Cognitive control in media multitaskers

Abstract

Chronic media multitasking is quickly becoming ubiquitous, although processing multiple incoming streams of information is considered a challenge for human cognition. A series of experiments addressed whether there are systematic differences in information processing styles between chronically heavy and light media multitaskers. A trait media multitasking index was developed to identify groups of heavy and light media multitaskers. These two groups were then compared along established cognitive control dimensions. Results showed that heavy media multitaskers are more susceptible to interference from irrelevant environmental stimuli and from irrelevant representations in memory. This led to the surprising result that heavy media multitaskers performed worse on a test of task-switching ability, likely due to reduced ability to filter out interference from the irrelevant task set. These results demonstrate that media multitasking, a rapidly growing societal trend, is associated with a distinct approach to fundamental information processing.

Full text

Cognitive control in media multitaskers
Eyal Ophir
a
, Clifford Nass
b,1
, and Anthony D. Wagner
c
aSymbolic Systems Program and bDepartment of Communication, 450 Serra Mall, Building 120, Stanford University, Stanford, CA 94305-2050;
and cDepartment of Psychology and Neurosciences Program, Jordan Hall, Building 420, Stanford University, Stanford, CA 94305-2130
Edited by Michael I. Posner, University of Oregon, Eugene, OR, and approved July 20, 2009 (received for review April 1, 2009)
Chronic media multitasking is quickly becoming ubiquitous, al-
though processing multiple incoming streams of information is
considered a challenge for human cognition. A series of experi-
ments addressed whether there are systematic differences in
information processing styles between chronically heavy and light
media multitaskers. A trait media multitasking index was devel-
oped to identify groups of heavy and light media multitaskers.
These two groups were then compared along established cognitive
control dimensions. Results showed that heavy media multitaskers
are more susceptible to interference from irrelevant environmental
stimuli and from irrelevant representations in memory. This led to
the surprising result that heavy media multitaskers performed
worse on a test of task-switching ability, likely due to reduced
ability to filter out interference from the irrelevant task set. These
results demonstrate that media multitasking, a rapidly growing
societal trend, is associated with a distinct approach to fundamen-
tal information processing.
attention 兩cognition 兩executive function 兩multitasking 兩
working memory
In an ever-more saturated media environment, media multi-
tasking—a person’s consumption of more than one item or
stream of content at the same time—is becoming an increasingly
prevalent phenomenon, especially among the young (1). Re-
searchers have examined the immediate effects of multitasking,
and of media multitasking in particular, on memory, learning,
and cognitive functioning (2–4). However, it is unknown
whether and how chronic heavy multitaskers process informa-
tion differently than individuals who do not frequently multitask
(viewing multitasking as a trait, not simply a state). This issue
seems especially pertinent in light of evidence that human
cognition is ill-suited both for attending to multiple input
streams (5, 6) and for simultaneously performing multiple tasks
(7–9). Is breadth-biased media consumption behavior mirrored
by breadth-bias in cognitive control? That is, are chronic mul-
titaskers more attentive to irrelevant stimuli in the external
environment and irrelevant representations in memory?
The present research addressed this question via a series of
cognitive control studies comparing chronic heavy media mul-
titaskers to those who infrequently multitask. The goal was to
examine whether there is a relationship between chronic media
multitasking and cognitive control abilities. One possibility is
that chronic media multitaskers exhibit advantages in cognitive
control, which would motivate future work to establish whether
heavy multitasking confers or reflects these advantages. Alter-
natively, if heavy media multitasking behavior is associated with
deficits in cognitive control, such a finding would offer important
prescriptive guidance irrespective of the direction of causality. If
chronic media multitasking is the cause, then a change in
multitasking behav ior might be warranted. Conversely, if chronic
media multitasking behavior is more frequently engaged in by
individuals least able to cope with multiple input streams, then
behavior change may confer particular benefits to these indi-
viduals as they would have to deal with fewer distractors.
Results
Media Multitasking Index. To identify people who are heavy vs.
light media multitaskers, we developed a questionnaire-based
media multitasking index to determine the mean number of
media a person simultaneously consumes when consuming me-
dia and selected those individuals who were heavy media mul-
titaskers (HMMs were one standard deviation or more above the
mean) or light media multitaskers (LMMs were one standard
deviation or more below the mean) on this index. We then
examined these groups’ abilities on cognitive control dimensions
that could indicate a breadth-bias in cognitive control at differ-
ent control loci: the allocation of attention to environmental
stimuli and their entry into working memory, the holding and
manipulation of stimulus and task set representations in working
memory, and the control of responses to stimuli and tasks.
