Professional Documents
Culture Documents
Working Memory Capacity and Its Relation To General Intelligence
Working Memory Capacity and Its Relation To General Intelligence
Early investigations of working memory capacity executive-attention control mechanism, which is mediated
(WMC) and reasoning ability suggested that WMC by portions of the prefrontal cortex.
might be the basis of Spearman’s g. However, recent The research to be reviewed draws upon methodology
work has uncovered details about the basic processes from both the differential and experimental traditions of
involved in working memory tasks, which has resulted psychology [7]. Because the Kyllonen and Christal data
in a more principled approach to task development. As provide an excellent point of departure for this review, we
a result, claims now being made about the relation begin with a discussion of the differential approach. We
between WMC and g are more cautious. A review of the then review experimental investigations into the basic
recent research reveals that WMC and g are indeed processes contributing to the performance of WMC tasks
highly related, but not identical. Furthermore, WM span and consider how those investigations inform the differ-
tasks involve an executive-control mechanism that is ential data. Finally, we review studies of the neural bases
recruited to combat interference and this ability is of WMC and g and consider how this work might further
mediated by portions of the prefrontal cortex. More constrain theories of working memory, intelligence, and
combined experimental –differential research is needed their relation. We conclude with problems that still exist
to understand better the basis of the WMC –g relation. with respect to the measurement of WMC and provide
suggestions for future research.
In the past decade, cognitive scientists have entertained
the notion that working memory capacity (WMC) is the Differential approach to the relation between WMC and g
‘Factor X’ that underlies individual differences in general The following review of individual differences studies of
intelligence (or Spearman’s g) [1– 5]. Much of the WMC and g (or general reasoning ability) is restricted to
speculation on this topic has been motivated by a series projects that used latent variable analyses (see Box 1).
of studies published in 1990 by Kyllonen and Christal [6], This criterion was adopted for two reasons. First, latent
in which they demonstrated strong correlations between variables provide a cleaner measure of the construct under
WMC and reasoning ability (r ¼ 0.80– 0.90). As impressive investigation and therefore these analyses provide the
as these correlations were, Kyllonen and Christal them- clearest picture of the true relation between WMC and g.
selves had serious reservations about the battery of WMC Second, the number of published papers containing at
tasks used in their 1990 studies: ‘We concede to a certain least one correlation between a single measure of WMC
degree of arbitrariness in creating tasks according to such and a single measure of intelligence (or reasoning) is
a broad and vague definition of their requirements, and simply too large to be adequately summarized in a brief
readers may find fault with the way we operationalized the review.
working memory factor. But without well-developed Kyllonen and Christal developed the first latent
models of information-processing requirements of the variable analysis of individual differences in WMC and
tasks, we can only proceed with what is available.’ g, and demonstrated strong correlations (near unity)
More recently, differential, experimental, and neuroi- between WMC and reasoning ability (see also [8]). Our
maging research projects have uncovered important main concern here is not with the magnitude of the
details about the information-processing requirements of correlations but rather with the characteristics of the
WMC tasks. The purpose of this paper is to review this battery of their WMC tasks.
recent progress, and in so doing, re-examine the relation Kyllonen and Christal adopted a ‘varied-content and
between WMC and g in light of these new findings. The varied-process’ approach to selecting their battery of WMC
main points to emerge from the review are: (1) the creation tasks. As mentioned above, they did this because well-
of WMC tasks is a much more principled and much less formulated theories of the task requirements of different
arbitrary endeavor than it was in 1990; (2) WMC and g are WMC tasks did not exist. It therefore made sense to select
indeed highly related, but are not the same construct; and a wide range of tasks, in terms of content and process, and
(3) the basis of the WMC– g relation is most likely to be an derive a latent variable from the battery of selected tasks.
The problem with this approach is that it is difficult to
Corresponding author: Andrew R.A. Conway (aconway@uic.edu). ascertain the key cognitive mechanisms underlying
http://tics.trends.com 1364-6613/$ - see front matter q 2003 Elsevier Ltd. All rights reserved. doi:10.1016/j.tics.2003.10.005
548 Review TRENDS in Cognitive Sciences Vol.7 No.12 December 2003
TREE IS (4 x 3) – 4 = 8 ?
TREE
PILL IS (3 x 2) + 2 = 6 ?
PILL
LOCK IS (1 x 5) – 2 = 3 ?
LOCK
??? ???
