See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/7797610

What makes working memory spans so predictive of high-level

cognition?

Article in Psychonomic Bulletin & Review · March 2005

DOI: 10.3758/BF03196363 · Source: PubMed

CITATIONS READS

167 190

3 authors, including:

Pierre Barrouillet Valérie Camos

University of Geneva Université de Fribourg
173 PUBLICATIONS 8,869 CITATIONS 213 PUBLICATIONS 6,567 CITATIONS

SEE PROFILE SEE PROFILE

All content following this page was uploaded by Valérie Camos on 23 May 2014.

The user has requested enhancement of the downloaded file.

Psychonomic Bulletin & Review
2005, 12 (1), 165-170

What makes working memory spans so predictive

of high-level cognition?
RAPHAËLLE LÉPINE and PIERRE BARROUILLET
Université de Bourgogne, Dijon, France
and
VALÉRIE CAMOS
Université René Descartes–Paris 5, Paris, France

Working memory (WM) span tasks involving a complex activity performed concurrently with item
retention have proven to be good predictors of high-level cognitive performance. The present study
demonstrates that replacing these complex self-paced activities with simpler but computer-paced pro-
cesses, such as reading successive letters, yields more predictive WM span measures. This finding sug-
gests that WM span tasks evaluate a fundamental capacity that underpins complex as well as elemen-
tary cognitive processes. Moreover, the higher predictive power of computer-paced WM span tasks
suggests that strategic factors do not contribute to the relationship between WM spans and high-level
cognition.

The theoretical construct of working memory has fluid intelligence (Engle, Tuholski, Laughlin, & Conway,
played an increasingly important role in accounting for 1999). In order to account for these relationships, it has
cognition, and especially for high-level cognition. Work- been proposed that WM measures assess a fundamental
ing memory (WM) usually refers to a cognitive system de- capacity required by complex activities, which is conceived
voted to the simultaneous maintenance and treatment of of as a capacity to control attention (Engle, Kane, & Tuhol-
information that is involved in the planning, coordination, ski, 1999) or to supervise and coordinate multiple-system
and control of high-level cognitive processes (Baddeley functioning (Baddeley, 1990). Consequently, the activi-
& Hitch, 1974; Daneman & Carpenter, 1980). This cen- ties included as processing components in WM span tasks
tral role in cognition led many authors to propose span are usually selected from those thought to require a high
tasks that aimed at assessing WM capacity in individuals. level of executive control (e.g., problem solving, reading
In contrast to traditional short-term memory span tasks, comprehension, reasoning, mental calculation). The un-
which require simple maintenance and recall of informa- derlying idea is that more controlled and complex activi-
tion such as digits, letters, or words, WM span tasks in- ties provide better WM span measures because complex
volve a processing component in addition to the mainte- activities tap the limited pool of cognitive resources suf-
nance of to-be-recalled items. For example, individuals are ficiently to disrupt maintenance and permit an accurate
asked to read sentences while maintaining their final words measure of WM capacity.
(reading span; Daneman & Carpenter, 1980) or to solve However, is the complexity of the processing compo-
arithmetic problems while maintaining words (operation nent of WM span tasks necessary to disrupt maintenance
span; Turner & Engle, 1989). and accurately assess WM capacity? Barrouillet, Bernar-
These WM span measures proved to reliably predict, din, and Camos (2004) have recently shown that very
better than short-term memory spans, performance in a simple activities included as processing components in
wide range of complex activities such as reading com- WM span tasks have an equally detrimental effect on re-
prehension (Daneman & Carpenter, 1980; Daneman & call as do complex activities, provided that they are not
Merikle, 1996), complex learning (Shute, 1991), and rea- self-paced but computer-paced. Remembering letters
soning (Barrouillet, 1996; Kyllonen & Christal, 1990). while solving running operations such as adding or sub-
Moreover, it has been demonstrated that, in contrast to tracting 1 to or from digits, or even merely reading dig-
short-term memory, WM shows a strong connection to its successively presented on a screen at a fast pace, was
very difficult, and adult participants exhibited WM spans
lower than 3 in these tasks. The authors accounted for this
This work was presented at the Thirteenth Conference of the European effect by proposing that simple but time-constrained ac-
Society of Cognitive Psychology in Granada, Spain, September 2003. The tivities capture a sufficient amount of attention to disrupt
authors thank Karine Petetin for collecting the data. Correspondence con-
cerning this article should be addressed to R. Lépine, Université de Bour-
the maintenance of items to be recalled.
gogne, LEAD-CNRS 5022, Pôle AAFE, Esplanade Erasme, B.P. 26513- Nonetheless, the value of the WM span tasks is to pre-
21065, Dijon, France (e-mail: [email protected]). dict performance on complex cognitive activities. Al-

