Bilingualism, Aging, and Cognitive Control: Evidence From the Simon Task

Text-only Preview

Psychology and Aging
Copyright 2004 by the American Psychological Association
2004, Vol. 19, No. 2, 290 –303
DOI: 10.1037/0882-7974.19.2.290
Bilingualism, Aging, and Cognitive Control:
Evidence From the Simon Task
Ellen Bialystok
Fergus I. M. Craik
York University
Rotman Research Institute
Raymond Klein
Mythili Viswanathan
Dalhousie University
York University
Previous work has shown that bilingualism is associated with more effective controlled processing in
children; the assumption is that the constant management of 2 competing languages enhances executive
functions (E. Bialystok, 2001). The present research attempted to determine whether this bilingual
advantage persists for adults and whether bilingualism attenuates the negative effects of aging on
cognitive control in older adults. Three studies are reported that compared the performance of mono-
lingual and bilingual middle-aged and older adults on the Simon task. Bilingualism was associated with
smaller Simon effect costs for both age groups; bilingual participants also responded more rapidly to
conditions that placed greater demands on working memory. In all cases the bilingual advantage was
greater for older participants. It appears, therefore, that controlled processing is carried out more
effectively by bilinguals and that bilingualism helps to offset age-related losses in certain executive
Research in cognitive aging has advanced enormously in the
establish the precise effects of bilingualism on cognitive process-
past few decades, producing detailed studies and sophisticated
ing and the way in which these effects are modulated by aging.
models of age-related changes in cognitive functions (see chapters
Studies involving adult bilinguals have focused largely on psy-
in Craik & Salthouse, 2000). Most of this research involves
cholinguistic aspects of language use, so most of these studies have
English-speaking participants, and conclusions have been drawn
investigated only bilingual participants to compare processing in
with little or no regard to the possibility that the participants might
the two languages. A few studies on lexical processing that have
also speak another language. Yet the existing evidence strongly
included between-groups comparisons have reported bilingual dis-
suggests that bilingualism has an effect on cognitive processing, at
advantages on some tasks, such as lexical decision (Ransdell &
least for children and younger adults (see chapters in de Groot &
Fischler, 1989) and semantic fluency (Gollan, Montoya & Werner,
Kroll, 1997, and Harris, 1992). What has not been examined is
2002). In a review of this literature, Michael and Gollan (in press)
whether these effects persist over the life span and continue to
point out that these deficits are quite limited, but they attribute the
influence changes in cognitive processing in bilingual older adults.
observed reduction in fluency to the bilingual’s need to maintain a
One current reality is that bilingualism is increasingly common in
vocabulary base approximately twice as large as that of a mono-
many countries. As an example, the 1996 Canadian Census re-
lingual and to the reduced frequency with which bilinguals access
ported that approximately 11% of Canadians spoke English or
any particular word. These conditions result in weaker links be-
French at home in addition to some other language; when only
tween words and concepts for bilinguals; semantic fluency tasks,
respondents over age 65 were considered, the figure was 13%
these authors argue, are a measure of the strength of these word–
(Canada Census 1996, n.d.). In the United States, 17.9% of Amer-
concept associations. Although some research has examined the
icans reported that they spoke a language other than English at
role of cognitive processes such as working memory in the acqui-
home, and it is a reasonable assumption that most of them also
sition of a second language (Harrington & Sawyer, 1992; Miyake,
speak English (U.S. Census Bureau, 2003). Given the prevalence
1998), very little research has investigated whether those processes
of bilingualism in North American society (and the prevalence is
are modulated by bilingualism.
certainly greater in most European countries), it is important to
Research with children has addressed the cognitive impact of
bilingualism more directly. Bilingual advantages have been re-
ported across a variety of domains, for example, creativity (Kessler
Ellen Bialystok and Mythili Viswanathan, Department of Psychology,
& Quinn, 1987), problem solving (Bain, 1975; Kessler & Quinn,
York University, Toronto, Ontario, Canada; Fergus I. M. Craik, Rotman
1980), and perceptual disembedding (Duncan & De Avila, 1979).
Research Institute, Toronto, Ontario, Canada; Raymond Klein, Department
Positive effects for bilinguals, however, have not always been
of Psychology, Dalhousie University, Halifax, Nova Scotia, Canada.
found; some studies reported negative effects (Macnamara, 1966),
This research was supported by a grant from the Canadian Institutes of
and others found no group differences (Rosenblum & Pinker,
Health Research to Ellen Bialystok and Fergus I. M. Craik.
1983). The disparate findings can be resolved by considering the
Correspondence concerning this article should be addressed to Ellen
Bialystok, Department of Psychology, York University, 4700 Keele Street,
cognitive processes implicated in the various tasks used to assess
Toronto, Ontario M3J 1P3, Canada. E-mail: [email protected]
the effects of bilingualism. In general, tasks showing a bilingual

advantage are characterized by the presence of misleading (usually
and refined by Hasher and Zacks themselves (Hasher, Zacks, &
perceptual) information and the need to choose between competing
May, 1999; Zacks, Hasher, & Li, 2000). What does seem clear is
response options; tasks based more heavily on analytic knowledge
that older adults show a decline in the effectiveness of executive
or detailed representations of knowledge presented without a mis-
control processes in many situations unless task performance de-
leading context are solved equally well by monolinguals and
pends on strongly ingrained habits (Hay & Jacoby, 1996, 1999) or
bilinguals. This difference corresponds to the difference between
is well supported by the environmental context (Craik, 1986). In
control and representational processes, respectively. The functions
summary, then, children’s cognitive development is characterized
contributing to control include selective attention to relevant as-
by a growth in both control of attention and representational
pects of a problem, inhibition of attention to misleading informa-
complexity, whereas aging leads to a decline in the effectiveness of
tion, and switching between competing alternatives. The functions
attentional control but not in the ability to utilize habitual proce-
involved with representation include encoding problems in suffi-
dures and representational knowledge. Bilingual children, there-
cient detail, accessing relevant knowledge, and making logical
fore, experience a boost in the development of the types of cog-
inferences about relational information. Research by Bialystok has
nitive processing that typically decline with aging.
