Working Memory(s)

Brain and Cognition 41, 1–8 (1999) Article ID brcg.1998.1092, available online at http://www.idealibrary.com on

Working Memory(s) James T. Becker Neuropsychology Research Program, Departments of Psychiatry and Neurology, University of Pittsburgh School of Medicine

and Robin G. Morris Neuropsychology Unit, Institute of Psychiatry, London, United Kingdom Working memory is variously defined as a set of linked and interacting information processing components that maintain information in a short-term store (or retrieve information into that store) for the purpose of the active manipulation of the stored items. The purpose of the this Special Issue is to present data relevant to the question of the functional organization of working memory. In this Introduction we review the two models of working memory and suggest that some of the similarities may be more apparent than real. We further suggest that the two models describe different systems that are specialized for different kinds of stimuli and for different kinds of information processing.  1999 Academic Press

INTRODUCTION

Working memory has been variously defined as a system of interacting components that maintain newly acquired and reactivated stored information, both verbal and nonverbal, and make it available for further information processing. Although the term was, perhaps, first used by Douglas (1967) in his influential paper on the functions of the hippocampal system, it was not until later that there developed two very influential, parallel, and distinct conceptual models of working memory. On the one hand there was Honig’s theory that grew out of his work with animals, and on the other Baddeley During the preparation of this paper, and of the Special Issue, J.T.B. was the recipient of a Research Scientist Development Award (K02) from the National Institute of Mental Health (MH01077). Address correspondence and reprint requests to James T. Becker, Neuropsychology Research Program, 502 Iroquois Building, 3600 Forbes Avenue, Pittsburgh PA 15213. Fax: 412383-1755. 1 0278-2626/99 $30.00 Copyright  1999 by Academic Press All rights of reproduction in any form reserved.

2

BECKER AND MORRIS

and Hitch’s working memory model, which developed from their research with humans. Both models assumed that information is maintained online for short periods of time, that material has a specific temporal context, and that rehearsal was critical for the active maintenance of the information in working memory. Beyond this there are important differences, such that the common use of the term can occasionally lead to confusion in the literature. The aim of this introduction to the Special Issue is to compare and contrast these two models and elucidate these differences. Honig Honig (1978) originally developed his concept of working memory based on data derived from his study of short-term memory in the pigeon. The key component, at its most basic, was the distinction between what were called working memory and reference memory. Working memory was information that the animal needed to remember in order to perform successfully on a single task trial. The information was critical for one trial, and one trial only, and the animal should, in fact, actively forget the information lest it interfere with performance on subsequent trials. Reference memory, by contrast, was information that was true on all trials of a behavioral task. By way of example, when a rat runs a single alternation procedure in a T-maze, the reference memory components of the task include the fact that there is food in the maze, that the animal must search to find it, and that the placement of the food on the second trial is a consequence of the placement on the first trial. The working memory component of the task is the location of the food on the first, or instruction, trial. That is, if the food is located in the lefthand arm on the first trial, the rat must remember this the next time they are placed in the start box so that they will know to enter the (correct) righthand arm. Honig’s model was used by Olton and colleagues (Olton, Becker, & Handelman, 1979) as they developed a working memory model of hippocampal system function. Using data derived from various radial arm maze procedures, they demonstrated that rats with lesions of the fimbria-fornix could retain the reference memory components of tasks (Olton & Papas, 1979), but would never relearn the working memory component (at least within the limits of extensive postoperative training). Rats were impaired regardless of whether the maze used spatial cues or visual/tactile cues to guide behavior (Olton & Feustle, 1981). This distinction between working memory and reference memory tasks can also be applied to tasks used in nonhuman primates. Thus, delayed response (e.g., Mahut & Cordeau, 1963; Mishkin & Pribram, 1954), delayed alternation (e.g., Correll & Scoville, 1965; Mahut, 1971; Mahut & Cordeau, 1963; Orbach, Milner, & Rasmussen, 1960; Pribram, Wilson, & Connors, 1962), and various match-to-sample procedures (e.g., Correll & Scoville, 1965; Drachman & Ommaya, 1964; Gaffan, 1972) all qualify as having sig-

WORKING MEMORY(S)