165 Copyright 2005 Psychonomic Society, Inc.

166 LÉPINE, BARROUILLET, AND CAMOS

though the new tasks designed by Barrouillet et al. (2004) operations [i.e., addition (subtraction) of 1 to (from) num-
have the expected detrimental effect on recall, it remains bers from 1 to 9], in contrast to the complex equations of
to be established whether they assess the same capacity the operation span task.
and have the same predictive power as traditional WM
span tasks. Indeed, two alternative hypotheses can be put METHOD
forward to account for the well-known relationship be-
tween WM span and performance in complex cognitive Participants
activities. According to the first hypothesis, traditional Ninety-three French sixth graders attending a suburban public
WM tasks would evaluate some general cognitive capac- middle school (43 girls and 50 boys; mean age ⫽ 11 years,
ity, described by Barrouillet et al. as an attentional ca- 11 months; SD ⫽ 8 months) participated as volunteers.
pacity, that is involved in any cognitive process requiring Materials and Procedure
access and maintenance of items of knowledge. This lim- Scholastic evaluation. The scholastic evaluation was provided by
ited attentional capacity would thus underpin and con- individual scores from the national academic achievement test that
strain each step of the complex activities involved in each French sixth grader takes at the beginning of the academic
high-level cognition. According to this view, complexity year. This test yields compound scores in literacy and mathematics
is an unnecessary characteristic of the processing com- as well as a global scholastic score in terms of percentage of suc-
cess. The literacy assessment involves language comprehension,
ponent of the WM tasks because what is required is an language tools, and written production. The mathematics assess-
activity that captures attention. Moreover, the temporal ment involves geometry, number knowledge and writing of num-
constraints of Barrouillet et al.’s computer-paced WM bers, arithmetic problem solving, word problem solving, and com-
tasks hamper the use of possible strategies for coping prehension of mathematical tools and concepts.2
with the specific demands of the dual-task paradigm— Traditional working memory span tasks. In the reading span
strategies that are allowed by most of the traditional self- task, the participants were asked to read aloud sentences containing
4–11 words (M ⫽ 6.9) while maintaining to-be-remembered num-
paced WM span tasks (Baddeley, Logie, Nimmo-Smith, &
bers. These sentences were presented in sets of increasing length
Brereton, 1985; Case, Kurland, & Goldberg, 1982; (from 1 to 7) with three sets of each length. Half of the 84 sentences
Daneman & Carpenter, 1980; Turner & Engle, 1989).1 were true, and the other half were false (e.g., “A cow lays eggs”).
According to the fundamental capacity hypothesis, these The participants were asked to evaluate the truthfulness of each
strategies may produce biased measures of the funda- sentence by pressing labeled keys on the keyboard. The to-be-
mental capacity. Thus, this hypothesis predicts that the remembered numbers were randomly drawn from between 1 and
spans provided by the new tasks will be more predictive 16, with the exception of 14 because it is bisyllabic in French; one
number was presented before each sentence. Each set of sentences
of performance in high-level cognition than the traditional started with a 1-sec “ready” signal. After 500 msec, a first number
tasks are. An alternative hypothesis would be that the tra- to be remembered was displayed on the screen for 1,500 msec and
ditional tasks help to evaluate some high-level executive was followed by a sentence to be read after a delay of 500 msec. The
capability that is essential both for the complex pro- children read the sentence at their own pace. When they responded
cesses that they involve and for the use of strategies for on the keyboard, the sentence was deleted and, after a 500-msec
dealing with the requirements of the dual task. High- delay, a 1,500-msec display of a new number was triggered. This
was followed by a new sentence to be read after a 500-msec delay,
WM-span individuals would be those who are better able and so on. At the end of the series, the word RAPPEL (“recall”) was
to strategically plan and monitor their activity in com- displayed on the screen and the participant had to recall the pre-
plex situations, achieving better performance in WM sented numbers in the correct order.
dual tasks and in higher level cognition. According to The operation span task had the same structure as the reading
this account, WM spans are predictive because the tradi- span task, except that sentences were replaced by equations to be
tional tasks mimic high-level cognitive activities. This hy- verified and numbers were replaced by consonants to be remem-
bered; each equation to be verified was preceded by a consonant to
pothesis predicts that the traditional spans will have a be remembered. The equations were three-operand additive prob-
greater predictive value because the new tasks involve fewer lems (e.g., 6 ⫹ 7 ⫹ 2 ⫽ 13?), the proposed result being correct for
strategic factors and only elementary processes. half of the problems and false for the others. The participants had
We tested these two hypotheses by comparing the cor- to read each equation aloud before solving it. All consonants were
relation between the academic achievements of 11-year- used except W, which is trisyllabic in French.
old children on the one hand, and WM spans evaluated The new working memory span tasks. The new tasks had ex-
either by traditional or by new WM span tasks on the actly the same structure as the traditional tasks except that the pro-
cessing component was computer paced and time constrained. As
other. The traditional WM tasks were a reading span task, in the reading span task described above, in the reading letter span
inspired by Baddeley et al. (1985), in which participants task the participants were presented with numbers to be remem-
had to read and understand sentences, and an operation bered. Instead of sentences, sets of 4–6 letters, to be read aloud, ap-
span task, inspired by Turner and Engle (1989), in which peared successively on the screen after each number to be remem-
participants verified complex equations. The new tasks bered. These letters were randomly drawn from the 26 letters of the
were (1) a reading letter span task in which, while main- alphabet, provided that a given set contained neither the same let-
ter twice nor two successive letters of the alphabet. Each letter re-
taining digits, the children had only to read letters in- mained on the screen for 1,000 msec after a delay of 350 msec.
stead of understanding sentences, and (2) the continuous The continuous operation span task was similar to the traditional
operation span task of Barrouillet et al. (2004), in which, operation span task. The participants were presented with series of
while maintaining letters, the children performed simple letters to be remembered, but the equations were replaced with run-
 WORKING MEMORY SPANS 167

