|
Chapter summaryAll human mental life and behaviour involves memory. Perceiving the world around you involves using your memory so that you can recognise and categorise what you see, hear, taste, touch and smell. Likewise, the study of memory underpins most of the other areas in cognitive psychology, and although a considerable amount of research has been devoted to memory, our understanding of memory is far from complete. In this section we will cover research into three main areas: immediate memory, episodic memory and semantic memory. In Chapter 10 recovered memory and false memory are covered through an example essay question. Memory in the short termIn this section you will learn about how short-term memory is studied, what the main findings are and what types of theories have been constructed. Several important techniques commonly used in short-term memory research will be outlined and a number of key theories will be explained. Memory metaphorsAn important question is how memories are stored. Philosophers and psychologists have used a variety of metaphors for thinking about memory and most of these are spatial, that is it is assumed that memories are stored in mental locations. Aristotle compared memory to a wax tablet, Plato compared it to an aviary and John Locke compared it to a cabinet. Later metaphors began to draw on the technology of the time. For example, in the 1950s memory was being likened to a telephone exchange system, and then when the computer was developed psychologists found its most sophisticated metaphor yet. More recently, the connectionist metaphor has come to the fore and this has gone hand in hand with the development of neural network technology in engineering and computer science. Cognitive models of memory are therefore embedded within either the traditional computer metaphor or the connectionist metaphor (see Chapter 9). The earliest models of memory in the 1950s focused on the dual store theory that is based on the computer metaphor: since computers consisted of short-term and long-term information storage, so human memory was conceived of as consisting of dual storage systems. Long before the computer was developed, the early psychologists had already speculated about dual storage systems. For example, William James (1890) divided all memory into primary memory and memory proper. Primary memory was considered to be that which we are immediately conscious of and memory proper was that which required ‘recollection' to ‘revive it' back into consciousness. This concept was then extended in the 1950s and 1960s and the terms short-term store (STS) and long-term store (LTS) were coined. Broadbent's modelBroadbent likened cognition to an electronic communications system through which information flows. The S-system receives external information, which it stores very briefly and then passes selected portions of the information into the P-system. This system contains that which we are consciously aware of. The S-system and the P-system collectively represent primary memory. Since primary memory is of limited capacity, information can only be retained through the process of rehearsal. Secondary or long-term memory was a more permanent store, and information is held there after being rehearsed in short-term memory. Rehearsal was conceived of as a form of inner speech. For example, if given a telephone number to remember you might repeat it to yourself mentally. EvaluationIn order to test the idea that rehearsal involves inner speech, Conrad (1964) tested recall for similar and dissimilar sounding items. Participants were presented with a list of letters that either sounded similar (such as T and V) or different (such as T and X). One group of participants read the letters aloud as they were presented and another group read them silently. It was found that the errors made by both groups were based on the sound of the letters (for example, they were more likely to confuse T with V than T with X during recall). Similar findings are reported in Baddeley (1966). These results imply the use of subvocal rehearsal when participants were reading the letters silently. Further evidence is provided by Murray (1967) who devised a procedure known as articulatory suppression. One group of participants were presented with a list of words and were required to say the word the repeatedly out loud (the articulatory suppression group). Another group were shown the words but not required to repeat the word the. Predictably, the errors made by the non-articulatory suppression group were based on phonemic similarity. However, the errors made by the articulatory suppression group were not based on phonemic similarity. This demonstrates that participants normally subvocalise the items as they are being learned, and when this is not possible, the information is not stored as an acoustic code. A second question concerning Broadbent's model is the issue of capacity limitation. Miller (1956) found across numerous experiments that primary memory appears to be limited to ‘The magic number seven, plus or minus two', which was also the title of his paper. His work identified a limitation on the ability to process information in the short term. This limitation he measured using the memory span technique. The experimenter begins by presenting a list of items aurally, beginning with amanageable number and then increasing the number of items by one in each phase. The participant has to recall the items in the correct order. As the number of items is increased so too are errors and the point at which the participant makes errors at least 50 per cent of the time reflects the number of items they can recall as their memory span. About 90 per cent of the adult population can recall at least five items in order but not more than eight.
