When Goals Conflict With Values: Counterproductive Attentional and Oculomotor Capture by Reward-Related Stimuli

When Goals Conflict With Values: Counterproductive Attentional and
Oculomotor Capture by Reward-Related Stimuli
Mike E. Le Pelley, Daniel Pearson, Oren Griffiths, and Tom Beesley
University of New South Wales
Attention provides the gateway to cognition, by selecting certain stimuli for further analysis. Recent
research demonstrates that whether a stimulus captures attention is not determined solely by its physical
properties, but is malleable, being influenced by our previous experience of rewards obtained by
attending to that stimulus. Here we show that this influence of reward learning on attention extends to
task-irrelevant stimuli. In a visual search task, certain stimuli signaled the magnitude of available reward,
but reward delivery was not contingent on responding to those stimuli. Indeed, any attentional capture by
these critical distractor stimuli led to a reduction in the reward obtained. Nevertheless, distractors
signaling large reward produced greater attentional and oculomotor capture than those signaling small
reward. This counterproductive capture by task-irrelevant stimuli is important because it demonstrates
how external reward structures can produce patterns of behavior that conflict with task demands, and
similar processes may underlie problematic behavior directed toward real-world rewards.
Keywords: visual attention, reward learning, reinforcement learning, eye movements, attentional capture
Attention describes the cognitive mechanisms used to allocate
mental resources to the processing of certain aspects of sensory
input. For example, when driving we might use controlled, goaldirected
attention to prioritize processing of events on the road
ahead, and to ignore conversation from the backseat. But a sudden
bang from the car’s rear will capture our attention in an automatic,
stimulus-driven fashion (Yantis, 2000). Beyond these goaldirected
and stimulus-driven processes, research (as described
below) has demonstrated that the extent to which stimuli capture
attention is also influenced by learning about the significance of
those stimuli (for recent reviews, see Awh, Belopolsky, & Theeuwes,
2012; Chelazzi, Perlato, Santandrea, & Libera, 2013). This
possibility of an interaction between attention and learning—
wherein stimuli with meaningful consequences might “stand
out”—is not a new idea. William James (1890/1983) introduced
the concept of derived attention; a form of attention to a stimulus
that “owes its interest to association with some other immediately
interesting thing” (p. 393). While the idea of derived attention has
been around for some time, convincing empirical support for the
concept has arisen only relatively recently, and has emerged in
parallel from two sources.
Predictiveness-Driven Attentional Capture
Researchers working in the associative learning and animal conditioning
tradition have tended to focus on how attention is influenced
by learning about the predictiveness of stimuli (also referred to as
validity or informativeness). A predictive stimulus is one that provides
information regarding other events that will occur, or actions that
should be taken. For example, a green traffic light provides the
information that pulling out into an intersection is safe, and a red light
provides the information that it is unsafe; hence these are predictive
stimuli. The color of the car in the neighboring lane does not provide
any information regarding whether pulling out is safe, and so is a
nonpredictive stimulus. A large body of work in both humans and
nonhuman animals has examined the influence of previous experience
of the predictiveness of a stimulus on the rate of subsequent learning
about that stimulus (for reviews, see Le Pelley, 2004, 2010; Pearce &
Mackintosh, 2010). In humans at least, the typical finding is that
predictive stimuli are learned about more rapidly in future than are
nonpredictive stimuli. This finding is consistent with the suggestion
that attention is allocated to stimuli as a function of learning about
their predictiveness, on the assumption that the rate of learning about
a stimulus provides a measure of attention to that stimulus. Recent
studies go further, by demonstrating that the learned predictiveness of
a stimulus influences the extent to which that stimulus automatically
captures our attention (Le Pelley, Vadillo, & Luque, 2013; Livesey,
Harris, & Harris, 2009). For example, Le Pelley et al. (2013) gave
participants an initial training phase in which certain stimuli (say,
colored squares) predicted which of two buttons would be the correct
response on each trial, whereas other stimuli (say, sets of oblique
lines) provided no information regarding the correct response and
hence were nonpredictive. After many trials of training on this task,
participants moved on to a test phase that involved a variant of the dot
probe procedure (MacLeod, Mathews, & Tata, 1986). On each trial of
this test phase, a target (a small white triangle) appeared either in a
location cued by a stimulus that had been predictive in the training
This article was published Online First November 24, 2014.
Mike E. Le Pelley, School of Psychology, University of New South
Wales; Daniel Pearson, School of Psychology, University of New South
Wales; Oren Griffiths, School of Psychology, University of New
South Wales; Tom Beesley, School of Psychology, University of New
South Wales.
This work was supported by Australian Research Council Grant
FT100100260. We thank Branka Spehar for help in creating the stimuli,
and Tom Whitford for helpful comments on the manuscript.
Correspondence concerning this article should be addressed to Mike E.
Le Pelley, School of Psychology, University of New South Wales, Sydney
NSW 2052, Australia. E-mail: m.lepelley@unsw.edu.au
This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
Journal of Experimental Psychology: General © 2014 American Psychological Association
2015, Vol. 144, No. 1, 158–171 0096-3445/15/$12.00 http://dx.doi.org/10.1037/xge0000037
158
phase (a colored square), or in a location cued by a stimulus that had
been nonpredictive (a set of oblique lines). Participants were required
to press the spacebar as quickly as possible when the target appeared.
Importantly, across these test trials, the location in which the target
appeared was independent of the type of cueing stimulus (predictive
or nonpredictive). Hence, participants gained no advantage by orienting
their attention to the location of either type of cueing stimulus in
advance of the target appearing. Indeed, participants were explicitly
informed that in order to respond to the target as quickly as possible,
their best strategy was to ignore the initially presented stimuli. Despite
this instruction, responses were significantly faster when the target
appeared in the location of the predictive stimulus than the location of
the nonpredictive stimulus.
The implication is that the predictive stimulus captured participants’
spatial attention and hence responses to events occurring in
that location were initiated earlier (cf. Posner, 1980). This attentional
capture occurred even though (a) it was not required by the
task, (b) it was not adaptive with regard to that task (since target
location was independent of the type of cueing stimulus), and (c)
there was little time for participants to consciously process and
respond to the cueing stimuli on each test trial. Le Pelley et al.
(2013) demonstrated that providing more time for participants to
consciously process the stimuli—by increasing the SOA on test
trials to 1,000 ms—significantly weakened the influence of predictiveness
on dot probe responding. This suggests that the pattern
observed at short SOA was not a result of conscious, controlled
processing but instead reflected a rapid and automatic effect of
predictiveness on attentional capture. A long SOA then provides
time for participants to use controlled processes to correct for the
automatic attentional capture caused by presentation of the stimuli,
returning attention to the center of the display (cf. Klauer,
Ronagel, & Musch, 1997).
Value-Driven Attentional Capture
As noted above, predictiveness refers to the extent to which a
stimulus provides information regarding subsequent events. In a
parallel stream, researchers working within the perception and
cognition traditions have demonstrated that attentional capture by
a stimulus is also influenced by the value of those subsequent
events (Anderson, Laurent, & Yantis, 2011a, 2011b; Anderson &
Yantis, 2012; Libera & Chelazzi, 2009; Kiss, Driver, & Eimer,
2009; Rutherford, O’Brien, & Raymond, 2010; Theeuwes & Belopolsky,
2012). Specifically, if attending to a stimulus is consistently
paired with a high-value reward, then that stimulus becomes
more likely to capture attention than an equally salient stimulus
paired with low-value reward.
The clearest demonstration of this value-driven attentional capture
comes from studies of visual search (Anderson et al., 2011a,
2011b). On each trial of an initial training phase, a target (either a
red or green circle) was presented together with distractors (similar
circles rendered in other colors). Participants were required to
respond according to the orientation of a line segment contained in
the target circle. During this training phase, a particular participant
might receive a large reward for making a rapid response when the
target was, say, a red circle, and a small reward for making a rapid
response when the target was a green circle. Hence for this participant,
red was the “high-value color,” and green was the “lowvalue
color.” After extensive training on this task—typically over
1,000 trials—participants moved on to a test phase, in which the
target on each trial was now defined by shape (say, a diamond
among circles; the diamond was never red or green). Critically,
during the test phase, people were slower to respond to this
diamond target if one of the nontarget circles was rendered in the
high-value color than if it was rendered in the low-value color; that
is, a nontarget in the high-value color was more distracting. This
attentional capture by a stimulus previously paired with large
reward occurred even though attending to color conflicted with
current task demands in the test phase (respond to the diamond),
suggesting that capture was automatic and involuntary.
Derived Attention, Addiction, and Psychosis
The interaction between attention and learning that is implicated
in predictiveness-driven and value-driven capture is notable because
it demonstrates that processing of sensory input is not a fixed
function of physical salience, but is instead malleable and based on
our previous experiences. This may bring adaptive advantages by
improving and speeding detection of informative and/or rewardrelated
stimuli. But it may also create problems. For example,
many drugs of abuse produce potent neural reward signals (Dayan,
2009; Hyman, 2005; Robinson & Berridge, 2001). Involuntary
attentional capture by stimuli associated with these drug rewards
(such as drug paraphernalia, or people and locations associated
with drug supply) is known to predict relapse in recovering addicts
(Cox, Hogan, Kristian, & Race, 2002; Marissen et al., 2006;
Waters et al., 2003). Interestingly, Anderson, Faulkner, Rilee,
Yantis, and Marvel (2013) have recently used the procedure described
above to demonstrate that value-driven attentional capture
is magnified in drug addicts, consistent with the idea of a link
between addiction and attention to reward-related stimuli. It has
also been argued that the psychotic symptoms of schizophrenia
reflect a dysfunction of the relationship between reward learning
and attention (Frank, 2008; Kapur, 2003; Morris, Griffiths, Le
Pelley, & Weickert, 2013). On this account, abnormal reward
learning results in patients attending to stimuli that would normally
be ignored, and ignoring those to which they should attend. This in
turn produces unusual sensory experiences (hallucinations) and
cognitive efforts to make sense of those unusual experiences
(delusions). A better understanding of the mechanisms underlying
derived attention in humans has the potential to shed light on
aspects of mental disorder, including addiction and psychosis, that
implicate a dysfunction of these mechanisms.