shown that bilingual children develop control processes more
A formidable obstacle to conducting research that allows mean-
readily than do monolingual children but that the two groups
ingful comparisons of young children and older adults is the
progress at the same rate in the development of representational
identification of a task that is suitable for all ages. Most of the
processes (for reviews, see Bialystok, 1993, 2001).
research with young children has been based on tasks that are
Why would bilingualism enhance the development of children’s
trivially easy even for older children, and studies of adult perfor-
control processes? Evidence from psycholinguistic studies of adult
mance typically require expertise and endurance beyond the ability
language processing shows that the two languages of a bilingual
of children. Therefore, a task is needed that is relatively content
remain constantly active while processing is carried out in one of
free but dependent on the cognitive processes proposed to charac-
them (Brysbaert, 1998; Francis, 1999; Gollan & Kroll, 2001; Kroll
terize the performance advantage of bilingual individuals. A task
& Dijkstra, 2002; Smith, 1997). The joint activity of the two
that meets these criteria is the Simon task (see review in Lu &
systems requires a mechanism for keeping the languages separate
Proctor, 1995). The task is based on stimulus–response compati-
so that fluent performance can be achieved without intrusions from
bility and assesses the extent to which the prepotent association to
the unwanted language. Green (1998) proposed a model based on
irrelevant spatial information affects participants’ response to task-
inhibitory control in which the nonrelevant language is suppressed
relevant nonspatial information. In our implementation of this task,
by the same executive functions used generally to control attention
colored stimuli were presented on either the left or the right side of
and inhibition. If this model is correct, then bilinguals have had
a computer screen. Each of the two colors was associated with a
massive practice in exercising inhibitory control, an experience
response key that was also on one of the two sides of the keyboard,
that may then generalize across cognitive domains. If the boost
aligned with the two stimulus positions. On congruent trials, the
given by childhood bilingualism is sufficiently strong, bilingual-
key that was the correct response for that color was on the same
ism may continue to influence certain control processes throughout
side as the stimulus; on incongruent trials, the correct response key
the life span. Two questions follow from this possibility. The first
was on the opposite side. Numerous studies with this task have
is whether the advantages found for young children in executive
confirmed that the irrelevant location information results in reli-
processes are also seen in adult bilinguals. The second is whether
ably longer reaction times (RTs) for incongruent items.
such advantages are maintained in older adulthood and protect
The increased time needed to respond to the incongruent items
bilingual adults from the normal decline of these processes that
is the Simon effect. Van der Lubbe and Verleger (2002) found a
occurs with age.
larger Simon effect in a group of older adults (mean age
With regard to aging, it is well established that the representa-
years) than in a comparable group of young adults (mean age
tional functions that depend on well-learned knowledge and ha-
years), even after correcting for the general slowing associated
bitual procedures (“crystallized intelligence”) hold up well in the
with aging. Therefore, the Simon task measures aspects of pro-
later adult years, whereas abilities that depend on executive control
cessing that decline with aging. The next question is whether the
processes (“fluid intelligence”) show a marked decline in effi-
ability to attend to the stimulus and ignore the irrelevant location
ciency. In the former category, vocabulary levels (Park, 2000;
information reflects the same type of cognitive control that is
Salthouse, 1991), general world knowledge (Salthouse, 1982), and
enhanced in development by bilingualism. If this is the case, then
language use (Wingfield & Stine-Morrow, 2000) all show little
the performance of young bilingual children should be less af-
age-related decline. In contrast, executive control functions un-
fected by the irrelevant spatial code of the target than the perfor-
dergo declining efficiency with aging. In perceptual processing,
mance of comparable monolingual children; bilinguals, that is,
older adults are less able to ignore irrelevant stimuli (Rabbitt,
should show a reduced Simon effect. Moreover, if the effects of
1965) and to attend selectively to important aspects of the envi-
bilingualism on cognitive processing persist through adulthood
ronment. Less effective attentional processes result in less efficient
and into aging, then this advantage should be found as well for
detection, discrimination, and selection of wanted stimuli, reduced
adult bilinguals. Finally, if lifelong bilingualism provides a de-
resistance to interference, and impaired inhibition of information
fense against the normal decline of these control processes, then
that is unimportant or irrelevant (McDowd & Shaw, 2000). Hasher
older bilinguals should show less decrement in control as mea-
and Zacks (1988) argued that much of the observed decline in
sured by the Simon task than should comparable older
cognitive functioning is the result of a decline in the effectiveness
of inhibitory processes, although that general conclusion has been
In two studies with 4-year-olds (Martin & Bialystok, 2003),
called into question by the results of more recent studies (e.g.,
bilinguals outperformed monolinguals on the Simon task, but
Kieley & Hartley, 1997; Kramer & Strayer, 2001) and modified
contrary to prediction, the advantage was found for both the

congruent and incongruent trials. The advantage for bilinguals,
and the older adults were recruited through flyers posted in community
therefore, may lie not in their enhanced ability to inhibit the
centers in both countries.
misleading spatial cue but in their ability to manage attention to a
complex set of rapidly changing task demands. The present studies
Tasks and Procedure
extend this paradigm into adulthood and aging. In three studies, we
had monolingual and bilingual younger and older adults perform
Language background questionnaire.
This questionnaire was filled out
versions of the Simon task to determine whether the processing
by the experimenter while interviewing the participant on language use and
differences shown by bilingual children would extend into adult-
fluency in his or her two languages. The language usage chart addressed
the percentage usage of each language at home, at work, with friends, and
hood and old age. If they did, the implication would be that the
overall. The responses indicate the extent to which each language is used
advantage in cognitive control goes beyond the management of
daily and the degree to which the participant is functionally bilingual.
language processing to cognitive processing in general and may
Peabody Picture Vocabulary Test—Revised (PPVT–R; Dunn & Dunn,
ameliorate the age-related declines seen in many cognitive tasks.