ning operations. After each to-be-remembered letter, a digit, called tradeoff between spans and verification performance in
the root, appeared for 1,500 msec, followed by a series of two, three, the traditional span tasks (r ⫽ .02, p ⫽ .86 for the reading
or four successive screens displaying either “⫹1” or “ –1.” The sets span task and r ⫽ .11, p ⫽ .30 for the operation span task).
of operations were randomly constructed with the constraint that
We computed the internal consistency reliability for each
no final or intermediate result was below 1 or above 9 (e.g.,
8/⫹1/⫹1). Each operation (i.e., ⫹1 or ⫺1) remained on the screen WM span task by considering the series to be recalled as
for 1,500 msec after a 500-msec delay. The participants were asked items scored as either 1 or 0, for correct or incorrect recall,
to read all the stimuli and to perform the operations aloud (e.g., for respectively. Cronbach’s (1951) alpha coefficients were
the series 5/⫹1/⫺1, the participants were to utter “five, plus one, .65 and .70 for the reading and operation spans, respec-
six, minus one, five”). tively, in comparison with .59 and .65 for the reading let-
For each WM task (new or traditional), two one-item and two ter and continuous operation spans, respectively. The in-
two-item training sets preceded the experimental series. For all
tasks, the children were presented with increasingly longer sets be- ternal consistency reliability was thus almost equivalent
ginning with length 1 and continuing until they failed to recall the for the four tasks.
memory items of all three sets at a particular length. Testing was
terminated at this point. Each correctly recalled set counted as one Correlation Analyses
third, and the thirds were added up to provide a span score (Kemps, All of the WM spans were highly correlated with each
De Rammelaere, & Desmet, 2000; Smith & Scholey, 1992). For ex-
ample, the correct recall of all the sets of one, two, and three items, other, particularly the new and traditional spans (see
of two sets of four items, and of one set of five items resulted in a Table 1). These correlations were higher when the pro-
span of 4 [(3 ⫹ 3 ⫹ 3 ⫹ 2 ⫹ 1) ⫻ 1/3]. cessing components of both new and traditional tasks
The four WM tasks were administered in the middle of the aca- were from the same domain (r ⫽ .551 between the read-
demic semester (March) in two sessions of 30 min each. In the first ing and the reading letter spans, r ⫽ .678 between the op-
session, the children were presented with the reading span task and eration and the continuous operation spans), but they
then the reading letter span task. In the second session, the opera-
tion span and the continuous operation span tasks were presented,
reached a high level of significance even when the new
in that order. and traditional tasks concerned different domains (r ⫽
.492 between the reading and the continuous operation
RESULTS spans, r ⫽ .460 between the operation and the reading let-
ter spans). These first results suggest that both types of
Descriptive Statistics span task aid in the evaluation of a common construct
The mean scholastic scores (71%, SD ⫽ 17; 73%, plus some specific capabilities related to the type of pro-
SD ⫽ 17; and 70%, SD ⫽ 18 of correct responses for the cessing component. A principal component analysis con-
global, literacy, and mathematics scales, respectively) firmed the existence of a factor for the commonality in all
were close to the mean national scores observed in com- four tasks, which accounted for 66% of the total variance,
parable middle schools (referenced as “collèges publics as well as a second factor that distinguished between
hors ZEP/REP” on the Web pages of the French Ministry “reading” and “arithmetic operation” tasks (15% of the
of Education; 68%, SD ⫽ 19; 68%, SD ⫽ 17; and 67%, variance) and a third that distinguished between new and
SD ⫽ 20 for global, literacy, and mathematics scales, re- traditional tasks (11.5% of the variance; see Table 2).
spectively), although our sample was slightly above the More important, and as previously reported in the lit-
national mean. The mean spans from the new WM tasks erature, both traditional spans were significantly corre-
(2.81, SD ⫽ 0.74 and 2.32, SD ⫽ 0.84 for the reading lated with the literacy score (rs ⫽ .296 and .316 for the
letter and the continuous operation spans, respectively) reading and the operation spans, respectively), the math-
did not significantly differ from the traditional mean ematics score (rs ⫽ .327 and .353), and the global score
spans (2.77, SD ⫽ 0.85 and 2.46, SD ⫽ 0.92 for the read- (rs ⫽ .335 and .359). The compound traditional WM
ing and the operation spans, respectively). There was no score obtained by averaging z scores on the reading and