Is Task-Relevance Essential for Learned
Attentional Capture?
All previous studies that have demonstrated learned attentional
capture in humans (be that predictiveness- or value-driven) have one
feature in common. In all cases, during the training phase that was
used to establish differences in the predictiveness or value of the
stimuli, these stimuli were task-relevant for participants; that is, they
were the stimuli that participants were required to identify in order to
perform the task. For example, in Le Pelley et al.’s (2013) study of
predictiveness-driven capture, the predictive stimuli (colored squares
in the example given above) were predictive precisely because they
defined the correct response on each training trial. Hence during the
training phase, participants needed to identify these stimuli in order to
This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
REWARD AND ATTENTION 159
make the correct response. In the initial training phase of Anderson et
al.’s (2011a, 2011b) studies of value-driven capture, the stimuli that
predicted reward magnitude (red and green circles) defined the targets
to which participants were required to direct their responses. In each
case, this raises the possibility that attentional capture by similar
stimuli in the subsequent test phase was simply a carryover of an
overlearned attentional orienting response to these stimuli, previously
established in the training phase. For example, the training phase of
Anderson et al.’s task essentially constitutes extensive training that
orienting rapidly to (say) a red circle leads to a large reward. Reinforcement
learning might therefore promote the extent to which red
stimuli automatically elicit such a rapid orienting response in the
future. If this automatic attentional orienting response persisted into
the test phase, this would explain why red stimuli were more distracting
during this test phase.1
In the real world, however, stimuli that signal rewards are not
always the goals that produce those rewards. For example, an addict
may typically take drugs in a particular room. This room signals the
drug’s rewarding effect, but has no instrumental relationship with
achieving that reward: entering the room does not itself elicit a drug
reward, and the drug would have a similar rewarding effect if ingested
elsewhere. In this sense, the room is task-irrelevant with respect to the
goal of achieving drug reward. This raises the question of whether
such task-irrelevant stimuli can nevertheless support attentional capture.
To investigate this possibility, the studies of value-driven capture
reported here used training in which the critical stimuli were never
task-relevant for participants. On every trial, participants searched for
a diamond-shaped target among circles. One of these nontarget circles
(the distractor) could be colored, and the distractor’s color predicted
the reward available on that trial. But crucially, reward was obtained
by responding to the diamond, not to the distractor circle. Hence the
stimuli predicting reward were not those to which participants were
required to direct their responses or attention. In fact, we ensured that
attending to reward-predictive distractors would, if anything, hinder
participants’ performance and hence the payoff they achieved. If
value-driven attentional capture by distractors were nevertheless observed,
it would imply the existence of an automatic attentional
process that prioritizes reward-related stimuli even when capture by
those stimuli is, and always has been, directly contrary to participants’
goals.
A finding of value-driven capture by task-irrelevant stimuli
would shed important light on the nature of the information that
underlies learned attentional capture, because it would suggest that
it is the simple correlation of stimuli with reward, rather than their
functional role in obtaining that reward, that determines capture.
Rephrased in the language of associative learning theory, this
would imply that derived attention is a product of Pavlovian
conditioning (i.e., learning about the extent to which a stimulus
signals reward), rather than instrumental conditioning (i.e., learning
about the relationship between a response and the reward that
it produces). As noted above, the color of the distractor in our
experiments signaled the size of the reward available on each trial.
Hence, the high-value color was a Pavlovian signal of large reward,
and the low-value color was a Pavlovian signal of small
reward. If value-driven capture arises because signals of large
reward are more likely to capture attention than signals of
small reward, then we would expect greater capture by the highvalue
color than the low-value color. But as noted above, participants
were not rewarded for responding to (or orienting attention
toward) the stimulus that was presented in the value-related color
since this stimulus was, by definition, a nontarget. Hence there was
no potential for instrumental conditioning to promote value-driven
capture. In fact, since our experiments were arranged so that
orienting toward the distractor resulted in loss of reward, participants
would receive larger rewards when they successfully suppressed
attention to this distractor (or oriented away from it). In
particular, the largest rewards would occur when participants successfully
refrained from orienting toward the high-value color.
Consequently, the instrumental relationships in force in this experiment
would, if anything, promote suppression of attentional
orienting to the high-value color relative to the low-value color.
This would result in less distraction by the high-value color; the
opposite pattern to that anticipated by the Pavlovian account.
Below we describe three experiments investigating the influence
of reward learning on attentional capture by task-irrelevant stimuli.
In Experiments 1 and 2, the primary measure of capture is response
time, on the assumption that attentional capture by the distractor
will result in slowing of responses to the target. Experiment 3 goes
on to use a gaze-contingent eye-tracking procedure to measure the
extent to which reward-related distractors capture eye gaze.
Experiment 1
Participants
Participants in Experiment 1 were 27 University of New South
Wales (UNSW) students, who received course credit and also
performance-based payment (M $19.92 AUD).
Apparatus
Participants were tested individually using a standard PC with a
23-in. monitor (1,920 1,080 resolution, refresh rate 120 Hz),
positioned 60 cm from the participant. Stimulus presentation was
controlled by MATLAB using Psychophysics Toolbox extensions
(Brainard, 1997; Kleiner, Brainard, & Pelli, 2007; Pelli, 1997).
Stimuli
The experiment used a variant of the additional singleton paradigm
(Theeuwes, 1991, 1992). Each trial consisted of a fixation
display, a search display, and a feedback display (Figure 1a). All
stimuli were presented on a black background. The fixation display
consisted of a white cross (subtending 0.5 degrees of visual angle;
1 It is worth mentioning here a recent study by Anderson et al. (2012).
This used the same training procedure as Anderson et al. (2011a, 2011b) in
which, say, red circles were the targets that predicted large reward and
green circles were the targets predicting small reward. In this case, however,
the subsequent test phase showed greater attentional capture by letters
rendered in red rather than letters in green. Strictly, then, Anderson et al.
(2012) demonstrated value-driven capture by stimuli (colored letters) that
had not been task-relevant during the training phase—since they had not
appeared during this training phase. Crucially, however, these stimuli
shared the critical, task-relevant feature that defined targets in the training
phase, namely red or green color. Hence, like the earlier results of Anderson
et al. (2011a, 2011b), these findings reflect value-driven attentional
capture by a feature (red or green color) that had been task-relevant. In
contrast, in the current studies, the feature that predicted reward magnitude
(again, color) was never task-relevant.
This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
160 LE PELLEY, PEARSON, GRIFFITHS, AND BEESLEY
dva) presented centrally. The search display comprised the fixation
cross surrounded by six shapes (2.3 2.3 dva) positioned at equal
intervals around an imaginary circle with diameter 10.1 dva (with
the first position directly above the fixation cross). Five of these
shapes were circles, and one was a diamond. The diamond and
four of the circles were rendered in gray. The remaining circle (the
distractor) was rendered either in red, blue, green, or the same
shade of gray as the other shapes (Commision Internationale
d’Elairage x, y chromaticity coordinates of .595/.360 for red,
.160/.116 for blue, .300/.611 for green, .304/.377 for gray). The
values of red, blue, and green had similar luminance (42.5
cd/m2), which was higher than that of the gray (36.5 cd/m2). The
target contained a white line segment (length 0.76 dva) oriented
either vertically or horizontally. Each nontarget contained a similar
line segment tilted 45° randomly to the left or right.
Design
Two colors from the set of red, green and blue were randomly
. . . assigned to act as high-value and low-value colors for each
participant.2 The experiment comprised 10 training blocks. Each
block contained 18 trials with a distractor rendered in the highvalue
color, 18 trials with a distractor in the low-value color, and
4 distractor-absent trials on which there was no color singleton in
the display. Trials occurred in random order.
Correct responses to the target that were faster than the participant’s
latency limit (see Procedure) were followed by feedback
indicating reward. If that trial had a high-value distractor, reward
was large (10¢); if it had a low-value distractor, reward was small
(1¢). If there was no distractor, the reward was equally likely to be
small or large. Correct responses that were slower than the participant’s
latency limit received no reward, though participants were
told how much they could have won. Errors resulted in loss of the
amount that had been scheduled as a reward on that trial. Target
location, distractor location, and target line segment orientation
(vertical or horizontal) were randomly determined on each trial.
Procedure
The single session began with a practice phase of 20 trials, with
a yellow distractor on each and no reward feedback. The first two
practice trials were discarded; the upper quartile of response latencies
for the remaining correctly responded-to practice trials
defined the latency limit for each participant. If response accuracy
over practice trials was below 75%, participants repeated the
practice phase. The mean latency limit was 964 ms (standard error
of the mean [SEM] 42 ms)
Instructions informed participants that for subsequent trials, they
would earn a reward only for correct responses that were faster
than their latency limit (this latency limit was given to them in
2 The remaining color was assigned to act as a rare color. Each training
block contained four trials with the distractor rendered in this rare color;
reward magnitude was equally likely to be small or large on these trials.
