Prof. Alan Baddeley

Revision as of 16:12, 17 March 2009

Working memory

Abstract

Working memory is a limited capacity part of the human memory system that combines the temporary storage and manipulation of information in the service of cognition. Short-term memory refers to information-storage without manipulation and is therefore a component of working memory. Working memory differs from long-term memory, a separate part of the memory system with a vast storage capacity that holds information in a relatively more stable form. According to the multi-component model, working memory includes an executive controller that interacts with separate short-term stores for auditory-verbal and visuo-spatial information. The concept of working memory has proved useful in many areas of application including individual differences in cognition, neuropsychology, normal and abnormal child development and neuroimaging.

The term working memory is used most frequently to refer to a limited capacity system that is capable of storing and manipulating temporary information involved in the performance of complex cognitive tasks such as reasoning, comprehension and certain types of learning. Working memory differs from short-term memory (STM) in that it assumes both the storage and manipulation of information, and in the emphasis on its functional role in complex cognition. A range of different approaches to the study of working memory have developed with differences reflecting the interests of the researcher, whether neuropsychological (Vallar, 2006), neurobiological (O’Reilly et al., 1999), psychometric (Engle et al., 1999) or oriented towards providing practical guidance on human factors (Kieras et al., 1999). Despite very different theoretical methods and styles, there is general agreement on a need to assume a central role for some form of executive controller, probably of limited attentional capacity, aided by temporary storage systems, with visual and verbal storage probably operating separately ( Miyake & Shah, 1999). Such a structure had in fact been proposed by Baddeley and Hitch (1974). While accepting that this is now one of range of models, the Baddeley and Hitch multicomponent model provides a convenient structure for summarising research on working memory over the 30 years since it was first proposed.

Before moving on it is important to note that the term working memory has also been used in other ways that are distinct from the concept we discuss here. One is as a technical term in the production systems, approach to artificial intelligence where working memory typically refers to the system that holds the ‘if-then’ rules that are basic to the operation of this computational approach. Working memory when used in this sense is assumed to be unlimited in capacity and not to have a direct counterpart in human cognition (Newell, 1990; Newell & Simon, 1972). However, production system architecture can be used to develop specific models of human cognition, including those which assume a working memory (Anderson, 1983). A further meaning of the term working memory comes from the study of animal learning and refers to the type of learning or memory thought to underpin tasks such as the radial arm maze, in which an animal has to remember which of several arms have already been visited on that day, a task which probably involves long-term memory (Olton, Becker & Handelmann, 1979).

The multicomponent model of working memory In the 1960s there was a short period of consensus among researchers that human memory consisted of a system that could be divided into two principal components. One was a short-term store capable of holding small amounts of information for a few seconds. This fed into a separate long-term store holding vast amounts of information over longer time intervals. This so-called modal model could account for a range of data and to give an account of selective effects of different types of brain damage on short- and long-term recall.

Baddeley and Hitch (1974) set out to test the hypothesis that STM also functioned as a working memory. They did so by requiring participants to perform reasoning, comprehension or learning tasks at the same time as they were holding in STM between 0 and 8 digits for immediate recall. If STM does function as a working memory, then loading it to capacity should lead to massive disruption of cognitive processing. It did indeed cause some disruption with time to perform a reasoning task increasing with load, but the effect was not huge, and there was no influence on error rate. Baddeley and Hitch (1974) therefore abandoned the modal model, according to which STM is a unitary store, proposing instead the multicomponent model shown in Figure 1. This assumes an attentional controller, the central executive aided by two subsystems, the visuo-spatial sketchpad concerned with visual storage and processing, and its acoustic/verbal equivalent, the phonological loop.

