What the Board expects you to know
Models of memory
The multi-store model of memory: sensory register, short-term memory and long-term memory. Features of each store: coding, capacity and duration. Types of long-term memory: episodic, semantic, procedural. The working memory model: central executive, phonological loop, visuo-spatial sketchpad and episodic buffer. Features of the model: coding and capacity. |
Forgetting and EWT
Explanations for forgetting: proactive and retroactive interference and retrieval failure due to absence of cues. Factors affecting the accuracy of eyewitness testimony: misleading information, including leading questions and post-event discussion; anxiety. Improving the accuracy of eyewitness testimony, including the use of the cognitive interview. |
Click on the links above to navigate to relevant pages.
Short term memory (STM)
Capacity
Capacity refers to the amount of information that can be held at any one time in memory.
Capacity of STM is limited. Experiments that investigate capacity traditionally use the serial digit span method in which numbers/letters etc have to recalled in the correct order. The most famous study was Miller’s (1956) ‘magic number seven plus or minus two.’ However, procedures for this study are notoriously difficult to find, so instead Jacobs’ (1887) study might be better to discuss. The two are very similar.
Jacobs (1887)
Participants were read lists of either words or numbers that they had to recall immediately after presentation. Jacobs gradually increased the length of these digits etc until the participant could only accurately recall the information, in the correct order, on 50% of occasions. Recall has to be in the correct order (serial recall).
X N J P T C B D L Y Q H
Findings
Jacobs found a difference between capacity for numbers and for letters. On average participants could recall 9 numbers but only 7 letters. He also noticed that recall seemed to increase with age. Eight year olds being able to recall an average of 7 digits whereas by the age of 19 recall had increased to 9 digits.
(NB later studies have suggested that capacity does eventually start to decrease in much older participants).
Conclusion
STM has a capacity of between 5 and 9 (7 +/-2) items of information and as age increases we appear to develop better strategies of recall.
Memory is a complex and varied phenomenon. Ideas about what constitutes memory and how it works can be traced back to ancient times. Plato constructed a simple analogy thousands of years ago and in some respects his ideas have remained little changed into the modern era.
Plato likened human memory to an aviary with memories (birds) flying around inside. A new bird can be captured and added to the aviary (placing a new memory into storage), and at a later date the bird can be captured in a net and removed (retrieval of a memory). Inability to capture a bird or its escape from the cage are useful analogies for the two basic processes of forgetting.
Some modern theories of memory still use this principle of storage and retrieval, however it is becoming more popular now to see memory as a process rather than simply a storage system. Research in recent years has shown that far from being a perfect recording of an event our memories do change over time and can be influenced by others and by later events.
Plato likened human memory to an aviary with memories (birds) flying around inside. A new bird can be captured and added to the aviary (placing a new memory into storage), and at a later date the bird can be captured in a net and removed (retrieval of a memory). Inability to capture a bird or its escape from the cage are useful analogies for the two basic processes of forgetting.
Some modern theories of memory still use this principle of storage and retrieval, however it is becoming more popular now to see memory as a process rather than simply a storage system. Research in recent years has shown that far from being a perfect recording of an event our memories do change over time and can be influenced by others and by later events.
Evaluation
Later studies (Miller 1956) have supported Jacobs’ findings and conclusions. Miller published his findings in an article entitled “The magic number seven, plus or minus two.” Miller and others have also discovered that chunking can increase capacity (BBC or 01858 becomes one chunk of information rather than 3 or 5 distinct chunks). However, there is no consistency across different stimulus material. It seems that with numbers we can indeed typically recall seven. However, this drops to six for letters and is nearer five for words. Simon (1974) found that length of the chunk is important. Not surprisingly we can recall fewer larger chunks. Simon however only used one participant… himself, so hardly a representative cross-section! George A Miller |

Factors affecting duration of STM
Obviously rehearsal will help with duration of STM. Try to remember the registration of that speeding car and you will repeat it to yourself over and over.
Amount of information: Murdock 1961 used a version of the Brown-Peterson technique to show that number of chunks affects duration. Participants were given either three letters that spelt a familiar word such as c, a, t or three unrelated three letter words such as sun, pat, lid. The latter deteriorated at the same rate as predicted by B-P so recall after 18 seconds was minimal. However, recall of the three letters was very stubborn to erase and after 18 seconds recall was still at over 90%.
