- Sensory register
- The first store in the multi-store model: holds raw, modality-specific sensory information for a very brief period (iconic memory: ~0.5 seconds for visual; echoic memory: ~2–4 seconds for auditory). Very large capacity but most information decays unless attended to. Sperling (1960) estimated iconic memory capacity at 9–12 items using the partial report technique.
- Short-term memory (STM)
- The second store: holds information that has been attended to for approximately 15–30 seconds without rehearsal (Peterson & Peterson, 1959). Capacity is approximately 7 ± 2 chunks (Miller, 1956). Information in STM is primarily encoded acoustically (Conrad, 1964) — even when input is visual, items tend to be recoded into sound-based representations. Maintenance rehearsal can keep information active; sufficient rehearsal transfers it to LTM.
- Long-term memory (LTM)
- The third store: potentially unlimited in capacity and duration. Information in LTM is primarily encoded semantically — meaning-based rather than sound-based (Baddeley, 1966). LTM stores both explicit (declarative) memories — episodic and semantic — and implicit (non-declarative) memories such as procedural skills and conditioned responses. The hippocampus is critical for transferring information from STM to LTM.
- Capacity
- The amount of information a memory store can hold. Sensory register: very large (~9–12 items). STM: approximately 7 ± 2 chunks (Miller, 1956); revised to ~4 chunks by Cowan (2001). LTM: effectively unlimited. Capacity in STM can be extended by chunking — grouping individual items into meaningful units.
- Duration
- How long information lasts in each store without rehearsal. Sensory register: ~0.5 seconds (iconic) to ~4 seconds (echoic). STM: approximately 15–30 seconds without rehearsal (Peterson & Peterson, 1959). LTM: potentially permanent, though retrieval failure can make memories inaccessible.
- Coding
- The format in which information is represented in each store. Sensory register: modality-specific (visual, auditory, etc.). STM: predominantly acoustic — sound-based — even for visually presented material (Conrad, 1964). LTM: predominantly semantic — meaning-based (Baddeley, 1966). Evidence: acoustic similarity causes confusion in STM recall; semantic similarity causes confusion in LTM recall.
- Maintenance rehearsal
- Repetitive, shallow rehearsal (the "inner voice" repeating items) that keeps information active in STM and can transfer it to LTM through sufficient repetition. Distinguished from elaborative rehearsal (processing meaning), which produces stronger long-term retention. Levels of processing research (Craik & Lockhart, 1972) showed that maintenance rehearsal alone produces weak long-term memories — challenging the model's assumption that rehearsal equals transfer.
- Chunking
- Organising individual items into meaningful groups or units, effectively extending STM capacity beyond the raw 7 ± 2 item limit. For example, the sequence 1-9-4-5-1-9-6-8 (8 items) can be chunked as "1945" and "1968" (2 chunks). Expert chess players can recall entire board positions by chunking pieces into familiar tactical patterns rather than storing each piece individually.