Canadian Journal of Experimental Psychology

2010, Vol. 64, No. 2, 124 128

2010 Canadian Psychological Association

1196-1961/10/$12.00 DOI: 10.1037/a0019109

Attentional Capture by Working Memory Contents

Yi Pan
Hangzhou Normal University
There has been controversy on whether working memory (WM) contents automatically guide attention.
The present study tried to replicate the effect of WM-based attentional capture using an adaption of
Downings (2000) paradigm, in which WM and attentional capture were combined. Subjects were
presented with an attention display containing two objects, one of which could be precued by a matching
item being held in WM. As measured by a probe discrimination task, the memory-matching object had
a privileged status to capture attention regardless of the stimulus onset asynchronies between the memory
cue and the attention display, even when there was absolutely no benefit for subjects to bias attention in
favour of the memory match. These results suggest that WM contents guide attention in an involuntary
manner. The implications of current findings for understanding of WM effects on visual selection are
discussed.
Keywords: working memory, attentional capture, automaticity

items match the task-irrelevant stimuli (memory-matching distractors)? Do the memory-matching distractors involuntarily capture
attention? Previous studies testing effects of WM on attention have
reported contradictory findings. Some researchers have reported
that the contents of WM guided attention in an involuntary fashion
(e.g., Downing, 2000; Soto, Heinke, Humphreys, & Blanco, 2005).
On the other hand, Woodman and Luck (2007) argued that WM
did not automatically guide attention, but rather, WM contents
could be used in a flexible manner according to their relationship
with the search task at hand. The present study aimed to replicate
the automatic effect of WM on visual selection using a task
paradigm similar to that used by Downing (2000).
In the first experiment of Downings (2000) study, participants
were shown a face at the beginning of each trial and were instructed to hold it in WM throughout the trial. After the offset of
the memory face, two prime faces were simultaneously presented,
one of which matched the memory face and the other novel. This
was followed by a probe briefly presented at the location of one of
the previous prime faces. The results showed that performance was
faster to the probe at the location of the memory-matching face
than to the probe at the location of the nonmatching face. Since
participants were told that the two prime faces were task irrelevant,
Downing suggested that this pattern of results reflected the automatic guidance of visual spatial attention from WM. In line with
this, Soto et al. (2005) also reported evidence that WM contents
could exert an involuntary influence over visual attention. Participants were asked to search for a left or right titled line amongst
vertical lines during the retention interval of WM. Each line was
surrounded by a colored shape that could match the WM contents.
In their Experiment 4, when there was a memory-matching shape
in the search display, it always contained a distractor rather than a
target. However, compared to a neutral baseline without a match
between WM contents and search arrays, performance was reliably
slower in the invalid condition where a distractor matched the
memory item, indicating that WM contents guided spatial attention
in an involuntary fashion.

Because there are usually much more stimuli in the scene than
the limited capacity visual system can process at any moment,
visual attention selects only a minority of them for further processing. Many models of attention in visual search suggest that
working memory (WM) contents play an important role in the
guidance of visual attention (e.g., Bundesen, 1990; Desimone &
Duncan, 1995; Duncan & Humphreys, 1989; Wolfe, 1994). The
target template held in WM serves to bias attention in a top-down
manner to process target-like objects in the visual scene. Singleunit recording studies provide the strongest evidence for the maintenance of target template in WM and its key role in the guidance
of attention during visual search (e.g., Chelazzi, Miller, Duncan, &
Desimone, 1993, 2001). For example, Chelazzi et al. (1993) recorded neural activity in the temporal lobe of monkeys while they
performed a delayed match-to-sample task. Chelazzi et al. found
that the neurons coded for the target maintained an elevated firing
rate during the delay interval. This was interpreted as evidence that
the target template was being actively held in WM during the
retention interval. Also, the target template provides a bias signal
to the neurons that perform perceptual analysis, inducing a competitive advantage for neurons that selectively respond to the
target.
Items held in WM should be deliberately used by observers to
bias the allocation of attention when they are related to the target
of the task at hand. However, what will happen when the memory

Yi Pan, Department of Psychology, Hangzhou Normal University.

