Memory — AQA GCSE Psychology Revision Notes

What you'll learn

• The three processes every memory depends on, and what goes wrong when each one fails
• The multi-store model, the evidence that supports it and the reasons it is now thought too simple
• Three distinct types of long-term memory, and how amnesia cases prove they are separate
• Why the serial position curve is such powerful evidence for more than one memory store
• The theory of reconstructive memory, and how schemas quietly rewrite what you remember
• Two competing explanations of forgetting, and how to tell which one a question is describing
• Why eyewitness testimony can be sincerely and confidently wrong

Key terms and definitions

Encoding — converting information into a form the memory system can hold.

Storage — retaining information in the memory system over time.

Retrieval — recovering stored information when it is needed.

Capacity — how much a store can hold. Short-term memory holds around 7 items; long-term memory is treated as unlimited.

Duration — how long a store holds information. Short-term memory lasts roughly 18–30 seconds without rehearsal; long-term memory potentially a lifetime.

Rehearsal — repeating information to maintain it in short-term memory or transfer it to long-term memory.

Episodic memory — memory for personal events from your own life, usually with a sense of when they happened.

Semantic memory — memory for facts, meanings and general knowledge, detached from when you learned them.

Procedural memory — memory for how to carry out a skill, often without conscious thought.

Schema — a mental framework of expectations built from previous experience.

Reconstructive memory — the view that recall rebuilds a memory rather than replaying it, so it can be altered.

Interference — forgetting caused by one memory disrupting another, usually because they are similar.

Retrieval failure — forgetting because the right cue is absent, even though the memory is still stored.

Retrograde amnesia — inability to recall events from before a brain injury.

Anterograde amnesia — inability to form new long-term memories after a brain injury.

Core concepts

The three processes of memory

Every memory question ultimately concerns encoding, storage or retrieval. Separating them matters because forgetting looks identical from the outside whatever the cause. A name you never caught was never encoded. A name you knew last year and cannot produce now may have decayed in storage, or may be intact but inaccessible — a retrieval problem. Deciding which process has failed is often the whole of a question.

The multi-store model

The multi-store model describes memory as three stores information passes through in sequence.

Sensory memory holds incoming sights and sounds very briefly. Almost all of it is lost; only what receives attention passes on.

Short-term memory holds about 7 items for 18–30 seconds. Information is lost by displacement, where new items push out old ones, or by decay if it is not rehearsed. Rehearsal both maintains information here and transfers it onward.

Long-term memory has unlimited capacity and duration. Retrieval brings information back into short-term memory to be used.

The model's strength is that it makes clear predictions which can be tested. Its weaknesses are now well established. It treats each store as a single unit, when long-term memory demonstrably has separate types. It also makes rehearsal the only route into long-term memory, yet a shocking or deeply meaningful event is remembered permanently without any rehearsal at all.

Types of long-term memory

Long-term memory is not one store but at least three.

Episodic memory covers events you experienced: a birthday, a first day at school. Semantic memory covers knowledge: that Paris is a capital, that water boils at 100 °C. Procedural memory covers skills: cycling, typing, tying a shoelace.

The strongest evidence for the distinction comes from amnesia. A patient unable to form new episodic memories may nonetheless learn a new physical skill through practice — and then deny ever having practised it. The skill was stored while the memory of acquiring it was not. A single undivided long-term store cannot produce that pattern.

The serial position curve

Ask someone to learn a list of words and recall them in any order, and the results form a U shape: items from the beginning and the end are recalled well, those in the middle poorly. Murdock's study established this pattern.

The multi-store model explains both halves. Early items had time to be rehearsed and transferred into long-term memory, producing the primacy effect. Late items are still sitting in short-term memory when recall begins, producing the recency effect. Middle items got neither advantage.

The decisive detail is what happens when the two effects are separated. Insert a distracting task — counting backwards for thirty seconds — between the list and recall, and the recency effect disappears while primacy survives untouched. The last items were displaced from short-term memory; the early ones were safe in long-term memory. Two effects behaving differently under the same manipulation is very hard to explain with one store.

Memory as an active process

Bartlett's alternative view is that memory reconstructs rather than records. In his 'War of the Ghosts' study, participants read an unfamiliar folk tale and retold it repeatedly over time. Their accounts grew shorter, and unfamiliar details shifted towards what the participants' own culture found expected — details that made no sense to them were dropped or replaced with something that did.

This is the theory of reconstructive memory. Recall is not retrieval of a fixed record; it is rebuilding, using fragments plus the schemas that tell you what such an event is normally like. Where the fragments are incomplete, the schema fills the gap, and the filled-in material feels exactly as real as the remembered part.

Schemas are usually helpful — they let you understand a restaurant or a classroom instantly. The cost is systematic error. A witness to a robbery in a shop may confidently report a till that was never there, because a shop schema supplied it.

Explanations of forgetting

Interference occurs when memories disrupt one another, and it is worst where they are similar. Learning a second list of words damages recall of the first; a new phone number makes the old one hard to retrieve.

Retrieval failure says the memory remains stored but the cue needed to reach it is missing. The evidence is the experience of sudden recall the moment a cue appears — impossible if the memory had gone. Context supports this: revising in the room where the exam will be held can improve recall, because the surroundings act as cues. This is context-dependent retrieval.