Table 1
Correlations Between Key Measures of WM Spans and Scholastic
Performance
Key Measures 1 2 3 4 5 6 7 8
1. Traditional compound span –
2. New compound span .709 –
3. Reading span .900 .610 –
4. Operation span .900 .625 .621 –
5. Reading letter span .562 .855 .551 .460 –
6. Continuous operation span .650 .855 .492 .678 .461 –
7. Literacy score .341 .497 .296 .316 .462 .385 –
8. Mathematics score .377 .521 .327 .353 .479 .411 .811 –
9. Global scholastic score .385 .538 .335 .359 .499 .421 .947 .957
Note—Correlations ⱖ.338 are significant at p ⫽ .001, and correlations ⱖ.267 are sig-
nificant at p ⫽ .01.
168 LÉPINE, BARROUILLET, AND CAMOS

Table 2 tributed by the new tasks. By contrast, forced regres-

Factor Loadings for the Principal Component Analysis sions, in which the compound new WM score was en-
Performed on the Four WM Span Tasks
tered after the compound traditional WM score, revealed
Factors that the new tasks still accounted for a substantial part of
WM Span Task 1 2 3 the remaining variance of the literacy score [F(2,90) ⫽
Reading span .823 .205 ⫺.483 15.38, DR2 ⫽ .13, p ⬍ .0005], the mathematics score
Reading letter span .748 .568 .330 [F(2,90) ⫽ 15.92, D R 2 ⫽ .13, p ⬍ .0005], and the
Operation span .859 ⫺.313 ⫺.129
Continuous operation span .814 ⫺.399 .321 global scholastic score [F(2,90) ⫽ 17.92, D R2 ⫽ .14,
Variance .659 .155 .115 p ⬍ .0001; see Table 3]. These results clearly indicate
that the new computer-paced tasks account for an addi-
tional part of variance that is not accounted for by the
traditional spans.
the operation spans correlated with the global scholastic Detailed forward stepwise regression analyses were
score at .385 ( p ⬍ .001). However, the spans given by also conducted on the subcomponents of the scholastic
the new WM tasks were even better predictors of acade- scale with the four WM spans as independent variables.
mic performance. The reading letter span was a better For each of these subcomponents, a significant contri-
predictor than the reading span (rs ⫽ .462, .479, and bution was never observed for traditional spans that were
.499 vs. .296, .327, and .335 for the literacy, mathemat- systematically preceded by the new spans. The reading
ics, and global scores, respectively), and the continuous letter span appeared in the first step for language com-
operation span was a better predictor than the operation prehension, language tools, geometry, number knowl-
span (rs ⫽ .385, .411, and .421 vs. .316, .353, and .359, edge and writing of numbers, and arithmetic problem
respectively). The compound new WM score reached a solving, whereas the continuous operation span was the
significantly higher correlation with the global scholas- best predictor for written production, word problem
tic measure (r ⫽ .538, p ⬍ .001) than did the compound solving, and comprehension of mathematical tools and
traditional WM score [r ⫽ .385, t(90) ⫽ 2.25, p ⬍ .05]. concepts. The literacy, mathematics, and global scores
We therefore performed stepwise regression analyses to were accounted for best by the reading letter span [R2s ⫽
evaluate the extent to which new and traditional WM .22, .23, and .25, respectively; ps ⬍ .001], the continu-
spans accounted for a shared variance. ous operation span contributing in each case to a signif-
icant part of the remaining variance [ΔR2s ⫽ .04, .05,
Regression Analyses and .05, respectively; ps ⬍ .05]. None of the traditional
In order to determine the predictive power of the two tasks contributed significantly to the residual variance.
types of WM tasks (traditional vs. new), stepwise regres-
sion analyses were conducted on the literacy, mathemat- DISCUSSION
ics, and global scholastic scores with the two compound
WM scores as independent variables. The compound new These results clearly indicate that, in comparison with
WM score was the best predictor of the global scholastic traditional tasks, the new WM span tasks designed by
score [F(2,90) ⫽ 37.12, R2 ⫽ .29, p ⬍ .0001], of the lit- Barrouillet et al. (2004) provide a more accurate evalua-
eracy score [F(2,90) ⫽ 30.93, R2 ⫽ .25, p ⬍ .0001], and tion of the capacity of the common mechanism reflected
of the mathematics score [F(2,90) ⫽ 33.85, R2 ⫽ .27, by WM span measures. This finding is all the more re-
p ⬍ .0001]. The part of residual variance accounted for markable when it is considered that those tasks that in-
by the compound traditional WM score never reached volve the simplest activities and hamper the planning of
significance (Fs ⬍ 1), the highest D R2 value observed sophisticated dual-task strategies turned out to be the
being .0001. Thus, it appears that traditional WM tasks most predictive of complex cognitive achievements. The
contributed no variance in addition to that already con- predictive power of the new WM span tasks echoes pre-

Table 3
Stepwise Regressions for Scholastic Performance
Forward Forced
Dependent Variable Factor Step DR2 Factor Step DR2
Literacy New WM 1 .25* Traditional WM 1 .13*
Traditional WM 2 .00 New WM 2 .13*
Mathematics New WM 1 .27* Traditional WM 1 .14*
Traditional WM 2 .00 New WM 2 .13*
Global New WM 1 .29* Traditional WM 1 .15*
Traditional WM 2 .00 New WM 2 .14*
Note—“New WM” and “traditional WM” refer to the compound new and compound
traditional WM scores, respectively. *p ⬍ .05.
 WORKING MEMORY SPANS 169