This is a standardized test of receptive vocabulary. The test
consists of a series of plates, each containing four pictures. The experi-
Study 1
menter names one of the pictures, and the respondent indicates which
picture illustrates that word. The items become increasingly difficult, and
In the first study, we investigated possible effects of adult aging
testing continues until the participant makes 6 errors in 8 consecutive
and language group on the Simon task by replicating the experi-
items. The score is determined by tables that convert the raw score to a
ment conducted with monolingual and bilingual children (Martin
standard score in terms of the age of the respondent. The test was admin-
istered in English to all participants.
& Bialystok, 2003). The parameters for this earlier experiment
Raven’s Standard Progressive Matrices (Raven, 1958).
Raven’s Ma-
were designed to be appropriate for young children—there were
trices is an untimed test that measures abstract nonverbal reasoning ability.
long delays between events and very few trials. Although this
The test consists of 60 items arranged in five sets (A, B, C, D, and E) of
design has many fewer trials than is typical in such studies, we
12 items each. Each item contains a picture with a missing piece. Below the
decided that this preliminary experiment should replicate the de-
picture, there are either six (Sets A and B) or eight (Sets C to E) possible
sign that had already produced language group differences in
pieces to complete the picture. Both the sets and the items within the sets
children. Nonetheless, it is worth noting that the original study by
are arranged in order of difficulty. Participants are given a score for each
Simon and Wolf (1963) in which the effect was discovered in-
correct answer, and these raw scores are converted into standardized ranks
cluded only 16 trials per condition.
through tables based on the participants’ ages.
Simon task.
The experiment was presented on a laptop Gateway Solo
2150 computer with a 12-in. monitor. The sequence of events and collec-
tion of data were controlled by a program running in DMDX (n.d.), which
is a Win 32-based display system. Each trial began with a fixation cross
( ) in the center of the screen, measuring x
0.48°, y
0.40°, that
remained visible for 800 ms and was followed by a 250-ms blank interval.
There were 40 participants who composed two language groups and two
At the end of this interval, a red or blue square appeared on the left (x
age groups. Twenty of the participants were younger adults ranging in age
0.02°, y
0.36°) or the right (x
0.82°, y
0.36°) side of the screen and
from 30 to 54 years (mean age
43.0 years), and 20 were older adults
remained on the screen for 1,000 ms if there was no response. Participants
ranging in age from 60 to 88 years (mean age
71.9 years). In each age
were instructed to press the left shift key (marked “X”) when they saw a
group, half the participants were monolingual English speakers living in
blue square and the right shift key (marked “O”) when they saw a red
Canada, and the other half were Tamil–English bilinguals living in India.
square. The timing began with the onset of the stimulus, and the response
Tamil is an alpha-syllabic language from the Southern Indian state of
terminated the stimulus; there was then a 500-ms blank interval before the
Tamil Nadu. The monolingual and bilingual participants in each group
onset of the next trial. The experiment began with eight practice trials, and
were matched on age so that a monolingual was included in the study if his
participants had to complete all eight trials successfully to proceed to the
or her age matched exactly that of one of the bilingual participants. There
experimental trials for that condition. If a mistake was made, participants
were equal numbers of men and women in each group. All participants
were given additional practice trials until all eight trials were completed
were tested by the same experimenter (Mythili Viswanathan) using the
without an error, but only 1 participant needed to repeat the practice set to
same equipment and the same instructional protocols, although the actual
achieve error-free performance. The 28 experimental trials, half of which
testing was carried out in two different countries.
presented the square on the same side as the associated response key
The bilingual participants learned Tamil as their first language and were
(congruent trials) and half of which presented the square on the opposite
educated in both languages beginning at the age of 6 years. Schooling was
side (incongruent trials), were presented in a randomized order.
conducted primarily in English, but Tamil was both taught as a subject and
used as the medium of instruction for some subjects. From the beginning
of schooling, the participants had used both Tamil and English on a daily
basis throughout their lives. Data from the language background question-
naire (described below) indicated that the average daily use of English was
The background measures of age, PPVT scores, and Ravens
56% and that of Tamil was 44%. Research with both bilingual adults (Kroll
scores are shown in Table 1. A two-way analysis of variance
& Stewart, 1994) and bilingual children (Bialystok, 1988) has revealed that
(ANOVA) on the Ravens scores found no differences for either
the cognitive and linguistic consequences of bilingualism are more salient
age or language group (both Fs
1), and a similar analysis on the
for those bilinguals who are relatively balanced in their proficiency, so the
PPVT scores also found no differences for either age, F(1, 36)
criterion of balanced bilingualism was used for the selection of the sample
1.51, p
.23, or language group, F(1, 36)
2.76, p
in the present studies. The monolingual English participants lived in
Canada and were not functionally fluent in any other language despite the
The mean accuracy scores and RTs for the congruent and
inevitable language courses in school. All the participants in both groups
incongruent trials in the Simon task as a function of age and
had bachelor’s degrees and shared similar middle-class socioeconomic
language group are shown in Table 2. For the accuracy scores, a
backgrounds. The younger adults were recruited through e-mail postings,
three-way ANOVA for age group (older, younger), language

Table 1
Mean Background Measures (and Standard Deviations) by Age and Language Group in Study 1
Age (in years)
43.0 (7.3)
43.0 (7.3)
71.6 (7.5)
72.3 (8.7)
91.0 (4.4)
91.8 (9.8)
85.8 (7.1)
91.9 (2.6)
1.4 (0.5)
1.3 (0.5)
1.5 (0.5)
1.3 (0.5)
Peabody Picture Vocabulary Test—Revised; Raven
Raven’s Standard Progressive
group (monolingual, bilingual), and congruency (congruent, in-
bilingual adults with that of children. All participants were com-
congruent) showed that there were more errors on incongruent
parable on measures of verbal and spatial intelligence and similar
trials, F(1, 38)
42.21, p
.01, and there were interactions of
in educational and social background, but bilinguals were consis-
language and congruency, F(1, 38)
31.86, p
.01, and of
tently faster in responding to the Simon task. The pattern of results
language, age, and congruency, F(1, 36)
7.34, p
.01, that
replicated that obtained with children in that the bilinguals were
confined the difference to the older monolinguals in the incongru-
faster overall; in addition, the bilinguals in the present study
ent condition. The high error rate in this condition (28%) suggests
showed a smaller Simon effect in that the incongruent items were
that the RT data should be treated with some caution. However,
less disrupting. For the older participants, the bilinguals also
mean RTs were also very high in this condition, so the result is not
avoided the increase in errors that characterized the performance
due to a speed–accuracy trade-off.
