Lesson 1 of 11 · 16 min
The multi-store model of memory
Describe Atkinson and Shiffrin's multi-store model, trace information from sensory memory to short-term and long-term memory, and use the serial position effect and case studies as evidence.
What you will be able to do
- Define memory as encoding, storage and retrieval.
- Describe the three stores of the multi-store model and the processes that move information between them: attention, rehearsal, encoding and retrieval.
- Explain the serial position effect with the multi-store model.
- Use evidence from case studies to support the idea of separate short-term and long-term stores, and identify the model's limitations.
Someone reads you a six-digit code to type into a website. You repeat it under your breath as you type, and a minute later it is gone. Your own birthday, on the other hand, you will never lose. Both are memories, but they behave so differently that psychologists in the 1960s concluded they must be held in different places.
Memory is the set of processes by which we encode information (change it into a form that can be stored), store it over time, and retrieve it when we need it. A model of memory is a simplified description of how those processes are organised, which researchers can test.
Definition
The multi-store model
The multi-store model (Atkinson & Shiffrin, 1968) describes memory as three separate stores, sensory memory, short-term memory (STM) and long-term memory (LTM), with information flowing between them in a fixed order. Each store has its own capacity, duration and way of encoding information.
The multi-store model (after Atkinson & Shiffrin, 1968). Information that is not attended to, or not rehearsed, is lost.
Follow one piece of information through the model.
- Sensory memory holds an exact, unprocessed copy of everything your senses receive, but only for a moment. Visual information is held in iconic memory, sounds in echoic memory. What you do not attend to fades (decay).
- Attention transfers the information you focus on into short-term memory, the store of what you are consciously thinking about. STM holds only a few items, and only for seconds.
- Repeating the information (maintenance rehearsal) keeps it in STM. Without rehearsal it decays, and a new item arriving in a full STM pushes an old one out (displacement).
- Rehearsal also encodes information into long-term memory: the more it is rehearsed, the more likely it is to be transferred. LTM is a large, lasting store.
- To use a long-term memory, you retrieve it back into STM.
Each store shows its capacity and duration. Let time pass to see unattended items decay from sensory memory.
According to the multi-store model, what moves information from sensory memory into short-term memory?
Key idea
The model's key claims
- Memory has three separate stores, each with a different capacity, duration and encoding.
- Information passes through them in order: sensory memory → STM → LTM.
- Rehearsal is the main way information is transferred to LTM: more rehearsal, better long-term memory.
- STM is a single (unitary) store.
Evidence: the serial position effect. Read someone a list of 15 words, one at a time, and ask them to write down as many as they can in any order (free recall). Plot the percentage recalled at each position in the list and you get a U-shaped curve: the serial position effect.
- The first few words are recalled well: the primacy effect. The multi-store model says they were rehearsed more often (the person had time to repeat them before the list got long), so they were transferred to LTM.
- The last few words are also recalled well: the recency effect. They are still in STM when recall begins, so they are simply read out.
- The middle words were not rehearsed enough to reach LTM and were displaced from STM by the words that followed.
The serial position effect in free recall. The values are illustrative, drawn to match the shape found by Murdock (1962) for immediate recall and Glanzer and Cunitz (1966) for recall after a 30-second filled delay.
The two ends of the curve can be changed separately, which is what you would expect if they come from two different stores.
- Make people count backwards for 30 seconds before recalling, and the recency effect disappears while primacy stays (Glanzer & Cunitz, 1966). The counting displaced the last words from STM; the first words were already in LTM.
- Present the words faster, so there is less time to rehearse, and the primacy effect shrinks while recency stays.
Recall each list straight away, then one after 30 seconds of counting backwards. Your results stay in this browser and are for learning only. Compare your two curves at the end of the list.
A participant hears a list of 15 words and must count backwards in threes for 30 seconds before recalling them. Compared with immediate recall, which part of the serial position curve changes most, and why?
Worked example
Score and explain a serial position curve
Twenty students heard a list of 12 unrelated words, one every 2 seconds, and then wrote down as many as they could in any order. The table shows how many students recalled the word at each position.
| Position | 1 | 2 | 3 | 4 | 5 | 6 | 7 | 8 | 9 | 10 | 11 | 12 |
|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Students recalling it | 17 | 14 | 11 | 8 | 7 | 6 | 7 | 6 | 8 | 12 | 16 | 19 |
(a) Calculate the mean percentage of students who recalled a word from the first three positions, the middle six and the last three. (b) Explain the pattern using the multi-store model. (c) Predict how the results would change if the students counted backwards for 30 seconds before recalling.
Step 1: First three positions
Mean count , and .
Step 2: Middle six positions
Mean count , and .
Step 3: Last three positions
Mean count , and .
Step 4: Explain with the multi-store model
The early words were rehearsed more often before the list grew, so they were transferred to LTM: the primacy effect. The last words were still in STM when recall began: the recency effect. The middle words were displaced from STM before they were rehearsed enough to reach LTM.