Filtering Environmental Distractions: Filter and AX-CPT Tasks. Inatest
of filtering ability (10)—an ability that can point to a breadth
orientation in allowing stimuli into working memory—
participants viewed two consecutive exposures of an array of
rectangles and had to indicate whether or not a target (red)
rectangle had changed orientation from the first exposure to the
second, while ignoring distractor (blue) rectangles (Fig. 1A). We
measured performance for arrays with two targets and 0, 2, 4, or
6 distractors. Repeated-measures ANOVA revealed a
group*distractor level interaction (Fig. 1B), F(1, 39) ⫽4.61, P⬍
0.04: HMMs’ performance was linearly negatively affected by
distractors, F(1, 18) ⫽9.09, P⬍0.01, whereas LMMs were
unaffected by distractors, demonstrating that LMMs have the
ability to successfully filter out irrelevant stimuli, F(1, 21) ⫽0.18,
P⬎0.68.
Further evidence for HMMs’ tendency to allow irrelevant
stimuli into working memory emerged on the AX-CPT variant
(11, 12) of the Continuous Performance Task (13). This task
examined whether HMMs and LMMs differ in their represen-
tation and maintenance of context. Participants viewed cue-
probe pairs of letters, and were to respond ‘‘yes’’ when they saw
the target cue-probe pair, ‘‘AX’’, that is, an ‘‘A’’ (cue) followed
by an ‘‘X’’ (probe). All other combinations—‘‘A’’ then not-‘‘X’’
(‘‘AY’’), not-‘‘A’’ then ‘‘X’’ (‘‘BX’’), and not-‘‘A’’ then not-‘‘X’’
(‘‘BY’’) were to be responded to with a ‘‘no’’ button press. In
addition to the standard version of the AX-CPT, we adminis-
tered a second version using distractor letters, identified by a
different color, that intervened between cue and probe (14). In
this version, participants were to ignore letters marked as
distractors and to perform the task as if they did not exist.
Performance analyses revealed no significant differences be-
tween HMM and LMM performance in the standard AX-CPT
in either accuracy (d⬘), t(28) ⫽0.51, P⬎0.61, or response times,
t(28) ⫽⫺0.16, P⬎0.88. However, in the AX-CPT with
distractors, HMMs were 77 ms slower to respond to the probes,
t(28) ⫽⫺3.33, P⬍0.002 (Fig. 2), even though there was again
no difference in accuracy, t(28) ⫽0.01, P⬎0.99. The response
time difference was driven by responses to those trials where an
Author contributions: E.O., C.N., and A.D.W. designed research; E.O. performed research;
E.O. and C.N. analyzed data; and E.O., C.N., and A.D.W. wrote the paper.
The authors declare no conflict of interest.
This article is a PNAS Direct Submission.
See Commentary on page 15521.
1To whom correspondence should be addressed. E-mail: nass@stanford.edu.
www.pnas.org兾cgi兾doi兾10.1073兾pnas.0903620106 PNAS
兩
September 15, 2009
兩
vol. 106
兩
no. 37
兩
15583–15587
PSYCHOLOGICAL AND
COGNITIVE SCIENCES
SEE COMMENTARY

X-probe was presented, and participants had to refer to the cue
they maintained in the face of distractors (AX and BX trials):
HMMs were 84 ms slower than LMMs to respond to AX trials,
t(28) ⫽⫺3.27, P⬍0.003, and 119 ms slower to respond to BX
trials, t(28) ⫽⫺3.25, P⬍0.003, yielding a significant LMM/
HMM status*presence of distractors interaction, F(1, 28) ⫽
5.21, P⬍0.03. These data replicate the results from the filter
task, again demonstrating that HMMs are less selective in
allowing information into working memory, and are therefore
more affected by distractors. As target trials comprised 70% of all trials in the standard
version of the AX-CPT, the task was also indicative of the
participants’ ability to withhold prepotent responses, i.e., their
ability to withhold a target response on the relatively rare BX or
AY trials, each of which constituted only 10% of trials. The lack
of significant differences between the groups, reinforced by the
absence of a group difference on the Stop-Signal task (15), t
(37) ⫽⫺0.15, P⬎0.88, suggests that the two groups do not differ
in their level of response control.
Filtering Irrelevant Representations in Memory: Two- and Three-Back
Tasks. In the two- and three-back tasks (16), which examine the
monitoring and updating of multiple representations in working
memory, HMMs showed a significantly greater decrease in
performance (d⬘) from the two- to the three-back task;
task*HMM/LMM status interaction, F(1, 28) ⫽4.25, P⬍0.05.
Interestingly, although both groups showed similar decreases in
hit-rates (the number of targets correctly identified) from the
two-back to the three-back task, F(1, 28) ⫽0.14, P⬎0.72 (Fig.
