AQA A-Level Psychology: The Multi-Store and Working Memory Models — Revision Notes

Condensed for the final weeks. For the full explanation, use the Multi-Store and Working Memory Models study guide.

Two different questions

Multi-store model (Atkinson and Shiffrin) — a sequence of separate stores information passes through. Working memory model (Baddeley and Hitch) — a detailed account of what happens inside STM specifically. The second refines the first’s STM box; they are not competing explanations of the same thing.

Multi-store model — comparison table

Store	Coding	Capacity	Duration
Sensory register	Modality-specific (iconic, echoic)	Very large	<0.5 seconds
Short-term memory	Mainly acoustic	7±2 items (Miller)	18–30 seconds unrehearsed
Long-term memory	Mainly semantic	Potentially unlimited	Up to a lifetime

Background, not specification content: types of LTM (episodic, semantic, procedural) were removed from the Memory topic in specification version 1.4 (first teaching September 2025). They are not required for exams from June 2026 (AS) or June 2027 (A-level).

Working memory model — four components

• Central executive — directs attention, coordinates the others; no storage of its own.
• Phonological loop — auditory/verbal information (split into phonological store + articulatory process).
• Visuo-spatial sketchpad — visual and spatial information.
• Episodic buffer — integrates the other components, links to LTM. Easy to forget since it was added later than the other three.

Worked example

A student struggles to hold a phone number in mind while reading a map.

Phone number:       phonological loop (verbal/acoustic, kept by
                    sub-vocal rehearsal)
Map reading:        visuo-spatial sketchpad for the layout -- but
                    reading street names to yourself also uses the
                    phonological loop
Why the struggle:   both tasks compete for the SAME limited-capacity
                    phonological loop, so rehearsal of the number is
                    displaced and the number is lost
Central executive:  limited capacity, no storage; loaded by working
                    out the route, so it cannot also keep directing
                    rehearsal of the number
Contrast:           a purely visual task (tracing a route by eye)
                    would use a separate component and interfere
                    far less

Identify which component handles which information type, then explain the resulting behaviour — that is the core applied skill this content is assessed against.

Why the models needed refining

The multi-store model’s account of STM as a single, unitary store ran into a specific problem: if STM really were one undifferentiated store, then two demanding tasks performed at the same time should always interfere with each other, since they would be competing for the same limited resource. In practice, people can often do two things at once relatively well — for example, holding a short phone number in mind while walking somewhere unfamiliar — provided the two tasks draw on different types of information. This everyday observation is exactly what motivated Baddeley and Hitch to propose that STM itself is not a single box but several interacting components, each handling a different type of information and therefore each with its own limited capacity.

MSM and WMM side by side

	Multi-store model	Working memory model
What it describes	The overall sequence of memory stores (sensory → STM → LTM)	The internal structure of STM specifically
Number of “stores” involved	Three (sensory register, STM, LTM)	Four components within STM (central executive, phonological loop, visuo-spatial sketchpad, episodic buffer)
Key evidence type	Coding, capacity and duration studies for each store	Dual-task studies showing selective interference

Treat the WMM as “zooming into” the STM box of the MSM, not as a rival account of memory as a whole — this framing answers most “compare the two models” exam questions directly.

Exam traps

• Treating MSM and WMM as rival explanations of the same phenomenon, instead of WMM refining MSM’s STM box.
• Mixing up coding types (acoustic for STM, semantic for LTM) or misstating STM’s capacity/duration figures.
• Naming a WMM component without describing its specific function.
• Forgetting the episodic buffer entirely.

Evaluating the two models

For exam evaluation marks, the multi-store model’s strength is its simplicity and the amount of supporting evidence for separate short-term and long-term stores (for example, case studies of amnesia patients whose STM and LTM are affected differently). Its weakness is oversimplifying STM as a single unitary store, which the working memory model directly addresses. The working memory model’s strength is its ability to explain dual-task performance in a way the multi-store model cannot; its weakness is that the central executive’s exact role remains vague and difficult to test directly, since it has no proposed structure of its own beyond “coordinating attention.” A strong evaluation paragraph states a specific strength or weakness and ties it to a specific feature of the model, rather than offering a generic “this model is useful/not useful” statement.

Keep both evaluation points and structural facts (coding, capacity, duration figures, component names) revision-ready separately, since exam questions may ask purely for description, purely for evaluation, or for both combined in a single extended-writing question worth significant marks.

Self-test

1. What is STM’s typical capacity and who identified it?
2. Name the three features on which the multi-store model’s stores differ.
3. Which WMM component has no storage capacity of its own?
4. Why can a person often talk and navigate a map at the same time, per the WMM?
5. Why is the working memory model not simply a rival to the multi-store model?

Answers: 1. 7±2 items, identified by Miller. 2. Coding, capacity and duration. 3. The central executive. 4. Because the two tasks draw on different WMM components (phonological loop for talking, visuo-spatial sketchpad for the map), so they interfere with each other only a little. 5. Because it is a more detailed account of what happens specifically within the multi-store model’s STM component, not a separate account of memory as a whole.