Long-term memory (LTM) is not one single “store”. Instead, psychologists argue that we keep different kinds of information in different systems, which helps explain why someone might forget life events but still play the piano perfectly. This page explores episodic, semantic and procedural memory, and the evidence that supports (and challenges) these distinctions.

Key terms

Long-term memory (LTM): A memory system with potentially unlimited capacity and duration, storing information for long periods.
Episodic memory: Memory for personal events (e.g., “my first day at secondary school”), including time, place, and emotion.
Semantic memory: Memory for facts, concepts, and general knowledge (e.g., “Paris is the capital of France”).
Procedural memory: Memory for skills and actions (e.g., riding a bike), often learned gradually through practice.
Declarative memory: “Knowing that” memory that can be consciously recalled and described (includes episodic + semantic).
Non-declarative memory: “Knowing how” memory that is often unconscious/automatic (includes procedural, and other implicit forms).
Amnesia: Memory loss; anterograde = difficulty forming new memories, retrograde = loss of past memories.
PET scan: A brain-imaging method that tracks metabolic activity to infer which brain areas are active during tasks.

The idea: LTM has multiple systems

A common model splits LTM into declarative (episodic + semantic) and non-declarative (procedural). The key claim is that these systems rely on partly different brain networks, so damage can affect one system more than another.

Episodic memory

What it is
Episodic memory stores autobiographical events and the context around them (time, place, feelings). It supports “mental time travel” — re-experiencing moments rather than just knowing facts.

Case evidence: Clive Wearing
Clive Wearing developed profound amnesia after herpes simplex encephalitis. Reports describe an extremely limited ability to retain new personal experiences, with his awareness often feeling like it “resets” after a short time. However, some abilities (especially skilled musical performance) remained strong, which suggests severe impairment to episodiclearning alongside preserved skill memory.

Neuroimaging evidence: Tulving and PET scans
PET research linked different patterns of brain activity to different memory demands, supporting the idea that remembering personal episodes recruits partly distinct networks (including frontal regions) compared with other forms of memory. This biological evidence strengthens the argument that episodic memory is not identical to other LTM types.

Semantic memory

What it is
Semantic memory stores facts and meanings that are not tied to one specific personal event. For example, you may know what “photosynthesis” means without remembering when you first learned it.

How it can differ from episodic memory
Some people can lose detailed autobiographical recollection while still keeping parts of their general knowledge, suggesting that semantic memory can be at least partly separable from episodic memory (even though they often work together in everyday life).

Procedural memory

What it is
Procedural memory stores learned skills and habits. It is often hard to put into words, and you usually show it through performance rather than explanation.

Case evidence: Henry Molaison (H.M.)
After surgery that removed parts of his medial temporal lobes, H.M. showed severe difficulty forming new declarativememories. However, evidence from repeated skill-learning tasks showed that he could still improve at motor skills even when he could not consciously remember practising them, supporting the distinction between declarative and non-declarative memory.

Case evidence: Clive Wearing (again)
Despite severe difficulties with forming new day-to-day memories, Wearing could still perform complex musical skills. This pattern supports the view that procedural memory can remain relatively intact when episodic memory is profoundly impaired.

Real-life application: memory training (Belleville)

A practical strength of separating memory systems is that it guides targeted interventions. Belleville and colleagues used cognitive training focused on episodic memory strategies with older adults (including individuals with mild cognitive impairment). The intervention group improved on episodic memory measures compared with groups who did not receive the training, suggesting that supporting episodic processes can produce measurable gains in real-world settings.
Longer-term follow-up work also reports sustained benefits for some participants years later, which strengthens the case for applied value beyond the lab.

Strengths

The key strength of the multi-system view of LTM is that it explains striking real-world patterns that single-store accounts struggle to handle. For example, H.M.’s inability to form new conscious memories alongside preserved skill learning provides a clear dissociation: if one memory system fails while another still works, it strongly suggests separate mechanisms. This supports the declarative vs non-declarative split and gives the theory explanatory power for neurological cases.

A second strength is the support from neuroscience methods, which add biological plausibility. PET evidence shows different activation patterns depending on the kind of memory processing taking place, suggesting that the brain does not treat “remembering” as one single operation. When behavioural patterns align with neuroimaging differences, the overall argument becomes more convincing than relying on one method alone.

A third strength is application value. Belleville’s intervention work suggests that understanding episodic memory as a distinct system can help psychologists design strategy-based training that improves performance in older adults, including those at risk of decline. This matters because it turns a theory about memory structure into a tool for improving everyday functioning and independence.

Weaknesses

A major weakness is that case-study evidence can be powerful but also hard to generalise. H.M. and Wearing are unusual individuals with rare and severe damage, so their memory profiles might not represent typical memory organisation. In addition, their brain injuries were not perfectly “neat”, so it becomes difficult to claim that one memory system alone caused each observed symptom. This limits how confidently we can apply one person’s pattern to everyone.

A second weakness is that the boundaries between systems can blur, which challenges a strict “separate stores” interpretation. Episodic and semantic memory often support each other in everyday cognition, and research reviews argue that the two forms can be interdependent rather than fully independent. This means the model may oversimplify memory by presenting categories as more distinct than they really are.

A further criticism targets the declarative vs non-declarative split itself: it can hide important complexity inside each category. For example, procedural learning can still involve conscious strategies at early stages, and declarative knowledge can become automatic with practice. If systems interact heavily during learning and recall, then the theory may describe useful labels rather than truly separate “stores” that operate independently.