Duration
Refers to the amount of time that a memory can be stored. In STM this is also limited. Traditionally duration in STM is measured by the Brown-Peterson technique. The technique was devised independently in 1959 by Brown and by Peterson and Peterson, hence the name.
Procedure
Participants are presented with trigrams (three consonants please Carol), e.g. ZTM. Vowels are avoided so words can’t be constructed. Participants are then asked to count backwards in 3s from various numbers. This prevents the participants thinking about the trigram letters so avoids the possibility of rehearsal. After a certain period of time, e.g. 3 seconds, 9 seconds etc. a tone sounds and the participant is asked to recall the trigram.
Findings:
When plotted it is clear that recall falls away very quickly over the first 20 seconds.
Conclusion: duration of STM is less than 18 seconds
Evaluation
This is one of the most repeated studies into human memory and other studies tend to replicate the original findings.
As with span measures, this technique is very artificial so again is lacking in ecological validity. Apart from in a psychology lesson, has anyone ever asked you to recall random groups of consonants?
Some studies have also discovered signs of interference. When being tested on later trigrams, participants have erroneously recalled earlier ones. This suggests that recall can be much longer than 18 seconds. As you probably discovered yourselves, it is also very easy to cheat when being tested.
Factors affecting duration of STM
Obviously rehearsal will help with duration of STM. Try to remember the registration of that speeding car and you will repeat it to yourself over and over.
Amount of information: Murdock 1961 used a version of the Brown-Peterson technique to show that number of chunks affects duration. Participants were given either three letters that spelt a familiar word such as c, a, t or three unrelated three letter words such as sun, pat, lid. The latter deteriorated at the same rate as predicted by B-P so recall after 18 seconds was minimal. However, recall of the three letters was very stubborn to erase and after 18 seconds recall was still at over 90%.
Encoding
Encoding refers to the format or code in which a memory is stored as a memory trace. Encoding can be by sound (acoustic) or by meaning (semantic) or by other means too.
The key study here can be used for STM and for LTM since it investigated both.
Baddeley (1966)
Participants were given four sets of words to recall in order. For the STM task they had to recall them immediately following presentation and for the LTM task they had to be recalled following a longer time interval. Set 1 were words that all sounded similar, for example: cat, mat, cap, map…
Set 2 were words that sounded differently for example: dog, bin, cup, pen….
Set 3 were words of similar meaning for example: big, large, huge, vast…
Set 4 were words of different meaning for example: huge, good, light, blue….
The researchers then recorded how many mistakes were made in recalling the sets of words.
STM
Findings
In the STM procedure participants made significantly more mistakes on words that sounded alike so for example would confuse cat and cap etc. Similarly with letters, S and X would be confused as would M and N and P and B etc.
Conclusion
It was concluded that in STM information is encoded by its sound (acoustically) so when we recall information from STM similar sounding words get confused.
LTM
Findings
In the LTM procedure participants were far more likely to confuse words of similar meaning replacing huge with vast or night and dark etc.
Conclusion
It was concluded that in LTM information is encoded by its meaning (semantically).
Evaluation
The usual lacks ecological validity can be used, but make sure you explain why!
Later research suggested that the situation is more complex than this. For example there appears to be some visual encoding in STM. We take longer to decide that Aa are different than we do to decide that AA are the same even though they both sound the same.
In LTM we appear to also encode words visually and acoustically as well as semantically. For example the fact that STM knows how ‘huge’ and ‘vast’ etc sound, must be stored in our LTM!
Long Term Memory (LTM)
Capacity is vast. As far as we know capacity has never been reached, but don’t worry the term is still young! However, it’s unlikely that you’ve ever heard anyone complain that they need to delete a few memories before they can store anything new! As far as I’m aware nobody has tried to estimate capacity, but Solso (1991) compares it favourably with the largest computers.
A stunning factoid for you to contemplate. In 1973 Petr Anokhin of Moscow University wrote:
“We can show that each of the 10 billion neurons in the human brain has the possibility of connections to 1 with 28 noughts after it; that’s 10,000,000,000,000,000,000,000,000,000 connections! It means that the total combinations in the brain, if written out, would be 1 followed by 10.5 kilometres of noughts!