3

nificant working memory components (see Olton et al. (1979) for discussion). Similarly, at a procedural level of analysis, various behavioral tasks used for testing memory in humans also meet the criteria as having a working memory component using Honig’s nomenclature. For example, the N-back task (Braver, Cohen, Jonides, Smith, & Noll, 1997) requires subjects to respond when a target stimulus is repeated after some specified number of intervening items (e.g., 1-, 2-, 3-back). Information must be held online for a specified number of items, but should then be actively forgotten so as not to interfere with accuracy on subsequent trials. One feature that is common to all of these Honig-type working memory tasks is that they are recognition procedures. That is, the subjects are never required (and in the case of the animal studies are not able to) recall the items that are held in working memory. The Olton Maze has been used extensively in animal research and the function assessed is commonly referred to as ‘‘spatial working memory.’’ Baddeley and Hitch Working memory, as originally defined by Baddeley and Hitch (Baddeley, 1986; Baddeley & Hitch, 1974; Baddeley, 1992), consists of a Central Executive System and at least two slave systems—one specialized for verbal information and the other for nonverbal information. One component of the verbal subsystem is a phonological store that contains active representation of verbal material. The phonological similarity effect in verbal recall tasks— wherein span for phonologically dissimilar words is greater than span for similar words (Baddeley, 1966; Conrad & Hull, 1964)—suggests that the store is phonemically based. The active representations in the store are thought to decay rapidly unless they are refreshed by a speech-based rehearsal process (phonological loop); blocking such rehearsal impairs recall (Brown, 1958; Glanzer & Cunitz, 1966; Peterson & Peterson, 1959). Access to the phonological store can be relatively direct for auditory verbal information, since it already exists as phonological code. Visually presented material must first be recoded by a phonological loop system that, among other things, converts the visual information into phonological code (i.e., grapheme to phoneme). Interference with this recoding process, for example by requiring concurrent and repetitive articulation of a meaningless sound (e.g., hiya, hiya, hiya,. . . .), eliminates the phonological similarity effect for visual but not auditory stimuli (Baddeley, Lewis, & Vallar, 1984; Baddeley, Thomson, & Buchanan, 1975). The nonverbal slave system, originally referred to as the Visuospatial Sketch Pad due to the conceptual similarity to an artist’s sketchbook (as a repository of visual ideas to be developed or discarded later), is less well developed and much less understood and is the focus of much less research than the verbal slave system. In Logie’s excellent monograph (Logie, 1995) he details the history of the VSSP, its conceptual development, and the state

4

BECKER AND MORRIS

of affairs current to 1995. Critical to his view of visuospatial short-term memory is the distinction between visual and spatial information processing—a distinction that is frequently lost or overlooked by many neuroscientists. This distinction is critical on anatomical grounds as well as (or more importantly than) cognitive psychological grounds. Why Two Models? As we noted above, the Honig and Olton model of working memory and the Baddeley and Hitch model were developed out of distinct conceptual backgrounds and needs. They both were important for the development of thinking about short-term memory and have been very influential in subsequent experimental psychology. However, other than a superficial similarity based on the common notion of a need for a brain system to hold information briefly online for information processing, the models have very different organizations, roles within the broader framework of cognitive science, and (perhaps most important) different anatomical loci. As noted above, Olton and colleagues emphasized the role of the hippocampus in working memory and Goldman-Rakic has similarly presented data regarding the critical role of the dorsolateral prefrontal cortex in this type of memory (Goldman-Rakic, 1996a, b). By contrast, the Baddeley and Hitch model has emphasized the importance of structures typically associated with language processing (e.g., Broca’s area, supramarginal gyrus) and with visual imagery (posterior parietal cortex). The articles included in this Special Issue bear on both of these models. The first paper, by Baddeley and colleagues, examined the role of the Central Executive System (CES) in visual maintenance rehearsal. They studied the vigilance performance of elderly subjects who were tested on perceptual or memory tasks over a 40-min period. A similar design was used to study patients with Alzheimer’s Disease, who have been previously shown (Baddeley, Bressi, Della Sala, Logie, & Spinnler, 1991; Baddeley, Bressi, Della Salla, Logie, & Spinnler, 1986) to have significant impairments in CES function. For both the elderly normal and demented subjects there was a significant interaction between task (i.e., perceptual vs. memory) and delay such that only the memory-based task showed a performance decrement. These results suggest that the CES has a role in visual rehearsal processes as well as vigilance performance. D’Esposito and colleagues investigated the role of the frontal cortex in the retention of verbal information (i.e., letters) over brief delays. Specifically, they were interested in determining whether there was a functional dissociation within the lateral prefrontal cortex based on either stimulus parameters (i.e., spatial vs. object memory) or processing demands (i.e., maintenance or manipulation of information to-be-remembered). Using the powerful techniques of event-related functional magnetic resonance imaging

WORKING MEMORY(S)