The phonological loop This is the subsystem that is assumed to hold digit sequences for immediate recall. The fact that reasoning was slowed as number of digits increased suggests that it does play a role in reasoning, but the unchanged error rate indicates that it is not essential. It is assumed to have two basic components, a temporary speech-related/acoustic store and a subvocal articulatory rehearsal process. The phonological store is indicated by the presence of the phonological similarity effect, whereby people are much less accurate in repeating back sequences of similar-sounding words such as MAN CAP CAT MAT CAN, than dissimilar words such as PIT DAY COW PEN TOP. Similarity of meaning (HUGE LARGE BIG WIDE TALL) has little effect on immediate recall. On the other hand if you are given a longer list of say 10 words and several trials to learn it, meaning becomes all-important and sound loses it power, consistent with different systems for short-term and long-term storage (Baddeley, 1966a; 1966b).

Evidence for the importance of rehearsal comes from the word length effect, whereby immediate recall of long words (e.g. REFRIGERATOR UNIVERSITY TUBERCULOSIS OPPORTUNITY HIPPOPOTAMUS) is much more error-prone than for short words (Baddeley, Thomson & Buchanan, 1975).

Baddeley and Hitch suggested that the memory trace of items in the short-term store would rapidly fade, but could be maintained by saying them to oneself. Long words take longer to say, allowing more fading and hence more forgetting to occur. Consistent with this interpretation, preventing subjects from saying words to themselves by requiring the continuous utterance of an item such as the word ‘the’, removes the word length effect. Since the initial demonstration of the word length effect (Baddeley, Thomson and Buchanan, 1975) other interpretations have been proposed, differing principally in the implications of the effect for whether items in the short-term store are forgotten as a result of spontaneous decay of the memory trace, or by disruption from later material (See Baddeley, 2007 Chapter 3 for a discussion).

The concept of the phonological loop has influenced a plethora of attempts to simulate human performance in verbal STM tasks using more detailed computational models. The first tranche of such models focused on specifying mechanisms for handling information about the serial order of items, an aspect that was left unspecified in the original model of the loop. These models tend to agree that serial ordering involves ‘competitive queueing’ (Grossberg, 1987) a process whereby items are simultaneously active and compete for serial selection, and differ principally with respect to the nature of the ordering cues that determine these activation levels (Burgess & Hitch, 1992; Page & Norris, 1998; Brown, Preece & Hulme, 2000). Recent attempts at computational modelling have gone further by specifying how the short-term phonological storage system interacts with long-term memory (Burgess & Hitch, 2006; see also Botvinick & Plaut, 2006), an essential step if one wants to understand the role of the loop in long-term learning.

Function of the phonological loop Given that one accepts the evidence for a temporary verbal or phonological memory system, the question of its evolutionary significance arises. One possibility is that the phonological loop supports the acquisition of language, providing a temporary means of storing new words, while they are consolidated in phonological LTM (Baddeley, Gathercole & Papagno, 1998). Evidence for this comes from the study of a patient with a very pure phonological STM deficit, who found it extremely hard to learn to link new foreign words to their meaning, while performing normally when learning to link pairs of words in her native language (Baddeley, Papagno & Vallar, 1988).

A series of studies followed up this hypothesis. In one study, eight year old children with specific language impairment (SLI), who had normal general intelligence but the language of six year olds were tested. They found it very difficult to repeat back nonwords such as SKITICULT. As they showed no sign of impaired hearing or speech production , their deficit was attributed to impaired phonological STM (Gathercole & Baddeley, 1990). Twin studies have shown that the deficit in nonword repetition in SLI is inheritable, but that other deficits also contribute to the disorder (Pennington & Bishop, 2009). Performance on nonword repetition also correlates highly with the level of vocabulary development in young normal children although as children get older, other factors such as intelligence and exposure to language become increasingly important (Baddeley, Gathercole & Papagno, 1998).

The visuo-spatial sketchpad The study of visuo-spatial STM has developed enormously in recent years and is very well described in the entry by Luck, who suggests that its principal function is to create and maintain a visuo-spatial representation that persists across the irregular pattern of eye movements that characterise our scanning of the visual world.