However, data from trials with the rare distractor are uninformative with
regard to the critical issue of value-driven capture addressed in this article
(indeed, the rare color was omitted from Experiments 2 and 3 for exactly
this reason). Hence, for the sake of brevity we do not present or analyze
these data here.
Figure 1. Sequence of trial events, Experiment 1 (a). Participants responded to the orientation of the line
segment (horizontal or vertical) within the diamond (target). One of the nontarget circles could be a color
singleton distractor (shown in red in online version, light gray in print version). Fast, correct responses to the
target received monetary reward, depending on the distractor color. A high-value distractor color reliably
predicted large reward; a low-value reliably predicted small reward; if no color singleton was present in the
display (distractor-absent trial), then large and small reward were equally likely. Mean response time (b) and
mean accuracy (proportion correct; c) across the 10 training blocks of Experiment 1, for trials with a high-value
distractor, a low-value distractor, and distractor-absent trials. Error bars show within-subjects standard error of
the mean (see Cousineau, 2005). Critically, response times were significantly slower on trials with a high-value
distractor than trials with a low-value distractor, but no more accurate. See the online article for the color version
of this figure.
This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
REWARD AND ATTENTION 161
milliseconds), that such responses would earn either 1¢ or 10¢
“depending on the shapes presented on the trial,” and that errors
would result in loss of the corresponding amount. Participants
were not informed of the relationship between distractor color and
reward magnitude.
Each trial began with presentation of the fixation display for a
random period of 400, 500, or 600 ms. The search display then
appeared until a response was made or the trial timed out (after 2
s). Participants responded to the orientation of the line segment in
the diamond by pressing the C and M keys for horizontally and
vertically oriented lines, respectively. For fast correct responses
(or errors), the feedback display appeared for 1,400 ms (or 2,000
ms), showing the reward earned (or lost) on the previous trial, and
total earnings so far. For correct responses slower than the latency
limit, feedback displayed participants’ response time (RT) and the
amount they would have received for a faster response. Intertrial
interval was 1,000 ms. Participants took a short break every two
blocks.
Data Analysis
The first two trials, and the first two trials after each break, were
discarded. Timeouts (0.06% of all trials) and trials with RTs below
150 ms (0%) were also discarded. RTs were then analyzed for
correct responses only.
Results
Figure 1b and 1c show RTs and accuracy across training. RTs
were analyzed using a 3 (distractor type: high-value, low-value,
absent) 10 (block) analysis of variance (ANOVA). This revealed
a main effect of block, F(9, 234) 10.8, p .001, p
2
.29, with mean RT falling as participants became more practiced at
the task. There was also a significant main effect of distractor type,
F(2, 52) 23.1, p .001, p
2 .47. The interaction was not
significant, F(18, 468) .70, p .82, p
2 .03.
Planned pairwise t tests, averaging across training blocks, were
used to further analyze the main effect of distractor type. Each type
of colored distractor slowed RT relative to distractor-absent
trials— high-value versus absent: t(26) 6.29, p .001, d 1.21,
95% CI of difference (CIdiff) [19.8, 39.1]; low-value versus absent:
t(26) 5.05, p .001, d .97, CIdiff [11.6, 27.4]. Critically, RT
on trials with the high-value distractor (M 647 ms) was significantly
greater than on trials with the low-value distractor (M
637 ms), t(26) 2.15, p .041, d .41, CIdiff [.42, 19.4]. Slower
RTs on trials with the high-value distractor meant participants
were more likely to miss out on the large rewards available on
these trials: correct RTs were slower than the latency limit (and
hence reward was omitted) on 4.3% of high-value distractor trials
versus 3.6% of low-value distractor trials. This difference was
significant, t(26) 1.94, p .032 (one-tailed, since direction is
anticipated by the RT difference), d .37, CIdiff [.00, .013].
For the accuracy data, the omnibus 3 10 ANOVA revealed a
main effect of block, F(9, 234) 8.98, p .001, p
2 .26, with
accuracy increasing across training, but no significant main effect
of distractor type, F(2, 52) 1.55, p .22, p
2 .06, or
interaction, F(18, 468) .92, p .55, p
2 .03. Notably,
averaging across blocks, accuracy on trials with the high-value
distractor (M 93.0%) was not significantly different from that
for trials with the low-value distractor (M 93.6%), t(26) 1.12,
p .27, d .22, CIdiff [–.013, .004]. This suggests that the critical
difference in RT observed on these trials did not reflect a speed–
accuracy tradeoff.
Discussion
The finding that each type of colored distractor slowed RT
relative to distractor-absent trials replicates the well-established
finding that search for a target defined by shape is slowed by the
presence of a color singleton distractor (Theeuwes, 1992, 1994).
The distractors were physically salient, because they were the only
colored stimuli in the display. The implication is that these salient
stimuli capture attention in a relatively automatic fashion (automatic,
because attending to the distractors will impede the participant’s
goal of responding to the target as fast as possible).
More importantly, responses were significantly slower (but no
more accurate) for trials with a high-value distractor compared
with a low-value distractor. This suggests that the high-value
distractor was more likely to capture attention than the low-value
distractor, even though these distractors were always taskirrelevant.
That is, attentional capture by the distractors was modulated
by the magnitude of the reward that they signaled. The
implication, then, is that value-driven attentional capture is a
consequence of Pavlovian conditioning (based on the extent to
which distractor color provides a signal of reward value), rather
than instrumental conditioning (based on the value of the reward
that is produced by responding [i.e., attentional orienting] to the
color). Notably, the greater capture by high-value distractors
meant that participants missed more high-value than low-value
rewards. Hence, this increased capture by the task-irrelevant highvalue
distractor was counterproductive to participants’ goal of
maximizing their payoff.
Experiment 2
Notably, however, mean correct RT in Experiment 1 (636 ms)
was considerably shorter than the mean latency limit (964 ms).
Given that the mean RT difference between high- and low-value
distractor trials was numerically small (10 ms), this meant that
the greater attentional capture by the high-value target would
rarely impact on the reward received. In other words, slower RTs
on high-value distractor trials did not always result in reward
omission; the relationship between RT and reward in Experiment
1 was relatively indirect. Experiment 2 therefore implemented a
direct relationship between RT and reward, so that slower responses
necessarily resulted in reduced reward. We also investigated
whether the counterproductive effect observed in Experiment
1 would persist across extended training, or whether
participants would eventually come to show a more adaptive
pattern of faster responses on high-value trials. This possibility
was based on data from conditioning studies with rats (e.g., Holland,
1979), which demonstrate that some overt behaviors (e.g.,
approaching a food magazine during an auditory stimulus that
terminates in delivery of food) that are initially driven by Pavlovian
conditioning can, with extended training, come under the
control of instrumental conditioning; though other behaviors (e.g.,
the startle response elicited by the same auditory stimulus) remain
under Pavlovian control even with extensive training. Finally,
This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
162 LE PELLEY, PEARSON, GRIFFITHS, AND BEESLEY
Experiment 2 investigated the relationship between value-driven
capture and explicit awareness of the color–reward contingencies.
Method
Participants. Twenty-four people took part. Four were students
participating for course credit, and 20 were recruited from
the UNSW community in exchange for $45. All participants also
received performance-related payment (M $78.21 AUD).
Apparatus and stimuli. Apparatus and stimuli were as for
Experiment 1, except Experiment 2 had no green stimuli.
Design. The experiment comprised 36 training blocks of 48
trials, giving 1,728 trials in total. Each block contained 20 trials
with the distractor in the high-value color, 20 trials in the lowvalue
color, and 8 distractor-absent trials. These trials occurred in
random order. Assignment of red and blue to high- and low-value
colors was counterbalanced across participants.
Correct responses with RTs slower than 1,000 ms earned no
reward. For faster correct responses, reward (in cents) was calculated
as (1,000 – RT) 0.002 bonus_multiplier, rounded to the
nearest 0.01¢. For high-value distractor trials, bonus_multiplier
was always 10; for low-value distractor trials, bonus_multiplier
was always 1; for distractor-absent trials, bonus_multiplier was
equally often 1 and 10. Errors resulted in loss of the corresponding
amount. Other aspects of design were as for Experiment 1.
Procedure. Participants completed three 1-hr sessions on consecutive
days. Each session began with 10 practice trials featuring
a yellow distractor, with no reward feedback. This was followed
by 12 training blocks (giving 576 training trials per session).
Participants were told they would earn 0.2¢ for every 100 ms their
RT was below 1,000 ms, that on “10 Bonus” trials this reward
would be multiplied by 10, and that errors would result in loss of
the corresponding amount. Instructions made no reference to the
relationship between bonus trials and distractor color. Trials were
as for Experiment 1, except that feedback was presented for 2,000
ms after correct responses and 2,500 ms after errors. On nonbonus
trials, the feedback display showed the amount earned and total
earnings; on bonus trials this was accompanied by a yellow box
labeled “10 Bonus trial!” (bonus trials were not signaled explicitly
until after participants had made their response).
After the final training session, participants were told that bonus
trials had been determined by the color in the display; that when
certain colors appeared it would be a bonus trial, and when other
colors appeared it would not be. In a final awareness test, they
were then shown a red circle and a blue circle in random order, and
for each selected whether it would be a bonus trial or not when this
color of circle appeared in the display.