5

(fMRI), they studied the functional dissociation between the dorsal and ventral regions of the lateral prefrontal cortex (PFC). Although activation was seen in both regions during delay intervals, dorsal PFC had greater activation when stimulus manipulation (i.e., reordering of stimuli) was required. This suggests that, in humans at least, the different information processing demands of the working memory task are performed by different regions of the PFC. In a study derived directly from the Honig/Olton model of working memory, Abrahams and colleagues compared and contrasted the working and reference memory of patients with unilateral temporal lobe epilepsy on object and spatial tasks. Spatial memory, both working and reference, was impaired in patients with TLE on the right. By contrast, on the object task, there was no impairment in working memory for any patient group. However, there was an impairment in reference memory in both right and left TLE patients. Volumetric analysis of structural MRI data revealed that these findings were due, in large part, to atrophy localized to the hippocampus and parahippocampal gyrus. This led the authors to conclude that there is a specialized role for the hippocampal formation on the right in spatial memory, but, contrary to Olton’s hypothesis, this involved both reference and working memory. Spatial working memory was also the topic of Morris and colleagues’ study of patients with Asperger’s syndrome relative to those with unilateral temporal or frontal lobectomies. Using their Executive Golf Task, which has performance requirements appropriate for the Honig/Olton model of working memory, they were also able to assess patients’ use of strategy. The Asperger’s syndrome patients were impaired in working memory, but not in strategy formation, as were patients with left frontal excisions. As noted by Morris and colleagues, the patients with right temporal excisions were impaired on the spatial memory component, but not on strategy formation. These data show consistency in the patterns of performance on spatial working memory tasks by patients with frontal and temporal lobe resections and suggest that Asperger’s syndrome causes a defect in accessing different types of cognitive representations guiding voluntary behavior. Finally, Becker, MacAndrew, and Fiez provide a brief summary of the functional neuroimaging data relevant to the neuroanatomical localization of the phonological store of working memory. Although not an exhaustive review, their paper makes the point that there is considerable difference between the putative loci of the phonological store depending on the nature of the activation task used. Of critical importance is the question of whether the anatomical discrepancy represents a simple difference in task demands or a fundamental difference in the information processing demands of the activation tasks and hence of the underlying information processing systems being activated. Thus, the five papers included in this Special Issue of Brain and Cognition

6

BECKER AND MORRIS

each make important independent contributions to our understanding of the information processing involved in working memory and their anatomical loci. However, they also emphasize the important distinctions between the two general model systems, which leads us to a final point of speculation. One Working Memory or Two? The papers included in this Special Issue clearly demonstrate the validity of both working memory models. That is, there is clear evidence that both a Honig and Olton and a Baddeley and Hitch type of working memory can be supported in human neuropsychology. Which of the models is best? We take the view that both memory systems are valid in human cognitive and brain systems. However, they differ in one important factor that has implications for both models. The model described by Honig and Olton has been shown to be valid and reliable in a variety of animal species and has clear evolutionary pressure in such behaviors as food foraging (Olton, 1979). By contrast, the Baddeley model has been demonstrated only in humans, in large part because of the strong language component to at least one important aspect of the model (and we intentionally do not discuss possible transition species such as the bonobo chimp). Additionally, the Baddeley and Hitch model refers to a complex and differentiated cognitive structure which appears to be holding information within a briefer time period. According to this view, the search for similarities between the models is fraught with difficulty. Instead, researchers should focus on comparing and contrasting the various aspects of the models to determine how they are expressed in animals and man. It is our position that both models are valid in the study of humans, but they are very different in their underlying information processing demands and have different neuroanatomical bases. The phylogenetically older system described by Honig and Olton holds information online for short periods of time, can perform limited processing on that information, and can actively forget the data as necessary. It seems to rely heavily on a circuit involving the DLPFC, the cingulate gyrus, and (perhaps) the hippocampal system of the temporal lobe. It does not appear highly language dependent, and it may be the case that there is little distinction between verbal and nonverbal information in this memory system, except within the temporal lobe component. In a similar vein, the Baddeley and Hitch model also holds information online, both newly acquired and previously learned. That information is available for active processing, but must be rehearsed in order to be maintained in working memory. In contrast to the Honig and Olton model, language is a critical determinant of how the information is to be remembered, with verbal information processed distinctly from nonverbal. In terms of verbal information, this system appears to function within an interconnected set of brain areas, including Broca’s Area and the supramarginal gyrus of the

WORKING MEMORY(S)