Clearly, if items are presented in a meaningful way (e.g. 1–9–7–8) then this will increase the memory span. Miller termed the grouping of information in this way as chunking. When chunked items are used, participants can still recall, on average, seven plus or minus two chunks. It was argued that seven items are maintained in primary memory through rehearsal. Another method of preventing rehearsal is the Brown-Peterson technique. The experimenter reads out a consonant trigram (such as DYB) and then a three-digit number. The participant's task is to count backwards in 3s or 4s from the three-digit number for a specified time. After this the participant had to recall the three letters. The time spent counting backwards was varied from 3 to 18 seconds. They found that the number of items correctly recalled declined as the time was increased, and this is taken as evidence that rehearsal is used to keep items in primary memory, and that without rehearsal information decays rapidly. Waugh and Norman 's modelWaugh and Norman (1965) extended the Broadbent model in an attempt to understand the relationship between primary and secondary memory. They introduced the notion of displacement in primary memory, which is that as a new item enters so a previously held item must be lost or displaced. Rehearsal of items transfers the information into secondary memory, which has no capacity limitation. Rather than decay as an explanation of the Brown-Peterson effect they preferred the explanation that the few items that were recalled were due to rehearsal and that the main reason for information loss was displacement of the items by the items entering primary memory during the counting task. Atkinson and Shiffrin's modelAtkinson and Shiffrin (1968) introduced the terms short-term store (STS) which closely resembles Broadbent's use of the term P-system or primary memory, and long-term store (LTS) which is equivalent to the S-system or secondary memory. STS is a temporary memory store and information in it is always lost eventually, but LTS is a permanent store although information may be modified through learning. Information enters LTS via the STS and an item retrieved from LTS has to pass through STS. They defined a number of control processes that manipulate information in STS and that are under the volition of the individual. Rehearsal, coding and retrieval are control processes of the individual. Of these retrieval was said to be most interesting, as although it seems like an effortless process, searching for information in a vast store of information predicts that it should take much longer than it does. Although the Atkinson and Shiffrin model includes rehearsal as a way of storing information, it is assumed that information can be stored even in the absence of any intention to do so. As evidence that this occurs Hebb (1961) had previously shown that in a serial recall task, when a particular list is repeated intermittently in a long series of trials, the participants showed improved recall for that list over the other lists, even though participants were unaware of the repetitions. Atkinson and Shiffrin argue that this occurs because each time the list is presented some of the information is transferred into LTS. Thus their model is not an all-or-none affair and assumes that partial information can be stored and retrieved. For example, the tip-of-the-tongue phenomenon is explained by their model as being caused by the activity of partial rather than complete traces in memory. EvaluationOne of the advantages of the Atkinson and Shiffrin model is that it was expressed as a mathematical model and hence yielded precise, testable predictions. Previously, theories of immediate memory were vague and described in brief terms. Much evidence used to support the model have come from research on the serial position curve. The serial position curve is characterised as an inverted-U and is obtained through tests of free recall for items presented for a brief period. Its main features are the high accuracy of recall of the first few items, known as the primacy effect, and the high accuracy of recall of the last few items, known as the recency effect. According to the Atkinson and Shiffrin model, in serial recall primacy occurs because of the better rehearsal of the first few items. It is better because initially there are fewer items to rehearse. Recency can be explained by the displacement effect in STS, as the earlier items are pushed out of STS by newer items. Middle items are then poorly recalled because (1) they receive fewer rehearsals, and (2) as they are more volatile than the earlier items that do receive rehearsal they are pushed out by the later items. The model predicted that if recall was delayed then primacy will remain but recency will be lost. This is due to the fact that rehearsal of the earlier items pushed them into LTS but this is not so for the most recent items. This prediction was confirmed in Glanzer and Cunitz (1966) and Postman and Phillips (1965). A second prediction is that recency should remain regardless of the length of the list, since the most recent items remain in STS as they push out or displace earlier items in the list. Many studies have shown this to be the case, e.g. Murdock (1962). However, there are many findings that the model cannot account for. For example, while Glanzer and Cunitz (1966) found a method whereby recency could be eliminated, they did this by using a distracter at the end of the list. However, when a distracter is used throughout the list, e.g. after each and every item, a strong recency effect appears (Bjork and Whitten, 1974). The model cannot explain this finding. Koppenaal and Glanzer (1990) suggest that with continuous distracters, participants learn to time-share, that is they begin to alternate their attention between rehearsing the items and performing the distracter task. The essential problem with the model is that there is too strong a separation between STS and LTS, and as Neath (1998) points out STS is dependent upon LTS: ‘Information recalled from long-term memory has to pass through short-term memory on the way in and on the way out. Any response from any task should reflect both stores, and, because of this inherent contamination, it would seem impossible to separate the types of code' (p. 77). Indeed, many studies show that many of the results found with short-term memory can be replicated over much longer timescales (Neath, 1998). Working memoryThe distinction between short-term and long-term memory stores is no longer deemed plausible. How then can memory over the short term be characterised? The most influential theory is that of working memory (Baddeley and Hitch, 1974; Baddeley, 1986), and it is conceptualised as a location where many cognitive operations are carried out. Its main features are: • central executive that coordinates activities in working memory; • visuo-spatial sketch pad for processing visual information; • phonological loop for processing auditory information.