This work was supported by the construct program of the key discipline
in Hangzhou, grants from Zhejiang Provincial Natural Science Foundation
(Y207628) and the Natural Science Foundation of China (30970893). This
article has benefitted from comments from Michael E. J. Masson for whom
I am very grateful. Special thanks to David Soto for his continued cooperation with me in our common area of research interest.
Correspondence concerning this article should be addressed to Yi Pan,
Department of Psychology, Hangzhou Normal University, Hangzhou,
310036, Peoples Republic of China. E-mail: panyirich@zju.edu.cn
124

WORKING MEMORY AND ATTENTION

However, other authors reported that they could not find any
evidence for the automatic influence over attention from WM.
Woodman and Luck (2007) asked participants to perform a search
task during the retention interval of WM. One of distractors in the
search array matched the colour of the memory cue on half of the
trials, and there was no match between the memory and search
displays on the other half. Woodman and Luck found no evidence
for attentional capture by WM. Of interest, they found that search
performance was even better when there was a memory-matching
distractor in the search display than when there was no match
between the memory cue and the search array (Experiment 3).
However, this effect was reversed when the search target occasionally matched the memory cue (Experiment 5B). Woodman and
Luck suggested that WM contents did not capture attention automatically, but they could be used flexibly for a more efficient
search. When the WM contents would never match the search
target, a template for rejection could be formed in the cognitive
control setting and it could be used to deviate attention away from
the location of the memory-matching distractor. In contrast, when
the WM contents could match the search target, the cognitive
control setting could form a template for selection to bias attentional allocation to the location of the memory match.
Therefore, there still has contradictory evidence on whether
WM contents involuntarily guide attention. In the present study, I
sought to replicate attentional capture by WM contents using an
adaptation of Downings (2000) paradigm, while also demonstrating the effect of partial matches between the WM contents and the
attention display on the top-down guidance of attention. Unlike the
design of Downing (2000), here the memory match was defined as
the object matching either the colour or the shape of the memory
cue. It is important to note that the probe never appeared at the
location of the memory match. Thusly, participants were provided with no explicit incentive for attending to the memory
match since doing this will delay the response latencies to
probes. If visual attention could be automatically guided to the
memory matches, then response times to probes should be much
slower on the trials containing the memory matches relative to
the neutral trials where there were no memory matches. In
addition, according to Woodman and Luck (2007), the stimulus
onset asynchronies (SOAs) between WM and attention displays
may be critical for the attentional capture by WM. The effect of
memory-driven capture of attention would disappear when the
SOAs were long enough for the cognitive control setting to formulate a template for rejection and exert its effect. To further
determine whether attentional capture by WM is confined to the
early stage of cognitive processing, the duration of the memory cue
and the interstimulus interval between WM and attention displays
were manipulated here.

Method
Participants
Twenty-two naive volunteers, aged 19 to 25 years, took part in
this study. All reported having normal or corrected-to-normal
visual acuity and normal colour vision. All participants were
right-handed. They received 15 for their participation.

125

Apparatus and Stimuli

The experiment was run on a Pentium IV computer with a
processor speed of 2.4 GHz. The stimuli were displayed on a
17-inch colour monitor with a resolution of 1,024 768 pixels and
a 85-Hz refresh rate. Responses were made on a standard keyboard. The visual stimuli were geometrical shapes and each was
filled of a colour (red, green, blue, yellow, or cyan). The shapes
could be a circle (3 3 of visual angle), a triangle (2.6 2.1),
a diamond (3 3), a pentagon (2.7 2.7), or a hexagon
(3 2.5). The response probe was a black 0.6 0.6 square
with a 0.5 gap at the top or bottom. All stimuli were presented on
a grey background.