One way psychologists have studied memory is to experiment on themselves. The beautifully named Marigold Linton kept a detailed diary recording daily events and facts about flowers etc on cue cards over many years. Each day was represented by a single word. When given the cue word she was able to recall with 70% accuracy all the events and information recorded for that day even 7 years later.
Duration: Bahrick et al Year Book study (1975)
Procedure
374 participants aged between 17 and 74 were tested on their memory of school friends. A number of different tests were carried out including a free recall of all the names of classmates they could remember, recognition of classmates from a selection of 50 photographs, a name recognition test and a photo matching test.
In order to check accuracy of recall the researchers used year books for the relevant year groups of the participants.
Findings
For participants that had left school up to 34 years previously, accuracy of recall on the face and name recognition tasks was still an amazing 90%. Even for participants who had left school 48 years previously it was 80%.
Conclusion
Recall can be accurate over a very long period of time, leading to the term vLTM (very long term memory) to describe this phenomenon.
Evaluation
The procedure used is a field experiment so is much higher in ecological validity since this is far more similar to the purpose we generally use our memories for.
However, the study was poorly controlled. The researchers assumed that last contact with their classmates would have been when they left school. Little consideration seems to have been made of participants seeing classmates in the intervening years or even of them having looked through yearbooks themselves!
Only one type of recall (visual) was tested. It could even be argued that recognition of this sort is not recall as such anyway!
Evidence for two memory stores (STM and LTM) This is a favourite on examination papers. Clearly you could talk about research into the two main memory stores mentioned above. In addition to this it would be essential to mention some, if not all, of the following:
Multistore model of memory
As its name suggests, this model believes that we have more than one memory store. The two main ones being STM and LTM as well as the sensory stores mentioned at the start. The following studies can be used to support and in some cases criticise the idea of two main memory stores as well as the multistore model itself.
Obviously rehearsal will help with duration of STM. Try to remember the registration of that speeding car and you will repeat it to yourself over and over.
Amount of information: Murdock 1961 used a version of the Brown-Peterson technique to show that number of chunks affects duration. Participants were given either three letters that spelt a familiar word such as c, a, t or three unrelated three letter words such as sun, pat, lid. The latter deteriorated at the same rate as predicted by B-P so recall after 18 seconds was minimal. However, recall of the three letters was very stubborn to erase and after 18 seconds recall was still at over 90%.
Duration
Refers to the amount of time that a memory can be stored. In STM this is also limited. Traditionally duration in STM is measured by the Brown-Peterson technique. The technique was devised independently in 1959 by Brown and by Peterson and Peterson, hence the name.
Procedure
Participants are presented with trigrams (three consonants please Carol), e.g. ZTM. Vowels are avoided so words can’t be constructed. Participants are then asked to count backwards in 3s from various numbers. This prevents the participants thinking about the trigram letters so avoids the possibility of rehearsal. After a certain period of time, e.g. 3 seconds, 9 seconds etc. a tone sounds and the participant is asked to recall the trigram.
Findings:
When plotted it is clear that recall falls away very quickly over the first 20 seconds.
Conclusion: duration of STM is less than 18 seconds
Evaluation
This is one of the most repeated studies into human memory and other studies tend to replicate the original findings.
As with span measures, this technique is very artificial so again is lacking in ecological validity. Apart from in a psychology lesson, has anyone ever asked you to recall random groups of consonants?
Some studies have also discovered signs of interference. When being tested on later trigrams, participants have erroneously recalled earlier ones. This suggests that recall can be much longer than 18 seconds. As you probably discovered yourselves, it is also very easy to cheat when being tested.
Factors affecting duration of STM
Obviously rehearsal will help with duration of STM. Try to remember the registration of that speeding car and you will repeat it to yourself over and over.
Amount of information: Murdock 1961 used a version of the Brown-Peterson technique to show that number of chunks affects duration. Participants were given either three letters that spelt a familiar word such as c, a, t or three unrelated three letter words such as sun, pat, lid. The latter deteriorated at the same rate as predicted by B-P so recall after 18 seconds was minimal. However, recall of the three letters was very stubborn to erase and after 18 seconds recall was still at over 90%.
