Executive Attention, Working Memory Capacity, and a Two-Factor Theory of Cognitive Control

https://doi.org/10.1016/S0079-7421(03)44005-XGet rights and content

Publisher Summary

This chapter describes the nature of working memory capacity (WMC), and addresses the nature of WMC limitations, their effects on higher order cognitive tasks, their relationship to attention control and general fluid intelligence, and their neurological substrates. Much of work explores these issues in the context of individual differences in WMC and the cause of those individual differences. Measures of WMC are highly reliable and highly valid indicators of some construct of clear relevance to feral cognition. Macroanalytic studies have demonstrated that the construct reflected by WMC tasks has a strong relationship with gF above and beyond what these tasks share with simple span tasks. The conflict might also arise from stimulus representations of competing strength. This two-factor model fits with current thinking about the role of two brain structures: the prefrontal cortex as important to the maintenance of information in an active and easily accessible state and the anterior cingulate as important to the detection and resolution of conflict.

Introduction

This paper is about the nature of working memory capacity (WMC), and it will address the nature of WMC limitations, their effects on higher order cognitive tasks, their relationship to attention control and general fluid intelligence, and their neurological substrates. Much of our work has explored these issues in the context of individual differences in WMC and the cause of those individual differences. However, our ultimate goal is to understand WMC in its most general sense. We have used individual differences much in the way suggested by classic papers by Underwood (1975), who urged that individual differences be used as a crucible in which to test theory (see also Kosslyn et al., 2002), and Cronbach (1957) who argued that the two schools of psychology based on experimental and psychometric methods could be synergistic of one another.

We report the status of a nearly two-decade pursuit of the nature and cause of the relationship between “span” measures of WMC and complex cognition. One of the most robust and, we believe, interesting, important findings in research on working memory is that WMC span measures strongly predict a very broad range of higher-order cognitive capabilities, including language comprehension, reasoning, and even general intelligence. In due course, we describe our current thinking about the nature of these relationships and the ramifications for theories of working memory, executive attention, intelligence, and the brain mechanisms underlying those constructs.

Let us first try to place WMC in a context of general theories of immediate memory. In the 1970s and 1980s, after 20 years of work on short-term memory (STM) from the information-processing perspective, many theorists questioned the value of that work, the methods used, and the importance of the findings. Crowder (1982), in a paper pointedly entitled “The demise of short-term memory,” argued against the idea that we needed two sets of principles to explain the results of tasks measuring immediate memory and tasks clearly reflecting long-term memory (LTM). He concluded, much as his mentor Arthur Melton did in 1963, that there was insufficient evidence to support the notion of multiple memories. Evidence for a long-term recency effect similar to that found with immediate recall seemed to nullify the relationship between the recency portion of the serial position curve and STM (e.g., Baddeley 1974, Roediger 1976). Studies from the levels-of-processing perspective (e.g., Craik 1973, Hyde 1973) demonstrated that length of time in storage had little or no impact on delayed recall, contrary to quite specific predictions of the Atkinson and Shiffrin (1968) model. These studies suggested that memory was the residual of perceptual processing of an event and that orienting tasks that drove different perceptions of the event would lead to different types of codes and, in turn, differential recall. Crowder (1982) also called attention to the fact that individual differences studies had shown an inconsistent relationship between simple STM measures and such complex tasks as reading (Perfetti & Lesgold, 1977). If STM exists and is as important to higher-order cognition as early models suggested—that is, if STM is the bottleneck of the processing system—then one would expect measures of STM to correlate with performance in complex tasks such as reading comprehension.

Baddeley and Hitch (1974) questioned the simple notion of STM on these very grounds, but rather than abandon the notion of an immediate memory that is separate from LTM, they proposed a “working memory” model to supplant STM. Unlike the modal model of STM, working memory theory stressed the functional importance of an immediate-memory system that could briefly store a limited amount of information in the service of ongoing mental activity. It is quite unlikely that immediate memory evolved for the purpose of allowing an organism to store or rehearse information (such as a phone number) while doing nothing else. Instead, an adaptive immediate-memory system would allow the organism to keep task-relevant information active and accessible during the execution of complex cognitive and behavioral tasks. The “work” of immediate memory is to serve an organism's goals for action. Therefore, Baddeley and Hitch were more concerned about the interplay of storage and processing of information than about short-term storage alone. Empirically, they demonstrated that requiring concurrent memory for one or two items had virtually no impact on reasoning, sentence comprehension, and learning. Even when the concurrent memory load approached span length, performance was not devastated as should have been the case if STM was crucial to performance in these tasks. This finding led Baddeley and Hitch to propose separate components of the working memory system that traded off resources in order to handle competing storage and processing functions.