of the older monolinguals.
The RTs were examined with a three-way ANOVA for age
For both age and language groups, incongruent items required
group, language group, and congruency. The younger adults were
longer response times than congruent items, and this difference
faster than the older adults, F(1, 36)
28.29, p
.01; bilinguals
(the Simon effect) was reliably smaller for the younger adults and
were faster than monolinguals, F(1, 36)
16.12, p
.01; and
for the bilinguals. The absence of a significant three-way interac-
congruent items elicited faster responses than did incongruent
tion among age, language, and congruency means that the age-
items, F(1, 36)
55.88, p
.01. There were two-way interactions
related increase in the Simon effect was as great for the bilinguals
between age and congruency, F(1, 36)
21.60, p
.01, and
as for the monolinguals. Thus, the older adults and the monolin-
between language and congruency, F(1, 36)
12.93, p
gual participants in both age groups were less able to inhibit the
indicating that the magnitude of the difference between congruent
negative influence of the incongruent spatial information, but
and incongruent trials (the Simon effect, also shown in Table 2)
bilingualism (against our prediction) did not attenuate the age-
was greater for the older adults and for the monolingual partici-
related decline in inhibitory effectiveness. Nevertheless, the age-
pants. The table shows that the age-related increase in the Simon
related increase in the Simon effect was substantially less for the
effect was less for the bilingual groups (748
708 ms) than
bilingual adults (708 ms) than for the monolingual adults (1,178
for the monolingual groups (1,713 – 535
1,178 ms), but the
ms), but the analyses are based on relatively small sample sizes
three-way interaction of age, language, and congruency was not
and involve high variance in the RTs. Therefore, we postpone a
significant, F(1, 36)
1.34, p
.25, indicating that, statistically,
final conclusion concerning the effects of bilingualism on this
the age-related increase in the Simon effect was as great for
inhibitory function until we consider the results of the next exper-
bilinguals as it was for monolinguals.
iment, which involved more participants and more experimental
The bilingual speed advantage was reliably larger for the incon-
The main purpose of Study 1 was to explore the feasibility of
gruent items but still present for the congruent items. There are
comparing the Simon task performance of older monolingual and
three possible reasons for this speed advantage: Bilinguals may
Table 2
Mean Accuracy and Reaction Time (RT; in Milliseconds) by Age and Language Group in Study 1
Age and language group
(in ms)
(in ms)
Simon effect
770 (132.8)
1,304 (273.0)
535 (231.2)
497 (252.5)
536 (273.0)
40 (32.2)
1,437 (560.6)
3,150 (1,309.6)
1,713 (971.7)
911 (374.2)
1,659 (1,151.0)
748 (806.6)
Standard deviations are in parentheses.

simply be faster, bilinguals may profit more from the facilitation
in Hong Kong (12 participants). Each of these bilinguals was matched for
on the congruent trials, which may disproportionately boost per-
age with one of the monolinguals, making the age ranges and the mean
formance on these items, or bilinguals may be less disrupted by
ages the same for the two groups. There were equal numbers of men and
interference on the incongruent trials. We investigated these alter-
women in each group. The second age group consisted of 30 older adults
natives in the next study. In addition, the RTs in the present study
ranging in age from 60 to 80 years (mean age
70.3 years), divided
between English-speaking monolinguals and bilingual speakers of English
were very long, even if one considers that the older adults had a
and Tamil living in India (9 participants) or of English and French living
mean age over 70 years. This may be due to the fact that the RTs
in Canada (6 participants). There were equal numbers of men and women
were measured at very early stages of practice. In the next study,
in each group. Participants were recruited using the same procedures as in
we investigated this possibility by using a more standard design
Study 1. The Tamil participants were tested in India, and the Cantonese
that included more trials.
participants were tested in Hong Kong, all by the same experimenter using
the same equipment.
Study 2
All of the bilinguals were educated in both languages from the age of 6
years and had continued to use both their languages daily. As in Study 1,
In Study 1, bilinguals in both age groups performed the Simon
the language background questionnaire was used to determine the daily use
task more quickly than comparable monolinguals and showed less
of each language by the bilinguals. The first language of the Tamil–English
interference from the position information in the incongruent trials.
bilinguals was Tamil, and they used English 55% of the time. The
In all conditions, however, both the absolute RTs and the differ-
Cantonese–English bilinguals’ first language was Cantonese, and members
of this group used English 48% of the time. The French–English bilinguals
ence scores indicating the Simon effect were unusually large. The
learned both French and English from childhood and used English 52% of
main reasons for this may be methodological: The small number of
the time. The monolingual participants lived in Canada and did not have
trials meant that participants were very unpracticed on the task,
functional command of any other language. All participants had bachelor’s
and the relatively slow presentation rate may have produced a slow
degrees and similar middle-class socioeconomic backgrounds.
overall rate of responding. The RTs obtained in Study 1 were
similar to those produced by children using the same program—
specifically, in the range of 1,000 to 2,000 ms. In addition, al-
Tasks and Instruments
though the means of the Simon effect values in Table 2 show that
Language background questionnaire and usage chart.
