2.3 Working Memory

A famous quote by Tennessee Williams goes: "Life is all memory except for the one present moment that goes by you so quick you hardly catch it going." Today, however, theoretical frameworks such as Baars' (1988) Global Workspace theory, suggest that even the immediate present involves memory. What we know about this present moment is what is stored in our working memory.

Most introductory textbooks in psychology present working memory as synonymous to short-term memory. The term 'working memory' (WM) however implies not just a temporary storage system, but also an active processing mechanism used in all cognitive tasks of any degree of complexity - thinking, reasoning, problem solving and consciousness itself. The short-term memory store was also regarded as unitary and was assumed to operate in a single, uniform fashion (Atkinson & Shiffrin, 1968). In contrast, the most developed theory of WM, first presented by Baddeley and Hitch (1974), implies a system consisting of three different components: a modality-free 'central executive', which closely resembles 'attention'; an 'articulatory' or 'phonological loop' which can hold a limited amount of phonological or speech-based information; and a 'visuo-spatial scratchpad' or 'sketchpad' which is devoted to spatial and/or visual coding (Eysenck, 1991). This WM system is also hierarchical. The central executive is at the apex of the hierarchy, and beneath it are its two 'slave' systems, namely the phonological loop and visuo-spatial scratchpad. Their salient points are described briefly:

• Phonological Loop: It comprises a passive phonological store and an articulatory rehearsal process. Verbal information presented auditorily gains direct access to the passive store, which retains information in a phonological form. Information in the store is subject to decay over time and to interference from new verbal material. Loss of information from the store can be prevented by means of sub-vocal rehearsal, and with continuing rehearsal the contents of the store could in principle be retained indefinitely. (Logie, 1996)

• Visuospatial Scratchpad: It is thought to serve a similar function for visual or spatial material and to play a role in visual-imagery tasks. There is now an argument (Logie, 1996) that the concept might be more useful if it were fractionated further into a system that provided passive storage for visual material and a companion system that retained movement sequences which might be used for some sort of visuospatial rehearsal. Visual input is thought to have obligatory access to the passive visual system. (Logie, 1996)

• Central Executive: This component of the model provides co-ordinating function for dual-task performance, and also appears to play a role in reasoning and problem solving (Logie, Gilhooly & Wynn, 1994). Baddeley suggests (1986, 1992) it is involved in the allocation of attention.

The roots of WM as a theoretical construct, however, are still firmly rooted in short-term memory, and a core characteristic of the short-term memory store, which is still relevant to all discussions on WM today, is its limited capacity. This core issue, stemming from Miller's (1956) pioneering work, is still influencing contemporary debates on WM, and consequently on the study of consciousness. Therefore the issue of WM capacity becomes particularly pertinent to this dissertation and will be discussed both from a historical perspective, and in reviewing the variety of literature, which is currently being developed.

Most WM research has been grouped and summarised into historical 'ages' by Logie (1996). He identified seven WM ages, which he describes as follows:

• WM as contemplation

• WM as primary memory

• WM as short-term memory

• WM as processor

• WM as a constraint on language comprehension

• WM as activation, attention and expertise

• WM as multiple components

The first four ages are today only relevant from a historical perspective, but the latter three models of WM are still being elaborated upon by researchers who are taking different approaches to this complex theoretical framework.

According to Logie (1996) one of the earliest recorded references to a concept similar to WM is found in the writings of the seventeenth-century British philosopher John Locke (1690 as cited in Logie, 1996). Locke had described how an idea could be retained in two ways, either by permanently storing it in a 'storehouse of ideas' called 'memory', or by keeping it temporarily 'in view' through a process called 'contemplation.' Two centuries later, a similar concept appeared again in the writings of James (1918), who coined the expression 'primary memory,' as distinct from the storehouse of 'secondary memory.' These terms were revived in an influential paper by Waugh and Norman (1965), who specified more detailed characteristics of primary memory, including its limited capacity and the fact that its information would be displaced unless it was maintained through verbal rehearsal.