Another function of the sketchpad is to create and maintain visual images of the type that we might for example, use in attempting to answer questions such as whether the ears of a collie dog are floppy or pricked, in describing the route from the station to your home, or used by an architect to imagine a building he is designing. It is has been shown that spatial tasks can interfere with a spatial skill such as driving a car, while a more purely visual activity such as seeing a sequence of pictures or colour patches may interfere with your capacity to remember objects or shapes (Logie, 1986, Klauer & Zhao, 2004). Such results, together with the occurrence of brain-damaged patients who show one deficit and not the other, suggests that information about space, and about objects and their visual characteristics, may be stored separately (See Luck’s entry for further detail).

The central executive This is assumed to be an attentional control system of limited processing capacity that has the role of controlling action. Baddeley (1986) adopted a model proposed originally by Norman and Shallice (1986) which suggested that actions are controlled in two ways. Behavior that is routine and habitual is controlled automatically by a range of schemas, well-learned processes that allow us to respond appropriately to the environment. An experienced driver on a routine trip would be a good example of this, sometimes arriving at the destination with no memory of the journey. When such procedures are no longer adequate, for example finding the normal route blocked by an accident, a second system, the Supervisory Attentional System (SAS) comes into operation. This is capable of using long-term knowledge in order to set up possible solutions, and reflect on them before choosing the best. In the case of our interrupted journey, this might involve the central executive of working memory, probably in connection with LTM, the visuo-spatial sketchpad, and possibly the phonological loop.

Executive functioning has been extensively investigated by Shallice (2002), particularly in connection with its disruption following damage to the frontal lobes of the brain, a deficit referred to as the dysexecutive syndrome. This may result in major problems of attentional control, including sometimes repeatedly perseverating on a single action, while at others failing to maintain a goal against possible distraction. In the case of memory, this may result in confabulation where, in attempting to retrieve a memory, the patient is captured by inappropriate associations, sometimes resulting in totally false memories (Baddeley & Wilson, 1986).

The episodic buffer The initial three-component model of working memory ran into problems in accounting for the way in which the various subsystems could work together and in particular how they could interface with long-term memory. To tackle this problem, Baddeley (2000) proposed a fourth component, the episodic buffer (See Figure 2). This was assumed to be a temporary store of limited capacity that was capable of combining a range of different storage dimensions, allowing it to colate information from perception, from the visuo-spatial and verbal subsystems and LTM. It was assumed to do so by representing them as multidimensional chunks or episodes, which were assumed to be available to conscious awareness. The capacity to bind a range of separate sensory channels into the perception of integrated objects is often regarded as an important function of consciousness (e.g. Baars, 2002). Our investigation of this binding function in recent years suggests that the episodic buffer should be regarded as a passive store, rather than an active processor, capable of representing bound features, but not of performing the act of binding. In this respect, the current model differs from the proposal by Baars (1997) that consciousness operates like a stage on which actors perform, replacing it with a concept more closely resembling that of a screen on which binding processes operating elsewhere can be projected and utilised by the central executive.

Individual differences in working memory Daneman and Carpenter (1980) were interested in the role of working memory in comprehension. They developed a task that involved simultaneously processing sentences and remembering the last word of each, which they called working memory span. They found a remarkably high correlation between performance on this task and measures of reading comprehension in their college student participants. This finding has been replicated many many times, and shown to operate across a wide variety of tasks that combine temporary storage with processing. Some of these tasks are quite complex for example, remembering words that are sandwiched between arithmetic items (Turner & Engle, 1989), but even quite simple operations may correlate with measures such as scholastic achievement, provided they involve combining memory and rapid processing (Lépine et al, 2005). There is furthermore a very high correlation between tasks of this type and performance on conventional intelligence tests based on reasoning capacity (Kyllonen & Christal, 1990). This has led a number of groups to search for the crucial capacity that allows these apparently simple tasks to predict such a wide range of cognitive skills. A summary of the current state of play in this area is given a book edited by Conway et al. (2008).