3A), HMMs showed a greater increase in their false alarm rate
(the number of nontargets incorrectly marked as targets), F(1,
28) ⫽5.02, P⬍0.03 (Fig. 3B). This effect was driven by target
letters that had previously appeared during the task, but were
outside the range participants were instructed to hold in mem-
ory. Specifically, in the three-back task, HMMs were more likely
to false alarm to letters that had more previous appearances, F
(1, 13) ⫽6.31, P⬍0.03. This indicates that the HMMs were more
Fig. 1. The filter task. (A) A sample trial with a 2-target, 6-distractor array.
(B) HMM and LMM filter task performance as a function of the number of
distractors (two targets). Error bars, SEM.
Fig. 2. AX-CPT mean response times in the no-distractors and the distractors
conditions (note that the overall decrease in response times from the no
distractors to the distractors condition is due to greater predictability of probe
onset as a result of the rhythmic nature of the distractors; the key data point
is the difference in the distractors condition between LMMs and HMMs). Error
bars, SEM.
Fig. 3. Two- and three-back task results. (A) Hit rates. (B) False alarm rates.
Error bars, SEM.
15584
兩
www.pnas.org兾cgi兾doi兾10.1073兾pnas.0903620106 Ophir et al.

susceptible to interference from items that seemed familiar, and
that this problem increased as working memory load increased
(from the two- to the three-back task). This problem also became
more acute for HMMs as the task progressed, because proactive
interference from irrelevant letters accumulated across the
experiment. Specifically, a general linear model of the likelihood
of a false alarm in the three-back task revealed: (a) no main
effect of HMM/LMM status nor of time, but (b) a significant
HMM/LMM status*time interaction, such that the number of
false alarms increased over time more rapidly for HMMs, B⫽
0.081, P⬍0.001. These data demonstrate that HMMs are not
only less capable of filtering out irrelevant stimuli from their
environment, but also are less capable of filtering out irrelevant
representations in memory.
Task Switching. To compare the two groups’ task-set switching
abilities, we used a task-cued stimulus-classification procedure
(17): participants were presented a number and a letter, and
performed either a letter (vowel or consonant) or a number
(even or odd) classification task depending on a cue presented
before the stimulus. Switch cost was calculated as the difference
in mean response time between trials preceded by a trial of the
other type (switch trials) vs. trials preceded by a trial of the same
type (nonswitch trials). HMMs’ switch cost was 167 ms greater
than that of LMMs, t(28) ⫽⫺2.62, P⬍0.01; specifically, HMMs
were 426 ms slower to respond to switch trials, t(28) ⫽⫺2.66,
P⬍0.01, and 259 ms slower to respond to nonswitch trials, t
(28) ⫽⫺2.27, P⬍0.03. The difference in switch cost was not
simply a result of a general difference in performance on
response-time decision tasks: as reported earlier, LMMs and
HMMs did not exhibit differences in response times on the
AX-CPT task when no distractors were present. In addition,
switch cost was still significant when calculated in proportion to
each participant’s nonswitch trial mean response time, t(28) ⫽
⫺2.04, P⬍0.05. Because switch costs have been attributed to
competition from activation of the irrelevant task-set (18–20),
these results suggest that HMMs are less capable of filtering out
the irrelevant task-set representation in memory on a given trial.
This conclusion is reinforced by HMMs’ slowing on nonswitch
trials: such slowing in performance of mixed-task blocks has
been termed the ‘‘mixing cost’’ and is attributed to interference
from the currently irrelevant task (21).
Collectively, the data suggest that HMMs are less likely to
filter irrelevant representations arising from either external or
internal sources. To ensure that this different cognitive control
profile was not driven by general cognitive differences between
members of the HMM and LMM groups, we compared HMMs
and LMMs from an independent sample of participants on a
number of broader measures. This analysis revealed no signifi-
cant differences between the groups in SAT scores, need for
cognition (22), performance on a creativity task (23), or ratings
on the Big Five Trait Taxonomy—extraversion, agreeableness,
conscientiousness, neuroticism, and openness (24)—all t(30) ⬍
1.24, P⬎0.22; the Media Multitasking Index (MMI) also did not
differ with gender, t(31) ⫽0.19, P⬎0.85.