These ideas were considerably extended and revised by Atkinson and Shiffrin (1968), who referred to short-term memory as a combination of storage and control processes. They focused on the concept of a short-term buffer of limited capacity, which stored and processed auditory-verbal-linguistic information. Craik and Lockhart (1972) on the other hand emphasised 'processing' rather than structure in their 'Levels of Processing Model'. The nature and level of initial processing was thought to determine the ease with which information could be later recalled. Deep levels of processing, such as semantic judgements, were suggested to lead to better retention than shallow levels of processing, such as maintenance rehearsal. Thus memory was seen as a by-product of cognitive processing rather than as a separate entity.

In this short description of Logie's (1996) first four ages of WM, the term 'working memory' as such was never used. Interestingly, however, John Richardson (1996) maintains that the expression 'working memory' was first coined in 1960 by Miller, Galanter, and Pribram, who described one particular component of the human information-processing system that was implicated in the executive control of cognition and behaviour and also served as a form of short-term storage. These contents were assumed to be directly available to conscious awareness, anticipating both modern-day theoretical accounts of WM and contemporary consciousness literature. Miller et al (1960) even anticipated neuropsychological research by suggesting that the frontal association areas, or prefrontal cortex, could offer the neural basis of WM. This has been shown today through research by Fuster (1989) and Goldman-Rakic (1993) who pinpoint to the lateral prefrontal cortex as responsible for the executive aspects of WM.

Contemporary WM theories will now be discussed.

Logie's (1996) last three WM ages describe theories which are still being researched by today's cognitive psychologists. In all three theoretical frameworks the issue of WM capacity is salient.

Language comprehension has been widely researched in view of WM capacity. In this branch of WM research, the recurring theme is of a single system responsible for temporary storage and processing. In this approach, it is proposed that a limited-capacity WM is an essential component of the processes underlying normal reading. From this it follows that individuals with a larger WM are able to handle more items for processing during reading, and achieve better semantic integration of the text as a whole (Kintsch & van Dijk, 1978; Just & Carpenter, 1980).

However, reading comprehension often turns out to be unrelated to performance in tests of short-term serial recall, which are usually taken as a measure of WM capacity. Perfetti and Goldman (1976) found no difference between skilled and less skilled readers in performance on a 'probe digit task,' where subjects were required to recall the number that immediately preceded or followed the probe in a list of digits. However, they did find that skilled readers were better than less skilled readers in performance on a 'probe discourse task,' where a similar procedure to the 'probe digit task' was set, but using words instead of digits.

Thus Perfetti and Goldman (1976), together with Daneman and Carpenter (1980) who obtained similar findings, concluded that skilled and less skilled readers differed not in terms of the capacity of WM, but in terms of the effectiveness with which they could make use of its limited capacity in processing discourse. This would leave a larger residual capacity in short-term memory to be devoted to storage.

Turner and Engle (1989), however, insisted that the findings obtained by Daneman and Carpenter (1980) were still consistent with the idea that good readers had a larger WM capacity than did poor readers, regardless of the task being performed. On this assumption, good readers just have more capacity available for both processing and storage. Consequently, Just and Carpenter (1992), on citing additional evidence that performance differences related to reading span were much more apparent if the reading comprehension task was relatively demanding, modified the earlier Daneman and Carpenter (1980) theory and proposed that the total processing capacity was determined by the total amount of activation available for maintaining elements in WM.

Engle, Cantor and Carullo (1992) proposed a similar account which focused on the relationship between WM and comprehension, called the General Capacity Theory. An important prediction of Engle and his colleagues was that individual differences in WM capacity would be important not merely in the acquisition of new knowledge, but also in the retrieval of well-learned information. If the total amount of activation available in the system is an abiding characteristic of that system, then the effects of individual differences in WM should be observed upon the retrieval of information, even if the subjects have been equated in terms of how well they know that information.

The results produced from a series of experiments, (e.g. Cantor & Engle, 1993; Rosen & Engle, 1994; Conway & Engle, 1994) did not fit Engle, Cantor and Carullo's (1992) original suppositions, however. The General Capacity Theory was shown to be shortcoming in accommodating the new data which was suggesting that individual differences in reading span and operation span reflect differences, not in memory representations, but in controlled attention (Engle, 1996).

Thus, in view of the new results, Engle (1996) suggested a new set of ideas, the Inhibition-Resource Hypothesis, which is a "hybrid that calls on concepts of individual differences in inhibition, as resulting from differences in attentional resources" (p.110). This approach seems to correspond with Baddeley's (1986) description of the central executive, but argues that limited attentional resources are necessary for the inhibition of distracting events or thoughts, as well as thoughts that are incompatible with the goals of the current task.