Data analysis. Data were analyzed as for Experiment 1.
Timeouts (0.01% of all trials) and trials with RTs below 150 ms
(0.03%) were discarded.
Results
Figure 2A shows RTs across training. These were analyzed
using a 3 (distractor type: high-value, low-value, absent) 36
(block) ANOVA. A main effect of block, F(35, 805) 67.1, p
.001, p
2 .74, reflected the reduction in mean RT across blocks.
There was a significant main effect of distractor type, F(2, 46)
45.9, p .001, p
2 .67, and a Distractor Type Block interaction,
F(70, 1610) 2.04, p .001, p
2 .08.
A follow-up 2 36 ANOVA used only the data for high- and
low-value distractors. The main effect of distractor type was significant,
F(1, 23) 25.7, p .001, p
2 .53, with slower
responses on trials with high-value distractors than low-value
distractors. The Distractor Type Block interaction was not
significant, F(35, 805) 1.06, p .37, p
2 .044.
Figure 2. Mean response time across the 12 training blocks of each of the
3 sessions of Experiment 2, for trials with a high-value distractor, a
low-value distractor, and distractor-absent trials (a). Point markers and
error bars have been omitted for the sake of clarity. Mean response time (b)
and accuracy (proportion correct; c) for each session of Experiment 2, for
high-value, low-value, and distractor-absent trials. Error bars show withinsubjects
standard error of the mean (Cousineau, 2005). Response times
were significantly slower on trials with a high-value distractor than trials
with a low-value distractor in all sessions. See the online article for the
color version of this figure. p .05. p .01. p .001.
This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
REWARD AND ATTENTION 163
Figure 2b shows mean RTs for each session. Critically, the
high-value distractor slowed RT relative to the low-value distractor
in each session—Session 1: t(23) 4.00, p .001, d
.82, CIdiff [3.98, 12.5]; Session 2: t(23) 2.78, p .011, d
.57, CIdiff [2.31, 15.8]; Session 3: t(23) 5.48, p .001,
d 1.12, CIdiff [7.39, 16.4]. This did not reflect a speed–
accuracy tradeoff. Figure 2c shows that mean accuracy was
similar for high- and low-value distractors in Sessions 1 and 2,
t 1 for each, and Session 3 showed a trend toward lower
accuracy for high- than low-value distractors, t(23) 1.75, p
.093, d .36, CIdiff [–.002, .018].
It is clear from Figure 2b that the RT difference between
high-value and low-value distractor trials did not decrease with
extended training; the numerical difference was roughly constant
across Sessions 1 to 3 (M 8.26, 9.05, and 11.87 ms, respectively),
despite the reduction in baseline RTs across sessions.
Indeed, when expressed as a proportion of baseline RT, the difference
between high- and low-value trials is significantly greater
in Session 3 than in Session 1, t(23) 2.11, p .046, d .43,
CIdiff [.000, .022] (and as noted above, this RT difference is
accompanied by a trend-level difference in accuracy in Session 3
but not in Session 1).
In the final awareness test, seven participants showed no awareness
of the color–reward contingencies, incorrectly selecting that
the low-value color signaled bonus trials and that the high-value
color did not. Across all trials, these “unaware” participants still
showed significantly slower RTs on high-value distractor trials
(M 527 ms) than on low-value distractor trials (M 518 ms):
t(6) 4.24, p .005, d 1.60, CIdiff [3.53, 13.2]; alternatively
Wilcoxon’s T(7) 0, p .016. For the remaining 17 participants,
whose responses in the awareness test were consistent with them
having become aware of the veridical relationships between the
different colors and reward levels, RTs were also significantly
slower on high-value distractor trials (M 531 ms) than on
low-value distractor trials (M 520 ms): t(16) 4.01, p .001,
d .97, CIdiff [4.85, 15.7]. The magnitude of the value-driven
capture effect (i.e., the difference in RT between high- and lowvalue
distractor trials) did not differ significantly for “aware” and
“unaware” participants, t(22) .46, p .65, d .23, CIdiff
[10.7, 6.85].
Discussion
Experiment 2 replicated the key finding of Experiment 1: RTs
were slower on trials with a high-value distractor rather than a
low-value distractor. Once again, this suggests greater attentional
capture by the high-value distractor. Importantly, because RTs
directly determined reward in Experiment 2, this enhanced capture
by the high-value distractor was directly counterproductive, because
it meant that participants earned less on high-value trials
than would otherwise have been the case. This counterproductive
pattern persisted across extensive training: even after 2,000 trials,
attentional capture remained under the control of the Pavlovian
signal-value of the colors. Interestingly, this pattern was observed
in participants who were unaware of the color–reward contingencies,
suggesting that awareness is not necessary for value-driven
attentional capture by task-irrelevant stimuli.
Experiment 3
It is well-established that stimuli that capture attention also tend
to capture eye movements, known as oculomotor capture (e.g.,
Anderson & Yantis, 2012; Ludwig & Gilchrist, 2002; Theeuwes &
Belopolsky, 2012; Theeuwes, Kramer, Hahn, & Irwin, 1998;
Theeuwes, Kramer, Hahn, Irwin, & Zelinksy, 1999). Experiment 3
used a gaze-contingent eye-tracking procedure to determine
whether reward learning about task-irrelevant distractors influenced
the extent to which those distractors elicited oculomotor
capture. Critically, online monitoring of eye movements as participants
performed the search task allowed us to ensure that oculomotor
capture by distractors was never rewarded.
Method
Participants. Twenty-four UNSW students participated for
course credit and received performance-related payment (M
$20.42 AUD).
Apparatus. Experiment 3 used a Tobii TX300 eye-tracker,
with 300 Hz temporal and 0.15° spatial resolution, mounted on a
23-in. widescreen monitor (1,920 1,080 resolution, refresh rate
60 Hz). Participants’ heads were positioned in a chinrest 60 cm
from the screen. As participants performed the task, the program
controlling stimulus presentation requested data from the eyetracker
every 10 ms. Because the eye-tracker recorded gaze location
at 300 Hz (i.e., one recording every 3.3 ms), the 10 ms sample
taken by the experiment program would typically contain three
recordings of gaze location. Participants’ current gaze location was
defined as the average of the locations contained in the most recent
10 ms sample. Invalid recordings (when the eye tracker failed to
detect a gaze location) were not included in this average; if the
current 10 ms sample contained no valid recordings, then the gaze
location from the previous 10 ms sample was used instead.
The eye-tracker was calibrated using a 5-point procedure prior to
the practice phase, prior to the first training block, and after 6
training blocks.
Stimuli. Each trial consisted of a fixation display, a search
display, and a feedback display (Figure 3a). The fixation display
was a white cross surrounded by a white circle (diameter 3.0 dva).
The search display was as for Experiment 2, except that: (a) all
shapes were filled, (b) there were no line segments, (c) the fixation
cross was absent, and (d) the shade of gray was darker than in
previous experiments (luminance 32 cd/m2). The feedback display
showed the reward earned on the previous trial and total
earnings.
Design. The training phase comprised 10 blocks, which were
structured as for Experiment 2. Target location and distractor
location were randomly determined on each trial, with the constraint
that the distractor could never appear adjacent to the target.
On each trial, a small circular region of interest (ROI) with
diameter 3.5 dva was defined around the diamond target; a larger
ROI (diameter 5.1 dva) was defined around the distractor. A
response was registered when participants had accumulated 100
ms of dwell time inside the target ROI. Responses with RTs slower
than 600 ms earned no reward. If any gaze fell inside the distractor
ROI prior to a response being registered, even for a single 10-ms
period, the trial was recorded as an omission trial and no reward
was delivered. On distractor-absent trials, one of the gray circles
(that was not adjacent to the target) was chosen at random; gaze
This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
164 LE PELLEY, PEARSON, GRIFFITHS, AND BEESLEY
falling inside an ROI around the selected gray circle caused an
omission trial in exactly the same way as if the selected circle had
been a distractor.3
If RT was faster than 600 ms and no gaze was registered in the
distractor ROI, then a reward was delivered: 10¢ if the high-value
distractor was present, 1¢ if the low-value distractor was present,
and an equal likelihood of 10¢ or 1¢ on distractor-absent trials.
Notably, in Experiment 3, participants never lost money (unlike in
Experiments 1 and 2 where participants lost money for erroneous
responses). Other design aspects were as for Experiment 2, except
that in Experiment 3 the distractor never appeared adjacent to the
target.
Procedure. The single session began with 8 practice trials
with a yellow distractor and no rewards. Participants were then
told that on subsequent trials they would earn either 0¢, 1¢, or 10¢,
depending on “how fast and accurately you move your eyes to the
diamond.”
Each trial began with the presentation of the fixation display.
Participants’ gaze location was superimposed on this display as a
small yellow dot. Once participants had recorded 700 ms dwell
time inside the circle surrounding the fixation cross, or if 5 s had
passed, the cross and circle turned yellow, and the dot marking
gaze location disappeared. After 300 ms, the screen blanked, and
after a random interval of 600, 700, or 800 ms, the search display
appeared. The trial terminated when a response was registered (see
Design), or after 2 s (timeout). The feedback display then appeared
for 1,400 ms. Intertrial interval was 1,400 ms.
Data analysis. As for previous experiments, the first two
trials, and the first two trials after each break, were discarded.