Furthermore, although individual differences on measures of
cognitive control may be driven by individual differences in
working memory capacity, evidence from the filter task suggests
that LMMs and HMMs do not differ in this regard. Specifically,
we examined LMM and HMM performance on the filter task for
arrays with 2, 4, 6, or 8 targets and no distractors (target-only
arrays); this is a direct measure of memory capacity (10). A
repeated-measures ANOVA clearly showed no main effect for
HMM/LMM status (F⬍1), nor an interaction between HMM/
LMM status and the number of targets (F⬍1). In addition, the
groups clearly did not differ with respect to any particular
number of targets [all t(39) ⬍1.04, all P⬎0.30]. Thus, we can
rule out differences in working memor y capacity between
HMMs and LMMs as the cause of differences in cognitive
control.
Finally, the analyses presented here focus on the approxi-
mately one-third of the population who would be conventionally
called ‘‘heavy’’ or ‘‘light’’ media multitaskers, that is, people who
were one or more standard deviations away from the mean. This
dichotomization should not lead to biases in the results, because
the distribution of multitasking is approximately normal, there
are no outliers, and we are not using the center of the distribution.
There was insufficient variance within conditions on the
cognitive control tasks to examine within-condition effects with
one exception (which reinforces our conclusion): among HMMs
performing the filter task, a regression of performance on MMI
scores and number of distractors yielded a significant and
negative MMI*number of distractors interaction, B⫽⫺0.04,
t(17) ⫽2.22, P⬍0.04, demonstrating that even among heavy
multitaskers, more intensive multitaskers are more susceptible
to distractors.
Discussion
The present research suggests that individuals who frequently
use multiple media approach fundamental information-
processing activities differently than do those who consume
multiple media streams much less frequently: their breadth-
biased media consumption behavior is indeed mirrored by
breadth-biased cognitive control. HMMs have greater difficulty
filtering out irrelevant stimuli from their environment (as seen
in the filter task and AX-CPT with distractors), they are less
likely to ignore irrelevant representations in memory (two- and
three-back tasks), and they are less effective in suppressing the
activation of irrelevant task sets (task-switching). This last result
is particularly striking given the central role attributed to task-
switching in multitasking (25).
These results suggest that heavy media multitaskers are dis-
tracted by the multiple streams of media they are consuming, or,
alternatively, that those who infrequently multitask are more
effective at volitionally allocating their attention in the face of
distractions. This may be a difference in orientation rather than
a deficit; that is, although the data reveal negative effects in
HMMs on performance of tasks that require cognitive control,
it remains possible that future tests of higher-order cognition will
uncover benefits, other than cognitive control, of heavy media
multitasking, or will uncover skills specifically exhibited by
HMMs not involving cognitive control.
The present data suggest that LMMs have a greater tendency
for top-down attentional control, and thus they may find it easier
to attentionally focus on a single task in the face of distractions.
By contrast, HMMs are more likely to respond to stimuli outside
the realm of their immediate task, and thus may have a greater
tendency for bottom-up attentional control and a bias toward
exploratory, rather than exploitative, information processing
(26, 27). If so, they may be sacrificing performance on the
primary task to let in other sources of information.
With the diffusion of larger computing screens supporting
multiple windows and browsers, chat, and SMS, and portable
media coupled with social and work expectations of immediate
responsiveness, media multitasking is quickly becoming ubiqui-
tous. These changes are placing new demands on cognitive
processing, and especially on attention allocation. If the growth
of multitasking across individuals leads to or encourages the
emergence of a qualitatively different, breadth-biased profile of
cognitive control, then the norm of multiple input streams will
have significant consequences for learning, persuasion, and
other media effects. If, however, these differences in cognitive
control abilities and strategies stem from stable individual
differences, many individuals will be increasingly unable to cope
with the changing media environment. The determination of
cause and effect and the implications of these differing strategies
Ophir et al. PNAS
兩
September 15, 2009
兩
vol. 106
兩
no. 37
兩
15585
PSYCHOLOGICAL AND
COGNITIVE SCIENCES
SEE COMMENTARY

for other types of information processing are critical issues for
understanding cognition in the 21st century.
Materials and Methods
This research was conducted in five parts: a media use questionnaire and
index, three sets of cognitive experiments (Parts I, II, and III), and a final set
of questionnaires administered online. All aspects of the study involved
informed consent of the participants and were approved by the Stanford
Human Subject Review Board.
Media Use Questionnaire and MMI.
Participants.
Two-hundred sixty-two uni-
versity students participated in the questionnaire for course credit. The ques-
tionnaire was administered online, and took approximately 20 min to
complete.
Questionnaire design.
