Neural Correlates of Attentional Capture in Visual Search
Jan de Fockert
1
, Geraint Rees
2
, Chris Frith
2
, and Nilli Lavie
2
Abstract
& Much behavioral research has shown that the presence of
a unique singleton distractor during a task of visual search
will typically capture attention and thus disrupt target search.
Here we examined the neural correlates of such attentional
capture using functional magnetic resonance imaging in
human subjects during performance of a visual search task.
The presence (vs. absence) of a salient yet irrelevant color
singleton distractor was associated with activity in the superior
parietal cortex and frontal cortex. These findings imply that
the singleton distractor induced spatial shifts of attention
despite its irrelevance, as predicted from an AC account. More-
over, behavioral interference by singleton distractors was
strongly and negatively correlated with frontal activity. These
findings provide direct evidence that the frontal cortex is in-
volved in control of interference from irrelevant but attention-
capturing distractors. &
INTRODUCTION
Many studies have demonstrated that attention can be
easily directed toward a subset of stimuli that are
defined as goal relevant, as long as thes e stimuli are
different from goal-irrelevant stimuli on the basis of a
simple visual feature. For example, subjects can be very
efficient in searching for targets with a curved feature
among angular distractors (see Duncan & Humphreys,
1989; Treisman, 1988, for review).
However, although most of the distractor objects that
carry an irrelevant feature (e.g., all angular stimuli in the
example of focusing on targets with curved features) can
be successfully ignored, an irrelevant distractor with a
unique feature that makes it a singleton in the visual
field (e.g., an irrelevant red distractor presented in an
array of green objects) will typically distract attention
from focusing entirely on relevant stimuli (see Yantis,
1996, 2000, for review). Such interruption of goal-driven
attention can be found even when the distractor object
forms a singleton on a dimension that is never relevant
to the task (e.g., a color singleton will interfere with
search on the basis of other features such as the shape
search task described above), suggesting that the single-
ton distractor has captured attention, rather than that
attention was voluntarily allocated to distractor pro-
cessing (e.g., Theeuwes, 1996).
This phenomenon of attentional captur e (AC) has
stimulated much behavioral res earch (see Egeth &
Yantis, 1997; Theeuwes, 1996; Yantis, 1996, 2000, for
reviews). In the present article, we examine the neural
correlates of AC. As behavioral research has demon-
strated that a salient, yet irrelevant, singleton distractor
will nevertheless capture attention, we anticipated that
neural systems known to be involved in the allocation of
attention to goal-relevant stimuli may also be associated
with AC by goal-irrelevant singleton distractors. Specifi-
cally, activity in the parietal cortex has been previously
associated with the allocation of attention in a variety of
tasks, including visual search (Corbetta, Shulman, Mie-
zin, & Petersen, 1995) and spatial cueing (e.g., Hop-
finger, Buonocore, & Mangun, 2000; Kastner, Pinsk, De
Weerd, Desimone, & Ungerleider, 1999; Rosen et al.,
1999; Nobre et al., 1997; Corbetta, Miezin, Shulman, &
Petersen, 1993; for reviews, see Corbetta & Shulman,
2002; Wojciulik & Kanwisher, 1999). We therefore ex-
pected that capture of attention by an irrelevant single-
ton distractor during visual search will also be associated
with parietal activity. Moreover, capture of atte ntion by a
goal-irrelevant distractor should also impose a greater
demand on top-down control mechanisms typically
associated with the frontal lobe (for review, see Duncan
& Owen, 2000; Miller, 2000), as these are needed in
order to resolve the competition between the target and
the irrelevant singleton distractor that has captured
attention. We thus expected that AC by an irrelevant
singleton will also implicate activity in frontal cortices
associated with such top-down control.
Finally, as AC by singlet on distractors produces clear
interference effects with visual search performance, we
sought to examine the relationship between the neural
activity related to AC by such distractors and the extent
to which they produce behavioral interference on visual
search. Previous imaging studies have successfully iden-
tified neural correlates of attention and top-down con-
trol, but have not been able to specify the implications
1
Goldsmiths College,
2
University College London
D 2004 Massachusetts Institute of Technology Journal of Cognitive Neuroscience 16:5, pp. 751–759
of this neural activity for the extent to which behavioral
interference is found when attention is captured by goal-
irrelevant distractors.
We measured brain activity using functional magnetic
resonance imaging (fMRI) in humans during perfor-
mance in a task of visual search for a unique shape target
(circle) among distractors of a different shape (diamonds;
see Figure 1). We assessed performance of this search for
shape in the presence versus absence of a salient, yet
irrelevant, color singleton distractor that is known to
produce AC (Theeuwes, 1991, 1992). We also assessed
performance of the shape search task in the presence (vs.
absence) of a color singleton on the shape target. Thus,
the effects of the presence (vs. absence) of a color
singleton distractor could be contrasted with the effects
of the presence (vs. absence) of a color singleton target.
