The multi-store model (MSM), proposed by Atkinson & Shiffrin (1968), describes memory as three separate, fixed stores through which information flows in one direction. It is a foundational AQA Memory topic and a favourite for extended-answer questions because it comes with rich evaluation.
The sensory register receives everything from the senses. It has a very large capacity but a duration of around half a second, and coding is modality-specific (e.g. iconic for vision, echoic for sound). Information only passes on if we pay attention to it.
Short-term memory (STM) has a limited capacity of about 7±2 items (Miller), a duration of roughly 18–30 seconds without rehearsal (Peterson & Peterson), and codes mainly acoustically (Baddeley). Maintenance rehearsal keeps material in STM and, if prolonged, transfers it to long-term memory.
Long-term memory (LTM) has a potentially unlimited capacity, can last up to a lifetime (Bahrick et al. found people recognised classmates' faces decades on), and codes mainly semantically (by meaning).
There is solid supporting evidence for separate stores. Baddeley showed STM tends to code acoustically while LTM codes semantically, implying they are distinct. Glanzer & Cunitz's serial position effect — better recall of the first items (primacy, from LTM) and last items (recency, from STM) — also supports two separate stores.
The biggest problem is that STM is not a single store. The case study of patient KF (Shallice & Warrington) showed someone with a poor digit span (impaired verbal STM) but intact ability to process visual information, which a unitary STM cannot explain. The Working Memory Model (Baddeley & Hitch) was developed precisely because STM has multiple components, not one.
The model also oversimplifies rehearsal. Craik & Lockhart argued it is the depth of processing — how meaningfully you engage with material — that predicts long-term retention, not the sheer amount of rehearsal as the MSM claims.
Finally, much of the supporting research used artificial stimuli such as lists of unrelated words, which may not reflect how memory works for meaningful, everyday information.