Step 5: Predict the effect of a delay
Counting backwards for 30 seconds would displace the last words from STM, so recall of positions 10–12 would fall to about the level of the middle words, removing the recency effect. Recall of positions 1–3 would stay about the same, because those words are in LTM.
This is what Glanzer and Cunitz (1966) found, and it is evidence that the two ends of the curve come from different stores.
Try it yourself first, then check the working one step at a time.
Answer
(a) First three 70%, middle six 35%, last three 78.3%. (b) Primacy: early words rehearsed into LTM; recency: last words still in STM; middle words displaced. (c) The recency effect would disappear and the primacy effect would remain.
Method
Average the counts for each group of positions, then convert to a percentage of the number of participants.
Explain primacy with rehearsal into LTM and recency with STM.
A filled delay removes recency only.
Watch out for
Dividing by 12 (the list length) instead of 20 (the number of students) when converting a count to a percentage.
Saying the delay removes primacy: the first words are already in LTM.
Evidence from brain injury. Henry Molaison (known in research as HM) had parts of both medial temporal lobes removed in 1953 to treat severe epilepsy, including about half of the hippocampal formation and most of the amygdala (Corkin et al., 1997). Afterwards his short-term memory was normal: he could repeat a string of digits and hold a conversation. But he could not form new explicit long-term memories, and forgot people he met within minutes. One store was damaged and the other was not, which supports the model's claim that STM and LTM are separate.
A second case complicated the picture. KF (Shallice & Warrington, 1970) had a brain injury that left his short-term memory for heard words very poor, yet he could still form new long-term memories. If information must pass through STM to reach LTM, that should have been impossible. His STM for visual material was also better than for heard words, which suggests STM is not one store.
Note
Strengths and limitations
Strengths. The model was the first to separate memory into stores with different capacities and durations, and many studies support that separation: the serial position effect, Peterson and Peterson's (1959) duration study, and case studies such as HM. It is simple and makes testable predictions.
Limitations.
- STM is not a single store: KF could hold visual information better than heard words. This led to the working memory model.
- Rehearsal is not the only way into LTM, and repetition alone is a poor one. We remember many things we never rehearsed (a shock, a joke), and how information is processed matters more than how often it is repeated. This led to the levels of processing model.
- Information flows one way only, from STM to LTM, yet we use what is in LTM to make sense of what enters STM (recognising letters as a word, chunking).
- LTM is treated as one store, but it holds very different kinds of memory (facts, events, skills).
Worked example
Trace a phone number through the multi-store model
Mei hears a radio advertisement read out a phone number while she is driving. She repeats it to herself until she can pull over, then types it into her phone. A week later she cannot remember any of it. Her own mobile number, which she has used for years, she can say instantly.
Use the multi-store model to explain what happened to each number.
Step 1: Sensory memory
The spoken number entered echoic memory (sensory memory for sound). Because Mei attended to it, it was transferred to short-term memory; the rest of the advertisement she ignored decayed within seconds.
Step 2: Short-term memory
In STM the number (ten digits, but held as a few chunks) was kept available by maintenance rehearsal: repeating it to herself. Without the rehearsal it would have been lost within about 18–30 seconds.
STM has a limited duration. Rehearsal resets it, so the number survived until she could type it.
Step 3: Why it was lost
Once she had typed it she stopped rehearsing, so the number decayed or was displaced from STM. It had been rehearsed only briefly, so it was not encoded into long-term memory, and a week later there is nothing to retrieve.
Step 4: Her own number
Her own number has been rehearsed and used many times over years, so it was encoded into long-term memory, where it can last a lifetime. When she needs it she retrieves it from LTM into STM.
Try it yourself first, then check the working one step at a time.
Answer
The advertised number went from echoic memory to STM by attention and was held there by maintenance rehearsal, but was never encoded into LTM, so it was lost once rehearsal stopped. Her own number was rehearsed into LTM long ago and is retrieved into STM when needed.
Method
Start in sensory memory and name the sense (iconic or echoic).
Name the process that moves information at each step (attention, rehearsal, encoding, retrieval).
Explain loss by decay or displacement, with the store's duration.
Watch out for
Skipping sensory memory: everything enters through it.
Saying the number was "forgotten from long-term memory" when it never reached LTM.
Which finding is the strongest evidence against the multi-store model's claim that short-term memory is a single store?
Summary
Memory is encoding, storage and retrieval.
The multi-store model (Atkinson & Shiffrin, 1968) has three stores: sensory memory, short-term memory and long-term memory.
Attention moves information from sensory memory to STM; rehearsal keeps it in STM and encodes it into LTM; retrieval brings it back to STM.
Information is lost from sensory memory by decay and from STM by decay and displacement.
The serial position effect supports separate stores: primacy comes from LTM, recency from STM, and a filled delay removes only recency.
HM supports separate stores; KF suggests STM is not a single store, and the role of rehearsal is overstated.
Finished the lesson?
Try the 3 quick checks above and the lesson completes itself, or mark it complete now.