The same ques-
the bilingual advantage was greater for older adults (965 ms) than
tionnaire used in Study 1 was used in Study 2.
for younger adults (495 ms), the interaction between age and
Peabody Picture Vocabulary Test—Third Edition (PPVT–III; Dunn &
language on the size of the Simon effect was not significant, F(1,
Dunn, 1997).
The PPVT–III is a more recent version of the PPVT in
1.34, p
which the norms are extended to standardize scores of individuals who are
Study 2 was designed to build on the preliminary results of
more than 70 years old. The task proceeds in the same manner as that in the
Study 1 in two ways. The first was by replicating the patterns of
PPVT–R, presenting participants with plates of four pictures and one word.
age and group differences using a more conventional design;
The starting item is set according to the participant’s chronological age. In
participants in Study 2 completed 192 trials of the Simon task, in
this version, testing terminates when the participant commits 8 errors out
contrast to the 28 trials in Study 1. The second was by including
of 12 items in a set. As in Study 1, this test was administered only in
English to all participants.
conditions that would help to isolate the source of the bilingual
Cattell Culture Fair Intelligence Test (Cattell & Cattell, 1960).
advantage. The first condition was a control condition, called
Cattell test is a nonverbal test of general intelligence. The raw scores are
center–2, in which speed of responding could be measured inde-
converted into IQ scores by a set of tables based on age.
pendently of the Simon interference by placing the colored squares
Alpha span task (Craik, 1986).
The alpha span task (AST) is a measure
in the center of the screen, thus eliminating conflict between the
of verbal working memory. Lists ranging in length from two to eight words
position of the target and the position of the response. Another
are presented auditorily at the rate of 1 word per second. Words are
concern was that the bilingual advantage might not reflect superior
presented in random order, and participants are required to repeat the words
skill in ignoring the irrelevant position information but rather a
back in alphabetical order. The task begins with a list of two words and
greater ability to remember the rules associating each color with
proceeds by presenting two trials at each list length and increasing the
the appropriate response key. If bilingualism conferred an advan-
length by one upon completion of the pair. After an error, testing continues
for two more list lengths. In the scoring system, 1 point is awarded for each
tage in this type of working memory ability, then bilinguals would
item recalled in a correctly ordered pair; the paired word can either precede
be more able to make rapid judgments about the correct response.
or follow the scored word. For example, if a list of four items is recalled
We addressed this possibility by including two conditions in which
correctly, the score is 4 points; if the correct recall sequence for a list of
the working memory demands were increased to determine
five items is “apple, car, hotel, rabbit, toy,” and the participant recalls
whether this manipulation also favored bilinguals. In these condi-
“apple, hotel, rabbit, toy,” he or she would receive 3 points—1 each for
tions, the stimuli were four colors, so participants had to keep four
hotel, rabbit, and toy. “Apple” does not receive a point because “apple–
rules in mind associating each color with a response.
hotel” is not a correct pair. The AST score is the total number of points
awarded across all presented lists.
Sequencing span task.
The sequencing span task (SST) is similar to the
AST but uses strings of double-digit numbers ranging from 10 to 99 that
are presented in random order; the participant’s task is to repeat back
increasingly long strings of numbers in the correct order. No number was
There were 94 participants composing two age and two language groups.
repeated across any of the strings, and no pairs of numbers in the presen-
The first age group consisted of 64 younger adults, ranging in age from 30
tation strings appeared in their correct ascending order. The responses were
to 58 years (mean age
42.6 years), divided evenly between monolingual
scored using the procedure described above for the AST.
speakers of English living in Canada and bilingual speakers of English and
Simon tasks.
All participants completed four conditions in one of four
Tamil living in India (20 participants) or of English and Cantonese living
preset orders consisting of 24 trials per condition. The entire set of four

conditions was then repeated in the reverse order, producing 48 trials for
blocks. For example, if a participant completed conditions in the order B,
each of the four conditions.
D, A, C, then the second set of experimental trials would proceed in the
Condition A: Center–2 (control).
A series of squares that were either
order C, A, D, B.
brown or blue appeared in the middle of the screen. Participants were
instructed to press the left shift key (marked “X”) when they saw a blue
square and the right shift key (marked “O”) when they saw a brown square.
The trial began with a sound (a computer “bing”) and a fixation cross ( )
The results for the background variables are presented in Table
that appeared in the center of the screen for 300 ms. Immediately after this
3. Two-way ANOVAs testing for age and language group differ-
cue, the stimulus appeared (x
0.43°, y
0.38°) and remained on the
ences were carried out on each of these measures. The PPVT
screen until a response was made. The response clock started at the onset
scores were the same for both age, F(1, 90)
2.75, p
.10, and
of the stimulus. The fixation cross (plus the sound) reappeared 500 ms after
language groups, F(1, 90)
0.03, p
.86, with no interaction.
the response to signal the next trial.
Similarly, Cattell scores were the same for both age, F(1, 90)
Condition B: Side–2.
The parameters were the same as those in the
control condition, but the blue and brown squares appeared on either the
1.69, p
.20, and language groups, F(1, 90)
2.37, p
.13, with
left or the right side of the screen. The order of trials was randomized and
no interaction. In contrast, younger participants scored higher than
divided equally between congruent and incongruent items. The DMDX
older participants on both the AST, F(1, 90)
34.90, p
.01, and
parameters from Study 1 were used.
the SST, F(1, 90)
4.86, p
.03, but there were no differences
Condition C: Center– 4.