The questionnaire addressed 12 different media forms:
print media, television, computer-based video (such as YouTube or online
television episodes), music, nonmusic audio, video or computer games, tele-
phone and mobile phone voice calls, instant messaging, SMS (text messaging),
email, web surfing, and other computer-based applications (such as word
processing). For each medium, respondents reported the total number of
hours per week they spend using the medium. In addition, they filled out a
media-multitasking matrix, indicating whether, while using this primary me-
dium, they concurrently used each of the other media ‘‘Most of the time,’’
‘‘Some of the time,’’ ‘‘A little of the time,’’ or ‘‘Never.’’ As text messaging could
not accurately be described by hours of use, this medium was discarded from
the analysis as a primary medium, although it still appeared as an option in the
matrix (meaning respondents could still report text messaging while being
engaged in other media).
Index creation.
To create the MMI, we assigned numeric values to each of the
matrix responses as follows: ‘‘Most of the time’’ (⫽1), ‘‘Some of the time’’
(⫽0.67), ‘‘A little of the time’’ (⫽0.33), and ‘‘Never’’ (⫽0). For each primary
medium, we summed the responses. This resulted in a measure of the mean
number of other media used while using each primary medium. To account for
the different amounts of time spent with each medium, the MMI was created
by computing a sum across primary media use weighted by the percentage of
time spent with each primary medium. Thus, the index is an indication of the
level of media multitasking the participant is engaged in during a typical
media-consumption hour. In summary, the formula is as follows:
MMI ⫽
冘
i⫽1
11
m
i
⫻h
i
h
total
where miis the number of media typically used while using primary medium
i,hiis the number of hours per week reportedly spent using primary medium
i, and htotal is the total number of hours per week spent with all primary media.
Index results.
The MMI produced a relatively normal distribution, with a mean
of 4.38 and standard deviation of 1.52. This suggests that there is not a
bimodal distribution of ‘‘heavy multitaskers’’ and ‘‘nonmultitaskers.’’ None-
theless, we can identify individuals who very frequently use multiple media
and those who tend to limit their use of multiple media. Media multitasking
was correlated with total hours of media use, r(260) ⫽0.46, P⬍0.001.
However, this is not an artifact of our measurement approach, because we
control for the total number of hours of media use in our computation of the
media multitasking index.
Part I: Filtering and Response Inhibition.
Participants.
Based on the question-
naire, those students with an MMI less than one standard deviation below the
mean (LMMs) or an MMI greater than one standard deviation above the mean
(HMMs) were invited to participate. The invitation yielded 22 LMMs and 19
HMMs who gave informed consent and participated in the study for course
credit.
Procedure.
Participants completed four tasks assessing different facets of
cognitive control: the Stroop Task, a task-switching procedure, a filtering task,
and a stop-signal task (data from the Stroop Task and this implementation of
the task-switching procedure are not included in this report). The tasks were
performed using a PST Serial Response Box and a Dell Poweredge computer
running EPrime 2.0 software, with stimuli presented on a Dell Trinitron 17’’
CRT display. All participants performed these tasks in the same order, taking
approximately 50 min to complete the entire study.
Filtering task.
Attention allocation and stimulus filtering refer to the ability to
willfully allocate attention to some stimuli in the environment, thus allowing
those stimuli to enter working memory, while preventing other, irrelevant
stimuli from entering working memory. In the filtering task, participants were
told they would view a number of different arrays of red and blue rectangles.
They were instructed to pay attention only to the red rectangles, and to ignore
the blue rectangles.
In each trial of the task, participants were presented with an array of red
and blue rectangles of differing orientations for 100 ms (Fig. 1A). After an
interval of 900 ms, a second array was presented, this time for 2,000 ms, and
participants were asked to indicate whether one of the red rectangles had
changed orientation (orientation changes consisted of rotation by 45° either
clockwise or counterclockwise, and no more than one red rectangle ever
changed orientation). Participants indicated that a change had taken place by
pressing a button marked ‘‘yes,’’ and that no change had taken place by
pressing a button marked ‘‘no.’’ Trials were separated by an interval of 200 ms.
To measure the participants’ filtering effectiveness, different numbers of
blue rectangles (distractors) were included in the arrays: 0, 2, 4, or 6. In
addition, the number of red rectangles (targets) also varied between 2, 4, 6,
and 8. The rectangles were evenly and randomly distributed within the display
area, and no two rectangles overlapped or were within one rotation of
overlapping. Target-distractor combinations were restricted so that the size of
the array never exceeded eight rectangles in total. For example, if there were
four target rectangles, there would only be zero, two, or four distractor
rectangles. Thus, there were 10 possible combinations.
After completing a practice session, participants performed a single block
of 200 trials, with an equal number of trials of each of the target-distractor
combinations, and an equal number of change and no-change trials within
each type of array. Trial order was randomized.
Presumably, if a person filters distractors effectively, an increase in the
number of distractors should have no effect on performance. Conversely, if a
person does not filter effectively, performance should decline as the number
of distractors rises. To test filtering ability, the effect of added distractors on
performance in the filtering task was examined in all trials with two targets.