Figure I. Examples of (a) a simple span task (word span), in which serially presented material has to be recalled, and (b) a working-memory span task (operation span),
in which the subject is given simple mathematical operations to solve at the same time as words to be recalled later.
different experimental contexts. Although it is not ideal to WMC. And importantly, the slopes did not differ at all
categorize a continuous variable in this fashion, the between span groups in the no-overlap condition. Fur-
approach has been productive for conducting exploratory thermore, for high WMC subjects, the slopes in the overlap
research to identify experimental manipulations that do and no-overlap conditions were equal, suggesting that
and do not interact with WMC (in fact, this is exactly the high-WMC individuals did not allow the overlap manipu-
type of treatment by organism approach endorsed by lation to impair retrieval, whereas low-WMC individuals
Cronbach [7]). were negatively affected.
The general conclusion from investigations that have The conclusion derived from these empirical findings
adopted this approach is that WMC is related to was that WMC is related to an executive attention ability,
performance in situations in which an executive attention which supports the active maintenance of goal-relevant
control mechanism is needed to combat some form of information in the face of interference. This attention
salient interference, be it proactive interference, response ability is most critical in interference-rich conditions
competition, or habitual but inappropriate responses. For because correct responding cannot be achieved via
example, WMC is related to the speed and accuracy of automatic spreading activation among memory represen-
retrieval of information from long-term memory (LTM), tations, or via habitual responding. In interference-rich
but only when a level of response competition is inherent conditions, such automatic activation or habitual respond-
in the task [16]. In a variant of the classic Sternberg ing leads to errors and, therefore, must be suppressed.
memory-scanning paradigm, subjects committed to mem- This general conclusion is supported by other LTM studies
orydifferent sets of letters of varying set size (e.g. set 2 ¼ R, W; finding WMC to predict encoding and retrieval success
set 6 ¼ Q, T, P, S, F, K). In one condition, there was no under conditions of proactive interference [17– 20].
overlap in set membership; that is, if a letter was a member Similar conclusions have been drawn from research on
of one set, it could not be a member of another set. In ‘low-level’ attention tasks that require little by the way of
another condition, each letter was a member of two memory, per se, such as dichotic listening [21], Stroop
different sets. WMC was related to retrieval speed and color-naming [22], visual orienting [23], and negative
accuracy in the overlap condition only. Specifically, the priming [24]. As an example, we will consider just one of
slope of the reaction-time/set-size function was steeper for these studies here – visual orienting. The pro-saccade task
individuals with lower WMC than for those with higher is perhaps one of the simplest cognitive tasks available. In
http://tics.trends.com
550 Review TRENDS in Cognitive Sciences Vol.7 No.12 December 2003
fact, it can hardly be described as ‘cognitive’ at all. The measured by fMRI [29]. In the study, 48 subjects
subject’s task is to orient to a visual cue presented in their performed Ravens Advanced Progressive Matrices
periphery, typically a flashing stimulus. In some cases (RAPM), a measure of g, outside the fMRI scanner and
subjects are also required to detect a target that is then performed the n-back task, a measure of WMC, inside
subsequently presented in the same location as the cue the scanner. The n-back task is a continuous performance
(in the simplest case subjects only make an eye-movement, test in which individual stimuli (typically letters or words)
in which case performance is monitored using an eye- are presented in rapid succession and the subject must
tracker). The anti-saccade version of the task is much more indicate whether the current stimulus matches the one
difficult. Here, subjects must make an eye-movement in presented n-back in the stream, and n varies, typically
the opposite direction of the visual cue. This is difficult from 1 – 3. N-back is thought to tap WMC because it
because orienting to a flashing stimulus in the periphery is involves not only the storage of the last n presented
a reflexive response. Thus, in the anti-saccade task the stimuli, but also the continuous updating of n stimuli,
habitual orienting response must be suppressed to make which involves the deletion of stimuli presented more
the correct, controlled eye-movement in the opposite than n back in the stream. Moreover, some trials within
direction. the n-back task are thought to require more executive
Consistent with the memory-retrieval data discussed attention than others. For example, in a 3-back task, the
earlier, WMC is related to anti-saccade but not to pro- second B in B-R-B-X is referred to as a ‘lure’ trial because it
saccade performance: Individuals with greater WMC are matches a recently presented stimulus but does not match
less likely to make erroneous saccades, and are faster to the one presented three back. Indeed, in the 3-back task,
make correct saccades and to correct erroneous ones, than control foils were rejected more accurately than lure foils
are individuals with lesser WMC. In the pro-saccade task, (accuracy ¼ 96% versus 75%, respectively) [29].