As developed by Baddeley 1986, Baddeley 1996, Baddeley 2000, the working memory model now arguably emphasizes structure over function. It consists of both speech-based and visual⧸spatial-based temporary storage systems (the phonological loop and visuo-spatial sketchpad), with associated rehearsal buffers, as well as an “episodic buffer” thought to maintain episodic information using integrated, multi-modal codes. Finally, a central executive component, analogous to Norman and Shallice's (1986) supervisory attention system, regulates the flow of thought and is responsible for implementing task goals. Much of the experimental and neuroscience research on working memory has been directed at the nature of the phonological loop and visual-spatial sketchpad (Baddeley 1986, Jonides 1997), and although these “slave systems” are easily demonstrated by a variety of lab-based experimental paradigms, their importance to real-world cognition appears to be rather limited in scope (but see Baddeley, Gathercole, & Papagno, 1998).

We take a functional approach to the study of immediate memory, which is more akin to the original Baddeley and Hitch (1974) work than to Baddeley's more recent proposals (Baddeley 1986, Baddeley 1996, Baddeley 2000, Baddeley 1999). Specifically, we emphasize the interaction of attentional and memorial processes in the working memory system, and we argue that this interaction between attention and memory is an elementary determinant of broad cognitive ability. Moreover, we endorse Cowan 1995, Cowan 1999 proposal that the coding, rehearsal, and maintenance processes of immediate memory work upon activated LTM traces, rather than retaining separate representations in domain-specific storage structures. As illustrated in our measurement model depicted in Figure 1, STM is represented as activated LTM, and this activation may be maintained or made accessible via a number of strategies or skills (e.g., chunking, phonological rehearsal) that may differ across various stimulus and⧸or response domains. Attentional or “executive” processes may also contribute to maintaining access to memory traces if routine rehearsal strategies, such as inner speech, are unavailable, unpracticed, or otherwise unhelpful for the task at hand, or if potent distractors are present in the environment. Our idea is that immediate memory, and executive attention in particular, is especially important for maintaining access to stimulus, context, and goal information in the face of interference or other sources of conflict.

By our view, then, working memory is a system of: (a) short-term “stores,” consisting of LTM traces in a variety of representational formats active above a threshold; (b) rehearsal processes and strategies for achieving and maintaining that activation; and (c) executive attention. However, when we refer to individual differences in WMC, we really mean the capability of just one element of the system: executive-attention. Thus, we assume that individual differences in WMC are not really about memory storage per se, but about executive control in maintaining goal-relevant information in a highly active, accessible state under conditions of interference or competition. In other words, we believe that WMC is critical for dealing with the effects of interference and in avoiding the effects of distraction that would capture attention away from maintenance of stimulus representations, novel productions, or less habitual response tendencies. We also believe that WMC is a domain general construct, important to complex cognitive function across all stimulus and processing domains.

To better illustrate our view, let us place WMC in a context of general cognition. We believe that much of what we need to know to function, even in the modern world, can be derived from retrieval of LTM—retrieval that is largely automatic and cue-driven in nature. Under those circumstances, WMC is not very important. Even in some putatively complex tasks such as reading, WMC is not required in all circumstances (Caplan 1999, Engle 1998). However, as we see in the following example, proactive interference can lead to problems from automatic retrieval. When the present context leads to the automatic retrieval of information, which in turn leads to an incorrect or inappropriate response in a task currently being performed, a conflict occurs between the automatically retrieved response tendency and the response tendency necessary for the current task. That conflict must often be resolved rather quickly, and so we need to have some way to keep new, novel, and important task-relevant information easily accessible.