Probably the most sustained attempt to solve the problem of why working memory span predicts so many cognitive tasks has come from Engle and his co-workers who have shown that the capacity to predict cognitive function is not limited to memory tasks, but can also be found in attentional control paradigms such as that involved in the antisaccade-task (See Kane et al. 2008). This involves participants moving their eyes from a fixation point to a target as rapidly as possible. This can be facilitated by a peripheral warning light occurring at the point where the target will appear, as there is a strong tendency for the eye to move automatically to a new stimulus. In a second condition however, rather than directly indicating the location of the target, the warning light signals participants to move their eyes to the opposite side. This information is still helpful to high span but not low working memory span participants. Based on this and a range of other studies Engle and colleagues argue that the crucial feature of working memory is to the capacity to maintain attention against distraction, whether this is perceptual or comes from other sources such as earlier memories.

However, while Engle makes a strong case for an association between working capacity and the ability to inhibit distracting stimuli, it is not clear that this is the only feature that characterises working memory capacity; it may indeed be just one example of a number of functions of a more basic attentional control system. Furthermore the concept of inhibition itself is open to a range of interpretations at both a psychological and physiological level.

The fact that we still do not fully understand how these complex working memory span tasks work does not, of course, mean that we can not use them profitably. Susan Gathercole and colleagues used the multicomponent working memory model to develop a working memory battery suitable for children of school age, using complex working memory span tasks as a measure of central executive capacity and other tasks to assess phonological and visuo-spatial subsystems. Factor analysis showed that the multi-component structure of the working memory system remains remarkably stable as children develop (Gathercole, Pickering, Ambridge & Wearing, 2004), and its capacity grows smoothly (Case, Kurland & Goldberg, 1982). However there are nevertheless marked developmental changes in the way working memory is utilised. For example the developmental progression whereby the child masters increasingly complex intellectual operations has been linked to the growth of central executive capacity (Halford, 1993). There are also significant developmental changes in subsystems of working memory, the best known being the expansion in the range of operation of the phonological loop, from development of the capacity for inner speech and rehearsal strategies in children (Hitch, 2006) to the involvement of a wider range of aspects of executive control in adults (Saeki & Saito, 2004).

Gathercole’s test battery is able to identify children who are at risk of encountering academic difficulties, with different patterns of working memory deficit associated with problems in different subject areas. Careful observation of children at school has showed that children with poor working memory skills tend to struggle because of a difficulty in following the sometimes surprisingly complex instructions provided by teachers. They also have trouble in coping with many of the techniques and strategies that are designed to help children cope, since these often require additional working memory capacity. Such children resemble those characterised as suffering from the attention deficit component of the ADHD syndrome. A programme has been developed that helps teachers identify such children and optimise teaching methods has been developed (Gathercole & Alloway, 2008).

Neural substrates of working memory A good deal of research has been carried out on this topic, initially through the study of patients with localised lesions and subsequently by using neuroimaging methods. Broadly speaking, the results fit the three-component model, with the phonological loop being represented in the left hemisphere where storage is associated a region in the temporoparietal junction (Brodmann area 40), and rehearsal with the more anterior Brodmann area (44) that is known to be associated with speech production (Paulesu, Frith & Frackowiak, 1993). The visuo-spatial sketchpad appears to involve a number of predominantly right hemisphere areas, one visual, presumably reflecting the processing and retention of objects and their visual features, a second area is more parietal, presumably involving spatial aspects, while two frontal areas of activation are presumably associated with control functions (Henson, 2001). There is general acceptance that the frontal lobes play an important role in executive control, although opinions differ as to the extent to which different executive functions may be separately localisable (Duncan & Owen, 2000; Shallice, 2002). There is as yet little evidence as to the localisation of the episodic buffer, which seems likely to reflect a broadly distributed system which may possibly not give rise to activation in any one specific area.

----

References

• Anderson, J. R. (1983). The architecture of cognition. Cambridge, MA: Harvard University Press.

• Baars, B. J. (1997). In the Theater of Consciousness. New York: University Press.

• Baars, B. J. (2002). The conscious access hypothesis: origins and recent evidence. Trends in Cognitive Sciences, 6(1), 47-52.