Figure 2. Behavioral performance during scanning. Bars represent
mean RT and % error for targets and distractors, with a color singleton
present or absent.
Figure 1. Example stimuli of
each of the four experimental
trial types. Participants were
requested to make a speeded
key-press response to the
orientation of the line segment
in the target circle. On color
singleton present trials, one of
the display items (the circle on
target color singleton trials,
and one of the diamonds on
distractor color singleton
trials) was presented in red.
On color singleton absent
trials, all display items were
presented in green, and one of
them (the circle on target
singleton trials, and one of the
diamonds on distractor
singleton trials) was slightly
reduced in size.
Figure 3. Activity related to
the presence (vs. absence) of a
color singleton distractor.
Shown are posterior (left
panel), left lateral (middle
panel), and dorsal (right panel)
views of a T1-weighted
anatomical template image in
Talairach space (Talairach &
Tournoux, 1988). For display
purposes, activity is shown at
p < .001, uncorrected, with an
extent threshold of 200 voxels.
752 Journal of Cognitive Neuroscience Volume 16, Number 5
RESULTS
Behavioral Responses
Behavioral data were collected during the scanning
sessions (see Figure 2). An ANOVA comparing target
reaction times (RTs) showed th at the presence (vs.
absence) of a color singleton distractor produced signif-
icant interference [96 msec; F (1,9) = 38.4, p < .001].
This result is consistent with previous behavioral find-
ings in similar visual search studies of AC (e.g.,
Theeuwes, 1991, 1992, 1994). When the shape singleton
target was also presented in a unique color, there was a
small and nonsignificant trend for facilitation (9 msec;
F < 1).
1
Thus, participants were clearly able to select the
target on the basis of its unique shape, gaining very little
from presentation of the target in a unique color as well.
Note that although salient singleton distractors have
been consistently found to produce interference on
performance of the feature search task used here, the
effects of adding another s ingleton feature (e.g., a
unique color) to the feature target (e.g., a unique shape)
in this task have not been previously examined (facilita-
tion effects have been previously found when a singleton
feature has been added to more complex targets such
as letters; see Yantis, 2000, for a review). For the present
purpose, the finding that the RT interference was
specific to the presence of a color singleton distractor
[as also shown by the highly significant interaction,
F(1,9) = 21.8, p < .001, between color singleton pres-
ence (present, absent) and singleton stimulus (target,
distractor)] indicates that this interference w as not
simply due to the mere presence of an odd color in
the array, but was associated with the presentation of a
distractor, rather than a target, in a singleton color.
Thus, despite efficient selection of the shape target,
presentation of a color singleton distractor produced
robust interference.
Imaging Data
Brain activity time-locked to the individual trials was
determined using an event-related analysis (see Meth-
ods) of the fMRI data. Figure 3 shows areas of activity
associated with the presence (vs. absence) of color
singleton distractors (see also Table 1 for stereotactic
locations of the peak voxels in these areas of activation).
As we anticipated, the presence (vs. absence) of a color
singleton distractor was associated with bilateral activa-
tion of the superior parietal lobe (Brodmann’s area [BA]
7), an area previously associated with voluntary alloca-
tion of attention in a variety of tasks (e.g., Kastner et al.,
1999; Corbetta et al., 1993, 1995), but with involuntary
allocation of attention (as in AC) only in spatial cueing
tasks so far (Rosen et al. 1999; Nobre et al., 1997;
Corbetta et al., 1993). Moreover, as can be seen in
Figure 3, color singleton presence (vs. absence) was
also associated with activity in an area in the left lateral
precentral gyrus (BA 6) of the frontal cortex (anterior,
inferior, and lateral to the frontal eye fields; Paus, 1996).
The finding that the presence of irrelevant singletons is
associated with frontal activity is in line with our sug-
gestion that target selection in the presence of a com-
peting, attention-capturin g singleton distractor would
place a greater demand on top-down frontal control.
No voxels showed significant activity (at p < .05 cor-
rected for multiple comparisons) related to the pres-
ence (vs. absence) of a singleton color when it coincided
with the shape target. This result is consistent with the
lack of a signif icant behavioral effect in these conditions
and clearly rules out the mere presence of a unique
color in the array as a possible cause for the activity
related to the presentation of a distractor in a unique
color. Moreover, the parietal and frontal cortices also
showed significant interactions, such that activity in the
presence (vs. absence) of a color singleton distractor
was greater than activity in the presence (vs. absence) of
a color singleton target (left superi or parietal cortex,
t = 4.2, p < .01; right superior parietal cortex, t = 3.73,
p < .01; left frontal cortex, t = 4.4, p < .01, using small
volume correction). This result further confirmed that
activity in the parietal and frontal cortices related to the
presence (vs. absence) of a color singleton distractor
could not be attributed to the mere presence of an odd
color in the array.