Encoding
Encoding refers to the format or code in which a memory is stored as a memory trace. Encoding can be by sound (acoustic) or by meaning (semantic) or by other means too.
The key study here can be used for STM and for LTM since it investigated both.
Baddeley (1966)
Participants were given four sets of words to recall in order. For the STM task they had to recall them immediately following presentation and for the LTM task they had to be recalled following a longer time interval. Set 1 were words that all sounded similar, for example: cat, mat, cap, map…
Set 2 were words that sounded differently for example: dog, bin, cup, pen….
Set 3 were words of similar meaning for example: big, large, huge, vast…
Set 4 were words of different meaning for example: huge, good, light, blue….
The researchers then recorded how many mistakes were made in recalling the sets of words.
STM
Findings
In the STM procedure participants made significantly more mistakes on words that sounded alike so for example would confuse cat and cap etc. Similarly with letters, S and X would be confused as would M and N and P and B etc.
Conclusion
It was concluded that in STM information is encoded by its sound (acoustically) so when we recall information from STM similar sounding words get confused.
LTM
Findings
In the LTM procedure participants were far more likely to confuse words of similar meaning replacing huge with vast or night and dark etc.
Conclusion
It was concluded that in LTM information is encoded by its meaning (semantically).
Evaluation
The usual lacks ecological validity can be used, but make sure you explain why!
Later research suggested that the situation is more complex than this. For example there appears to be some visual encoding in STM. We take longer to decide that Aa are different than we do to decide that AA are the same even though they both sound the same.
In LTM we appear to also encode words visually and acoustically as well as semantically. For example the fact that STM knows how ‘huge’ and ‘vast’ etc sound, must be stored in our LTM!
Long Term Memory (LTM)
Capacity is vast. As far as we know capacity has never been reached, but don’t worry the term is still young! However, it’s unlikely that you’ve ever heard anyone complain that they need to delete a few memories before they can store anything new! As far as I’m aware nobody has tried to estimate capacity, but Solso (1991) compares it favourably with the largest computers.
A stunning factoid for you to contemplate. In 1973 Petr Anokhin of Moscow University wrote:
“We can show that each of the 10 billion neurons in the human brain has the possibility of connections to 1 with 28 noughts after it; that’s 10,000,000,000,000,000,000,000,000,000 connections! It means that the total combinations in the brain, if written out, would be 1 followed by 10.5 kilometres of noughts!
One way psychologists have studied memory is to experiment on themselves. The beautifully named Marigold Linton kept a detailed diary recording daily events and facts about flowers etc on cue cards over many years. Each day was represented by a single word. When given the cue word she was able to recall with 70% accuracy all the events and information recorded for that day even 7 years later.
Duration: Bahrick et al Year Book study (1975)
Procedure
374 participants aged between 17 and 74 were tested on their memory of school friends. A number of different tests were carried out including a free recall of all the names of classmates they could remember, recognition of classmates from a selection of 50 photographs, a name recognition test and a photo matching test.
In order to check accuracy of recall the researchers used year books for the relevant year groups of the participants.
Findings
For participants that had left school up to 34 years previously, accuracy of recall on the face and name recognition tasks was still an amazing 90%. Even for participants who had left school 48 years previously it was 80%.
Conclusion
Recall can be accurate over a very long period of time, leading to the term vLTM (very long term memory) to describe this phenomenon.
Evaluation
The procedure used is a field experiment so is much higher in ecological validity since this is far more similar to the purpose we generally use our memories for.
However, the study was poorly controlled. The researchers assumed that last contact with their classmates would have been when they left school. Little consideration seems to have been made of participants seeing classmates in the intervening years or even of them having looked through yearbooks themselves!
Only one type of recall (visual) was tested. It could even be argued that recognition of this sort is not recall as such anyway!
Evidence for two memory stores (STM and LTM) This is a favourite on examination papers. Clearly you could talk about research into the two main memory stores mentioned above. In addition to this it would be essential to mention some, if not all, of the following:
Multistore model of memory
As its name suggests, this model believes that we have more than one memory store. The two main ones being STM and LTM as well as the sensory stores mentioned at the start. The following studies can be used to support and in some cases criticise the idea of two main memory stores as well as the multistore model itself.