Procedure and Design

Participants initiated each trial by pressing the space bar. A
white central fixation cross (0.2 0.2) was displayed for 1,000
ms, followed by the memory item presented at the centre of the
screen for 100 ms or 1,000 ms. Here, participants were instructed
to memorize both the colour and the shape of the memory item and
to keep it in mind through the entire trial. After a delay of 200 ms
or 1,506 ms, two prime stimuli (colored shapes) were simultaneously presented to the left and right of the central fixation for
187 ms, separated by approximately 18of visual angle from centre
to centre. Participants were told that the two prime stimuli were
task-irrelevant and they should not respond to them. After a 40-ms
delay, a small black square with a gap at the top or bottom was
displayed for 100 ms at the centre of the location previously
occupied by one of the two prime stimuli. Participants were
instructed to discriminate immediately the orientation of the small
square, by pressing the 1 key or the 2 key with the right
hand. Participants had 1,500 ms to respond to the probe after its
offset, followed by a memory-test object presented at the centre of
the screen and remained visible until response. Participants were
required to press the V key with their left hand if the object
matched the memory item in both dimensions of colour and shape,
and the N key with their left hand if they had just their colour in
common, just their shape in common, or neither attribute in common. Participants were encouraged to perform both tasks as accurately as possible.
If the memory item was presented for 100 ms, then the two
prime items would be displayed after a 200-ms delay. However, if
the memory item was presented for 1,000 ms, the primes would be
displayed after a 1,506-ms delay. The SOAs between the memory
item and the prime items were thusly 300 ms and 2,506 ms, and the
two SOAs occurred equally in the experiment. The two prime
items were always different from each other in both dimensions of
colour and shape on each trial. The memory match was defined as
the object sharing only one feature with the memory item. As
illustrated in Figure 1, there were three different match conditions.
In the colour match condition, one of the prime items shared its
colour but not its shape with the memory item. In the shape match
condition, one of the prime items shared its shape but not its colour
with the memory item. Colour matches and shape matches occurred on 30% of trials, respectively. The remaining 40% of trials
were the neutral match condition, in which neither of the features
of the memory item was shared by any of the prime items. These
were the same for each level of SOAs. Match condition and SOAs

PAN

126

Figure 1.

Schematic illustration of the experimental procedure and example stimuli.

varied randomly across trials. Of note, the attention probe never

appeared at the location of the memory match. Participants were
instructed that memory cues were always detrimental to the attention and memory tasks. The locations of the memory match and the
probe as well as its orientations were counterbalanced across trials.
Participants were first familiarised with the tasks and performed 20
practise trials. Then, they performed four blocks of 60 trials each.

Results
Errors averaged 2.3% on the attention probe task and 5.3% on
the memory task (see Table 1). An analysis of variance (ANOVA)
over probe errors with SOA and matching as factors showed no
significant effects (all ps .298). Memory errors were also
entered in an ANOVA with SOA and matching as factors. There
was an significant main effect of SOA, with more memory errors
when SOA was 300 ms than when SOA was 2,506 ms, F(1, 21)
29.37, p .001, 2 .586. This drop in performance at the
shorter SOA could be because of either an inability to identify
the stimuli or an inability to form durable WM representations.
The main effect of matching and its interaction with SOA did not
approach significance ( ps .658).
Analyses of reaction times (RTs) in the attention probe task included only trials on which both responses were correct. First, to