The phonological loop is further divided into the phonological store, which stores speech based information, and the articulatory control process, which translates visual information into speech-based information. Memory traces are assumed to decay rapidly over one or two seconds unless refreshed by the articulatory control process which controls subvocal rehearsal. The phonological loop was devised to account for the following: • The phonological similarity effect. Items that have similar sounds are confused. For example, the list BCDEPT is more difficult to recall than the list QSXKGN (Baddeley, 1966). Furthermore, phonemic similarity results in a large number of transposition errors for example, the sequence BFCUTJ being recalled as BFTUCJ with the phonemically similar C and T exchanging places in the sequence recalled). The effect occurs in the model due to covert rehearsal as the list is being presented.
• Articulatory suppression. According to the model, when rehearsal is prevented the phonological similarity effect should disappear. The evidence supports this prediction (e.g. Baddeley, Lewis and Vallar, 1984).
• The irrelevant speech effect. When irrelevant speech is played in the background serial recall is severely disrupted (Colle and Welsh, 1976). The model explains this finding as intrusions in the phonological store from the irrelevant speech. It further predicts that articulatory suppression should remove the effects of irrelevant speech and that irrelevant non-speech-based sounds in the background will not produce the same reduced performance as irrelevant speech. These predictions were supported by the findings of Salame and Baddeley (1982).
• The word-length effect. In serial recall, short words are recalled better than long words (Watkins, 1972). Baddeley, Thomson and Buchanan (1975) found that if one set of words takes less time to pronounce than another set of words, then memory for the shorter items is better. The model accounts for these findings by supposing that memory span is determined by the time each item takes to be subvocally rehearsed, and predicts removal of the effect with articulatory suppression. Indeed, when subvocal rehearsal is prevented throughout a trial there is no word length effect (Baddeley, Lewis and Vallar, 1984). EvaluationThere are several findings that contradict the working memory model. First, the irrelevant speech effect is not restricted to speech but can be produced with pure tones (Jones and Macken, 1993). Second, the word-length effect is more pronounced when the items are presented visually than aurally (Watkins and Watkins, 1973), and further, that the opposite effect can also occur (that shorter words can sometimes be recalled less well than longer words). A more general problem with the model is that it is expressed as a series of statements rather than as a mathematical or computational model (although there have been attempts to devise connectionist models of working memory, e.g. Burgess and Hitch, 1992). Activation theoriesRather than conceiving of short-term memory as a special location in memory where information leaves and enters (as in the Atkinson and Shiffrin and working memory models), immediate memory may merely be those areas of memory that are currently active (Cowan, 1993). The idea is that immediate memory is that part of our knowledge that is currently in a heightened state of activation, and that only a small portion of knowledge may be in the state of activation at any one time. Activations decay through time unless rehearsed and a central executive directs attention and controls which information will be at the centre of attention. EvaluationCowan's activation model disposes of the need for a specialist, discrete short-term storage system, and is one that is also consistent with the connectionist metaphor (see Chapter 9). There are a number of problems: • Such a model is too general to make precise predictions (about the effects of serial recall, for example). • Activation is not clearly defined. What does it mean to say that some bits of knowledge are in a heightened state of activation? One possibility is that activation may reflect increased activation of neurons that represent particular memories, but this is not clear. • The notion of decay of information through time. As pointed out by Neath (1998) ‘time is not a causal agent: iron rusts over time, but time should not be given a causal role; some other activity (usually defined as interference) that unfolds over time should be the causal agent.' Therefore, models of immediate memory should consider the role of inhibition or interference.