We focused on the two-target arrays as these contained the widest range of
target-distractor combinations. In addition, two targets should be well within
the memory capacity of most participants, allowing for high initial perfor-
mance (in trials with target-only arrays) and thus more easily discernible
distractor effects. The comparison thus examines four possible arrays of
rectangles: two targets, with 0, 2, 4, or 6 distractors. We computed the
performance measure in terms of memory capacity (10, 28): K⫽S(H-F), where
Kis the memory capacity, Sis the size of the array, and Hand Fare the hit (a
correct indication that a rectangle had rotated) and false alarm (an incorrect
indication that a rectangle had rotated) rates, respectively. Because partici-
pants were explicitly told to ignore distractors, S, the size of the relevant array,
was set to the number of targets, that is, two.
Stop-signal task.
Response inhibition was measured using a stop-signal task. In
this task, participants were first presented 24 words (balanced for familiarity),
one at a time, in random order, and instructed to categorize them as either
animal or nonanimal as quickly as possible by pressing one of two buttons.
Each word appeared twice, for a total of 48 trials, constituting an initial timing
block. The participants were then presented the same words and asked to
make the same categorization—but to withhold their response if a tone (the
stop-signal) was heard before they entered their response.
After a practice session, the participants performed three blocks of 96 trials
each. The stop-signal was present on 25% of trials, with trials presented in
random order. When present, the stop-signal was presented 225 ms before
the mean response time as calculated based on the participant’s performance
in the initial timing block.
Part II: Two- and Three-Back Tasks.
Participants.
Thirty Stanford students took
part in this second study for course credit. Participants were recruited based on
their MMI, calculated from their responses to the Media Use Questionnaire, in
the same manner used for Part I. This time, 15 LMMs and 15 HMMs responded
to the invitation, and took part in the study after having given their informed
consent.
Procedure.
These studies used the same tools and setting used in the first study.
The N-Back tasks were administered after a Stroop Task (excluded from this
report). The entire study took approximately 40 min to complete.
Two- and three-back tasks.
To examine individual differences in the ability to
monitor and update multiple representations in working memory, we used
the two- and three-back tasks. In these tasks, participants were presented a
series of individual letters in the middle of the screen. Each letter was pre-
sented for 500 ms, followed by a white screen for 3,000 ms. Upon presentation
of a letter, participants were to indicate whether or not the present letter was
a ‘‘target,’’ meaning that it matched the letter presented two (for the two-
back task) or three (for the three-back task) trials ago. They pressed one
button for ‘‘target’’ and another for ‘‘nontarget.’’
Participants completed a practice session and then three blocks of the
15586
兩
www.pnas.org兾cgi兾doi兾10.1073兾pnas.0903620106 Ophir et al.

two-back task, with each block consisting of 30 trials, including 10 target and
20 nontarget trials. After completing the two-back task, participants per-
formed a training session and three blocks of the three-back task. Perfor-
mance was calculated using d⬘.
Part III: Task-Switching and AX-CPT.
Participants.
Thirty-two Stanford students
took part in this second study for course credit. Participants were recruited
based on their MMI, calculated from their responses to the Media Use Ques-
tionnaire, in the same manner used for Part I. Again, 15 LMMs and 15 HMMs
took part in the study after having given their informed consent.
Procedure.
These studies used the same tools and setting used in the first two
studies. The task-switching procedure was administered first, followed by the
AX-CPT and finally the AX-CPT with distractors. The entire study took approx-
imately 60 min to complete.
Task switching.
To measure the cost of switching between task sets, we used a
number-letter task. In this task, participants switched back and forth between
classifying numbers and classifying letters, according to a cue presented at the
outset of each trial. Participants were presented with one of two cues (‘‘NUM-
BER’’ or ‘‘LETTER’’) for 200 ms, followed by a stimulus that consisted of a
digit-letter pair (such as ‘‘2b’’ or ‘‘b2’’). Participants classified the stimuli using
two buttons, depending on the task indicated by the cue. If shown the
NUMBER cue, participants were to press the left button for an odd number
and the right button for an even number. If, conversely, participants were
shown the LETTER cue, they were to press the left button if the letter in the
stimulus was a vowel and the right button if it was a consonant.
The set of letters consisted of the vowels a, e, i, and u, and the consonants
p, k, n, and s. The set of even numbers consisted of 2, 4, 6, and 8, whereas the
odd numbers were 3, 5, 7, and 9. The relative positions of the number and
letter were counterbalanced across trials. The interval between cue offset and
stimulus onset was set to 226 ms, and the intertrial interval was set to 950 ms.