assess whether there was an overall effect of the memory matches, I

pooled the RTs in both conditions of the colour and shape match, and
compared this average with the performance in the neutral match
condition. An ANOVA with SOA and matching as factors showed
that response times were slower when SOA was 300 ms than when
SOA was 2,506 ms, F(1, 21) 11.599, p .003, 2 .356.
Response times were slower to probes on the trials containing the
memory matches than to probes on the neutral trials, F(1, 21)
25.389, p .001, 2 .547, and this did not vary as a function of
SOA (F 1).
Second, I carried out separate ANOVAs with SOA and matching
as factors on the colour match and the shape match conditions,
respectively. In the colour match condition, response times were
slower when SOA was 300 ms than when SOA was 2,506 ms, F(1,
21) 10.56, p .004, 2 .335. Response times were slower to
probes on the trials containing the colour matches than to probes on
the neutral trials, F(1, 21) 32.089, p .001, 2 .604, and this did
not vary as a function of SOA (F 1). Similarly, in the shape match
condition, the effect of SOA was significant, F(1, 21) 12.386, p
.002, 2 .371. Response times were slower on the trials containing
the shape matches than on the neutral trials, F(1, 21) 5.718, p
.026, 2 .214, but its interaction with SOA did not reach significance (F 1). Figure 2 illustrated this pattern of the results.

Table 1
Proportion of Probe and Memory Errors
Probe errors

Memory errors

Stimulus onset asynchronies

Color match

Shape match

Neutral match

Color match

Shape match

Neutral match

300 ms
2,506 ms

.031
.015

.029
.018

.026
.020

.078
.031

.076
.032

.071
.029

WORKING MEMORY AND ATTENTION

Figure 2. Mean reaction times (RTs) to attention probes for accurate

trials, as a function of matching and stimulus onset asynchronies (SOAs).
Error bars represent SEs.

Last, I assessed whether colour and shape information in WM

differed in terms of their efficacy at guiding visual attention. Here,
I compared the colour effect (colour-match RTneutral-match RT)
and the shape effect (shape-match RTneutral-match RT). A 2
(colour effect, shape effect) 2 (SOA) ANOVA showed that the
colour effect (27 ms) was larger than the shape effect (12 ms), F(1,
21) 4.984, p .031, 2 .192, and this did not vary as a
function of SOA (F 1).

Discussion
The present study, using the similar paradigm to Downing
(2000), replicated the effect of attentional capture by WM contents. Attention was automatically deployed to items that matched
the current contents of WM. According to the strategic perceptual
reexampling account of Woodman and Luck (2007), participants
in Downings (2000) study may have voluntarily deployed attention to the memory-matching item in order to refresh their memories, since the matching item was presented during the retention
interval on every trial. In the present study, however, the memory
match was presented on only 60% of trials, and importantly, the
upcoming probe never appeared at the location of the memory
match. Therefore, participants have no incentive for attending to
the memory match, because they were informed that strategically
shifting attention toward the matching item would be detrimental
to the probe task. Despite this, response times were slower to
probes in the memory match condition than to probes in the neutral
match condition, suggesting that WM contents guided spatial
attention in an involuntary manner.
The present study showed that partial matches between WM and
attention displays could guide attention, while colour had a stronger effect than shape did. This finding supports the idea that colour
can be more efficient than shape in biasing attention (Pan, 2009;
Soto et al., 2005). For example, Pan (2009) had participants
discriminate whether the colors (or the shapes) of two objects
simultaneously presented to the left and right side of the fixation
were the same. On half of the trials, the colors of the two objects