Activation models, such as the feature model (Nairne, 1990) take account of the many criticisms of previous models, but the feature model too is not without its problems. Distinguishing between episodic and semantic memoryTulving (1986) is usually credited with distinguishing between episodic memory (memory of when something happened) and semantic memory (memory for facts and knowledge). He raised five issues concerning the two types of memory: • Time Memory can either be time-based or independent of time. For example, recalling your first day at college is dependent upon time (i.e. you remember the events as occurring on the first day), but the information you learned in your first lecture that day (such as Baddeley's theory of working memory) is not dependent upon knowing when you learned it. • Types of associations Episodic memory is closely related to personal events and is autobiographical in nature. Semantic memory concerns the associations between concepts (e.g. that between restaurants and eating, for example) and is knowledge that is not necessarily of a personal nature. • Retrieval Retrieval of semantic memory seems not to be dependent upon the learning situation. However, for episodic memory, the learning context is important during retrieval, and retrieval can strengthen the information. • Interference Semantic memory appears to be less susceptible to interference than episodic memory. • Independence Although the two systems are sometimes dependent (you need semantic memory in order to recall a particular episode in your life, but you don't need semantic memory to recall the Kings and Queens of England ), they can be thought of and treated as two independent memory structures. Memory dissociationsHow can we test whether two memory systems, such as semantic and episodic memory, are truly independent? The dissociation technique is to attempt to manipulate one memory type without affecting the other. So, if an experiment shows that episodic memory can be affected but not semantic memory then this could be taken as evidence of the distinction. For example, suppose the recall of time-related information of an experimental task was weak when tested one month later but the recall of the materials studied was intact then this might be evidence of the distinction. (This method is not without its critics, and one problem for the approach is that if the two measures used are different – and in many ways they have to be – then one type of information might merely be more difficult to recall than the other. This might be because the participant paid more attention to one aspect of the materials than other aspects.) An important distinction in memory that has been extensively studied is that between explicit and implicit memory. Explicit memory refers to stored information that we can consciously recollect and talk about. Implicit memory refers to stored information that affects our behaviour but which is difficult to verbalise (such as being able to speak grammatically correctly, but not being able to state the rules of grammar). Dissociation tasks have been used to draw a distinction between explicit and implicit memory. Tasks used are those that affect explicit memory but leave implicit memory intact. An example is the observation of Clarapede (reported in Baddeley, 1992) on individuals with amnesia. The doctor gives the patient a pinch while shaking hands when being introduced. The patient flinches and asks why he did that. Some days later, the doctor returns to the patient and attempts to shake hands. Since the patient is amnesic he or she has no recollection of ever meeting the doctor; however, the patient is reluctant to shake hands and cannot explain why. This demonstrates the possibility that explicit memory (in this case memory of the initial meeting) and implicit memory (feeling reluctant to shake hands the second time around) are separate. Episodic memoryTwo main areas of episodic memory which have been extensively studied are the effects of context on retrieval and memory for when something happened, which includes autobiographical memory. Context effectsWhen we use our memory, we are said to be retrieving information. Various cues seem to affect the ability to retrieve information, and many of these are dependent upon the circumstances of when the information was acquired. The encoding specificity hypothesis formulated by Tulving and Thomson (1973) is that the best conditions for retrieval are those that are most similar to those during encoding or learning. An example of encoding specificitySuppose that three groups of participants are required to learn the same list of words that were presented in different categories, such as a list of fruits, place names and so on. One group are required to free recall as many items as possible, the second group are given the category labels (e.g. fruit, place names) of the words presented earlier and asked to recall as many of each (congruent cued recall), while the third group are given cued recall but using categories of words that did not appear in the list (incongruent cued recall). The group given congruent cued recall do better than the group given free recall, and the group given incongruent cued recall perform the worst (Roediger and Payne, 1983). According to the encoding specificity principle recall in these cases is dependent upon the presence and quality of appropriate cues. As well as particular cue words, environmental cues can improve recall. Godden and Baddeley (1975) gave divers lists of words in one of two conditions, underwater or at the surface. They were then tested either in the same or different setting. The divers who recalled the most words were those whose learning and test conditions were the same. Smith et al. (1978) found similar effects when the context was the same or a different room at learning and at recall. A distinction is made between context alpha, which refers to aspects of the immediate environment, and context beta (or interactive context), which refers to things close to the ongoing activity (Wickens, 1987). For example, while reading the sentence ‘The man robbed the bank' at home in your study, context alpha is the study environment, and context beta is the word ‘robbed' that helps you distinguish the meaning of the word ‘bank' as a place to keep money from its alternative meanings, such as ‘river bank'. The above examples refer to context alpha.
