The Working Memory Model

The multi-store model treats short-term memory (STM) as one unitary store that simply holds information before it passes to long-term memory. Baddeley and Hitch (1974) argued this was too simple. If STM were a single store, doing two tasks at once should always overload it equally, yet people can often do a visual task and a verbal task at the same time without much trouble.

Their solution was the working memory model (WMM), an explanation of short-term memory only. Crucially, it describes STM not as a passive store but as an active processing system made of several separate components, each with its own job, its own type of coding, and its own limited capacity.

The working memory model explains short-term memory as an active system of multiple components, not a single store. It says nothing about long-term memory as a whole.

Because the components are separate, two tasks only interfere with each other when they compete for the same component. That single idea is what the rest of the model, and most of its supporting evidence, is built on.

Working memory has a controlling component, the central executive, which supervises three subordinate "slave systems": the phonological loop, the visuo-spatial sketchpad, and the episodic buffer. The episodic buffer, and the system as a whole, links to long-term memory (LTM) so that stored knowledge can be pulled into current processing.

The solid arrows show the central executive directing each slave system. The dashed arrows show information being integrated by the episodic buffer, which draws on the other components and on long-term memory. Keep this shape in your head: one controller, three subsystems, one link out to LTM.

The specification asks specifically for the coding (the type of information a component handles) and the capacity (how much it can hold) of each part. Learn this table; the differences between the components are exactly what earns marks.

Component	Coding	Capacity	Key feature
Central executive	Modality-free (any type)	Very limited; no storage of its own	Directs attention; allocates resources to slave systems
Phonological loop	Auditory / verbal	About 2 seconds of speech	Preserves the order of words
Visuo-spatial sketchpad	Visual and spatial	Around 3–4 objects (Baddeley)	Stores what things look like and where they are
Episodic buffer	Integrated (all types)	About 4 chunks	Combines information into one time-sequenced record

"Modality-free" means the central executive is not tied to one kind of information. It can process visual, verbal or spatial input, but it does not store any of it — a point examiners reward and students often miss.

Notice that only the central executive lacks storage. The three slave systems each store a specific kind of information, which is why interference depends on which system a task uses.

The central executive is the most important component and the hardest to pin down. It is the attentional control system: it decides what working memory attends to and allocates processing resources to the slave systems. It is modality-free and has a very limited capacity, and it has no storage capacity of its own — it directs traffic rather than holding cargo.

The phonological loop deals with auditory and verbal information and preserves the order in which words arrive. Baddeley subdivided it into two parts:

• Phonological store — the "inner ear": holds words you have recently heard.
• Articulatory process — the "inner voice": allows maintenance rehearsal by silently repeating words to keep them in the loop.

Its capacity is limited to roughly 2 seconds of spoken material. This is shown by the word-length effect: people recall short words better than long words, because only about two seconds' worth of speech fits in the loop, and long words take longer to rehearse.

Phonological loop = inner ear (store) + inner voice (rehearsal). It is verbal and time-limited, holding about two seconds of sound.

The visuo-spatial sketchpad, or "inner eye", stores visual and spatial information — what things look like and where they are in relation to each other. Its capacity is limited to around 3–4 objects (Baddeley). Logie subdivided it into:

• Visual cache — stores visual data such as form and colour.
• Inner scribe — records the arrangement of objects in space and passes information to the central executive.

The episodic buffer was added by Baddeley in 2000 because the original 1974 model had no general store and no way to combine different types of information. The buffer is a temporary store that integrates information from the phonological loop, the visuo-spatial sketchpad and long-term memory into a single, time-sequenced record (an "episode"). Its capacity is limited to about 4 chunks.

The episodic buffer is the model's "glue". It binds visual, verbal and spatial information into one coherent episode and provides the bridge between working memory and long-term memory.

Because it was a later addition, remember that the original 1974 model had only three components; the buffer completes the version you should draw and describe today.

A working memory answer is only worth top marks if you can evaluate the model, not just describe it. Develop three or four points fully, each with a point, evidence and an explanation of what it shows.

Dual-task studies support separate stores. Baddeley and Hitch found participants struggled to perform two visual tasks at once (for example, tracking a moving light while describing the angles of a letter) because both compete for the visuo-spatial sketchpad. But they could do one visual and one verbal task together with little loss of performance, because these use different components. This dissociation is strong evidence that the phonological loop and the sketchpad are genuinely separate systems.

Clinical evidence supports a separate phonological loop. Patient KF (Shallice & Warrington, 1970) had brain damage that left his auditory short-term memory very poor while his visual short-term memory stayed much better. If STM were a single store, damage should affect all input equally. The selective impairment fits a model with a separate phonological loop and visuo-spatial sketchpad.

The central executive is under-specified. Critics argue the central executive is too vague to be a proper explanation. It may itself consist of several separate subcomponents rather than one system, and it remains the least understood and least tested part of the model. A theory whose most important component cannot be clearly measured is weaker than it first appears.

The model explains only short-term memory. The WMM says nothing about long-term memory or memory as a whole, so it is a partial account. Much of its support also comes from brain-damaged case studies such as KF, whose injuries are unique and may have affected other cognitive processes, making it hard to generalise the findings to healthy people.

1. Saying the central executive stores information

The central executive has no storage capacity of its own. It directs attention and allocates resources to the slave systems; the storage happens in the phonological loop, the visuo-spatial sketchpad and the episodic buffer. Writing that it "stores" information loses the mark.

2. Confusing the phonological store with the articulatory process

The phonological store (inner ear) holds words you have heard; the articulatory process (inner voice) rehearses them by silent repetition. One receives sound, the other maintains it — do not swap the two.

3. Treating the WMM as a model of all memory

Working memory explains short-term memory only. It is not a rival to the multi-store model for long-term memory, and it does not describe how information is stored permanently. Framing it as a complete model of memory misreads its scope.

4. Forgetting the episodic buffer was a later addition

The original 1974 model had three components. The episodic buffer was added by Baddeley in 2000 to explain how information is integrated and linked to long-term memory. Presenting the buffer as part of the original model is a common date error.

5. Describing dual-task studies without saying what they show

Do not just state that people can do two tasks at once. Explain the logic: performance stays high only when tasks use different components, and drops when they compete for the same one, which is why the results support separate stores.