127

were the same and their shapes were different; the reverse was the
case on the other half. The results showed that response times were
faster to colour than to shape, indicating that colour has a competitive advantage to capture attention.
The current results are not likely due to the bottom-up priming,
since there has been clear evidence that merely processing stimuli
without encoding anything into WM leads no repetition priming
effects (e.g., Downing, 2000; Soto et al., 2005; Soto & Humphreys, 2007). This suggests that objects need to be encoded/
maintained in WM to guide visual attentional deployment. In fact,
WM effect and repetition priming are very different in terms of
neural mechanisms. A recent functional magnetic resonance imaging (fMRI) study by Soto, Humphreys, and Rotshtein (2007)
showed that the reappearance of an object held in WM enhanced
activity in superior frontal gyrus, midtemporal, and occipital areas,
but mere repetition of objects elicited suppressive responses in the
same regions (see also, e.g., Chelazzi et al., 1993; Desimone,
1996). The evidence from these studies suggests that content-based
WM-driven capture of visual attention reflects a top-down modulation of visual selection, whereas repetition priming operates in a
bottom-up manner.
The current findings seem to be inconsistent with those of
Woodman and Luck (2007). Woodman and Luck also provided
participants with an incentive for not attending to items that
matched WM contents, since the matching items were never search
targets. However, their results showed that items held in WM
guided attention away from rather than toward the matching distractors. While I believe that WM contents bias attention to the
matching stimuli by default, there appears to be several factors that
can modulate this default effect of WM-based attentional capture
(see Pan & Soto, 2009). Cognitive control may be the critical
factor that induced the lack of attentional capture by WM in
Woodman and Lucks (2007) study. When observers knew that the
target would never match the item held in WM, the cognitive
system could set the memory representation as a template for
rejection to direct attention away from the matching distractors.
However, it needs sufficient time for this cognitive control based
on top-down knowledge that target never matched WM contents to
be properly implemented (Han & Kim, 2009). The reason for the
results of Woodman and Lucks (2007) study could be that their
task allowed enough time for cognitive control to come into effect,
overriding the default prioritization of the memory matches. On
the other hand, there could be no enough time in the current study
for cognitive control to come into play, allowing the default effect
of WM-driven attentional capture to occur. Although the factors
that determine whether there is sufficient time for cognitive control
to be implemented are unclear, it appears that a long SOA between
the memory cue and the search display cannot guarantee the
effective influence of cognitive control on visual search. The
present study showed that WM contents captured attention regardless of the durations of the memory cue (100 and 1,000 ms) and the
interstimulus intervals between WM and attention displays (200
and 1,506 ms). In line with this, several other studies also provided
evidence for attentional capture by WM using the long SOAs
between the memory cue and the search array (e.g., Olivers,
Meijer, & Theeuwes, 2006; Soto & Humphreys, 2007, 2008).
Future study is needed for determination of the factors that influence the operation of cognitive control based on top-down knowledge.

PAN

128
In conclusion, although there may be boundary constraints on
attentional capture by WM (e.g., Woodman & Luck, 2007), recent
studies have shown that WM contents guide attention based on
exact visual matching (e.g., Olivers et al., 2006; Soto et al., 2005;
Soto, Humphreys, & Heinke, 2006), semantic links (e.g., Huang &
Pashler, 2007; Koivisto & Revonsuo, 2007; Moores, Laiti, &
Chelazzi, 2003; Soto & Humphreys, 2007) and even abstract
dimensional matching (Pan, Xu, & Soto, 2009) between the memory cue and the search display. The present study provides converging evidence for WM-based attentional capture using a simple
spatial probe task to measure attentional deployment. Here, I
demonstrated that attention was biased to a perceptual object that
contained a feature of the memory cue, even when doing this was
absolutely not beneficial for the task at hand.

Resume
Il existe une controverse a` savoir si le contenu en memoire de
travail (MT) guide automatiquement lattention. La presente etude
visait a` repliquer leffet de capture attentionnelle fondee sur la MT
en utilisant une adaptation du paradigme de Downing (2000), dans
lequel la MT et la capture attentionnelle sont combinees. Les sujets
se voyaient presenter une sce`ne attentionnelle contenant deux
objets, dont un pouvait etre preindice par un item associe retenu en
MT. Tel que mesure a` laide dune tache de discrimination de
cibles, lobjet associe en memoire joue un role important dans la
capture attentionnelle, peu importe lasynchronie de presentation
entre lindice en memoire et la sce`ne attentionnelle, et ce, meme
quand les sujets ne retirent aucun avantage a` biaiser leur attention
en faveur de lassociation en memoire. Ces resultats sugge`rent que
le contenu en MT guide lattention de facon involontaire. Les
repercussions de ces resultats sur la comprehension des effets de
MT sur la selection visuelle sont discutees.
Mots-cles : memoire de travail, capture attentionnelle, automaticite