Participants performed practice sessions for number categorization, letter
categorization, and switching. Participants then performed the recorded
session, consisting of four blocks. Each block consisted of 80 trials, with an
equal frequency of 1, 2, 3, and 4 same-trial sequences, yielding 40% switch
trials and 60% nonswitch trials. The cost of switching between task sets was
computed by comparing mean response times in trials preceded by the same
type of trial (a nonswitch trial—the second of two consecutive ‘‘NUMBER’’ or
‘‘LETTER’’ trials) with mean response times in trials preceded by a different
trial type (a switch trial—a ‘‘NUMBER’’ trial preceded by a ‘‘LETTER’’ trial, or
vice versa).
AX-CPT.
To both measure HMMs and LMMs ability to maintain contextual
information and to conceptually replicate our results from the filter task, we
used the AX-CPT both without and with distractors. In the AX-CPT, partici-
pants viewed a sequence of letters presented for 300 ms each in red on a black
screen. The letters formed cue-probe pairs, such that 4,900 ms elapsed be-
tween presentation of cue and probe, and 1,000 ms between successive trials.
Participants were to maintain the cue in memory, which could be either the
letter ‘‘A’’ or some other letter (other than ‘‘X,’’ the target probe, and ‘‘K’’ and
‘‘Y,’’ for their similarity to ‘‘X’’), until they saw the probe, which could be ‘‘X’’
or some other letter (not including ‘‘A’’, ‘‘K,’’ or ‘‘Y’’). If they saw the cue ‘‘A’’
followed by the probe ‘‘X,’’ they were to press a button marked ‘‘YES.’’ For all
other cue-probe combinations, they were to press a button marked ‘‘NO.’’
Participants were also instructed to respond ‘‘NO’’ to all cue letters.
The AX-CPT with distractors has one difference from the previous task:
between every cue and probe, presented in red, participants saw three
additional, distractor letters (which could be any letter but ‘‘A,’’ ‘‘X,’’ ‘‘K,’’ or
‘‘Y’’) presented in white. Distractors were also presented for 300 ms, with a
1,000-ms interval between the cue and the first distractor, between each of
the distractors, and between the last distractor and the probe (thus maintain-
ing the 4,900-ms interval between cue and probe from the original AX-CPT).
Participants were instructed to respond with a ‘‘NO’’ button press to the
distractors, but to otherwise completely disregard them, such that a red ‘‘A’’
cue followed by a red ‘‘X’’ probe, no matter what white distractor letters were
in between, called for a ‘‘YES’’ response, and all other cue-probe pairs called
for a ‘‘NO’’ response.
MMI and Gender, SAT Scores, Need for Cognition, Creativity, Extraversion,
Agreeableness, Conscientiousness, Neuroticism, and Openness. To examine the
relationship of media multitasking and the above traits and measures, we
conducted a final online questionnaire. Here, a new group of 110 participants
filled out the MMI questionnaire as well as their SAT scores, gender, the Need
for Cognition index questionnaire, the Big Five Trait Taxonomy (including
measures of extraversion, agreeableness, conscientiousness, neuroticism, and
openness), and a creativity task derived from the Torrance Tests of Creative
Thinking (TTCT). The data from participants with an MMI one standard
deviation or more above and below the mean (16 HMMs and 17 LMMs) were
compared. All creativity task responses were coded by a single coder blind to
the participants’ MMI scores; to verify the reliability of the coding, a second
coder coded a subset consisting of the responses of 10 participants. Correla-
tions of scores from the two coders were high: r(100) ⫽0.97, P⬍0.001 for
scores of fluency, r(100) ⫽0.93, P⬍0.001 for scores of uniqueness of ideas, and
r(100) ⫽0.96, P⬍0.001 for scores of flexibility.
ACKNOWLEDGMENTS. We thank R. Poldrack, S. McClure, and two anonymous
reviewers for comments on earlier versions of this manuscript; S. Tandon for
her assistance in collecting the data, B. Robinson and B. Fuller at University of
Maryland for their E-Prime implementation of the two- and three-back tasks;
and the Stanford CHIMe Lab and the Stanford Memory Lab for their input
throughout this research. This work was supported by Stanford Major Grant
1093864–2007-AABSK (to E.O.), Volkswagen Grant 1114143–100-UBBEH (to
C.N.), Nissan Grant 1122033–100-UDUPP (to C.N.), and an Alfred P. Sloan
Foundation grant (to A.D.W.).
1. Roberts D-F, Foehr U-G, Rideout V (2005) Generation M: Media in the lives of 8-18
year-olds (HJKF Foundation, Menlo Park, CA).