3 Allowing for omissions on distractor-absent trials is useful, because it
permits a valid test of the influence of stimulus salience on oculomotor
capture, by comparing the rate of omissions on trials featuring a salient
distractor with the rate on distractor-absent trials. This comparison controls
for causes of omission trials that are not related to distractor salience (e.g.,
inaccuracy in the recording of gaze location, random eye movements by the
participant, etc.), since these will be equal on trials with a salient distractor
and distractor-absent trials.
Figure 3. Sequence of trial events, Experiment 3 (a). Participants responded by moving their eyes to the
diamond target. One of the nontarget circles could be a color singleton distractor (shown in blue in online
version, light gray in print version). Dotted lines (not visible to participants) indicate the region of interest around
the target and distractor within which eye gaze was defined as falling on the corresponding stimulus. Fast, correct
responses received monetary reward, depending on the distractor color. A high-value distractor color reliably
predicted large reward; a low-value color reliably predicted small reward; on distractor-absent trials, large and
small rewards were equally likely. If any gaze fell within the distractor region of interest (or, on distractor-absent
trials, an equivalent region of interest positioned around a randomly chosen circle), the trial was deemed an
omission trial, and no reward was delivered. Mean proportion of omission trials (b) and mean response times (c)
across the 10 training blocks of Experiment 3, for high-value, low-value, and distractor-absent trials. Reward was
more likely to be omitted, and response times were slower, on trials with the high-value distractor than trials with
the low-value distractor. Mean saccade latencies on omission and non-omission trials, averaged across training
blocks (d). Saccade latencies were generally slower for non-omission trials than omission trials. Latencies were
also slower for non-omission trials featuring a high-value distractor than featuring a low-value distractor. All
error bars show within-subjects standard error of the mean (Cousineau, 2005). See the online article for the color
version of this figure.
This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
REWARD AND ATTENTION 165
Timeouts (1.3% of all trials) were also discarded. Finally, we also
excluded all trials on which valid gaze location was registered in
less than 25% of the 10-ms samples between presentation of the
search display and registering of a response (1.4% of all trials). For
remaining trials, averaging across participants, valid gaze location
was registered in 97.8% (SEM 0.8%) of samples, suggesting very
high fidelity of the gaze data on these trials.
For analysis of saccade latencies, we used the raw data from the
eye-tracker (sampled at 300 Hz, rather than the 100 Hz used for
gaze-contingent calculations). For these analyses, in addition to the
exclusions described above, we further excluded all trials on which
no eye gaze was recorded within 5.1 dva (100 pixels) of the
fixation point during the first 80 ms after presentation of the search
display. Saccade latency was then found by identifying the first
point at which five consecutive gaze samples lay more than 5.1 dva
from the fixation point. Saccades faster than 80 ms were excluded
from further analyses. The extra exclusions described in this paragraph
resulted in loss of an additional 6.1% of trials.
Results
Our primary measure in Experiment 3 was the proportion of
omission trials across training, shown in Figure 3b. These data
were analyzed using a 3 (distractor type: high-value, low-value,
absent) 10 (block) ANOVA. The main effect of block was not
significant, F(9, 207) 1.25, p .26. There was a significant
main effect of distractor type, F(2, 46) 24.5, p .001, p
2 .52,
and the Distractor Type Block interaction approached significance,
F(18, 414) 1.52, p .079, p
2 .062.
On the basis of this trend toward a change in the effect of the
distractor type on proportion of omissions across the course of
training, smaller ANOVAs were used to further analyze these data
for the different distractor types (as opposed to t tests collapsing
across training blocks). A 2 (distractor type) 10 (block)
ANOVA, using only the data for high-value and low-value distractors,
revealed a main effect of distractor type, F(1, 23) 11.4,
p .003, p
2 .33; that is, high-value distractors produced
significantly more omission trials than low-value distractors
(16.4% vs. 8.9%, collapsing across blocks). The Distractor Type
Block interaction approached significance, F(9, 207) 1.89, p
.055, p
2 .076, with the difference in proportion of omission
trials for high-value and low-value distractors tending to increase
as training progressed (see Figure 3b). Similar ANOVAs comparing
the data for high-value versus distractor-absent trials, and
low-value versus distractor-absent, revealed a significant main
effect of distractor type in each case: for high-value versus absent,
F(1, 23) 33.8, p .001, p
2 .59; for low-value versus absent,
F(1, 23) 36.5, p .001, p
2 .61. Unsurprisingly, then, color
singleton distractors generally produced more oculomotor capture
than if no color singleton was present in the display; just 2.4% of
distractor-absent trials produced omissions, collapsing across
blocks. For the comparison of high-value and distractor-absent
trials, the Distractor Type Block interaction was nonsignificant,
F(9, 207) .90, p .52, p
2 .04. For the comparison of
low-value and distractor-absent trials, this Distractor Type
Block Interaction approached significance, F(9, 207) 1.87, p
.058, p
2 .075. Figure 3b shows that the difference between
low-value and distractor-absent trials in proportion of omissions
reduced as training progressed.
Similar patterns were seen in RTs (Figure 3c; as noted earlier, a
response was registered in this task when participants had accumulated
100 ms of dwell time inside the target ROI). For these
data, 3 10 ANOVA revealed a significant main effect of
distractor type, F(2, 46) 24.1, p .001, p
2 .51, but no main
effect of block, F(9, 207) .47, p .89, p
2 .02, or interaction,
F(18, 414) 1.22, p .24, p
2 .05. Follow-up t tests, averaging
across training blocks, revealed that RTs were slower with highvalue
distractors (M 442 ms) than low-value distractors
M 429 ms, t(23) 4.00, p .001, d .82, CIdiff [6.31, 19.80],
and fastest on distractor-absent trials: for high-value versus absent,
t(23) 5.68, p .001, d 1.16, CIdiff [20.3, 43.5]; for low-value
versus absent, t(23) 3.94, p .001, d .80, CIdiff [8.93, 28.70].
Figure 3d shows saccade latencies on omission trials, and nonomission
trials (i.e., trials on which participants did not look at the
distractor), averaged across training blocks. Saccade latencies for
distractor-absent omission trials are not shown—even though
omissions could occur on these trials (see Design)—because there
were so few of these trials (8 of 24 participants registered zero
trials in this category, so mean saccade latencies could not be
calculated for these participants). Saccade latency was generally
shorter on omission trials than nonomission trials: this was true for
trials with a high-value distractor, t(23) 6.91, p .001, d
1.41, CIdiff [.021, .038], and with a low-value distractor, t(23)
8.72, p .001, d 1.78, CIdiff [.020, .033]. For nonomission
trials, saccade latency was significantly longer for high-value than
low-value trials, t(23) 2.29, p .031, d .47, CIdiff [.0003,
.0051]. While latency was numerically shortest for distractorabsent
trials, the relevant differences failed to reach significance:
for high-value versus absent, t(23) 1.47, p .15, d .30, CIdiff
[–.0012, .0072]; for low-value versus absent, t(23) .17, p .87,
d .03, CIdiff [–.0035, .0041]. For omission trials, saccade latency
did not differ significantly between high- and low-value trials,
t(23) .04, p .97, d .007, CIdiff [–.0088, .0091].
Finally, the duration of total dwell time on the distractor on
omission trials did not differ significantly between trials with
high-value distractors (M 112 ms, SEM 7.5 ms) and lowvalue
distractors (M 110 ms, SEM 8.2 ms), t(23) .30, p
.77, d .06, CIdiff [–.015, .020].
Discussion
Experiment 3 replicated the value-related effect on RTs observed
in Experiments 1 and 2. Moreover, high-value distractors
produced greater oculomotor capture than low-value distractors.
This was counterproductive because if oculomotor capture occurred,
reward was omitted. Experiment 3 thus provides an interesting
example of reward learning promoting an oculomotor response
that was never rewarded.
Mean saccade latency in Experiment 3 was generally shorter on
omission trials (i.e., trials on which participants looked at the
distractor before looking at the target) than on nonomission trials
(i.e., trials on which participants did not look at the distractor prior
to looking at the target). We interpret this to suggest that the salient
distractor had a tendency to elicit rapid oculomotor capture in a
stimulus-driven fashion, but that participants could use goaldirected
(controlled) processes to overcome this tendency to make
an initial saccade toward the distractor. The longer saccade latency
on nonomission trials then reflects the cost of engaging this con-
This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
166 LE PELLEY, PEARSON, GRIFFITHS, AND BEESLEY
trolled process. On this account, one would also expect longer
saccade latencies on nonomission trials for distractor-present than
distractor-absent displays. While the mean latencies showed this
pattern numerically, the relevant differences did not reach significance.
This could, however, be a consequence of the relatively
low number of distractor-absent trials (80 over the whole experiment,
as compared with 400 distractor-present trials), such that the
mean latency on distractor-absent trials is likely to be a relatively
poor estimate of the population mean.
Most interestingly, saccade latencies on nonomission trials featuring
a high-value distractor were significantly longer than on
nonomission trials featuring a low-value distractor. This implies
that greater cognitive effort was required to suppress the tendency
to saccade toward the high-value distractor than the low-value
distractor.
On omission trials, saccade latency was not significantly different
for high-value and low-value trials. This could be taken to
suggest that—while the high-value distractor was more likely to
produce oculomotor capture (Figure 3b)—when either distractor
actually produced capture, it did so with the same degree of
“force.” Also, the mean dwell time on high-value and low-value
distractors on omission trials was not significantly different. This
could be taken to suggest that, after capture, the ease of attentional
disengagement did not depend on distractor value. However, the
interpretations advanced in this paragraph are speculative. This is
because both rely on null results; in particular, because the number
of omission trials for each participant was relatively small, the
experiment may well have lacked sensitivity to detect differences
on these measures. Future studies could use more extensive training
to generate more omission trials, which would allow these
questions to be studied in greater depth.