7

inferior parietal lobe. The nature of the nonverbal circuits in this model are less clear, and Logie’s monograph (Logie, 1995) is particularly helpful in terms of disentangling some of the differences in information processing between verbal and nonverbal information. Nevertheless, it remains the case that our understanding of the VSSP is considerably less well developed than that of the phonological storage system. Summary We are pleased to be able to present to the readers of Brain and Cognition this series of papers on working memory. The topic is timely, as there has been an upsurge in interest in the topic of short-term, primary, or working memory. REFERENCES Baddeley, A. D. 1996. Short-term memory for word sequences as a function of acoustic, semantic and visual similarity. Quarterly Journal of Experimental Psychology, 18, 362– 365. Baddeley, A. D. 1986. Working memory. Oxford: Claredon. Baddeley, A. D., Bressi, S., Della Sala, S., Logie, R., & Spinnler, H. 1991. The decline of Working Memory in Alzheimer’s Disease: A longitudinal study. Brain, 114, 2521–2542. Baddeley, A. D., Bressi, S., Della Salla, S., Logie, R., & Spinnler, H. 1986. Senile dementia and working memory. Quarterly Journal of Experimental Psychology, 38A, 603–618. Baddeley, A. D., & Hitch, G. 1974. Working memory. In G. H. Bower (Ed.), The psychology of learning and motivation (Vol. 8, pp. 47–90). San Diego: Academic Press. Baddeley, A. D., Lewis, V. J., & Vallar, G. 1984. Exploring the articulatory loop. Quarterly Journal of Experimental Psychology, 26, 233–252. Baddeley, A. D., Thomson, N., & Buchanan, M. 1975. Word length and the structure of shortterm memory. Journal of Verbal Learning and Verbal Behavior, 14, 575–589. Baddeley, A. L. 1992. Working memory. Science, 255, 556–559. Braver, T. S., Cohen, J. D., Jonides, J., Smith, E. E., & Noll, D. C. 1997. A parametric study of prefrontal cortex involvement in human working memory. Neuroimage, 5(1), 49–62. Brown, J. 1958. Some tests of the decay theory of immediate memory. Quarterly Journal of Experimental Psychology, 10, 12–21. Conrad, R., & Hull, A. J. 1964. Information, acoustic confusion, and memory span. British Journal of Psychology, 55, 429–432. Correll, R. E., & Scoville, W. B. 1965. Performance on delayed match following lesions of medial temporal lobe structures. Journal of Comparative and Physiological Psychology, 60, 360–367. Douglas, R. J. 1967. The hippocampus and behavior. Psychological Bulletin, 67, 416–442. Drachman, D. A., & Ommaya, A. K. 1964. Memory and the hippocampal complex. Archives of Neurology, 10, 411–425. Gaffan, D. 1972. Recognition impaired and association intact in the memory of monkeys after transection of the fornix. Journal of Comparative and Physiological Psychology, 86, 1100–1109.

8

BECKER AND MORRIS

Glanzer, M., & Cunitz, A. R. 1966. Two storage mechanisms in free recall. Journal of Verbal Learning and Verbal Behavior, 5, 351–360. Goldman-Rakic, P. S. 1996a. The prefrontal landscape: implications of functional architecture for understanding human mentation and the central executive. Philosophical Transactions of the Royal Society of London, Series B, Biological Sciences, 351(1346), 1445–1453. Goldman-Rakic, P. S. 1996b. Regional and cellular fractionation of working memory. Proceedings of the National Academy of Sciences USA, 93(24), 13473–13480. Honig, W. K. 1978. Studies of working memory in the pigeon. In S. H. Hulse, H. Fowler, & W. K. Honig (Eds.), Cognitive processes in animal behavior (pp. 211–248). Hillsdale, NJ: Erlbaum. Logie, R. H. 1995. Visuo-spatial working memory. Hove, UK: Erlbaum. Mahut, H. 1971. Spatial and object reversal learning in monkeys with partial temporal lobe ablation. Neuropsychologia, 9, 409–429. Mahut, H., & Cordeau, J. P. 1963. Spatial reversal deficit in monkeys with amygdalohippocampal ablations. Experimental Neuropsychology, 7, 426–434. Mishkin, M., & Pribram, K. H. 1954. Visual discrimination performance following partial ablations of the temporal lobe: 1. Ventral surface versus hippocampus. Journal of Comparative and Physiological Psychology, 47, 187–193. Olton, D. S. 1979. Mazes, maps, and memory. American Psychologist, 34, 583–596. Olton, D. S., Becker, J. T., & Handelman, G. E. 1979. Hippocampus, space, and memory. Behavioral and Brain Sciences, 2, 313–365. Olton, D. S., & Feustle, W. A. 1981. Hippocampal function required for nonspatial working memory. Experimental Brain Research, 41,(3–4), 380–389. Olton, D. S., & Papas, B. C. 1979. Spatial memory and hippocampal function. Neuropsychologia, 17, 669–682. Orbach, J., Milner, B., & Rasmussen, T. 1960. Learning and retention in monkeys after amygdala-hypocampua resection. Archives of Neurology, 3, 230–251. Peterson, L. R., & Peterson, M. J. 1959. Short-term retention of individual verbal items. Journal of Experimental Psychology, 58, 93–198. Pribram, K. H., Wilson, W. A., & Connors, J. 1962. Effect of lesions of the medial forebrain on alteration behavior of rhesus monkeys. Experimental Neurology, 6, 36–47.