To ensure that the interaction between singleton
presence and singleton stimulus was consistent across
participants, we extracted the neural signal ( blood oxy-
genation level dependent [BOLD] signal, expressed as
percent departure from a global mean of 100) for each of
the three significant areas of activation in the compari-
son between presence and absence of a color singleton
distractor (bilateral superior parietal cortex and left
frontal area BA 6) for each subject, and entered these
into an ANOVA with participants as the random factor.
For all three areas of activity, there was a highly sig-
nificant interaction, left superior parietal cortex, F(1,9) =
21.84, p < .001; right superior parietal cortex, F(1,9) =
18.85, p < .01; left frontal cortex, F(1,9) = 19.68, p < .01,
between the presence of a color singleton (vs. absence)
and the singleton stimulus (target vs. distractor). As can
be seen in Figure 4, activity in these areas was greater in
the color singleton distractor present conditions than in
the color singleton distractor absent con ditions, left
superior parietal cortex, t(9) = 3.6, p < .01; right
superior parietal cortex, t(9) = 2.9, p < .02; left frontal
cortex, t(9) = 2.8, p < .025, all two-tailed, whereas there
was no such difference for the presence (vs. absence)
of a color singleton target (left superior parietal cortex,
t = .07, p = .95; right superior parietal cortex, t = .65,
p = .53; left frontal cortex, t = 1.9, p = .092, all two-tailed).
These results confirm that the activity in the parietal and
frontal cortices, which is specifically related to the
presence of color singleton distractors rather than the
de Fockert et al. 753
presence of color singletons per se, was c onsistent
across participants.
These analyses included all (both correct and incor-
rect) trials. The activity revealed, however, could not
have been due to processes occurring only on error
trials for two reasons. First, the same contrast of activity
in distra ctor present versus absent conditions, excluding
trials on which an incorrect response was made, pro-
duced the same pattern of activity in the parietal and
frontal cortices (as when error trials are included).
Second, the areas revealed by a contrast of activity in
error trials versus correct trials did not include any of the
areas found in the comparison between distractor pres-
ent and absent conditions. Thus, activity o n error trials
could not explain the results due to the presence (vs.
absence) of the color singleton distractor.
In order to investigate the relationship between be-
havioral interference by single ton distractors and the
accompanying pattern of neural activity, we performed
a correlational analysis between the RT data and the fMRI
signal. First, to control for between-subject differences in
the overall fMRI signal and in the overall RT, we derived
an index of interference by representing the interference
effect (color distractor present minus color distractor
absent) per participant as a proportion of their average
fMRI signal and average RT in the distractor color single-
ton present and absent trials. We then computed Pear-
son’s correlation coefficient between RTs and the fMRI
signal in each of the three areas of significant activity.
No significant correlat ion was found in the two clusters
of activation in the b ilateral superior parietal cortex
(R = .247, p = .49 and R = .103, p = .78 for the left
and right superior parietal cortex, respec tively). How-
ever, there was a significant negative correlation between
activity in the left frontal cortex and the magnitude of the
interference effect in RT, R = .712, p = .021 (two-
tailed). The sign of this correlation is important, as it
indicates that greater activity in the frontal cortex (when
a color distractor was present vs. absent) is associated
with smaller interference effects by the irreleva nt dis-
tractors. Further analysis confirmed that greater interfer-
ence effects on RTs were not significantly correlated with
greater overall variance in RTs (assessed by the magni-
tude of standard deviation from mean RT per subject,
R = .46, p = .18). Thus, the negative correlation between
the interference effects on RT and activity in the frontal
cortex cannot be attributed to greater RT variability in
subjects with greater interference effects (compared to
those with smaller behavioral interference).
Eye Position Data
Subjects were request ed to maintain fixation at the
center of the d isplay. During scanning, eye p osition
was monit ored continually to ensure that participants
indeed succeeded to maintain fixation throughout the
experimental sessions. Figure 5 presents the frequency
of the vertical and horizontal eye positions across all
subjects, plotted as a function of trial type. Eye position
Figure 4. Activity associated with the interaction between color
singleton presence (present, absent) and singleton stimulus
(distractor, target). Bars represent BOLD signal change, averaged
across voxels in each cluster and across participants. Shown is the
difference in mean activity between color singleton present versus
absent, plotted separately for left superior parietal cortex (L SPL),
right superior parietal cortex (R SPL), and left lateral precentral
gyrus, and for distractor and target singletons. Error bars represent
interparticipant standard error.