2. Armstrong G-B, Chung L (2000) Background television and reading memory in context:
Assessing TV interference and facilitative context effects on encoding versus retrieval
processes. Communic Res 27:327–352.
3. Foerde K, Knowlton B-J, Poldrack R-A (2006) Modulation of competing memory
systems by distraction. Proc Natl Acad Sci USA 103:11778–11783.
4. Furnham A, Bradley A (1997) Music while you work: The differential distraction of
background music on the cognitive test performance of introverts and extraverts. Appl
Cogn Psychol 11:445–455.
5. Cherry E-C (1953) Some experiments on the recognition of speech, with one and with
two ears. J Acoust Soc Am 25:975–979.
6. Wood N, Cowan N (1995) The cocktail party phenomenon revisited: How frequent are
attention shifts to one’s name in an irrelevant auditory channel? J Exp Psychol Learn
Mem Cogn 21:255–260.
7. Dux P, Ivanoff J, Asplund C, Marois R (2006) Isolation of a central bottleneck of
information processing with time-resolved fMRI. Neuron 52:1109–1120.
8. Hein G, Alink A, Kleinschmidt A, Mu¨ ller N, He S (2007) Competing neural responses for
auditory and visual decisions. PLoS ONE 2:e320.
9. Marois R, Ivanoff J (2005) Capacity limits of information processing in the brain. Trends
Cogn Sci 9:296–305.
10. Vogel E-K, McCollough A-W, Machizawa M-G (2005) Neural measures reveal individual
differences in controlling access to working memory. Nature 438:500–503.
11. Cohen J-D, Barch D-M, Carter C, Servan-Schreiber D (1999) Context-processing deficits
in schizophrenia: Converging evidence from three theoretically motivated cognitive
tasks. J Abnorm Psychol 108:120–133.
12. Servan-Schreiber D, Cohen J-D, Steingard S (1996) Schizophrenic deficits in the pro-
cessing of context. A test of a theoretical model. Arch Gen Psychiatry 53:1105–1112.
13. Beck L-H, Bransome E-D, Jr, Mirsky A-F, Rosvold H-E, Sarason I (1956) A continuous
performance test of brain damage. J Consult Psychol 20:343–350.
14. Braver T, Barch D-M, Keys B-A, Carter C-S, Cohen J-D (2001) Context processing in older
adults: Evidence for a theory relating cognitive control to neurobiology in healty
aging. J Exp Psychol Gen 130:746–763.
15. Logan G-D (1994) Spatial attention and the apprehension of spatial relations. J Exp
Psychol Hum Percept Perform 20:1015–1036.
16. Cohen J-D, et al. (1994) Activation of the prefrontal cortex in a nonspatial working
memory task with functional MRI. Hum Brain Mapp 1:293–304.
17. Monsell S (2003) Task switching. Trends Cogn Sci 7:134–140.
18. Badre D, Wagner A-D (2006) Computational and neurobiological mechanisms under-
lying cognitive flexibility. Proc Natl Acad Sci USA 103:7186–7191.
19. Wylie G, Allport A (2000) Task switching and the measurement of ‘‘switch costs.’’
Psychol Res 63:212–233.
20. Yeung N (2006) Between-task competition and cognitive control in task switching.
J Neurosci 26:1429–1438.
21. Rubin O, Meiran N (2005) On the origins of the task mixing cost in the cuing task-
switching paradigm. J Exp Psychol Learn Mem Cogn 31:1477–1491.
22. Cacioppo J-T, Petty R-E (1982) The need for cognition. J Pers Soc Psychol 42:116–131.
23. Torrance E-P (1974) Norm-technical manual: Torrance Tests of Creative Thinking
(Personnel, Lexington, MA).
24. John O-P, Srivastava S (1999) in Handbook of Personality Theory and Research,ed
Pervin L-A (Guilford, New York), pp 102–138.
25. Pashler H (2000) in Attention and Performance XVIII: Control of Mental Processes, eds
Monsell S, Driver J (MIT, Cambridge, MA), pp 277–309.
26. Aston-Jones G, Cohen J-D (2005) Adaptive gain and the role of the locus coeruleus-
norepinephrine system in optimal performance. J Comp Neurol 493:99–110.
27. Daw N-D, O’Doherty J-P, Dayan P, Seymour B, Dolan R-J (2006) Cortical substrates for
exploratory decisions in humans. Nature 441:876– 879.
28. Pashler H (1988) Familiarity and visual change detection. Percept Psychophys 44:369–
378.
Ophir et al. PNAS
兩
September 15, 2009
兩
vol. 106
兩
no. 37
兩
15587
PSYCHOLOGICAL AND
COGNITIVE SCIENCES
SEE COMMENTARY