Atkinson and Shiffrin's multi-store model (1968) was the first complete model of memory and remains the GCSE benchmark. It depicts memory as three connected stores arranged in a sequence, with information flowing from one to the next provided it is attended to and rehearsed.
The model
- Sensory register receives all sensory input. Most decays in <1 second. Selective attention transfers a tiny fraction to STM.
- Short-term memory (STM) holds 7±2 items for ~18–30 seconds. Maintenance rehearsal keeps material circulating in STM and, with enough repetition, transfers it to LTM. Without rehearsal, items decay or are displaced by new input.
- Long-term memory (LTM) is a permanent store of unlimited capacity. Material is retrieved by transferring it back to STM.
The model treats each store as separate — different capacity, duration and code (acoustic for STM, semantic for LTM). It is sometimes called a structural model because it focuses on the architecture of memory.
Evidence for the model
- Murdock's serial position curve (1962): people remember the first (primacy) and last (recency) items on a list better than the middle. Primacy reflects rehearsal into LTM; recency reflects items still in STM. Two stores, two effects.
- Glanzer & Cunitz (1966): a counting-back task between presentation and recall destroyed the recency effect (STM lost) but left primacy intact (already in LTM). Strong support for separate stores.
- Patient H.M.: hippocampal damage left STM intact but blocked transfer to LTM.
- Patient K.F. (Shallice & Warrington, 1970): impaired STM (especially auditory) but intact LTM — opposite dissociation.
Together, these provide a "double dissociation" — STM and LTM can be selectively damaged, supporting the claim they are separate stores.
Limitations
- Oversimplifies STM as a single store. K.F. could remember visual but not auditory material — suggesting STM has multiple components, leading to Baddeley & Hitch's working memory model.
- Oversimplifies rehearsal. Craik & Lockhart (1972) showed that type of rehearsal matters: deep, semantic ("elaborative") rehearsal transfers material to LTM far better than shallow repetition.
- Treats LTM as one store — but episodic, semantic and procedural LTM seem distinct.
- Mostly lab-based evidence with low ecological validity.
Quick exam tips
- Always name the model (Atkinson & Shiffrin, 1968).
- Always state encoding, capacity and duration of each store.
- Always include at least one piece of evidence (Murdock, Glanzer & Cunitz, or H.M.) and one limitation.
AI-generated · claude-opus-4-7 · v3-deep-psychology