there?

The first can be semantic knowledge; the second requires event-specific details and context. A test that scores both as simply “remote memory” can conceal different neural demands and mislead theories about consolidation.

6. Multiple memory systems: organize by the task

Broad category	Form	Example	Major neural contributions
Declarative / explicit	Episodic	Recalling a particular practice session.	Hippocampal and medial temporal systems interacting with distributed cortex.
Declarative / explicit	Semantic	Knowing what a tool is used for.	Distributed conceptual networks; medial temporal contributions to much new acquisition.
Nondeclarative	Skills and habits	Improved tracking or learned action tendencies.	Corticostriatal, motor, cerebellar, and other networks, depending on the task.
Nondeclarative	Repetition priming	Faster identification following prior exposure.	Relevant perceptual or conceptual processing networks.
Nondeclarative	Conditioning	An acquired autonomic response to a predictive cue.	Amygdala for some emotional conditioning; other circuits support other responses.

Explicit tests request conscious remembering. Implicit tests measure an influence of prior experience without requiring deliberate recollection. These test categories do not guarantee perfectly pure processes: participants may deliberately remember on an implicit task, and several systems can influence one response.

Teach it back: Classify remembering a typing lesson, knowing a keyboard shortcut, and becoming quicker at using it. Why might one test involve more than one category?

7. Skill learning: an improvement can outlast awareness of practice

H.M.’s improvement in mirror-tracing showed that some learning can proceed without ordinary recollection of training. It did not establish that every procedural task is normal in every amnesic person or that skill learning depends only on basal ganglia.

Patient comparisons have found impaired pursuit-rotor learning in Huntington’s disease alongside different patterns in amnesia and Alzheimer’s disease. Other research found preserved mirror-tracing but impaired rotary-pursuit learning within Huntington’s disease itself. [7]

This is a useful correction to the phrase “procedural memory system”: skills differ in feedback, sequence learning, motor demands, and strategies. A disorder affecting basal-ganglia circuits does not selectively remove every skill while leaving all declarative memory intact.

A double dissociation strengthens a functional distinction when one damage pattern disproportionately affects ability A and another disproportionately affects B. Compare learning rates and task demands, not just one patient’s pass and fail. Motor difficulty can lower performance without uniquely identifying a learning deficit.

Teach-back check: why compare improvement across trials rather than only the final tracing score?

An initial motor impairment can produce a poor score even when learning occurs. Change with practice helps assess acquisition, although it must still be interpreted against speed, accuracy, fatigue, and baseline differences.

8. Priming: exposure can influence performance without recollection

In word-stem completion, studying STAMP might increase the probability of completing STA___ as STAMP later. The same stem admits other words, so measure the change against an appropriate unstudied baseline rather than treat one completion as evidence of memory.

Amnesic patients can show priming despite poor explicit recall. To examine implicit influences, distinguish an instruction to complete with the first word that comes to mind from an instruction to retrieve a studied word.

Some comparisons found different patterns of priming and motor learning in Alzheimer’s and Huntington’s disease. [8] But Alzheimer’s disease does not uniformly abolish every form of priming. Perceptual versus conceptual demands, disease severity, baseline performance, and task instructions matter. “Neocortex” names a broad set of contributions rather than one priming centre.

Teach it back: Why would explicitly telling participants to recall the studied list change the interpretation of word-stem performance?

9. Emotional conditioning: bodily learning and explicit knowledge

An influential study compared patients with amygdala and hippocampal damage. The ability to acquire a conditioned autonomic response and the ability to describe the relationship between stimuli showed different impairment patterns. [9]

This supports a distinction between particular emotional-conditioning processes and declarative knowledge. The amygdala contributes to some autonomic threat-learning responses and can modulate other memory processes. It is not the sole site of emotion or a complete storage system for every emotional experience.

A person may know which cue predicts an event without expressing the typical measured autonomic response. Conversely, a learned bodily response can occur without a normal explicit account of the contingency. Neither response directly measures every aspect of felt experience.

10. Worked comparison: choose the evidence before the label

Imagine an examiner presents a word list, teaches a tracking task, and pairs a neutral cue with an outcome. These are teaching scenarios, not an individual diagnostic protocol.

Observation	What it suggests testing	What to avoid concluding
Poor delayed recall but adequate immediate span.	New declarative learning versus limited maintenance.	All short-term processes are intact.
Tracking improves despite no recollection of practice.	A dissociation between skill acquisition and event recollection.	Every implicit system is normal.
Studied stems are completed more often despite poor recall.	Exposure-related facilitation, with awareness controls.	Every completion was unconscious.
The person describes a cue relationship but lacks a measured bodily response.	Separate declarative and autonomic-conditioning contributions.	They feel no emotion.

The strongest account connects behaviour, task structure, lesion evidence, and alternative explanations. A single score cannot establish a full memory-system profile.

Retrieval workshop and spaced review

Draw a timeline marking before and after injury; use it to distinguish retrograde and anterograde deficits. Reconstruct the memory-systems table from memory. Then explain a double dissociation without naming any brain structures before adding the anatomical evidence.

Repair: “H.M. proved memory is localized in the hippocampus”

His case demonstrated a major medial temporal contribution to new declarative learning while revealing spared abilities. The surgery affected multiple structures, and memory representations and other forms of learning depend on distributed networks. Localization of a contribution is different from localization of every memory.

Repair: “A remote memory is permanent because it no longer needs the hippocampus”

Remote memories can change or be forgotten. Hippocampal dependence varies with the type and detail of retrieval, and theories differ. Reduced dependence in some tasks is not proof of permanent, unchanging storage.

Apply: why is a famous-face test not equivalent to recalling a personal event?

Recognizing a famous person can rely strongly on semantic familiarity and repeated exposure. Recollecting a unique personal event requires contextual and experiential details. The tasks differ in more than memory age, so anatomical findings cannot be transferred automatically.

Review tomorrow and in a week. Score each explanation: 0 for a label, 1 for a definition, 2 for a mechanism plus example and limitation. Repair the weakest one before adding terminology.

One-minute explanation: Amnesia separates abilities that ordinarily cooperate. Medial temporal systems contribute strongly to new declarative learning, while some skill, priming, and conditioning effects can remain. Retrograde patterns inform consolidation theories, but memory age alone does not determine brain dependence. Systems are partly specialized and interconnected, so conclusions must specify the task, memory content, and evidence.

Sources and further reading

The supplied numbered citations did not include a bibliography. These primary studies and theory papers support the clinical comparisons; the examiner scenarios and exercises are teaching aids.

1. Scoville, W. B., and Milner, B. (1957). Loss of recent memory after bilateral hippocampal lesions.
2. The impaired learning of semantic knowledge following bilateral medial temporal-lobe resection.
3. Evidence for semantic learning in profound amnesia: an investigation with patient H.M..
4. The effects of electroconvulsive therapy on memory of autobiographical and public events.
5. Consolidation of human memory over decades revealed by functional magnetic resonance imaging.
6. Multiple trace theory of human memory: computational, neuroimaging, and neuropsychological results.
7. Intact mirror-tracing and impaired rotary-pursuit skill learning in patients with Huntington’s disease.
8. Neuropsychological evidence for multiple implicit memory systems.
9. Bechara, A., et al. (1995). Double dissociation of conditioning and declarative knowledge relative to the amygdala and hippocampus in humans.