Take a simple example obvious to every American walking the streets of London for the first time. While driving in a country such as England can lead to potentially dangerous effects of proactive interference, there are numerous cues such as the location of the steering wheel, the cars on your side of the road, etc., prompting the maintenance of the proper task goals. However, in walking the streets of England, the cues are much like those present when walking the streets of any large American city and the temptation—shall we say prepotent behavior—is to look to the left when crossing the street. This can be disastrous. So much so that London places a warning, written on the sidewalk itself, on many busy crosswalks used by tourists. This is a situation in which the highly-learned production, “if crossing street then look left,” must be countered by a new production system leading to looking to the right when crossing streets. This task seems particularly problematic when operating under a load such as reading a map or maintaining a conversation. For individuals that travel back and forth between England and America, they must keep the relevant production in active memory to avoid disaster.

Section snippets

The Measurement of WMC

The construct, WMC, is tied to a sizable number of complex span tasks that we will detail in the following text. We will describe these in some detail because measures of WMC and STM, like all other measures used by psychologists, reflect multiple constructs or influences. Simple span measures of STM (such as word, letter, and digit span) require subjects to recall short sequences of stimuli immediately after their presentation. We believe that these tasks tell us primarily about

Alternative Explanations of the WMC × Higher-Order Cognition Correlation

Measures of WMC are reliable and valid, but what are the psychological mechanisms responsible for the fact that they correlate with such a wide array of higher-level cognitive tasks? First, we need to make a methodological point here that is probably obvious but needs to be stated. We need to constantly remind ourselves about the difficulty of attributing cause-effect relationships in psychology. Further, all readers will certainly understand the difficulty of attribution about cause and effect

Macroanalytic Studies of WMC: Its Generality and Relation to Other Constructs

Our large-scale, latent-variable studies have addressed questions about WMC at the construct level. Specifically, these studies have assessed the relationship between WMC and other constructs such as STM and general fluid intelligence, and they have also tested whether WMC should be thought of as a unitary, domain-general construct or whether separate verbal and visuo-spatial WMC constructs are necessary.

Before discussing this research in more detail, however, let us briefly note the advantages

Microanalytic Studies of WMC: Its Relation to Executive Attentional Control

We have argued, based on our large-scale macroanalytic studies, that the critical element of complex WMC span tasks for higher-order cognition and general fluid abilities, whether spatial or verbal, is the domain-general ability to control attention. That conclusion was inferential at the time we proposed it (Engle, Kane, & Tuholski, 1999; Engle et al., 1999), but we had no direct evidence for support. There is now considerable data to support that thesis and we will describe it here.

A Two-Factor Theory of Executive Control

Our antisaccade findings support a two-factor model of the executive control of behavior, which also seems to explain the Stroop results we will describe in the following. We propose one factor of control to be the maintenance of the task goals in active memory, and that low-span subjects are simply less able to maintain the novel production necessary to do the task (“Look away from the flash”) in active memory. All subjects clearly knew what they were supposed to do in the task, and they could

Implementation of WMC in the Brain

We have so far discussed our dual-process view of executive control as if it was entirely new, but this is not really the case. The behavioral and neuroscience research programs of both John Duncan and Jonathan Cohen have heavily influenced our thinking about WMC and executive attention, at least insofar as they relate to the idea of goal maintenance. These views also provide suggestions for how our ideas might be mechanistically implemented in the wetware of the brain. Duncan 1993, Duncan has

Conclusions

Measures of STM such as digit and word span correlate very poorly with real-world cognitive tasks but measures of WMC correlate with a wide array of such tasks. Measures of WMC are highly reliable and highly valid indicators of some construct of clear relevance to feral cognition. Our macroanalytic studies have demonstrated that the construct reflected by WMC tasks has a strong relationship with gF above and beyond what these tasks share with simple span tasks. Further, this construct is

Acknowledgements

Address correspondence to: Randall Engle, School of Psychology, Georgia Institute of Technology, 654 Cherry Street, Atlanta, GA 30332-0170. Email requests to randall.engle@ psych.gatech.edu.

References (133)

  • M Daneman et al.

    Individual differences in comprehending and producing words in context

    Journal of Memory and Language

    (1986)
  • J Duncan et al.

    Fluid intelligence after frontal lobe lesions

    Neuropsychologia

    (1995)
  • J Duncan et al.

    Intelligence and the frontal lobe: The organization of goal-directed behavior

    Cognitive Psychology

    (1996)
  • T.S Hyde et al.

    Recall of words as a function of semantic, graphic, and syntactic orienting tasks

    Journal of Verbal Learning and Verbal Behavior

    (1973)
  • R Kail et al.