CONTEXT EFFECTS AND RECOGNITION Light and Carter-Sobell (1970) presented sentences to participants that included two capitalised and underlined words, as in ‘The STRAWBERRY JAM tasted great'. Participants were forewarned to expect a recall test of the underlined and capitalised words. The recall test consisted of old sentences (ones presented in the list), new sentences (ones that did not appear in the list), and modified sentences that consisted of an old and a new item, as in ‘The TRAFFIC JAM was terrible'. Participants were told to identify any of the underlined and capitalised words as old or new, irrespective of whether they appeared in a different sentence or with a different word next to it. The results show that when JAM was presented in the exact same context it was recognised 65 per cent of the time and when it was in a new context it was recognised about 25 per cent of the time. This study demonstrates the effects of context beta, as the sentence (i.e. the task) provides the cue for recognition. Internal states as contextsJust as environmental or task-related stimuli can elicit context-dependent memory effects, it appears that one's internal state can produce similar effects (state-dependent memory effects). For example, Bartlett and Santrock (1979) changed the states of participants so that they were in either a happy or sad mood, and found better recall when the moods matched between learning and recall than when they were different. In Chapter 8, we discuss similar results that have been reported by Bower (1981) and the implications they have for understanding emotional disorders. Goodwin et al. (1969) reported the state-dependent memory effects of alcohol. They found some interesting and unexpected results. As you would predict, the best performance occurred for participants who were sober during learning and recall; however, the worst performance occurred when the participants were ‘tipsy' during learning and sober during recall – even worse than when participants were tipsy during learning and during recall. Very similar findings have been reported with marijuana (Eich, et al., 1975), and with nicotine (Peters and McGee, 1982). Autobiographical and temporal aspects of memoryWhen we recall something from the past, we not only recall aspects of the event itself but also we have a sense of when it occurred and how long ago this was. A feature of my own memory that I find quite intriguing occurs when I leave my office and go to my car. I can park my car in several places but most of the time I instantly know where I parked it in the morning. Surprisingly, where I parked it yesterday and all of the days before that do not seem to interfere with my knowing where it is today. It is as though previous memories are automatically overridden. Memory for when something occurred is one of those areas of research aimed at understanding an aspect of mental life that we seem to achieve effortlessly. However, research shows that (1) this is a difficult area to theorise in, and (2) people are not very good at locating an event in time accurately. Some findings on memory for when something occurred are as follows: • We know (from the previous section) that recall is better for the beginning and ends of a sequence (the primacy and recency effects), but this is also true for autobiographical memory (Baddeley, Lewis and Nimmo-Smith, 1978). • When people estimate when something happened they tend to make the error that it happened more recently than it did (Thompson et al. 1988). This is known as forward telescoping. • If there is a long interval between two events then estimates of when they occurred are better than if the interval is short (Underwood, 1977). • Memory for the time, day of the week, time of the month, the month, the year of a single event can be independent of each other (Friedman and Wilkins, 1985). That is, you may be able to recall the time of your first visit to the union bar at university, but may not recall the day of the week or time of the month very accurately. • For rare events, and these tend to be particularly special events for us such as our wedding day or the birth of a child, temporal memory can be very accurate from the actual time of the day to the day of the week and so on (Friedman, 1993). Models of temporal memoryMany explanations have been offered for the way in which we are able (or unable) to recall the time of a particular event. The main types of theories are as follows: • Strength theory People make judgments about when an event occurred by assessing the strength of its memory trace. One problem for this theory is that it cannot account for primacy effects where events that happened earlier can sometimes be recalled better. • Organisation theories Estimating the time of an event is based on its distinctiveness, and distinctiveness automatically occurs because information is stored spatially in memory (rather like a tape recorder stores information sequentially). Three problems with this approach arise in explaining primacy effects, telescoping and the