The approach also bridges to the sixth WM age as described by Logie (1996).

Although Engle (1996) seems to have abandoned the idea of WM as dependant on activation, the general theme of a single, flexible WM being the host to activated representation and activated procedures from long-term memory, continues to be a key feature of contemporary models of WM.

Ericsson and Kintsch (1991; 1995; Ericsson & Pennington, 1993) developed such a model of extended WM or 'long-term working memory' from research on the development and employment of cognitive expertise. There are numerous examples of impressive memory performance associated with expertise including: the recall of chess positions by expert chess players (e.g. Charness, 1976; De Groot, 1965); expert memory of meal orders in a restaurant (Ericsson & Polson, 1988), and the recall of the physical features of houses by expert burglars (Logie, Wright, & Decker, 1992). Ericsson and Kintsch (1995) argue that information in the domain of expertise is at a generally higher base level of activation, giving it privileged access to WM over domains that are outside the individual's areas of expertise. It is even suggested that WM capacity for our areas of expertise is actually much larger than WM capacity for other items.

One main path for verifying specific hypotheses about the way in which limited capacity affects cognitive functioning is to compare individuals or groups that are presumed to differ in WM capacity. This approach has been used to study group differences among, for example, younger and older adults, children of various ages, and patients of Alzheimer's disease. With regards to ageing, Stoltzfus, Hasher and Zacks (1996) state that the basic premise of capacity theories is that older adults exhibit poor performance across a wide variety of cognitive tasks because of the reduction in WM that occurs with increasing age. This viewpoint predicts that older adults will have lower average scores on measures of WM capacity. Researchers like Hartley (1986), however, have shown that this is not always the case.

Consequently the failure of the capacity model of WM to account for differences between older and younger adults' performance, led Hasher and Zacks (1988) to shift their attention from an emphasis on the capacity of WM to emphasising the contents of a WM that might or might not be limited in size. This alternative view was based on the idea that it was not the size of WM that determined performance, but how well the contents of WM (or activated memory set) represented the current task goals (Stoltzfus, Hasher & Zacks, 1996).

If the activated information in memory were closely tied to the goals of the ongoing task, then performance would be good. If on the other hand, the set of activated information included thoughts that were irrelevant to the task at hand, the simultaneously processed relevant and irrelevant trains of thought would create a situation analogous to the divided-attention task. Moreover, because the nature of on-line processing ensures that information once relevant will sometimes become irrelevant as processing proceeds, the comprehender had to work to remove that currently activated information from WM in a routine, quick and efficient manner. Consequently, Hasher and Zacks (1988) suggested that the functioning of WM was intimately tied to mechanisms of 'attentional selection.' In particular they postulated that inhibitory processes could accomplish the dual task of screening access to WM and suppressing previously relevant, but currently irrelevant, information from WM.

Recent work by Hasher, Zacks and their colleagues has indeed shown that elderly adults are impaired in one particular task that is thought to be a direct indicator of the inhibitory processes of attention (Stoltzfus, Hasher & Zacks, 1996). The task requires a simple selection response, but on some trials the item that previously served as a distracter becomes the current target. On these trials younger adults show slower response times compared with trials on which successive targets and distracters are unrelated, and this difference in responding has been attributed to the suppression of distracters, which results in making them temporarily less available as a response. Elderly adults, however, do not show this suppression effect, suggesting that they suffer from an inhibitory deficit (Hasher, Stoltzfus, Zacks, & Rypma, 1991).

Whilst the models of WM described so far focus on a general concept of WM, there has also been evolving throughout these years a theoretical model that describes a complex system of WM with its own internal structures and processes. This model which was first presented in the classic paper by Baddeley and Hitch (1974), has been generally adopted and elaborated upon by most cognitive psychologists.