This condition was similar to the control
between the language groups and no interactions (Fs
1). No
condition except that the stimulus was one of four colors: pink, yellow, red,
norms are available for the AST and SST scores, but the values
or green. Participants were instructed to press the left shift key when they
shown in Table 3 are typical for participants of these ages who
saw a green square, the right shift key when they saw a red square, the left
have been tested in our laboratory.
shift key when they saw a pink square, and the right shift key when they
The mean accuracy scores for the Simon task ranged from 97%
saw a yellow square. The instructions were presented as four individual
to 99% and are reported in Table 4. The error rates were higher for
rules (i.e., “press the left shift key for green”; “press the left shift key for
pink”) and not as two paired rules (i.e., “press the left shift key for green
the younger participants than the older participants, F(1, 90)
or pink”). All stimuli appeared in the center of the screen. This condition
13.94, p
.01. There was no difference between the language
placed greater demands on working memory for the assignment of colors
groups (F
1), but there was an interaction of language and age,
to responses than did the Center–2 condition.
F(1, 90)
8.62, p
.01, because the higher accuracy rate for the
Condition D: Side– 4.
In this condition, the stimuli were the same four
older participants was stronger in the bilinguals.
colors, but they appeared in one of two side positions. The order of trials
The mean RTs for the Simon task organized by position of the
was randomized and again divided equally between congruent and incon-
stimulus (center or side) and number of colors (2 or 4) are also
gruent items.
reported in Table 4. Before examining the Simon effect for the
A set of practice trials preceded each condition. The two-color condi-
different conditions, we conducted a preliminary four-way
tions had four practice trials, and the four-color conditions had eight
ANOVA involving age (2), language group (2), color (2), and
practice trials, demonstrating each unique stimulus configuration for the
condition. The parameters of these trials were identical to those of the test
position collapsed across congruency (2). This analysis explored
trials. Participants had to complete all practice trials correctly to proceed
the effect of position uncertainty (always in the center versus on
with testing. If a mistake was made during a practice trial, the program
one of two sides) on the different groups. The ANOVA revealed
recycled until all trials were completed without error. Two participants
significant effects for all four factors (younger participants, bilin-
repeated the set of practice trials.
guals, central position, and two-color conditions were faster), and
all interactions were also significant. Therefore, we analyzed each
condition separately in a series of two-way ANOVAs to examine
the effects of age and language group (the means are shown in
Test sessions began with the language background questionnaire and
Table 5). For all four analyses, younger adults were faster than
chart, the PPVT–III, and the AST, all administered in English. The RT
older adults: center–2, F(1, 90)
687.58, p
.01; side–2, F(1,
tasks were administered in one of four pseudorandom orders that presented
one block from each of the four conditions. After this, participants were
338.91, p
.01; center– 4, F(1, 90)
477.32, p
given a break in which they completed the Cattell Culture Fair Intelligence
side– 4, F(1, 90)
230.15, p
.01. The two language groups
Test and the SST. These tasks were followed by the remaining blocks of
performed the same in the center–2 condition (F
1), but bilin-
the Simon task, administered in the reverse order from that used for the first
guals were faster than monolinguals in the other three conditions:
Table 3
Mean Background Measures (and Standard Deviations) by Age and Language Group in Study 2
Age (in years)
42.6 (8.8)
42.6 (8.8)
70.4 (5.6)
70.2 (6.9)
85.4 (5.6)
86.0 (4.7)
79.7 (6.9)
81.4 (5.0)
109.1 (6.1)
109.5 (6.7)
108.5 (7.6)
109.7 (7.9)
28.8 (4.6)
28.0 (4.5)
22.4 (1.9)
24.0 (2.9)
25.1 (4.8)
24.0 (4.8)
21.6 (2.8)
23.0 (5.1)
Peabody Picture Vocabulary Test—Third Edition; Cattell
Cattell Culture Fair Intelli-
gence Test; AST
alpha span task; SST
sequencing span task.

Table 4
Mean Reaction Time (in Milliseconds) and Accuracy for Simon Task by Age and Language
Group in Study 2

Age and
No. of
language group
Simon effect
337 (16.4)
544 (42.2)
667 (76.2)
123 (88.8)
583 (61.8)
802 (69.5)
890 (33.9)
88 (80.1)
343 (27.0)
375 (42.1)
382 (39.9)
8 (27.3)
456 (66.4)
509 (84.6)
509 (90.4)
0 (29.9)
1,012 (216.2)
1,012 (280.6)
1,595 (384.4)
583 (174.9)
1,716 (320.6)
1,336 (334.2)
2,210 (547.9)
874 (280.9)
1,046 (204.0)
889 (231.2)
1,101 (267.8)
212 (180.6)
1,256 (368.9)
1,002 (212.5)
1,266 (284.2)
264 (249.0)
Standard deviations are in parentheses.
side–2, F(1, 90)
57.58, p
.01; center– 4, F(1, 90)
32.09, p
among age, language, and position, F(1, 90)
4.05, p
.01; side– 4, F(1, 90)
111.88, p
.01. There were interactions
Thus, larger working memory costs were associated with older
of language and age group for the conditions based on four colors:
adults, with monolingualism as opposed to bilingualism, and with
center– 4, F(1, 90)
14.16, p
.01; side– 4, F(1, 90)
central as opposed to peripheral (side) stimuli. As shown by
.01. In both cases, the age variable was associated with a
Figure 1a, the age-related increase in working memory costs was
larger increase in RT for monolingual participants: 1,133 ms
much smaller for bilingual participants; that is, bilingualism atten-
versus 800 ms for the center– 4 condition; 927 ms versus 625 ms
uates the negative effect of aging on working memory costs.
for the side– 4 condition. Language and age did not interact in the
In our view, the difference between RTs to congruent and
center–2 and side–2 conditions.