vious observations by Fry and Hale (1996), who ob- good readers’ use of more sophisticated, more complex
served that performance on WM tasks as simple as re- comprehension strategies that result in the generation of
porting the colors of items while maintaining their iden- more extensive retrieval structures and hence a larger ef-
tities or locations in view of their subsequent recall was fective WM” (Ericsson & Kintsch, 1995, p. 229). More-
highly correlated with fluid intelligence. A first conclu- over, the present results strengthen and extend the theo-
sion, therefore, is that the complexity of the processing retical interest of the WM construct by going beyond the
component in most of the traditional WM span tasks rather circular observation that performance in span tasks
(reading span, operation span, alphabet recoding, and that involve complex processes (e.g., reading sentences
ABCD) is a superfluous characteristic. In fact, self- while remembering words) is correlated with perfor-
paced WM span tasks require complex activities to in- mance in high-level cognition (e.g., reading comprehen-
duce the necessary time pressure that is inherent to their sion). The results demonstrate that WM span tasks mea-
structure. The second conclusion is that the predictive sure a fundamental capacity and that WM measures are
value of the traditional WM spans does not stem from central to any assessment of cognitive abilities. Computer-
their capacity to assess an ability to strategically cope paced WM span tasks could thus constitute useful tools.
with the demands of complex span tasks—which would Apart from being fast and easy to administer, they per-
also be involved in any complex activity—because when mit better control of the strategies that undermine the ra-
the possibility of dealing strategically with the task is re- tionale of WM span tasks. Moreover, they do not suffer
duced by computer-paced presentation, the predictive from the undesirable effect of large individual differ-
value is increased. This does not mean that mnemonic ences in complex skills. Individuals can differ greatly in
strategies for encoding and maintaining memory items literacy and mathematics skills, but reading letters and
are unimportant in complex spans (McNamara & Scott, browsing the number chain are universal skills in literate
2001). However, our results show that complexity and groups even in rather young children. The range of tasks
strategies do not contribute to the predictive value of WM that permit a valid measure of the WM capacity is thus
spans but rather introduce more noise than information considerably extended.
into their relationship with high-level cognition. This
conclusion has theoretical and practical implications. REFERENCES
The predictive power of span tasks with simple pro-
cessing components suggests that WM tasks measure Anderson, J. R. (1993). Rules of the mind. Hillsdale, NJ: Erlbaum.
Baddeley, A. D. (1990). Human memory: Theory and practice. Hills-
some fundamental and general capacity involved in both dale, NJ: Erlbaum.
elementary and complex cognitive processing. This re- Baddeley, A. D., & Hitch, G. J. (1974). Working memory. In G. A.
sult echoes models of WM that conceive of cognitive re- Bower (Ed.), Recent advances in learning and motivation (Vol. 8,