Baddeley and Hitch (1974) focused upon the limited capacity of WM, as evidenced in the limited span of immediate memory. They investigated the effects of a concurrent serial-recall task on performance in reasoning, comprehension and free recall and found that a concurrent memory load of six items impaired performance in all three sorts of task. They also observed, however, that a concurrent memory load of up to three items had little or no effect on the ability to carry the executive functions. To explain this unexpected finding, they suggested that the maintenance of items from the concurrent serial-recall task could be carried out by a separate supporting or 'slave' component, hence releasing the central core of the system for performance of the criterion task. Baddeley and Hitch (1974) described this slave component as "a phonemic response buffer which is able to store a limited amount of speech-like material in the appropriate serial order." This buffer store is known today as the 'phonological loop' (Baddeley & Logie, 1992).

The success of this work urged Baddeley to explore the possibility of other systems whose role might be to support a single common central processor. In particular the idea of a peripheral memory component based upon the visual system was studied by Baddeley and Lieberman (1980) and their research tended to confirm the existence of this store, which was described as a 'visuospatial scratchpad' or 'sketchpad'.

In contrast, the operating characteristics of the central component of the WM system had hitherto been left much less clearly defined, until Baddeley (1986) adopted an existing account developed by Norman and Shallice (1982), which incorporated a limited-capacity supervisory attentional system that served as the locus of conscious control in tasks involving planning or decision making. However, according to Richardson (1996), a limiting consequence of adopting this account of the 'central executive' was that it became explicitly regarded as a general attentional resource that co-ordinated the contributions of different storage subsystems, rather than a general workspace that itself had both processing and storage properties.

Indeed, Baddeley (1993) has recently restated his position that the central executive component of the WM system does not itself involve storage, and even considered that the whole system might more accurately be termed 'working attention' rather than 'working memory'. However, Baddeley (1993) concluded that since the temporary storage of information was an absolutely essential feature of the working memory system as a whole, this warranted the retention of the term 'working memory.' Indeed the alternative models of WM put forward by Hasher and Zacks (1988), Just and Carpenter (1992), and Engle (1996), all appear to refer to a WM consisting of a central executive that operates upon the activation of long-term memory traces.

In conclusion, although the different approaches to WM described illustrate a lack of consensus amongst researchers, a convergence point is evident in recognising the concept of attention as being central to WM. The role of attention is also pivotal to Baars' (1988) Global Workspace theory and in inducing mystical states: thus the role of attention in WM is believed to be the core bridge which mediates the main hypothesis being suggested in this dissertation. The role of attention in WM will now be explored further.

The central role of attentional processes in WM has been further explored by Engle, Kane and Tuholski (in press) who have described WM as a system consisting of:

• a store in the form of long-term memory traces active above threshold;

• processes for achieving and maintaining that activation; and

• controlled attention.

In this regard WM capacity, refers to the capacity of just one element of the system: controlled attention. Therefore Engle, Kane and Tuholski (in press) do not focus on the entire WM system, but rather on the capabilities of the limited-capacity attention mechanism described by Baddeley and Hitch (1974) as the central executive. Thus, WM capacity is not really about storage or memory per se, but about the capacity for controlled, sustained attention in the face of interference or distraction (Engle, Kane & Tuholski, in press).

Engle, Kane and Tuholski (in press) argue that this attention capability is domain free and therefore individual differences in this capability reveal themselves in a wide variety of tasks. Indeed Conway and Engle (1996) emphasise that the correlation between measures of WM capacity and higher-order cognitive tasks is not a result of skill in the specific tasks - as Ericsson and Kintsch (1995) suggest with their studies on expertise - but rather of the underlying critical feature of controlled attention which is inherently different in each individual.

At this point, where the role of controlled attention in WM capacity is becoming clearer, it becomes pertinent to ask where in the brain these cognitive processes are taking place. This question is put forward in view of the research presented by Newberg et al (1997a; 1997b) where experienced meditators showed a high increase of prefrontal activity whilst in deep meditation, which was suggested to be due to the effort of controlled attention which was being put into meditating. By presenting literature which suggests that the neorobiological substrate which sustains WM capacity, or controlled attention, is the prefrontal cortex (PFC), a stronger physiological link between the processes which induce mystical experience and WM, could be speculated.