The two are not really rivals. Interference explains confusion between similar material; retrieval failure explains inaccessible but intact memories. Good answers say which situation each fits rather than choosing a winner.

Amnesia

Retrograde amnesia removes access to events from before the injury. Anterograde amnesia prevents new long-term memories forming, while memories laid down beforehand often remain. The two can occur together, and the pattern of what is lost against what is spared is what makes such cases informative about how memory is organised.

Worked examples

Example 1: Identifying the failed process

A student reads a phone number, is interrupted, and cannot recall it. Explain the forgetting. (3 marks)

The number was held in short-term memory, which has a duration of only around 18–30 seconds without rehearsal (1). The interruption prevented rehearsal, so the information was not maintained and was not transferred to long-term memory (1). This is a failure of storage in short-term memory rather than of encoding, because the number was initially taken in (1).

Note how the third mark comes from ruling out the alternative. Examiners reward that.

Example 2: Explaining the serial position curve

Explain how the multi-store model accounts for the recency effect. (2 marks)

The final items in a list are still being held in short-term memory at the moment recall begins (1), so they can be reported directly without needing to be retrieved from long-term memory (1).

A common half-answer says only "they were heard most recently". That describes the effect instead of explaining it, and earns nothing.

Example 3: Applying reconstructive memory

A witness reports seeing a knife that was not present. Use the theory of reconstructive memory to explain this. (4 marks)

Recall is an active process in which the memory is rebuilt rather than replayed (1). Where the witness's memory of the event is incomplete, the gaps are filled using a schema — a framework of expectations about what happens in a robbery (1). That schema includes a weapon, so the detail is supplied during reconstruction and becomes part of the memory (1). The witness is not lying: the knife now genuinely features in what they recall, which is why they report it confidently (1).

The final point is the one most students miss, and it is usually where the evaluation marks sit.

Example 4: Distinguishing types of long-term memory

A stroke patient can still play the piano but cannot remember learning any of the pieces. What does this show about long-term memory? (3 marks)

Procedural memory for the skill is intact while episodic memory for the events of learning is impaired (1). Because one type has been damaged and the other spared, they cannot be parts of a single store (1). This supports the view that long-term memory consists of separate episodic, semantic and procedural types (1).

Common mistakes and how to avoid them

Confusing capacity with duration. Capacity is how much, duration is how long. Questions ask about one or the other and the wrong figure earns nothing.

Describing the serial position curve instead of explaining it. Saying which items are recalled best is description. The marks are for why — rehearsal into long-term memory, and items still present in short-term memory.

Treating episodic and semantic memory as the same. Knowing a fact is semantic; remembering the occasion you learned it is episodic. Questions often hinge on exactly this split.

Saying a witness "lied" or "made it up". Reconstructive memory explains sincere error. Suggesting deliberate deception shows the theory has been misunderstood.

Claiming interference and decay are the same thing. Interference is disruption by other memories; decay is the fading of the trace. Retrieval failure is a third option again.

Forgetting to evaluate. Any question worth more than three marks on the multi-store model almost certainly wants its limitations as well as its content.

Using one amnesia case as proof. A single patient is rich evidence but unique. Always add that findings from one brain injury may not generalise.

Exam technique for Memory

Match the length of your answer to the marks. A 1-mark definition needs one accurate sentence, not a paragraph. A 9-mark question needs content, evidence and evaluation with a conclusion.

Name the study when you use it. Murdock for the serial position curve, Bartlett for reconstructive memory. Naming it and stating what it found is usually worth more than describing the procedure at length.

Use the command word. "Describe" wants content only. "Explain" wants the reason or mechanism. "Evaluate" wants strengths and weaknesses plus a judgement. An evaluation that lists only weaknesses cannot reach the top band.

Apply the theory to the scenario given. Application questions supply a specific person and situation. Refer to them by name and to what they actually did — a generic answer about memory in general loses the application marks.

For evaluation, lead with the strongest point. For the multi-store model, that is usually the evidence from amnesia and the serial position curve; for the weaknesses, the existence of separate long-term stores.

Watch for questions that seem to be about memory but are about research methods. A question on a memory study may ask about sampling, ecological validity or ethics. Answer the question that was asked.

Quick revision summary

• Three processes: encoding in, storage held, retrieval out — decide which one has failed
• Short-term memory: about 7 items, 18–30 seconds, lost by displacement or decay
• Long-term memory: unlimited capacity and duration, reached through rehearsal
• Long-term memory divides into episodic (events), semantic (facts) and procedural (skills)
• Serial position curve: primacy from rehearsal into long-term memory, recency from items still in short-term memory
• A filled delay destroys recency but not primacy — the strongest single argument for separate stores
• Reconstructive memory: recall rebuilds, and schemas fill the gaps with what is expected
• Bartlett's 'War of the Ghosts': accounts shortened and shifted towards the participants' own culture
• Interference is disruption by similar memories; retrieval failure is an intact memory with no cue
• Context-dependent retrieval: matching the surroundings at learning and recall improves performance
• Retrograde amnesia looks backwards, anterograde forwards
• Multi-store model criticisms: stores are not unitary, and rehearsal is not the only route in
• Eyewitness error is usually sincere, not dishonest — say so explicitly