sources as a kind of mental energy required to produce pp. 647-667). New York: Academic Press.
activation (Engle, Cantor, & Carullo, 1992; Just & Car- Baddeley, A. D., Logie, R., Nimmo-Smith, I., & Brereton, N.
(1985). Components of fluent reading. Journal of Memory & Lan-
penter, 1992; Lovett, Reder, & Lebière, 1999). This re- guage, 24, 119-131.
sult is also in line with the time-based resource-sharing Barrouillet, P. (1996). Transitive inferences from set-inclusion rela-
model (Barrouillet et al., 2004), which assumes, follow- tions and working memory. Journal of Experimental Psychology:
ing Anderson (1993) and Cowan (1995), that this funda- Learning, Memory, & Cognition, 22, 1408-1422.
Barrouillet, P., Bernardin, S., & Camos, V. (2004). Time constraints
mental capacity is attentional. This attention is required and resource sharing in adults’ working memory spans. Journal of
for retrieval of knowledge from long-term memory dur- Experimental Psychology: General, 133, 83-100.
ing reading of letters or solution of simple operations, Case, R., Kurland, M., & Goldberg, J. (1982). Operational effi-
but also for reactivation of the decaying memory traces ciency and the growth of short-term memory. Journal of Experimen-
of the items to be recalled. The time constraints that re- tal Child Psychology, 33, 386-404.
Cowan, N. (1995). Attention and memory: An integrated framework.
sult from the computer-paced presentation imply that New York: Oxford University Press.
even simple secondary tasks almost continuously cap- Cronbach, L. J. (1951). Coefficient alpha and the internal structure of
ture attention and impede the refreshing of memory tests. Psychometrika, 16, 297-334.
traces. Thus, the ability of individuals to prevent the decay Daneman, M., & Carpenter, P. A. (1980). Individual differences in
working memory and reading. Journal of Verbal Learning & Verbal
of these traces reflects their available amount of atten- Behavior, 19, 450-466.
tion—that is, their WM capacity. By contrast, the pre- Daneman, M., & Merikle, P. M. (1996). Working memory and compre-
dictive power of the new tasks is at odds with theories of hension: A meta-analysis. Psychonomic Bulletin & Review, 3, 422-433.
WM such as that of Ericsson and Kintsch (1995), in Engle, R. W., Cantor, J., & Carullo, J. J. (1992). Individual differ-
which any notion of cognitive resources is denied and by ences in working memory and comprehension: A test of four hy-
potheses. Journal of Experimental Psychology: Learning, Memory,
which it is assumed that the relationship between WM & Cognition, 5, 972-992.
and complex cognition is due to strategic factors and Engle, R. W., Kane, M. J., & Tuholski, S. W. (1999). Individual dif-
knowledge. Indeed, when discussing Just and Carpen- ferences in working memory capacity and what they tell us about
ter’s (1992) capacity theory of comprehension, Ericsson controlled attention, general fluid intelligence, and functions of the
prefrontal cortex. In A. Miyake & P. Shah (Eds.), Models of working
and Kintsch assumed that their proposal for long-term memory: Mechanisms of active maintenance and executive control
WM “can do without the somewhat slippery notion of (pp. 102-134). Cambridge: Cambridge University Press.
cognitive resources altogether” but rather “emphasizes Engle, R. W., Tuholski, S. W., Laughlin, J. E., & Conway, A. R.
170 LÉPINE, BARROUILLET, AND CAMOS