In recent years, the role of the PFC and more specifically its dorsolateral region (Brodmann areas 9, 10 and 46) has been studied extensively by various researchers. Fuster and Baur (1974), for instance, show how macaques with surgical lesions in their dorsolateral PFC, or with temporary 'lesions' from cortical cooling or electrical stimulation, show chance levels of recall on delay tasks (such as 'delayed response' and 'delayed matching-to-sample'). Deterioration in performance is found even with delays of only a few seconds, and even if the lesion is in certain parietal, hippocampal and sub-cortical regions, which are significantly interconnected with the PFC at a neural level (Koch & Fuster, 1989; Kojima & Goldman-Rakic, 1982, 1984).

Humans with PFC damage also show more forgetting in delay tasks than do healthy subjects, or control patients with comparable damage to posterior brain areas (Chorover & Col, 1966; Verin et al., 1993). Imaging studies with healthy human subjects indicate that WM tasks elicit significant increases in dorsolateral PFC activity compared to non-WM tasks. For instance, studies that required subjects to shift attention between storage and processing functions, showed PFC activation (Cohen et al, 1994).

Memory interference paradigms also suggest that the PFC is important to the control of selective attention. In proactive interference tasks, in which subjects must recall only the most recent of a series of memory lists, patients with PFC damage have difficulty in limiting recall to the target list. Frontal lesion patients recall fewer target words as the successive lists proceed, and they make more intrusion errors from prior lists than do controls (e.g. Freedman & Cermak, 1986; Van der Linden et al., 1993). Imaging data also points to the PFC as crucial to interference resistance, as dorsolateral and anterior PFC areas show increased activity in healthy subjects when they attempt to recall words under high-interference conditions but not under low-interference conditions (Uhl, Podreka, & Deecke, 1994).

Not surprisingly, patients with PFC damage also show problems with complex, everyday cognitive activities (e.g. Lezak, 1983; Shallice & Burgess, 1991). Schizophrenia, for instance, whose symptoms include disorganised speech and loss of normal functions such as language and goal-directed behaviour, has been shown to be linked to a dysfunction of the PFC (Vogeley et al., in press). Moreover schizophrenics seem to have marked deficits in WM (Goldman-Rakic, 1994), and show significantly low blood flow in prefrontal regions whilst carrying out the Wisconsin Card Sorting Test, which tests for WM (Rubin et al., 1994).

Therefore considerable evidence suggests that the dorsolateral PFC is involved in the executive or general-purpose aspects of WM, most especially controlled attention^. Experiments with brain-damaged humans have shown that patients with PFC lesions show marked difficulty with vigilance tasks. On those tasks that require subjects to maintain attention to long series of stimuli while in search of rare targets, patients with PFC damage miss more targets and commit more false alarms than do control patients (e.g. Wilkins, Shallice, & McCarthy, 1987). In visual and auditory detection tasks with healthy humans, PFC areas (along with some parietal areas) show increased activation compared to baseline conditions (Cohen et al., 1988). Furthermore, PFC activation changes are also seen during habituation. Warach et al. (1992) found decreased PFC and inferior parietal blood flow across three 14-minute rest periods in which the testing context became less novel.

On this point of habituation, it is important to note that during the advanced form of Zen, the meditator fails to show habituation but shows a consistent response to a continuous stimulus. In Yoga meditation, however, yogis who are advanced in the practice, show a complete shutting down of awareness to all external stimuli, but when the yogi is not in meditation, there is no habituation to a repetitive stimulus (Ornstein, 1972). This is consistent with the differences in philosophical outlook of Zen and yoga: yogis seek to disengage themselves as much as possible from the phenomenal world, whereas the disciples of Zen seek not so much to withdraw from the objects that surround them, as to see these objects as they are (Wulff, 1997). These findings also sustain the hypothesis of this dissertation that attention exercises like meditation, alter the normal cognitive process of controlled attention, which is an executive function of WM.

Throughout this section controlled attention as a function of working memory, and the prefrontal cortex as the seat of WM have been explored, supplementing the research presented in the previous section regarding the central roles of attention and the PFC in practices which can induce a mystical experience. Consequently it is becoming clearer that there is scope for investigating Baars' (1988) Global Workspace theory of consciousness from the perspective of an altered state of consciousness.

At this point it becomes imperative to focus on specific practitioners whose disciplines can induce a mystical experience. These will be the main focus of study in this research. The two groups identified are meditators and contemporary actors. Their disciplines, and the specific elements in them, which can lead to an altered state of consciousness, will be explored in the following section.