individuals with high and low WMC are just as likely, and More importantly, significant correlations were
just as fast, to make the correct orienting response and to observed between RAPM, n-back, and activity in DLPFC
detect the target (if required). and ACC. In fact, the correlation between RAPM and
In conclusion, investigations into the basic processes event-related activity to lure trials was particularly
that contribute to WM span tasks suggest that they striking (r ¼ 0.54) and remained significant even when
critically tap an executive attention-control process performance on non-lure trials was partialled out. Finally,
recruited in situations where an inappropriate yet highly the correlation between lure performance and RAPM was
accessible response must be selected against in favor of the almost completely accounted for by lure-trial activity in
appropriate response. DLPFC. That is, the percent-signal-change in DLPFC
associated with lure trials accounted for 92% of the
Neuroimaging studies of WMC and g covariance between lure performance and RAPM,
The distinction between tasks that require storage versus suggesting that the WMC/executive attention/g relation
those that require storage and some form of processing can is mediated by activity in DLPFC.
also be found in the neuroimaging literature [12]. For
instance, several studies have shown that storage-only Future issues: the current range of WMC tasks is still too
tasks reveal activation primarily in areas related to the wide
content of the to-be-remembered material (e.g. Broca’s It is clear that the creation and selection of WMC tasks is
area for verbal material, right-hemisphere pre-motor more principled today than it was in 1990. However, the
cortex for spatial material) [12], whereas storage-plus- fact remains that a wide range of tasks are still being used
processing tasks reveal content-specific activation but also to measure WMC and it is not clear how all these tasks
domain-free activation in areas such as dorsolateral relate to one another or to g. Here we consider tasks other
prefrontal cortex (DLPFC) and anterior cingulate (ACC) than WM span and call for future research that addresses
[12,25,26]. Indeed, recent theories and models of prefron- (1) the basic processes involved in these tasks, and (2) the
tal function suggest that WMC, executive attention, and g relation between these tasks, WM span tasks, and g.
are all highly related constructs, all heavily reliant upon One problem with WM span tasks is that they clearly
the DLPFC [27,28]. require processes above and beyond WMC or executive
Indirect evidence in support of the notion that WMC, control. For example, operation span recruits mathemat-
executive attention, and g share a common neurological ical ability [30], reading span involves verbal ability [31],
basis comes from localization studies. As mentioned above, and strategy-training has been shown to improve scores in
investigations into the brain regions invoked by WMC WM span tasks [32] (however, differential skill and
tasks, for example, suggest that DLPFC and ACC are strategy-use does not affect the correlation between WM
involved. Importantly, the same conclusion has been span and higher-order cognition [30,33,34]). Although no
drawn from the literatures on localization of executive task is process-pure, it does seem possible to create
attention and g [28]. However, this evidence remains measures of WMC that are less contaminated by other
indirect because little attempt has been made to determine cognitive skills and strategies than WM span tasks. In a
whether variation in behavioral performance on tasks that recent review, Cowan [35] argued that the limitation on
purportedly tap these constructs is mediated by variation immediate memory and attention can be estimated if the
in brain activity within these regions. stimuli are presented in a manner that eliminates the
A recent study has examined the extent to which possibility of strategy recruitment, such as grouping or
behavioral correlations are mediated by brain activity as rehearsal. According to this perspective, the signature of
http://tics.trends.com
Review TRENDS in Cognitive Sciences Vol.7 No.12 December 2003 551
WM tasks is not the extent to which they require storage conducted to determine the relationship among differ-
and processing but rather how well they prevent the ent WMC tasks and between these tasks and g.
recruitment of domain-specific skills and strategies.
For example, consider the visual-array comparison task References
[36]. In the task, an array of colored squares is presented 1 Kyllonen, P.C. (1996) Is working memory capacity Spearman’s g? In
briefly, followed by an inter-stimulus interval, followed by Human Abilities: their Nature and Measurement (Dennis, I. and
a second array that is identical or similar to the first. One Tapsfield, P., eds), pp. 49 – 75, Erlbaum
square in the second array is marked by a circle, and if the 2 Jensen, A.R. (1998) The g Factor: the Science of Mental Ability, Praeger
3 Conway, A.R.A. et al. (2002) A latent variable analysis of working
two arrays were different, the difference was in the color of memory capacity, short term memory capacity, processing speed, and
the encircled square. The key manipulation is the number general fluid intelligence. Intelligence 30, 163– 183
of squares presented in the arrays. When the array is small 4 Deary, I.J. (2001) Human intelligence differences: towards a combined
(i.e. 4 or fewer), same/different judgments are very experimental-differential approach. Trends Cogn. Sci. 5, 164 – 170
accurate. However, accuracy deteriorates when the array 5 Plomin, R. and Spinath, F.M. (2002) Genetics and general cognitive
ability (g). Trends Cogn. Sci. 6, 169 – 176
size is greater than 4, suggesting that approximately 4
6 Kyllonen, P.C. and Christal, R.E. (1990) Reasoning ability is (little
independent locations can be actively maintained in more than) working-memory capacity?!. Intelligence 14, 389– 433
immediate memory without the aid of a grouping or 7 Cronbach, L.J. (1957) The two disciplines of scientific psychology. Am.