incongruent stimuli (the Simon effect) reflects the efficiency of
The relative effects of increasing the number of possible stimuli
inhibitory processes. That is, the participants’ task is to press the
from two to four—referred to here as working memory costs—
key associated with the stimulus color regardless of spatial posi-
were assessed by subtracting two-color RTs from four-color RTs
tion; therefore, smaller Simon effects reflect less inhibition cost
in all conditions and groups. The resulting means are shown in
and more efficient inhibitory processes. These costs are shown in
Figure 1a. The corresponding ANOVA (Age Group
Figure 1b. Larger costs are associated with older adults, with
Position) revealed main effects of age group, F(1, 90)
monolinguals, and with four-color conditions. A three-way
71.1, p
.01; language, F(1, 90)
129.0, p
.01; and position
ANOVA on the data shown in Figure 1b revealed significant
(center vs. side), F(1, 90)
17.6, p
.01. In addition, all
effects of age, F(1, 90)
307.3, p
.01; language, F(1, 90)
interactions were significant: Age
Language, F(1, 90)
146.6, p
.01; and number of stimuli (two or four), F(1, 90)
.01; Age
Position, F(1, 90)
20.9, p
.01; Language
17.8, p
.01. In addition, the following interactions were signif-
Position, F(1, 90)
7.13, p
.01; and the three-way interaction
icant: Age
Language, F(1, 90)
63.3, p
.01; Age
F(1, 90)
29.4, p
.01; Language
Number, F(1, 90)
.01; and the three-way interaction among age, language, and
Table 5
number, F(1, 90)
14.18, p
.01. Figure 1b shows that the
Mean Reaction Time (in Milliseconds) by Age and Language
age-related increase in the Simon effect was less when only two
Group for Each Experimental Condition
colors were involved and was less for bilingual participants. Fur-
ther analyses showed that the interaction between age and lan-
Condition and language group
guage group was reliable for both the two-color, F(1, 90)
.01, and four-color, F(1, 90)
57.04, p
.01, conditions
even though the effect was smaller for the two-color conditions.
That is, in both color conditions, the age-related increase in the
Simon effect was smaller for the bilingual participants.
Finally, we divided participants into decades of age to obtain a
more complete picture of the transition across this age span. The
numbers of participants in each decade were as follows: 30s, n
24; 40s, n
22; 50s, n
18; 60s, n
15; 70s, n
15. Figure 2a
displays the RTs for both language groups in the control condition
(center–2) and shows that the response times in the simplest

Figure 1.
Mean reaction time (RT) cost for working memory and inhibition by age and language group in
Study 2. (a): Working memory cost calculated as RT difference between four- and two-color conditions for
central (Condition C – Condition A) and side (Condition D – Condition B) presentations. (b): Inhibition cost
calculated as RT difference between incongruent and congruent trials for two-color (Condition B) and four-color
(Condition D) presentations. SE 2
Simon effect, 2 colors; SE 4
Simon effect, 4 colors.
condition did not distinguish between the language groups at any
samples in the performance of a straightforward choice RT task. In
age. Figure 2b shows the working memory costs, calculated as the
all other conditions, bilinguals achieved faster response times than
RT difference between the two- and four-color presentations av-
did monolinguals of the same age.
eraged across the central and peripheral display presentations.
The striking finding, shown in Figure 1, is that the costs for both
Figure 2c shows the inhibitory costs, calculated as the RT differ-
inhibition and working memory were greater for the monolinguals
ence between congruent and incongruent trials (i.e., the Simon
than the bilinguals in both age groups, and the increased RT
effect) for the two language groups averaged across the side–2 and
associated with aging for each of these factors was greater for the
side– 4 conditions. Although no formal analyses were conducted
monolinguals than the bilinguals. The age-related processing de-
on these data, it is clear from the figures that performance re-
cline associated with these factors, in other words, was more
mained constant until age 60 and then RTs increased over the next
severe for the monolinguals than for comparable bilinguals.
20 years. Figures 2b and 2c further show that the age-related
In the Simon conditions, the bilinguals were faster than the
increase in costs was greater for monolingual participants.
monolinguals on both the congruent and incongruent trials, but
more important, as in Study 1 for which practice levels were much
lower, the bilinguals showed a reliably smaller Simon effect than
As in Study 1, monolingual and bilingual adults who were either
the monolinguals. For the younger bilinguals, up to the age of 60
younger (approximately 40 years old) or older (approximately 70
years the Simon effect was very small (only 4 ms overall), repli-
years old) were equivalent on background measures of cognitive
cating the results of Study 1. The older bilinguals in the present
performance and working memory as well as on a number of social
study (and in Study 1) did show a Simon effect, but its magnitude
and educational factors, which made the lifelong bilingualism of
remained significantly smaller than that for monolinguals of the
one group the only notable difference between them. In addition,
same age (see Figure 1b).
in both the younger and older groups, the monolinguals and
The four-color conditions added a surprising amount of diffi-
bilinguals performed identically in the control condition, in which
culty to the task. For the younger participants in both language
two colored squares were presented in the center of the screen (see
groups, the cost of remembering and processing four colors rather
Figure 2a). This important result underlines the fact that there were
than two was greater than the cost of inhibiting the misleading
no inherent differences between the monolingual and bilingual
position cues (compare Figures 2b and 2c). Also surprisingly, the

Figure 2.