independence in the ability to recall the time, day, month, year and so on. • Time tagging The time is automatically recorded in memory when something happens and is tagged in with the memory of the event (Hasher and Zacks, 1979). An obvious problemwith this theory is that it cannot predict when time estimates will be accurate and when they will be inaccurate, such as in telescoping. • Perturbationtheory Item and order information about an event are stored separately, but associations are formed between items and their order in a sequence (Estes, 1972).The model is expressed as a relatively simple equation and can account for many observations in memory for sequential information. However, although it describes the findings well it is weak in offering psychological explanations of the findings. • Inference model People use multiple sources of information about the time of an event and then make an inference. In other words, if you were to recall the time of your first lecture at university, you might infer that it was a Monday (since this seems quite likely), it was either 9 a .m. or 2 p.m. (since these are typical lecture starting times), and it was late September or early October (since this is when the first term of university usually starts). So, multiple sources of information can give rise to quite an accurate estimation of time. The theory can explain recency effects (more cues are available for recently occurring events than events long ago) and also primacy effects (by relating them to landmarks in our personal history). The theory can also explain the finding that events more separated in time are easier to put a date to than events close in time, through the general cognitive principle that information that is more separated in time can be easier to discriminate. The strengths of the theory are that it can account for the findings and it can offer a plausible psychological explanation. Memory for specific events: flashbulb memoryFor some events we can recall the exact time and date it occurred because we deliberately make a note of it. For example, we remember the date of the terrorist attacks in the US as 11 September because that is one way in which it is recorded and labelled. For events like these we seem to be able to recall other more minor details. Can you remember where you were when you heard the news that a plane had hit the north tower of the World Trade Center in New York ? Can you recall how you heard the news, what you were wearing, whom you were with and what activity you were engaged in? Do you remember this moment like a photograph? According to Brown and Kulik (1977), when a major event occurs people record many important and not so important details of it in memory rather like the way a photograph preserves information indiscriminately. They coined the term flash bulb memory to refer to the photograph metaphor for when the details of an extremely surprising or shocking event are printed accurately in memory. Their studies were based on events such as the assassination of President John F. Kennedy in Dallas , Texas , and it seems that people are able to recall clearly not only the event itself but also other minor details of the moment they heard the news. They postulated that the emotional impact such an event has promotes rapid and durable imprints of the situation. EvaluationStudies of flashbulb memory have focused on major events such as the explosion of the Space Shuttle Challenger, the resignation of the British Prime Minister Mrs Thatcher, the death of Diana, Princess of Wales, and the assassination of the Swedish Prime Minister Olof Palme. Many studies claim to have found evidence for a special flashbulb memory mechanism. However, other studies suggest that memory for such events can be explained by normal memory mechanisms. For example, when such events occur people tend to talk about them and tell people how they heard the news (rehearsal). Secondly, people are continually reminded of the event through the news media, and this may remind them of the time they heard the news. Finally, many studies show that memory for the details of how such news was heard can be very inaccurate (see Crucial Study below).Weaver (1993) argues that when the experience of such an event occurs people are compelled to remember the event, but their memory is no better than that of more innocuous events.
FLASHBULB MEMORY (NEISSER AND HARSCH , 1992) Eye-witness testimonyAn important and well studied aspect of episodic memory is the memory of the eyewitness. Since a person can be convicted on the basis of the testimony of a single eyewitness, the reliability of this kind of memory has major implications. One example is that of Jennifer Thompson who identified Ronald Cotton as the man who had raped her at knifepoint. Cotton was convicted and given a life sentence. Some years later new DNA evidence revealed that it could not have been him and identified another man as the rapist. Her testimony had been convincing enough for a jury but her memory was erroneous (O'Neill, 2000).