Table 1. Regions of Activation Related to the Presence (vs. Absence) of Color Singleton Distractors
Talairach Coordinates
x y z t Value p Value (Corrected)
Left
Superior parietal lobe (BA 7) 24 66 50 5.85 .001
Lateral precentral gyrus (BA 6) 46 4 36 4.79 .018
Right
Superior parietal lobe (BA 7) 26 68 50 4.67 .030
Shown are voxels representing the peak activity in areas in which activity was greater than p < .05, corrected for multiple comparisons.
754 Journal of Cognitive Neuroscience Volume 16, Number 5
was consistently maintained within two degrees of fixa-
tion, less than the eccentricity of the search array (which
subtended 3 .18 from fixation to the center of each
display item). Moreover, there were no significant differ-
ences between any of the conditions (including fixation,
as well as the presence vs. absence of a color singleton
distractor), both in terms of individual subjects’ mean
eye position, and their standard deviations (all F < 1), in
either vertical or horizontal eye position. Thus, none of
the activations resulting from the comparisons of the
experimental conditions could be explained in terms of
eye movements (or, conversely, inhibition of eye move-
ments), as such accounts would predict a change in the
mean or the variance (or both) of the eye positions
between the e xperimental conditions.
DISCUSSION
The present results show that the neural correlates of
AC by an irrelevant color singleton in visual search are
the bilater al supe rio r pa rie tal cort ex an d left la ter al
precentral gyrus in the frontal lobe. Moreover, our
results also show a strong negative correlation between
the strength of the neural signal in the frontal cortex and
the magnitude of singleton distractor interference ef-
fects on behavior. These findings are in line with our
expectationthatanirrelevantsingletonwillcapture
attention and thus compete with the goal-relevant target
for selection, as we discuss below.
The Role of the Superior Parietal Cortex in AC
Activity in the superior parietal cortex has been typically
associated with spatial shifts of attention (see Corbetta &
Shulman, 2002, for a review, but see Hopfinger et al.,
2000, for an exception, emphasizing the involvement of
inferior parietal cortex in spatial attention shifts). The
finding that superior parietal cortex activity is associated
with the presence of an irrelevant singleton distractor in
visual search suggests that spatial attention was allocated
to the singleton distractor, consistent with an AC ac-
count of the behavioral effects. Specifically, many previ-
ous behavioral studies have shown that capture of
attention by an irrelevant singleton involves spatial shifts
of attention to the singleton position (e.g., Yantis &
Jonides, 1990; see Yantis, 2000, for a review).
2
More recently , Theeuwes, Kramer, Hahn, and Irwin
(1998) have found that AC by a singleton distractor not
only involves covert shifts of attention, but can also
involve triggering an eye movement to the location of
the singleton distractor. The superior parietal cortex has
been associated with both covert shifts of atte ntion (that
do not involve eye movements) and overt shifts of
attention (that do involve eye movements; Corbetta,
1998; Corbetta et al., 1998). In the present study, we
focused on the potential effects of singleton distractors
on covert attention, and therefore requested subjects to
maintain fixation at the center of the display, while
monitoring their eye position during scanning. The eye
position data confirmed that the presence (vs. absence)
of a color singleton distractor did not result in any change
in the number or variance of eye movements. Activity
related to the presence of a color singleton distractor in
our study therefore cannot be attributed to eye move-
ments. Instead, it suggests that such distractors triggered
involuntary covert shifts of spatial attention.
It is perhaps worth noting that although serial spatial
shifts of attention may not be required for the search
process in this feature-search task (Treisman, 1988),
shifts of focused attention to the target position are
required for the orientation discrimination aspect of this
task (in order to discriminate the orientation of the
small line [0.58 of visual angle] within the target shape,
among the competing orientat ions in the nontarget
shapes). Thus, in the absence of a singleton distractor,
although the target will initially pop out, focused atten-
tion will be shifted to it in order to perform the orien-
tation discrimination task. When the singlet on distractor
is present, however, it will pop out more readily than the
target (due to its greater salience; see Theeuwes, 1992),
and thus may be wrongly selected for a spatial shift of
attention. Thus, the presence of a singleton distractor
should involve an extra shift of spatial attention (as
Figure 5. Frequency plot of
vertical and horizontal eye
position. Data are shown for
each trial type, averaged across
subjects. Arrows indicate
eccentricity of the visual array.
In both the vertical and the
horizontal eye position data,
there was no difference
between trial types, both in
terms of individual subjects’
mean eye position, and their
standard deviations (all F < 1).
de Fockert et al. 755