(1999). Working memory, short term memory, and general fluid in- Smith, M. M., & Scholey, K. A. (1992). Determining spatial span: The
telligence: A latent variable approach. Journal of Experimental Psy- role of movement time and articulation rate. Quarterly Journal of Ex-
chology: General, 128, 309-331. perimental Psychology, 45A, 479-501.
Ericsson, K. A., & Kintsch, W. (1995). Long-term working memory. Turner, M. L., & Engle, R. W. (1989). Is working memory task-
Psychological Review, 102, 211-245. dependent? Journal of Memory & Language, 28, 127-154.
Fry, A. F., & Hale, S. (1996). Processing speed, working memory, and
fluid intelligence: Evidence for a developmental cascade. Psycho- NOTES
logical Science, 7, 237-241.
Just, M. A., & Carpenter, P. A. (1992). A capacity theory of compre- 1. By traditional self-paced WM span tasks, we do not intend to refer
hension: Individual differences in working memory. Psychological to self-presented tasks in which the entire procedure is driven by the
Review, 99, 122-149. participant, but to tasks in which the participant performs the process-
Kemps, E., De Rammelaere, S., & Desmet, T. (2000). The develop- ing component at his or her own pace. In some traditional tasks, such as
ment of working memory: Exploring the complementarity of two Daneman and Carpenter’s (1980) reading span or Case et al.’s (1982)
models. Journal of Experimental Child Psychology, 77, 89-109. counting span, it is not the participant but the experimenter who dis-
Kyllonen, P. C., & Christal, R. E. (1990). Reasoning ability is (little plays the next sentence to be read or the next array to be enumerated.
more than) working-memory capacity? Intelligence, 33, 1-64. However, these tasks remain self-paced in that there is no time con-
Lovett, M. C., Reder, L. M., & Lebière, C. (1999). Modeling work- straint in reading the sentence or in counting the array, the participant
ing memory in a unified architecture: An ACT–R perspective. In being free to perform these activities at his or her own pace.
A. Miyake & P. Shah (Eds.), Models of working memory: Mecha- 2. The definitions of these subdomains by the French Ministry of Ed-
nisms of active maintenance and executive control (pp. 135-182). ucation and the national results can be found on the Web at http://
Cambridge: Cambridge University Press. evace26.education.gouv.fr/.
McNamara, D. S., & Scott, J. L. (2001). Working memory capacity
and strategy use. Memory & Cognition, 29, 10-17.
Shute, V. J. (1991). Who is likely to acquire programming skills? Jour- (Manuscript received December 11, 2003;
nal of Educational Computing Research, 7, 1-24. revision accepted for publication March 23, 2004.)

View publication stats