Mean reaction time (RT) by decade of age for monolinguals and bilinguals. (a): Mean RT for control
condition (Condition A). (b): Mean RT cost for working memory (WM) as the average of the RT difference
between the two-color (Condition C – Condition A) and four-color (Condition D – Condition B) conditions. (c):
Mean RT cost for inhibition (Simon effect) as the average of the RT difference between congruent and
incongruent trials for two-color (Condition B) and four-color (Condition D) conditions.
increase from two to four colors was handled better by the bilin-
One result that should be noted is that the age-related decreases
guals than the monolinguals, for both the younger and older
in alpha span and sequencing span were not modulated by bilin-
participants (see Figure 1a). It seems, then, that the bilingual
gualism. Our purpose in including these two span measures was to
advantage appears in situations with high processing demands
establish the equivalence between the two language groups on a
(e.g., four colors vs. two colors) and is not restricted to conditions
measure of cognitive processing capacity or working memory.
necessitating inhibition (e.g., incongruent vs. congruent), although
Given that we found an unexpected bilingual advantage for work-
the advantage is also found in such situations. We return to this
ing memory costs and concluded that bilingual participants per-
important point in the General Discussion.
form more efficiently in tasks with high processing demands, it is

perhaps surprising that the bilingual advantage was not also found
thus represent the processes of major interest. As the experiment
for these complex span tasks. It is known that performance on span
was exploratory, we tested a group of younger adults, comparable
tasks reflects (in part) the ability to inhibit interference from
to the younger group in Study 2.
previous trials (Lustig, May, & Hasher, 2001; May, Hasher, Kane,
1999), and it might therefore be expected that bilinguals should
show an advantage on such tasks. Speculatively, the bilingual
advantage may be greater on tasks requiring speeded responses, or
less on tasks involving the manipulation of words in one of their
The participants were 20 adults ranging in age from 30 to 55 years. Half
two languages, but a proper understanding of tasks showing the
were French–English bilinguals (mean age
40.6 years), and half were
bilingual advantage must await the results of further exploratory
English-speaking monolinguals (mean age
38.8 years) living in the same
Canadian community. Participants in the two groups were matched on age,
The analysis by decades provides a more detailed description of
with the exception of the oldest member of each group. There were equal
the processing changes associated with this task for the two lan-
numbers of men and women in each group. Participants were recruited
guage groups. Although the groups maintained a constant differ-
through advertisement in a local community center. The bilinguals had
ence in performance until 60 years of age, the pattern of RT
been exposed to both languages at home from childhood and were educated
in both languages. The language background questionnaire indicated that
increase beyond that age was different for the three effects. The
they used English 50% of the time in their daily lives. All participants had
control condition indicated a constant increase over the last two
bachelor’s degrees and were from similar socioeconomic backgrounds.
decades that was identical for both language groups. This presum-
ably reflects the general slowing associated with normal aging. For
Tasks and Procedures
working memory costs, the bilinguals maintained their perfor-
mance levels until 70 years of age and then showed an increase in
Five tasks from Study 2 were repeated in this study: the language
RT for the last decade tested. The monolinguals, in contrast, began
background questionnaire, the PPVT–III, the Cattell Culture Fair Intelli-
to decline in their 60s and continued to decline in their 70s. For the
gence Test, the AST, and the SST. The procedures used to administer and
Simon effect, both groups revealed a sharp increase in RTs be-
score all these tests were the same as those described in Study 2.
tween 60 and 70 years, after which the monolinguals continued to
Two conditions of the Simon task used in Study 2 were administered, the
side–2 condition and the center– 4 condition. Each of these was presented
show increased RTs but the bilinguals remained constant. There-
in 10 consecutive blocks of 24 trials with a short break between each
fore, in spite of significant slowing in response to both working
repetition. The order of the two conditions was counterbalanced across
memory and inhibition demands, bilinguals beyond 60 years of age
participants. Testing began with the language background questionnaire,
continued to maintain an advantage over monolinguals in respond-
the PPVT–III, and the AST. These were followed by the first Simon task
ing to both these factors.
condition. Between the two Simon conditions, the Cattell Culture Fair
In summary, three main results were found in Study 2. First, the
Intelligence Test and the SST were administered.
bilinguals showed a reliably smaller Simon effect than the mono-
linguals—the difference between congruent and incongruent trials
was not even significant for the younger bilingual participants.
Second, bilingualism reduced the age-related increase in process-
The results for the background variables are presented in Table
ing costs associated with four stimulus colors (working memory
6. One-way ANOVAs testing for language group differences
costs). Third, both the Simon effect, taken here to indicate the
found no group difference for any of these measures.
efficiency of inhibitory processing, and working memory costs,
The performance accuracy in the side–2 condition was 97.5%
reflecting the ability to deal with increasing task complexity,
for the monolinguals and 95.4% for the bilinguals; in the center– 4
increased reliably with age, but these age-related increases were
condition, accuracy was 90% for the monolinguals and 91.7% for
also reliably attenuated by bilingualism.
the bilinguals. The difference between conditions was significant,
F(1, 18)
5.11, p
.04, but there was no difference between the
groups and no interaction (Fs
Study 3
The mean RTs on the center– 4 task for the two language groups
In Study 2, there were large differences in RT between the
are shown across blocks in Figure 3a. In a three-way ANOVA for
monolinguals and bilinguals for three of the conditions in spite of
equivalent performance on the control, or center–2, condition.
However, even that study included fewer trials than are usually
Table 6
used in this type of research. Therefore, in the final study, we
Mean Background Measures (and Standard Deviations) by
verified the group effects by presenting two of the conditions from
Language Group in Study 3
Study 2 to a new group of monolingual and bilingual adults but
repeating the blocks of trials 10 times. The purpose was to deter-
mine whether the two language groups would eventually converge
Age (in years)
38.8 (8.5)
40.6 (8.1)
in their performance after sufficient practice.
89.1 (6.1)
91.0 (5.8)
The two conditions chosen were side–2 and center– 4, the first
110.0 (6.7)
111.5 (7.4)
representing the classic Simon task and the second representing a
27.8 (3.8)
28.1 (3.4)
straightforward working memory task uncomplicated by congruity
25.4 (4.2)
28.0 (4.0)
effects. That is, the side–2 condition requires inhibitory processes
Peabody Picture Vocabulary Test—Third Edition;
but no working memory load, whereas the center– 4 condition
Cattell Culture Fair Intelligence Test; AST
alpha span task;
involves working memory but not inhibition; these two conditions
sequencing span task.