• Baddeley, A. D. (1966a). Short-term memory for word sequences as a function of acoustic, semantic and formal similarity. Quarterly Journal of Experimental Psychology, 18, 362-365.

• Baddeley, A. D. (1966b). The influence of acoustic and semantic similarity on long-term memory for word sequences. Quarterly Journal of Experimental Psychology, 18, 302-309.

• Baddeley, A. D. (1986). Working Memory. Oxford: Oxford University Press.

• Baddeley, A. D. (2000). The episodic buffer: A new component of working memory? Trends in Cognitive Sciences, 4(11), 417-423.

• Baddeley, A. D. (2007). Working memory, thought and action. Oxford: Oxford University Press.

• Baddeley, A. D., Gathercole, S. E., & Papagno, C. (1998). The phonological loop as a language learning device. Psychological Review, 105(1), 158-173.

• Baddeley, A. D., & Hitch, G. J. (1974). Working memory. In G. A. Bower (Ed.), The psychology of learning and motivation: Advances in research and theory. (Vol. 8, pp. 47-89). New York: Academic Press.

• Baddeley, A. D., Papagno, C., & Vallar, G. (1988). When long-term learning depends on short-term storage. Journal of Memory and Language, 27, 586-595.

• Baddeley, A. D., Thomson, N., & Buchanan, M. (1975). Word length and the structure of short-term memory. Journal of Verbal Learning and Verbal Behavior, 14, 575-589.

• Baddeley, A. D., & Wilson, B. (1986). Amnesia, autobiographical memory and confabulation. In D. Rubin (Ed.), Autobiographical memory (pp. 225-252). Cambridge: Cambridge University Press.

• Botvinck, M., & Plaut, D. C. (2006). Short-term memory for serial order: A recurrent neural network model. Psychological Review, 113, 201-233.

• Brown, G. D. A., Preece, T., & Hulme, C. (2000). Oscillator-based memory for serial order. Psychological Review, 107, 127-181.

• Burgess, N., & Hitch, G. J. (1992). Toward a network model of the articulatory loop. Journal of Memory and Language, 31, 429-460.

• Burgess, N., & Hitch, G. J. (2006). A revised model of short-term memory and long-term learning of verbal sequences. Journal of Memory and Language, 55, 627-652.

• Case, R. D., Kurland, D. M., & Goldberg, J. (1982). Operational efficiency and the growth of short-term memory span. Journal of Experimental Child Psychology, 33, 386-404.

• Conway, A. R. A., Jarrold, C., Kane, M. J., Miyake, A., & Towse, J. N. (2008). Variation in working memory. New York: Oxford University Press.

• Daneman, M., & Carpenter, P. A. (1980). Individual differences in working memory and reading. Journal of Verbal Learning and Verbal Behaviour, 19, 450-466.

• Duncan, J., & Owen, A. M. (2000). Common regions of the human frontal lobe recruited by diverse cognitive demands. Trends in Neurosciences, 23, 475-483.

• Engle, R. W., Kane, M. J., & Tuholski, S. W. (1999). Individual differences in working memory capacity and what they tell us about controlled attention, general fluid intelligence, and functions of the prefrontal cortex. In A. Miyake & P. Shah (Eds.), Models of working memory (pp. 102-134). Cambridge: Cambridge University Press.

• Gathercole, S., & Baddeley, A. (1990). Phonological memory deficits in language-disordered children: Is there a causal connection? Journal of Memory & Language, 29, 336-360.

• Gathercole, S. E., & Alloway, T. P. (2008). Working memory & learning: A practical guide. London: Sage Press.

• Gathercole, S. E., Pickering, S. J., Ambridge, B., & Wearing, H. (2004). The structure of working memory from 4 to 15 years of age. Developmental Psychology, 40, 177-190.

• Grossberg, S. (1987). Competitive learning: from interactive activation to adaptive resonance. Cognitive Science, 11, 23-63.

• Halford, G. S. (1993). Children's understanding: The development of mental modals. Hillsdale, NJ: Lawrence Erlbaum.