rehearsal strategy and, therefore, that this task provides Psychol. 12, 671 – 684
an estimate of WMC. 8 Kyllonen, P.C. (1993) Aptitude testing inspired by information
Cowan [35] reviewed several tasks with similar processing: a test of the four-sources model. J. Gen. Psychol. 120,
375 – 405
characteristics to the visual-array comparison task and 9 Baddeley, A.D. and Hitch, G. (1974) Working memory. In The
found a consistent capacity estimate of around 4 ^ 1. Psychology of Learning and Motivation (Vol. 8) (Bower, G.A., ed.),
Thus, it is possible that these tasks tap the limit that pp. 47– 89, Academic Press
others have referred to as WMC. However, it is not clear 10 Daneman, M. and Carpenter, P.A. (1980) Individual differences in
how these tasks relate to other WM tasks, such as WM working memory and reading. J. Verbal Learn. Verbal Behav. 19,
450– 466
span tasks, how stable and robust individual differences
11 Engle, R.W. et al. (1999) Working memory, short-term memory and
are in these tasks, or how these tasks relate to measures of general fluid intelligence: a latent variable approach. J. Exp. Psychol.
complex cognitive ability, such as g. Gen. 128, 309 – 331
Another class of tasks used to measure WMC is the 12 Smith, E.E. and Jonides, J. (1999) Storage and executive processes in
n-back task [37 – 39]. As mentioned above, n-back tasks the frontal lobes. Science 283, 1657– 1661
are widely used in neuroimaging studies, being preferred 13 Fiez, J.A. (2001) Bridging the gap between neuroimaging and
neuropsychology: using working memory as a case-study. J. Clin.
over WM span tasks in the fMRI environment because Exp. Neuropsychol. 23, 19 – 31
their presentation and response requirements are less 14 Süb, H.M. et al. (2002) Working-memory capacity explains reasoning
complex. Most n-back tasks, like the one described earlier, ability – and a little bit more. Intelligence 30, 261 – 288
not only challenge memory maintenance, but also contain 15 Ackerman, P.L. et al. (2002) Individual differences in working memory
lure trials, which are stimulus matches n 2 1 or n þ 1 within a nomological network of cognitive and perceptual speed
abilities. J. Exp. Psychol. Gen. 131, 567– 589
back. It is particularly difficult to reject these lure trials
16 Conway, A.R.A. and Engle, R.W. (1994) Working memory and retrieval:
because familiarity comes into conflict with the task goal a resource-dependent inhibition model. J. Exp. Psychol. Gen. 123,
[29]. One intriguing possibility is that performance on lure 354– 373
trials is related to WM span performance (given the 17 Cantor, J. and Engle, R.W. (1993) Working-memory capacity as long-
findings that span is particularly important under con- term memory activation: an individual-differences approach. J. Exp.
ditions of interference). However, at this point little work Psychol. Learn. Mem. Cogn. 19, 1101– 1114
18 Lustig, C. et al. (2001) Working memory span and the role of proactive
has been done to examine the relation between n-back, interference. J. Exp. Psychol. Gen. 130, 199 – 207
other measures of WMC, and higher-order cognitive tasks. 19 Kane, M.J. and Engle, R.W. (2000) Working memory capacity,
proactive interference, and divided attention: limits on long-term
Conclusion memory retrieval. J. Exp. Psychol. Learn. Mem. Cogn. 26, 336– 358
In summary, several recent latent variable analyses 20 Rosen, V.M. and Engle, R.W. (1998) Working memory capacity and
suppression. J. Mem. Lang. 39, 418– 436
suggest that WMC accounts for at least one-third and
21 Conway, A.R.A. et al. (2001) The cocktail party phenomenon revisited:
perhaps as much as one-half of the variance in g. What the importance of working memory capacity. Psychon. Bull. Rev. 8,
seems to be important about WM span tasks is that they 331– 335
require the active maintenance of information in the face 22 Kane, M.J. and Engle, R.W. (2003) Working-memory capacity and the
of concurrent processing and interference and therefore control of attention: the contributions of goal neglect, response
recruit an executive attention-control mechanism to competition, and task set to Stroop interference. J. Exp. Psychol.