EYEWITNESS TESTIMONY (LOFTUS , 1979) The study by Loftus (1979) shows that the wording of a question can influence the recall of a memory, but other studies have also shown that an object mentioned after an event can often be mistakenly recalled as having been there (Dodson and Reisberg, 1991). It appears that when a new object is mentioned it is integrated into the old memory and subsequently indistinguishable from what was originally seen. Eyewitnesses may provide detailed accounts of events and the amount of detail has been found to correlate with the likelihood of a conviction (Bell and Loftus, 1989), even when the detail can be irrelevant. In addition jurors tend to be impressed by the confidence of an eyewitness. Participants watched a traffic accident and were then asked one of the following: • How fast were the cars going when they contacted each other? • How fast were the cars going when they hit each other? • How fast were the cars going when they smashed into each other?
Eyewitness responses were influenced by the verb used in the question. Smashed produced the highest estimates of speed and a week later participants who had heard the word smashed recalled the accident as being more violent than participants who had heard the words contacted or hit. The study demonstrates not only that memory for an event is not necessarily a true record of an event, but that different methods of questioning can yield different responses for the same memory. However, we have seen in previous sections that confidence and memory accuracy do not always correlate well. There are now strict guidelines on how evidence should be abstracted from eyewitnesses in the US. Semantic memorySemantic memory is our knowledge of the world, from general to more specific and specialised knowledge. In this section you will be studying three influential views of semantic memory: schemas, semantic networks and connectionist networks. The discussion of the last is brief since it is covered more fully in Chapter 9. Schemas‘Would you like to try the wine, madam?' It is most likely that you know what this sentence refers to. It is the moment at a restaurant when someone at the table is asked to try the wine to check that it is not a bad bottle. Your knowledge of the sequence of events that takes place when you visit a restaurant is probably very detailed. Knowledge may be organised as schemas, thus we may have schemas for eating out at restaurants, schemas for football, schemas for college life, and so on. A feature of schematic memory is that given partial information, we are able to recall the entire schema, as in the restaurant example. Memory organised in this way may serve a cognitive economy: that we do not need to be told about every detail, we can infer much of the information. Indeed, people often say things like ‘You know when . . .' as in ‘You know when your phone goes off at the cinema. . . 'When we talk in this way we are announcing the schema we are about to refer to and it saves having to explicitly state a large amount of information. A schema is ‘an organised knowledge structure that reflects an individual's knowledge, experience, and expectations about some aspect of the world' (Neath, 1998, p. 328). When we hear someone telling us something or when we begin reading text, we often automatically infer the schema that is being referred to. Several candidate schemas may become active in semantic memory until further information is received, whereupon only one schema remains active.
ACTIVE SCHEMAS (BRANSFORD AND JOHNSON, 1972) Participants were presented with a piece of text and were asked to rate the text on how easy or difficult it was to understand. An abstract of this text is: The procedure is actually quite simple. First you arrange items into different groups. Of course, one pile may be sufficient depending on how much there is to do. If you have to go somewhere else due to lack of facilities that is the next step, otherwise you are pretty well set. It is important not to overdo things. It is difficult to foresee any end to the necessity for this task in the immediate future, but then one never can tell. After the procedure is completed one arranges the materials into different groups again. Then they can be put into their appropriate places. . . (p. 722) One group of participants rated the passage as very difficult to understand and another group rated it easy to understand. The only difference was that the second group were told beforehand that the text they were about to read was about doing the laundry. The experiment shows that by informing participants about the subject of the text, they can then activate their schema for doing the laundry and hence better comprehend the passage. EvaluationThe most obvious criticism about schema theory is that it is quite vaguely specified and cannot yield many precise predictions. The concept is not new and Plato had a version of memory that is consistent with the notion of the schema. However, this does not mean that the concept cannot form the basis of a more detailed theory of semantic memory, and semantic and connectionist networks attempt to do this. Semantic networksOne way in which schematic knowledge may be organised