• Henson, R. (2001). Neural working memory. In J. Andrade (Ed.), Working memory in perspective. (pp. 151-174.). Hove: Psychology Press.

• Hitch, G. J. (2006). Working memory in children: A cognitive approach. In E. Bialystock & F. I. M. Craik (Eds.), Lifespan Cognition: Mechanisms of Change. (pp. 112-127). New York: Oxford University Press.

• Kane, M. J., Conway, A. R. A., Hambrick, D. Z., & Engle, R. W. (2008). Variation in working memory capacity as variation in executive attention and control. In A. R. A. Conway & C. Jarrold & M. J. Kane & A. Miyake & J. N. Towse (Eds.), Variation in Working Memory (pp. 22-48). New York: Oxford Univesity Press.

• Kieras, D. E., Meyer, D. E., Mueller, S., & Seymour, T. (1999). Insights into working memory from the perspective of the EPIC architecture for modeling skilled perceptual-motor and cognitive human performance. In A. Miyake & P. Shah (Eds.), Models of working memory (pp. 183-223). Cambridge: Cambridge University Press.

• Klauer, K. C., & Zhao, Z. (2004). Double dissociaions in visual and spatial short-term memory. Journal of Experimental Psychology: General, 133, 355-381.

• Kyllonen, P. C., & Christal, R. E. (1990). Reasoning ability is (little more than) working memory capacity. Intelligence, 14, 389-433.

• Logie, R. H. (1986). Visuo-spatial processing in working memory. Quarterly Journal of Experimental Psychology, 38A, 229-247.

• Lépine, R., Barrouillet, P., & Camos, V. (2005). What makes working memory spans so predictive of high-level cognition? Psychonomic Bulletin & Review, 12, 165-170.

• Miyake, A., & Shah, P. (Eds.). (1999). Models of working memory: Mechanisms of active maintenance and executive control. New York: Cambridge University Press.

• Newell, A. (1990). Unified theories of cognition. Cambridge, MA: Harvard University Press.

• Newell, A., & Simon, H. A. (1972). Human problem solving. Englewood Cliffs, N.J.: Prentice-Hall.

• Norman, D. A., & Shallice, T. (1986). Attention to action: Willed and automatic control of behaviour. In R. J. Davidson & G. E. Schwartz & D. Shapiro (Eds.), Consciousness and self-regulation. Advances in research and theory (Vol. 4, pp. 1-18). New York: Plenum Press.

• O'Reilly, R. C., Braver, T. S., & Cohen, J. D. (1999). A biologically based computational model of working memory. In A. Miyake & P. Shah (Eds.), Models of Working Memory (pp. 375-411). Cambridge: Cambridge University Press.

• Olton, D. S., Becker, J. T., & Handelmann, G. E. (1979). Hippocampus, space, and memory. Behavioral and Brain Science, 2, 313-365.

• Page, M. P. A., & Norris, D. (1998). The primacy model: A new model of immediate serial recall. Psychological Review, 105, 761-781.

• Paulesu, E., Frith, C. D., & Frackowiak, R. S. J. (1993). The neural correlates of the verbal component of working memory. Nature, 362, 342-345.

• Pennington, B. F., & Bishop, D. V. M. (2009). Relations among speech, language and reading disorders. Annual Review of Psychology, 60, 283-306.

• Saeki, E., & Saito, S. (2004). Effect of articulatory suppression on task-switching performance: Implications for models of working memory. Memory, 12, 257-271.

• Shallice, T. (2002). Fractionation of the supervisory system. In D. T. Stuss & R. T. Knight (Eds.), Principles of frontal lobe function. (pp. 261-277). New York: Oxford University Press.

• Turner, M. L., & Engle, R. W. (1989). Is working memory capacity task-dependent? Journal of Memory and Language, 28, 127-154.

• Vallar, G. (2006). Memory systems: The case of phonological short-term memory. A festschrift for Cognitive Neuropsychology. Cognitive Neuropsychology, 23(1), 135-155.