Gen. 132, 47 – 70
combat interference. Furthermore, this ability seems to 23 Kane, M.J. et al. (2001) A controlled-attention view of working-
be mediated by portions of the prefrontal cortex. memory capacity. J. Exp. Psychol. Gen. 130, 169– 183
Although progress has been substantial in recent 24 Conway, A.R.A. et al. (1999) The effect of memory load on negative
years, much more work remains to be done, particu- priming: an individual differences investigation. Mem. Cogn. 27,
larly with respect to the measurement of WMC. 1042– 1050
25 Fiez, J.A. et al. (1996) A positron emission tomography study of the
Specifically, experimental task analyses need to be
short-term maintenance of verbal information. J. Neurosci. 16,
conducted on all the different types of WMC tasks that 808– 822
are currently being used in the field. And, latent 26 Jonides, J. et al. (1998) Inhibition in verbal working memory revealed
variable analyses of different WMC tasks need to be by brain activation. Proc. Natl. Acad. Sci. U. S. A. 95, 8410 – 8413
http://tics.trends.com
552 Review TRENDS in Cognitive Sciences Vol.7 No.12 December 2003
27 Duncan, J. (1995) Attention, intelligence, and the frontal lobes. In The reconsideration of mental storage capacity. Behav. Brain Sci. 24,
Cognitive Neurosciences (Gazzaniga, M., ed.), pp. 721– 733, MIT Press 87 – 185
28 Kane, M.J. and Engle, R.W. (2002) The role of prefrontal cortex in 36 Luck, S.J. and Vogel, E.K. (1997) The capacity of visual working
working-memory capacity, executive attention, and general fluid memory for features and conjunctions. Nature 390, 279 – 281
intelligence: an individual-differences perspective. Psychon. Bull. 37 Kirchner, W.K. (1958) Age differences in short-term retention of
Rev. 9, 637 – 671 rapidly changing information. J. Exp. Psychol. 55, 352– 358
29 Gray, J.R. et al. (2003) Neural mechanisms of general fluid 38 Mackworth, J.F. (1959) Paced memorizing in a continuous task. J. Exp.
intelligence. Nat. Neurosci. 6, 316 – 322 Psychol. 58, 206– 211
30 Conway, A.R.A. and Engle, R.W. (1996) Individual differences in 39 Jonides, J. et al. (1997) Verbal working memory load affects
working memory capacity: more evidence for a general capacity theory. regional brain activation as measured by PET. J. Cogn. Neurosci.
Memory 4, 577 – 590 9, 462 – 475
31 Copeland, D.E. and Radvansky, G.A. (2001) Phonological similarity in 40 Bollen, K.A. (1989) Structural Equations with Latent Variables, Wiley
working memory. Mem. Cogn. 29, 774– 776 41 Kline, R.B. (1998) Principles and Practices of Structural Equation
32 McNamara, D.S. and Scott, J.L. (2001) Working memory capacity and Modeling, Guilford Press
strategy use. Mem. Cogn. 29, 10– 17 42 Turrner, M.L. and Engle, R.W. (1989) Is working memory capacity task
33 Engle, R.W. et al. (1992) Individual differences in working memory and dependent? J. Mem. Lang. 28, 127 – 154
comprehension: a test of four hypotheses. J. Exp. Psychol. Learn. Mem. 43 Case, R. et al. (1982) Operational efficiency and the growth of short-
Cogn. 8, 972 – 992 term memory span. J. Exp. Child Psychol. 33, 386– 404
34 Turley-Ames, K.J. and Whitfield, M.M. Strategy training and working 44 Shah, P. and Miyake, A. (1996) The separability of working memory
memory task performance. J. Mem. Lang. (in press) resources for spatial thinking and language processing: an individual
35 Cowan, N. (2001) The magical number 4 in short-term memory: a differences approach. J. Exp. Psychol. Gen. 125, 4 – 27
Our experienced in-house team is under the guidance of a panel of experts from across the life sciences
who offer suggestions and advice to ensure that we have high calibre authors and have spotted
the ‘hot’ papers.
Visit the Research Update daily at http://update.bmn.com and sign up for email alerts to make sure you don’t miss a thing.
This is your chance to have your opinion read by the life science community, if you would like to contribute, contact us at
research.update@elsevier.com
http://tics.trends.com