You walk into a room to fetch a book, notice the cat on your bed, move the cat, and completely forget why you came in. That is not carelessness. It is a memory store with a limited capacity doing exactly what the model says it will do.
📚 What you need to know
The multi-store model (MSM) was devised by Atkinson & Shiffrin (1968) to conceptualise how memories are encoded, stored and lost.
The sensory register (SR) holds input from the five senses for half a second at most. Sounds are stored as echoic memory, images as iconic.
Anything given attention passes to short-term memory (STM): about 7 items, plus or minus 2, for up to 30 seconds, encoded acoustically.
Without rehearsal, material is lost from STM through displacement.
Long-term memory (LTM) is permanent, with capacity and duration researchers agree may be limitless, encoded semantically.
The study you need is Glanzer & Cunitz (1966).
The three stores
Everything you have ever sensed passed through the left box. Almost none of it made it to the right one.
Getting things into long-term memory
Maintenance rehearsal is shallow. You repeat something verbally just long enough to use it — telling yourself ‘don’t forget the milk’ on the way to the shop.
Elaborate rehearsal is deep. You learn the information in several ways: repeating it, reframing it, writing it down, reinterpreting it.
Chunking improves STM by grouping items. A long phone number becomes three or four chunks instead of eleven separate digits.
LTM is harder to research than STM, which can be studied in lab experiments. Information may decay in LTM if it is not revisited, or if the cues needed to retrieve it are missing.
Notice the encoding difference: STM works acoustically, by sound, and LTM works semantically, by meaning. That single contrast explains why cramming sounds fail and understanding sticks.
The serial position effect
Serial position just means where an item sits in a list. Give people a list of words and they tend to remember the first few and the last few, while the middle disappears. That pattern is the serial position effect (SPE).
The primacy effect is better recall of items from the beginning. Those items had time to be rehearsed, so they reached LTM.
The recency effect is better recall of items from the end. Those items are the most recent, so they are still sitting in STM.
Items in the middle were neither rehearsed nor recent — they were displaced by later items.
This matters because it gives separate evidence for two different stores, which is exactly what the MSM claims.
Key study: Glanzer & Cunitz (1966)
Part of the study
What happened
Aim
To investigate the serial position effect as evidence for two separate memory stores, supporting the MSM.
Participants
46 males, all enlisted in the US army.
Procedure
Each participant was shown 15 lists of 15 words individually. Three conditions: immediate recall; recall after a 10-second interference task; and recall after a 30-second interference task. The interference used the Brown-Peterson technique of counting backwards aloud in threes from a given number.
Results
Immediate recall showed the expected SPE, with both primacy and recency effects. After a 10-second delay, primacy was similar but recency was much reduced. After 30 seconds, primacy remained high but the recency effect had disappeared.
Conclusion
Preventing rehearsal removes the recency effect, because items from the end of the list cannot be held in STM long enough to transfer to LTM. There appear to be two separate storage facilities.
Why the interference task is the point. Counting backwards in threes is not hard, but it occupies STM completely. The words at the end of the list get pushed out, while the words at the start — already in LTM — are untouched. One task, two different outcomes, two different stores.
Evaluating the model
What is strong about it
It was the first model to attempt to conceptualise memory, and it paved the way for everything that followed.
Separate stores are supported by lab experiments and by case studies of brain-damaged patients such as HM.
What is weaker about it
As a linear model it lacks flexibility and dynamism. It presents STM and LTM as static stores without explaining how they operate.
Later research identified more than one type of LTM: episodic memory for your own history, procedural memory for actions such as riding a bike, and semantic memory for facts about the world.
It cannot explain how some memories bypass rehearsal entirely and go straight into LTM.
Link to concepts
Causality: MSM research is largely lab-based with a clearly operationalised IV and a quantitatively measured DV, so it is objective and suited to hypothesis testing. That fits the features of science. But humans are not easily quantified, and lab research loses a lot in translation — it cannot convey real experience.
Perspective: the MSM is an example of machine reductionism. Comparing human memory to a computer is arguably artificial and forced. Human memory is unpredictable, multi-faceted and shaped by emotion, as flashbulb memories show. A computer does not respond that way, so the model has limited application to real human memory.
EXAM PRACTICE
Evaluate one model of memory. [22]
Step 1: outline the model
MSM (Atkinson & Shiffrin, 1968). SR to STM via attention, STM to LTM via rehearsal. Capacities and durations for each store.
Step 2: the evidence
Glanzer & Cunitz (1966). 46 soldiers, 15 lists of 15 words, three conditions.
Step 3: the finding that matters
After 30 seconds of interference the recency effect disappeared while primacy held up. Two stores, behaving differently.
Step 4: strengths
First model of its kind, supported by lab work and by case studies such as HM.
Step 5: limitations
Linear and static, ignores types of LTM, machine reductionism, and lab tasks lack real-world texture.
Model + study + key finding + strengths + limitationsExplain why the recency effect vanished. That is the whole argument for two stores.
💡 Exam tip
Learn the numbers: half a second, 7 plus or minus 2, 30 seconds. They are quick marks.
Say acoustic for STM and semantic for LTM. Precision counts.
Link primacy to LTM and recency to STM explicitly. Do not leave the examiner to infer it.
Use HM as a second source of support — two kinds of evidence is stronger than one.
Episodic, procedural and semantic LTM is the cleanest limitation available.
⚠ Common mix-up
Swapping primacy and recency. Primacy is the beginning; recency is the end.
Saying the 30-second delay wiped out all recall. Primacy stayed high; only recency went.
Confusing maintenance and elaborate rehearsal. One is repetition; the other is deeper processing.
Saying LTM has a known capacity. Researchers think it may be limitless, which is not the same as measured.
Calling the interference task difficult. It was simple, but it occupied STM, which is the point.
Up next: Working Memory and Cognitive Load — the model built to fix what the MSM left as a single grey box.
Want this explained one-to-one?
Book a free session with an experienced IB Psychology tutor and get your trickiest topics made simple.