Primacy and Recency Effects
Murdock (1962) gave 103 psychology students lists of words to free recall (in any order) in 90 seconds.
Typically words at the start of the list and especially those at the end tended to be recalled most often. This was explained by words at the start being rehearsed from STM into LTM creating a stronger trace and those at the end still being present in STM when recall begins.
Note: Glanzer and Cunitz did something very similar in 1966 and seems to be the most cited study in texts.
Murdock (1962) gave 103 psychology students lists of words to free recall (in any order) in 90 seconds.
Typically words at the start of the list and especially those at the end tended to be recalled most often. This was explained by words at the start being rehearsed from STM into LTM creating a stronger trace and those at the end still being present in STM when recall begins.
Note: Glanzer and Cunitz did something very similar in 1966 and seems to be the most cited study in texts.
Amnesiac case studies.
Most people with memory problems have either impairment of their STM or LTM, not usually both. This suggests that they are different systems.
Most people with memory problems have either impairment of their STM or LTM, not usually both. This suggests that they are different systems.
HM (or Henry Molaison)
The classic case is that of H.M who at the age of 27 underwent surgery in an attempt to cure his epilepsy apparently triggered by a cycling accident when he was nine. A surgeon, William Scoville removed both his temporal lobes including a structure known as the hippocampus (Latin for sea horse) and an area now known to be crucial to memory
Henry Molaison died in 2008, aged 82
The docudrama below gives a good feel for the level of impairment Henry suffered following his surgical procedure.
The classic case is that of H.M who at the age of 27 underwent surgery in an attempt to cure his epilepsy apparently triggered by a cycling accident when he was nine. A surgeon, William Scoville removed both his temporal lobes including a structure known as the hippocampus (Latin for sea horse) and an area now known to be crucial to memory
Henry Molaison died in 2008, aged 82
The docudrama below gives a good feel for the level of impairment Henry suffered following his surgical procedure.
HM’s pattern of memory loss is not unusual. Clive Wearing suffered a similar form of amnesia following a herpes simplex (cold sore) infection that spread to areas of his temporal lobes.
Episodic and Semantic
On the face of it, the cases of HM and Clive Wearing support the idea of two memory stores. Each has lost their LTM but their STM remains intact. However, things are more complex.
Clive Wearing has no recollection of any event in his life. His wife Deborah explains that he knows that he has a wife but doesn’t remember getting married. His memory for faces, other than his own and his wife’s his minimal. His episodic memory (for episodes in his life) has been damaged beyond repair. However, he can still walk, talk, read, play the piano, conduct; he can still use a knife and fork and a telephone. His LTM for skills and his understanding of how the World works is intact. This seems to be the usual pattern with patients that have catastrophic memory loss.
HM is similar in that his semantic memory is still intact and he can learn and improve new skills. Like Clive HM is unable t for new episodic memories. Unlike Clive, he did retain some episodic memories of his earlier life, up to two years prior to the surgery.
It seems therefore that the multistore model is overly simplistic. It considers only one LTM when there are at least two types.
STM
KF who suffered damage to his STM following a motorcycling accident still retained a near normal LTM. Again this suggests two different memory stores, with STM and LTM being distinct.
However, yet again the situation wasn’t quite that simple since KF could still recall visual information using his STM but struggled with auditory and verbal information, making conversation difficult. This seems to suggest that STM is also more complex.
We therefore have a situation were cases of amnesia can both support the idea of two memory stores whilst at the same time question the idea. This is particularly good stuff to include in a discussion of the existence of two memory stores. To summarise:
Two memory stores
Amnesiacs tend to lose only one store (either STM or LTM) supporting the idea of two separate stores presumably located in different brain areas or structures.
More than two stores
However, HM and Clive Wearing both provide us with evidence that LTM is more complex and seems to comprise at least two components (semantic and episodic LTM). Similarly KF suggests that STM is also more complicated having separate stores for both auditory and visual information. See later section on working memory.
|
There is lots of video footage of Clive's condition and of his long-suffering wife Deborah. The BBC did a full length documentary on him about a year following his illness and various others have been made since.
There are many clips on Youtube but this three minute exert provides an idea of how short his memory is. |