It is perhaps worth noting that the omission contingency implemented
in Experiment 3 (wherein if participants looked at the
distractor, the reward was omitted) means that participants must
have learned the signal-value of the distractor colors (high-value
color signals high-value reward and low-value color signals lowvalue
reward) on trials on which they did not look at the distractor.
That is, participants must have encoded the presence of a particular
distractor color in the array using peripheral vision, and this
supported learning about the relationship between the presence of
that color and the reward value obtained on that trial.
General Discussion
Considered in the most general terms, our findings replicate
recent demonstrations that involuntary attentional and oculomotor
capture in visual search is influenced by reward learning (Anderson
et al., 2011a, 2011b, 2012; Anderson & Yantis, 2012; Theeuwes
& Belopolsky, 2012). More importantly, however—and unlike
all previous studies of derived attentional capture in humans—
the current experiments demonstrate value-driven attentional
capture by stimuli that were never task-relevant for participants.
That is, participants were never required to direct their responses
(or attention) toward these stimuli. Indeed, if participants did direct
attention toward the critical distractor stimuli, their payoff was
reduced. Nevertheless, we still found evidence of greater capture
by stimuli that predicted high-value rewards than those predicting
low-value rewards. Moreover, in Experiment 2, this maladaptive
pattern of capture was observed in participants who were seemingly
unaware of the stimulus–reward relationships.
Unlike in previous studies, then, the value-driven capture observed
in the current experiments cannot be a hangover from
participants’ previous experience of being rewarded for directing
attention to the critical stimuli. The clearest evidence for this
comes from Experiment 3, where the oculomotor capture produced
by the distractor stimuli was never rewarded at any point in the
experiment. These findings therefore imply that the crucial determinant
of capture is not instrumental learning about the reward
value produced by orienting attention to a stimulus (responsevalue).
Instead capture seems dependent on Pavlovian learning
about the reward value signaled by the presence of a stimulus
(signal-value). Specifically, our findings suggest that signals of
large reward become more likely to capture attention than signals
of small reward.
The difference in behavior on high-value versus low-value trials
must reflect learning about signal-value, since (across participants)
this was the only difference between distractor stimuli. However,
the size of this difference did not interact significantly with training
block in any of the experiments (although the relevant interaction
for the proportion of omission trials in Experiment 3 approached
significance). While this null result may simply reflect
noise in the block-by-block data, it suggests that the influence of
reward learning on attentional capture developed early and did not
change greatly over the course of training. Notably, Experiment 2
demonstrated that the maladaptive pattern of greater capture by
high-value distractors persisted over extended training. Even with
extensive experience, capture did not come under the control of
response-value; that is, participants did not come to suppress
attention to the high-value distractor (which would have increased
their payoff). The implication is that value-driven attentional capture
remains an effect of Pavlovian conditioning, and that instrumental
learning about the relationship between orienting attention
to distractors and the resulting reduction in payoff never comes to
exert a comparable influence.
At this point, we should acknowledge a subtle issue of interpretation,
and a caveat to some of the arguments set out above. In all
of the current experiments, the value-predictive distractors were
physically salient, since they were the only colored stimuli in the
display. It is well-established that salient stimuli such as these “pop
out,” and capture covert attention automatically (Theeuwes, 1992,
1994). Thus we might expect that covert attention will shift to the
distractors on a certain proportion of trials due to their physical
salience; we term this a distractor-shift. Now let us focus on the
procedure of Experiment 1, in which participants were rewarded if
their response to the target was faster than a certain criterion value
(the latency limit). Suppose that, after a distractor-shift, participants
are still generally able to respond to the target faster than the
latency limit. Hence, the distractor-shift is followed by reward;
more specifically, a distractor-shift to a high-value distractor is
followed by high reward, and a distractor-shift to a low-value
distractor is followed by low reward. Thus distractor-shifts to the
different distractor types are differentially rewarded, even though
shifting attention to the distractor plays no causal role in achieving
that reward. This provides circumstances under which instrumental
conditioning could differentially promote distractor-shifts to the
high-value distractor in the future (in the terminology of learning
theory, this would be an example of superstitious conditioning: cf.
This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
REWARD AND ATTENTION 167
Skinner, 1948). A similar argument could apply to the current
Experiment 2 if, after a distractor-shift, participants were still able
to respond to the target fast enough that the rewards achieved on
high-value trials were (on average) greater than those on low-value
trials.
The procedure of Experiment 3 mitigates against this line of
reasoning to an extent. In this experiment, the overt attentional
behavior that was measured (shifting eye gaze to the distractor)
was never rewarded at any point in the experiment, since making
such a shift would produce omission of reward. Consequently, it is
less straightforward to see how the value-driven modulation of
oculomotor capture observed in this study could be driven by
instrumental conditioning. However, one possibility remains. As
noted above, the distractors in Experiment 3 were physically
salient and hence likely to produce automatic capture of covert
attention. Suppose that there are a subset of trials on which (a)
participants shift covert attention toward the distractor, and (b)
they do not look at the distractor, and (c) their response to the
target is still fast enough to fall within the 600-ms deadline for
earning a reward. Under these circumstances, covert distractorshifts
could then be subject to superstitious instrumental conditioning
as described in the previous paragraph. If we then assume
that the greater likelihood of covert distractor-shifts to the highvalue
distractor occasionally translates into making overt (oculomotor)
shifts to this distractor, then this could account for the
observed pattern of more omission trials for high-value than lowvalue
distractors. On this account, then, the oculomotor bias is a
product of an instrumentally conditioned bias in covert attention.
However, it is important to remember that, if participants do make
an oculomotor shift to the distractor, then they lose a reward that
would otherwise have been earned. Hence if instrumental conditioning
of covert attentional shifts to the distractor builds up to
such a level that it produces an oculomotor shift to this distractor,
then the resulting loss of reward would immediately drive instrumental
conditioning to reduce the tendency of making such oculomotor
shifts in future. Essentially, the “covert bias produces
oculomotor bias” argument is subject to a negative feedback loop,
which reduces the likelihood that it is responsible for the effects
observed in Experiment 3.
More generally, previous research has demonstrated that learning
about rewards modulates attentional capture by task-relevant
stimuli, both when those stimuli are physically salient (Anderson
et al., 2011a), and when they are not (Anderson et al., 2011b). The
current experiments demonstrate that learning about rewards also
modulates attentional capture by task-irrelevant stimuli, in the case
in which they are physically salient. However, the physical salience
of these stimuli opens up the possibility (at least in theory)
that instrumental conditioning might contribute to the value-driven
modulation of attention observed in these experiments. It is noteworthy,
then, that a very recent study by Failing and Theeuwes
(2014) has demonstrated value-driven capture by task-irrelevant
distractors that are not distinguished by their physical salience.
This experiment used a procedure based on the current Experiment
1, but crucially each of the outline shapes in the search display was
uniquely colored (see Figure 4). Participants were required to
report the letter (S or P) inside the shape singleton (the diamond in
Figure 4) as rapidly as possible. As in the current experiments, for
half of participants, the presence of a red distractor in the display
signaled that a fast response to the target would receive high
reward, and a blue distractor signaled low reward; for the other
half of participants this was reversed. As these critical valuepredictive
distractors were no longer color singletons (since all
stimuli were colored), they would not be expected to elicit attentional
capture based on their physical salience. Nevertheless, Failing
and Theeuwes still found slower responses to the target on
trials with a high-value distractor than on trials with a low-value
distractor, implying that the high-value distractor became more
likely to capture attention than the low-value distractor.
Although the current experiments demonstrate that reward
learning can modulate attentional capture by salient stimuli, the
data of Failing and Theeuwes (2014) suggest that reward learning
is sufficient to produce attentional capture even by nonsalient
stimuli. And since the study by Failing and Theeuwes used nonsalient
stimuli, it rules out the account via instrumental conditioning
advanced in the previous paragraphs. Taken together, our
findings and those of Failing and Theeuwes therefore strongly
suggest that value-driven capture by task-irrelevant stimuli is a
product of Pavlovian, not instrumental, conditioning.
Finally in this section, we note an interesting complementary
relationship between previous studies of value-driven capture and
the current findings. Previous studies demonstrated value-driven
capture by stimuli that had previously been task-relevant, but this
capture was observed in a test phase in which these stimuli no
longer predicted reward (e.g., Anderson et al., 2011a, 2011b). The
current studies demonstrate value-driven capture by task-irrelevant
stimuli, during a training phase in which rewards were provided
throughout. The implication, then, is that the influence of value-
Figure 4. Example search display from the study by Failing and Theeuwes
(2014). Participants were required to report the letter (S or P) inside the
shape singleton (the diamond in this example) as rapidly as possible. Each
of the outline shapes in the display was uniquely colored (shown in
different colors in online version, and different grayscale tones in print
version). For half of participants, the presence of a red distractor in the
display signaled that a fast response to the target would result in high
reward and a blue distractor signaled low reward; for the other half of
participants, this was reversed. See the online article for the color version
of this figure.
This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
168 LE PELLEY, PEARSON, GRIFFITHS, AND BEESLEY
learning on attentional capture is quite general. Pairing with highvalue
reward increases the likelihood of capture by a stimulus that
has been task-relevant but no longer predicts reward (Anderson et
al., 2011a, 2011b), and by a stimulus that predicts reward but is
task-irrelevant (current data and Failing & Theeuwes, 2014).