is through hierarchical arrangements of concepts. In other words, units of knowledge may be associated with each other through meaningful links. A semantic network consists of a collection of nodes (Collins and Loftus, 1975). Each node represents a concept, such as animals and categories of livings things. Nodes are connected according to their relationship. For example, the node Animal is connected to both Living thing and Bird nodes. They are connected by the semantic association is a, since an animal is a living thing and a bird is an animal. Furthermore, the nodes can be arranged hierarchically according to categorical information, and in this example, Living thing is the top node, Animal is at the next level and Bird is at the lowest level. However, Bird can have nodes at even lower levels, as in Penguin since it is a type of bird, and also Wings since a bird has wings (the connection in this case is has). According to the theory, when information is retrieved one node becomes active and activity then spreads throughout the network. An example of spreading activation is that if you heard the word Bird then nodes such as Animal and Wings will become active. Furthermore, if you were asked, true or false, whether a penguin is a bird, the relevant nodes in your semantic network would become active and enable you to answer the question. Some information can be recalled ‘directly' and other information has to be inferred. For example, ‘Is a robin a bird?' can be answered directly because it is most likely that this has been learned directly. However, ‘Is a penguin a mammal?' might require an inference since we may have never directly learned that a penguin is not a mammal. In these cases, response times to the former questions will be shorter than those to the latter types of questions. Collins and Loftus (1975) predict that the time taken to respond to these types of questions will directly relate to their distance in a semantic network. This prediction has been confirmed. EvaluationOne of the main problems with semantic network theory is the concept of spreading activation itself. How does the system know when to stop spreading its activation? Suppose a word, like Bird, was associated with 10 other words. Then those 10 words become activated when the word Bird is presented. Those 10 words may each be associated with a further 10 words, and those further 10 words with another 10 words, and so on. After only a brief moment, activation of the word Bird would activate tens of thousands of other words in semantic memory. Connectionist networksIn a connectionist network there is a collection of units or nodes where each node represents a concept. Connections between nodes represent learned associations. Activation of a node will activate other nodes associated with it. Connections between nodes are not programmed into the network. Rather, the network learns the association by ‘exposure' to the concepts. This model of semantic memory has a number of psychologically plausible concepts, such as the following: • Generalisation Given partial information the network can retrieve a whole memory. For example, if we overhear someone talking about a key political figure in the UK named Tony (partial information), we can infer that the person being referred to is Tony Blair the British prime minister. Likewise, given a few details of a complete memory, the network can retrieve the memory in its entirety. • Fault tolerance In humans neuron loss is mostly constant, but this does not seem to have a dramatic effect (unless the amount is great or the damage is located in a particular area of the brain). Indeed, as neuron loss increases behaviour shows a graceful degradation – performance on a task gradually declines rather than disappears all together. The advantage of the connectionist network is that the loss of a handful of units will not result in the complete inability to recall the memory whole. Chapter 9 discusses this approach in more detail. Connectionist networks are an increasing influence in theoretical cognitive psychology, so it is a useful topic to study. Typical Examination Questions1. Describe and evaluate the Atkinson and Shiffrin model of short-term memory. 2. Outline ways in which episodic memory differs from semantic memory. What cognitive mechanisms have been suggested that help people recall the time of an event? 3. Outline and evaluate one theory of semantic memory. Further readingNeath, I. (2002) Human Memory: An Introduction to Research, Data, and Theory. Belmont , CA : Brooks/Cole Publishing. Baddeley, A. D. (1990) Human Memory: Theory and Practice. Hillsdale , NJ : Erlbaum & Associates. Andrade, J. (2002) Working Memory in Perspective. Hove: Psychology Press. |
||
This book was first published in 2003 by Crucial, a division of Learning Matters Ltd [ISBN 1 903337 13 5] © 2003 Eamon Fulcher; © 2009 GEFT Consultance Services (geft.co.uk). All rights reserved. No part of this publication may be reproduced, stored in a retrieval system, or transmitted in any form or by any means, electronic, mechanical, photocopying, recording, or otherwise, without prior permission in writing from Geft Consultancy Services, who may be contacted via www.geft.co.uk.
|
|||