References
Bundesen, C. (1990). A theory of visual attention. Psychological Review,
97, 523547.
Chelazzi, L., Miller, E. K., Duncan, J., & Desimone, R. (1993). A neural
basis for visual search in inferior temporal cortex. Nature, 363, 345347.
Chelazzi, L., Miller, E. K., Duncan, J., & Desimone, R. (2001). Responses
of neurons in macaque area V4 during memory-guided visual search.
Cerebral Cortex, 11, 761772.
Desimone, R. (1996). Neural mechanisms for visual memory and their role
in attention. Proceedings of the National Academy of Sciences, USA, 93,
13494 13499.
Desimone, R., & Duncan, J. (1995). Neural mechanisms of selective visual
attention. Annual Review of Neuroscience, 18, 193222.

Downing, P. E. (2000). Interactions between visual working memory and

selective attention. Psychological Science, 11, 467 473.
Duncan, J., & Humphreys, G. W. (1989). Visual search and stimulus
similarity. Psychological Review, 96, 433 458.
Han, S. W., & Kim, M.-S. (2009). Do the contents of working memory
capture attention? Yes, but cognitive control matters. Journal of Experimental Psychology: Human Perception and Performance, 35, 1292
1302.
Huang, L., & Pashler, H. (2007). Working memory and the guidance of
visual attention: Consonance-driven orienting. Psychonomic Bulletin &
Review, 14, 148 153.
Koivisto, M., & Revonsuo, A. (2007). How meaning shapes seeing. Psychological Science, 18, 845 849.
Moores, E., Laiti, L., & Chelazzi, L. (2003). Associative knowledge
controls deployment of visual selective attention. Nature Neuroscience,
6, 182189.
Olivers, C. N. L., Meijer, F., & Theeuwes, J. (2006). Feature-based
memory-driven attentional capture: Visual working memory content
affects visual attention. Journal of Experimental Psychology: Human
Perception and Performance, 32, 12431265.
Pan, Y. (2009). Research on content-based working memory-driven capture of visual attention and its automaticity (in Chinese). Doctoral
dissertation, Zhejiang University.
Pan, Y., & Soto, D. (2009). The modulation of perceptual selection by
working memory is dependent on the focus of spatial attention. Vision
Research, doi: 10.1016/j.visres. 2009.10.016
Pan, Y., Xu, B., & Soto, D. (2009). Dimension-based working memorydriven capture of visual selection. Quarterly Journal of Experimental
Psychology, 62, 11231131.
Soto, D., Heinke, D., Humphreys, G. W., & Blanco, M. J. (2005). Early,
involuntary top-down guidance of attention from working memory.
Journal of Experimental Psychology: Human Perception and Performance, 31, 248 261.
Soto, D., & Humphreys, G. W. (2007). Automatic guidance of visual
attention from verbal working memory. Journal of Experimental Psychology: Human Perception and Performance, 33, 730 737.
Soto, D., & Humphreys, G. W. (2008). Stressing the mind: The effect of
cognitive load and articulatory suppression on attentional guidance from
working memory. Perception & Psychophysics, 70, 924 934.
Soto, D., Humphreys, G. W., & Heinke, D. (2006). Working memory can
guide pop-out search. Vision Research, 46, 1010 1018.
Soto, D., Humphreys, G. W., & Rotshtein, P. (2007). Dissociating the
neural mechanisms of memory-based guidance of visual selection.
Proceedings of the National Academy of Sciences, USA, 104, 17186
17191.
Wolfe, J. M. (1994). Guided search 2.0: A revised model of visual search.
Psychonomic Bulletin & Review, 1, 202238.
Woodman, G. F., & Luck, S. J. (2007). Do the contents of visual working
memory automatically influence attentional selection during visual
search? Journal of Experimental Psychology: Human Perception and
Performance, 33, 363377.

Received August 9, 2009

Accepted December 16, 2009