A Previous Study of Capture by Distractors
One previous study by Della Libera and Chelazzi (2009) examined
the influence of reward learning on attention to distractors in
humans. In a complicated procedure, when critical stimuli appeared
as distractors, they signaled reward magnitude (with 80%
validity). Evidence from Libera and Chelazzi’s Experiment 1
suggested that this training led to reduced capture by distractors
that signaled large reward compared to small reward. This is the
opposite of our findings, and suggests response-value was the
critical variable in their case. The reason for this discrepancy
remains unclear; however, we note that: (a) The effect for distractors
in Libera and Chelazzi’s Experiment 1 occurred on only one
of two response measures, at p .04, and did not replicate in
Experiment 2. Our effect replicated across Experiments 1–3 with
medium-to-large effect sizes, in both RTs and oculomotor capture.
(b) Libera and Chelazzi had no consistent distinction between
targets and distractors: a given stimulus acted as a target on some
trials and as a distractor on others, but signaled reward magnitude
only when it appeared in one of these roles. Thus all stimuli in their
study were task-relevant on 50% of their appearances. In our
experiments, colored stimuli only ever appeared as distractors, and
so were task-irrelevant throughout. (c) Attentional capture by
distractors in Libera and Chelazzi’s procedure had no influence on
rewards and hence was not maladaptive; in our experiments,
capture resulted in reduced reward.
Leaving aside for a moment the specific issue of the taskrelevance
or task-irrelevance of stimuli, we would argue that our
task has advantages over previous techniques (e.g., Anderson et
al., 2011a, 2011b, 2012; Libera & Chelazzi, 2009; Theeuwes &
Belopolsky, 2012) as a general procedure for investigating valuedriven
attentional capture in humans. All of these previous procedures
require a lengthy training period during which differences in
value are established, before a shorter, unrewarded test phase in
which value-related differences in attentional capture might be
observed. Since test phase trials are unrewarded, any value-related
effects are liable to dissipate as reward learning extinguishes (e.g.,
see Anderson et al., 2011a). In contrast, our procedure involves
only a single phase. Every trial is both a training trial (on which
reward learning can occur) and a test trial (on which the effects of
that learning on capture can be measured). Consequently, this
procedure is more efficient, allows the influence of value-driven
capture to be observed over prolonged training, and allows the
development of this capture to be tracked online on a session-bysession
or block-by-block basis.
Neural Loci of Value-Driven Attentional Capture
The findings reported here are somewhat similar to those of a
study by Peck, Jangraw, Suzuki, Efem, and Gottlieb (2009) using
monkeys. On each trial of that study, a peripheral visual reward
cue (RC) predicted whether the trial outcome would be juice
reward (RC ) or no reward (RC–). However, to achieve this
outcome, monkeys were required to make a saccade to a target
whose location was independent of the RC. Even though RCs had
no operant role, the RC became more likely to attract attention
and the RC– to repel attention (measured using eye tracking). This
suggests that, as in the current experiments, attention was under
the control of learning about the signal-value of the RC rather than
its response-value. One difference is that in Peck et al.’s (2009)
study, the independence of RC and target location meant that on
50% of trials, the target would appear in the same location as the
RC. Hence, monkeys would often be rewarded for making a
saccade toward a location in which the RC had recently appeared.
In the current experiments, the distractor and target on a
given trial never appeared in the same location, thus producing a
clearer distinction between signal-value and response-value. Nevertheless,
these findings suggest an interesting parallel between
value-driven attention in humans and nonhuman animals. Using
single-unit recording, Peck et al. showed that attentional modulation
in their task was encoded in posterior parietal cortex, specifically
in the lateral intraparietal area.
This latter finding is particularly interesting, because of its
relation to previous data from humans. In a study in humans using
task-relevant stimuli, Kiss et al. (2009) used electroencephalography
to demonstrate that the magnitude of the reward that participants
received for responding to target stimuli modulated eventrelated
potential (ERP) signatures of attentional selection elicited
by those stimuli. Specifically, the N2-posterior-contralateral
(N2pc) ERP component occurred earlier, and had greater magnitude,
for targets rendered in a high-value color than targets in a
low-value color. The N2pc is an early, lateralized component
emerging around 200 ms after display onset, and extensive study
of singleton visual search has identified it as an important correlate
of visual target selection (see Eimer, 1996; Woodman & Luck,
1999). Importantly, neural source analyses based on magnetoencephalography
highlight the posterior parietal cortex as contributing
to the N2pc induced by task-relevant items in visual search
(e.g., see Hopf et al., 2000). Thus, we have two studies implicating
the posterior parietal cortex in value-driven attentional capture:
Kiss et al.’s (2009) study (in humans) using task-relevant stimuli,
and Peck et al.’s (2009) study (in monkeys) using task-irrelevant
stimuli. Although the species and procedures used in these studies
are clearly very different, these findings are at least consistent with
the possibility that the same brain regions—and potentially the
same neural mechanisms—underlie value-driven attentional capture
by both task-relevant and task-irrelevant stimuli. Future work
in humans using the procedures developed in this article will probe
this question more directly, by comparing neural signatures of
value-driven capture by task-relevant and task-irrelevant stimuli
within a single species, and with a single procedure.
Sign-Tracking and Goal-Tracking of Attention
The contrast between signal- and response-value of stimuli that
is made by the current experiments relates to animal conditioning
studies distinguishing between sign-tracking and goal-tracking
behavior (Boakes, 1977). Suppose a neutral cue (e.g., insertion of
a lever into a conditioning chamber) is repeatedly paired with food.
A rat showing goal-tracking behavior learns to approach the location
of food delivery when the lever is inserted. In contrast, a rat
showing sign-tracking behavior will approach the lever itself and
This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
REWARD AND ATTENTION 169
grasp, lick, or gnaw it, indicating that this signal of reward has
become attractive in its own right. Sign-tracking thus involves the
neutral cue taking on “incentive salience” and hence coming to
exert control over behavior, even though it has no operant value in
producing reward. The parallel with the current procedure is clear.
Specifically, our data indicate that reward learning promotes attentional
capture by the distractor (signal) more than by the target
(goal). These findings therefore provide evidence of a signtracking
process modulating attentional capture in humans.
Notably, it has been argued that the attribution of salience to
predictive cues that is implicated in sign-tracking, and the subsequent
ability of those cues to gain inordinate control over behavior,
bears striking similarities to prominent symptoms of drug abuse
(Flagel, Akil, & Robinson, 2009; Tomie, Grimes, & Pohorecky,
2008; Uslaner, Acerbo, Jones, & Robinson, 2006). In support of
this, animal studies suggest that individual differences in the
tendency to engage in sign-tracking versus goal-tracking behavior
may confer vulnerability or resistance to compulsive behavioral
disorders, including addiction (Flagel, Watson, Akil, & Robinson,
2008; Flagel, Watson, Robinson, & Akil, 2007). This raises the
possibility that our procedure might be used to assay the extent to
which an individual’s attentional capture is under the control of
sign-tracking, and hence to identify people who are predisposed
toward showing maladaptive patterns of attentional bias that are
associated with addiction and other compulsive disorders.
Conclusions
As noted in the introduction, research on derived attention has
proceeded largely independently and in parallel in the associative
learning and perception– cognition literatures. We hope that the
research reported here might go some way toward bringing these
literatures closer together. Our results demonstrate value-driven
attentional capture by task-irrelevant stimuli. More generally,
however, these studies show that the well-established techniques
of perceptual– cognitive research (such as visual search and eye
tracking) can be used to shed light on the associative learning
mechanisms underlying derived attentional capture. And as a complement,
a consideration of associative learning theory (such as the
distinction between Pavlovian and instrumental conditioning) can
shed light on the cognitive information structures that support
changes in capture of visual attention. This improved understanding
of the psychology of derived attention can then be used to
investigate the brain mechanisms supporting these processes, and
the mechanisms by which some dysfunctional cognitive features
(e.g., heightened salience of drug paraphernalia in the case of
addiction) are sustained.
References
Anderson, B. A., Faulkner, M. L., Rilee, J. J., Yantis, S., & Marvel, C. L.
(2013). Attentional bias for nondrug reward is magnified in addiction.
Experimental and Clinical Psychopharmacology, 21, 499–506. http://
dx.doi.org/10.1037/a0034575
Anderson, B. A., Laurent, P. A., & Yantis, S. (2011a). Learned value
magnifies salience-based attentional capture. PLoS ONE, 6, e27926.
http://dx.doi.org/10.1371/journal.pone.0027926
Anderson, B. A., Laurent, P. A., & Yantis, S. (2011b). Value-driven
attentional capture. Proceedings of the National Academy of Sciences of
the United States of America, 108, 10367–10371. http://dx.doi.org/
10.1073/pnas.1104047108
Anderson, B. A., Laurent, P. A., & Yantis, S. (2012). Generalization of
value-based attentional priority. Visual Cognition, 20, 647–658. http://
dx.doi.org/10.1080/13506285.2012.679711
Anderson, B. A., & Yantis, S. (2012). Value-driven attentional and oculomotor
capture during goal-directed, unconstrained viewing. Attention,
Perception, & Psychophysics, 74, 1644 –1653. http://dx.doi.org/
10.3758/s13414-012-0348-2
Awh, E., Belopolsky, A. V., & Theeuwes, J. (2012). Top-down versus
bottom-up attentional control: A failed theoretical dichotomy. Trends in
Cognitive Sciences, 16, 437–443. http://dx.doi.org/10.1016/j.tics.2012
.06.010
Boakes, R. A. (1977). Performance on learning to associate a stimulus with
positive reinforcement. In H. Davis & H. Hurwirtz (Eds.), Operant-
Pavlovian interactions (pp. 67–97). Hillsdale, NJ: Erlbaum.