    Processing speed as a mental capacity

    Acta. Psychologica

    (1994)
  • K.A Kiewra et al.

    The relationship between information processing ability and notetaking

    Contemporary Educational Psychology

    (1988)
  • J King et al.

    Individual differences in syntactic processing: The role of working memory

    Journal of Memory and Language

    (1991)
  • P.C Kyllonen et al.

    Reasoning ability is (little more than) working-memory capacity?

    Intelligence

    (1990)
  • P.C Kyllonen et al.

    Cognitive abilities as determinants of success in acquiring logic skill

    Learning and Individual Differences

    (1990)
  • D.J Law et al.

    Training effects and working memory contributions to skill acquisition in a complex coordination task

    Learning and Individual Differences

    (1995)
  • C.M MacLeod et al.

    Interdimensional interference in the Stroop effect: Uncovering the cognitive and neural anatomy of attention

    Trends in Cognitive Sciences

    (2000)
  • P.L Ackerman et al.

    Individual differences in working memory within a nomological network of cognitive and perceptual speed abilities

    Journal of Experimental Psychology: General

    (2002)
  • D.A Allport et al.

    Shifting attentional set: Exploring the dynamic control of tasks

  • P.A Arnett et al.

    Depression in multiple sclerosis: Relationship to working memory capacity

    Neuropsychology

    (1999)
  • J.R Anderson

    The architecture of cognition

    (1983)
  • R.C Atkinson et al.

    Human memory: A proposed system and its control processes

  • A.D Baddeley

    Working memory

    (1986)
  • A.D Baddeley

    Exploring the central executive

    Quarterly Journal of Experimental Psychology

    (1996)
  • A.D Baddeley et al.

    Recency re-examined. In S. Dornic (Ed)

  • A.D Baddeley et al.

    Working memory: The multiple component model

  • A.D Baddeley et al.

    The phonological loop as a language learning device

    Psychologial Review

    (1998)
  • P Barrouillet

    Transitive inferences from set-inclusion relations and working memory

    Journal of Experimental Psychology: Learning, Memory, and Cognition

    (1996)
  • S.L Benton et al.

    Cognitive capacity differences among writers

    Journal of Educational Psychology

    (1984)
  • M.M Botvinick et al.

    Conflict monitoring and cognitive control

    Psychological Review

    (2001)
  • T.S Braver et al.

    On the control of control: The role of dompamine in regulating prefrontal function and working memory

  • L.R Brooks

    The suppression of visualization by reading

    Quarterly Journal of Experimental Psychology

    (1967)
  • M.F Bunting et al.

    Individual differences in the fan effect and working memory capacity

    (2003)
  • J Cantor et al.

    Working-memory capacity as long-term memory activation: An individual-differences approach

    Journal of Experimental Psychology, Learning, Memory, and Cognition

    (1993)
  • D Caplan et al.

    Verbal working memory and sentence comprehension

    Behavioral and Brain Sciences

    (1999)
  • P.A Carpenter et al.

    The role of working memory in language comprehension

  • R Case

    Intellectual development: Birth to adulthood

    (1985)
  • R.B Cattell

    Measuring intelligence with the Culture Fair tests

    (1973)
  • L Clarkson-Smith et al.

    The game of bridge as an exercise in working memory and reasoning

    Journal of Gerontology

    (1990)
  • A.R.A Conway et al.

    Working memory and retrieval: A resource-dependent inhibition model

    Journal of Experimental Psychology: General

    (1994)
  • A.R.A Conway et al.

    The cocktail party phenomenon revisited: The importance of working memory capacity

    Psychonomic Bulletin and Review

    (2001)
  • A.R.A Conway et al.

    The effect of memory load on negative priming: An individual differences investigation

    Memory & Cognition

    (1999)
  • A.R.A Conway et al.

    Is Spearman's g determined by speed or working memory capacity? Book review of Jensen on Intelligence-g-Factor

    Psycoloquy

    (1999)
  • N Cowan

    Attention and memory: An integrated framework

    (1995)
  • N Cowan

    An embedded-process model of working memory

  • L.J Cronbach

    The two disciplines of scientific psychology

    American Psychologist

    (1957)
  • Cited by (873)

    View all citing articles on Scopus
    View full text