Brainard, D. H. (1997). The psychophysics toolbox. Spatial Vision, 10,
433–436. http://dx.doi.org/10.1163/156856897X00357
Chelazzi, L., Perlato, A., Santandrea, E., & Libera, C. D. (2013). Rewards
teach visual selective attention. Vision Research, 85, 58–72. http://dx
.doi.org/10.1016/j.visres.2012.12.005
Cousineau, D. (2005). Confidence intervals in within-subject designs: A
simpler solution to Loftus and Masson’s method. Tutorials in Quantitative
Methods for Psychology, 1, 42–45.
Cox, W. M., Hogan, L. M., Kristian, M. R., & Race, J. H. (2002). Alcohol
attentional bias as a predictor of alcohol abusers’ treatment outcome.
Drug and Alcohol Dependence, 68, 237–243. http://dx.doi.org/10.1016/
S0376-8716(02)00219-3
Dayan, P. (2009). Dopamine, reinforcement learning, and addiction. Pharmacopsychiatry,
42, S56 –S65. http://dx.doi.org/10.1055/s-0028-
1124107
Della Libera, C., & Chelazzi, L. (2009). Learning to attend and to ignore
is a matter of gains and losses. Psychological Science, 20, 778–784.
http://dx.doi.org/10.1111/j.1467-9280.2009.02360.x
Eimer, M. (1996). The N2pc component as an indicator of attentional
selectivity. Electroencephalography & Clinical Neurophysiology, 99,
225–234. http://dx.doi.org/10.1016/0013-4694(96)95711-9
Failing, M. F., & Theeuwes, J. (2014). Don’t let it distract you: How
information about the availability of reward affects attentional selection.
Manuscript submitted for publication.
Flagel, S. B., Akil, H., & Robinson, T. E. (2009). Individual differences in
the attribution of incentive salience to reward-related cues: Implications
for addiction. Neuropharmacology, 56, 139 –148. http://dx.doi.org/
10.1016/j.neuropharm.2008.06.027
Flagel, S. B., Watson, S. J., Akil, H., & Robinson, T. E. (2008). Individual
differences in the attribution of incentive salience to a reward-related
cue: Influence on cocaine sensitization. Behavioural Brain Research,
186, 48–56. http://dx.doi.org/10.1016/j.bbr.2007.07.022
Flagel, S. B., Watson, S. J., Robinson, T. E., & Akil, H. (2007). Individual
differences in the propensity to approach signals vs goals promote
different adaptations in the dopamine system of rats. Psychopharmacology,
191, 599–607. http://dx.doi.org/10.1007/s00213-006-0535-8
Frank, M. J. (2008). Schizophrenia: A computational reinforcement learning
perspective. Schizophrenia Bulletin, 34, 1008–1011. http://dx.doi
.org/10.1093/schbul/sbn123
Holland, P. C. (1979). Differential effects of omission contingencies on
various components of Pavlovian appetitive conditioned responding in
rats. Journal of Experimental Psychology: Animal Behavior Processes,
5, 178–193. http://dx.doi.org/10.1037/0097-7403.5.2.178
Hopf, J. M., Luck, S. J., Girelli, M., Hagner, T., Mangun, G. R., Scheich,
H., & Heinze, H. J. (2000). Neural sources of focused attention in visual
search. Cerebral Cortex, 10, 1233–1241. http://dx.doi.org/10.1093/
cercor/10.12.1233
Hyman, S. E. (2005). Addiction: A disease of learning and memory. The
American Journal of Psychiatry, 162, 1414–1422. http://dx.doi.org/
10.1176/appi.ajp.162.8.1414
This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
170 LE PELLEY, PEARSON, GRIFFITHS, AND BEESLEY
James, W. (1890/1983). The principles of psychology. Cambridge, MA:
Harvard University Press.
Kapur, S. (2003). Psychosis as a state of aberrant salience: A framework
linking biology, phenomenology, and pharmacology in schizophrenia.
The American Journal of Psychiatry, 160, 13–23. http://dx.doi.org/
10.1176/appi.ajp.160.1.13
Kiss, M., Driver, J., & Eimer, M. (2009). Reward priority of visual target
singletons modulates event-related potential signatures of attentional
selection. Psychological Science, 20, 245–251. http://dx.doi.org/
10.1111/j.1467-9280.2009.02281.x
Klauer, K. C., Roßnagel, C., & Musch, J. (1997). List-context effects in
evaluative priming. Journal of Experimental Psychology: Learning,
Memory, and Cognition, 23, 246–255. http://dx.doi.org/10.1037/0278-
7393.23.1.246
Kleiner, M., Brainard, D. H., & Pelli, D. G. (2007). What’s new in
Psychtoolbox-3? Perception, 36, ECVP Abstract Supplement.
Le Pelley, M. E. (2004). The role of associative history in models of
associative learning: A selective review and a hybrid model. The Quarterly
Journal of Experimental Psychology B: Comparative and Physiological
Psychology, 57, 193–243. http://dx.doi.org/10.1080/
02724990344000141
Le Pelley, M. E. (2010). Attention and human associative learning. In C. J.
Mitchell & M. E. Le Pelley (Eds.), Attention and associative learning:
From brain to behaviour (pp. 187–215). New York, NY: Oxford University
Press.
Le Pelley, M. E., Vadillo, M., & Luque, D. (2013). Learned predictiveness
influences rapid attentional capture: Evidence from the dot probe task.
Journal of Experimental Psychology: Learning, Memory, and Cognition,
39, 1888–1900. http://dx.doi.org/10.1037/a0033700
Livesey, E. J., Harris, I. M., & Harris, J. A. (2009). Attentional changes
during implicit learning: Signal validity protects a target stimulus from
the attentional blink. Journal of Experimental Psychology: Learning,
Memory, and Cognition, 35, 408 – 422. http://dx.doi.org/10.1037/
a0014525
Luck, S. J., & Woodman, G. F. (1999). Electrophysiological measurement
of rapid shifts of attention during visual search. Nature, 400, 867–869.
http://dx.doi.org/10.1038/23698
Ludwig, C. J. H., & Gilchrist, I. D. (2002). Stimulus-driven and goaldriven
control over visual selection. Journal of Experimental Psychology:
Human Perception and Performance, 28, 902–912. http://dx.doi
.org/10.1037/0096-1523.28.4.902
MacLeod, C., Mathews, A., & Tata, P. (1986). Attentional bias in emotional
disorders. Journal of Abnormal Psychology, 95, 15–20. http://dx
.doi.org/10.1037/0021-843X.95.1.15
Marissen, M. A. E., Franken, I. H. A., Waters, A. J., Blanken, P., van den
Brink, W., & Hendriks, V. M. (2006). Attentional bias predicts heroin
relapse following treatment. Addiction, 101, 1306–1312. http://dx.doi
.org/10.1111/j.1360-0443.2006.01498.x
Morris, R., Griffiths, O., Le Pelley, M. E., & Weickert, T. W. (2013).
Attention to irrelevant cues is related to positive symptoms in schizophrenia.
Schizophrenia Bulletin, 39, 575–582. http://dx.doi.org/10.1093/
schbul/sbr192
Pearce, J. M., & Mackintosh, N. J. (2010). Two theories of attention: A
review and a possible integration. In C. J. Mitchell & M. E. Le Pelley
(Eds.), Attention and associative learning: From brain to behaviour (pp.
11–40). New York, NY: Oxford University Press.
Peck, C. J., Jangraw, D. C., Suzuki, M., Efem, R., & Gottlieb, J. (2009).
Reward modulates attention independently of action value in posterior
parietal cortex. The Journal of Neuroscience, 29, 11182–11191. http://
dx.doi.org/10.1523/JNEUROSCI.1929-09.2009
Pelli, D. G. (1997). The VideoToolbox software for visual psychophysics:
Transforming numbers into movies. Spatial Vision, 10, 437–442. http://
dx.doi.org/10.1163/156856897X00366
Posner, M. I. (1980). Orienting of attention. The Quarterly Journal of
Experimental Psychology, 32, 3–25. http://dx.doi.org/10.1080/
00335558008248231
Robinson, T. E., & Berridge, K. C. (2001). Incentive-sensitization and
addiction. Addiction, 96, 103–114. http://dx.doi.org/10.1046/j.1360-
0443.2001.9611038.x
Rutherford, H. J. V., O’Brien, J. L., & Raymond, J. E. (2010). Value
associations of irrelevant stimuli modify rapid visual orienting. Psychonomic
Bulletin & Review, 17, 536–542. http://dx.doi.org/10.3758/PBR
.17.4.536
Skinner, B. F. (1948). Superstition in the pigeon. Journal of Experimental
Psychology, 38, 168–172. http://dx.doi.org/10.1037/h0055873
Theeuwes, J. (1991). Cross-dimensional perceptual selectivity. Perception
& Psychophysics, 50, 184–193. http://dx.doi.org/10.3758/BF03212219
Theeuwes, J. (1992). Perceptual selectivity for color and form. Perception
& Psychophysics, 51, 599–606. http://dx.doi.org/10.3758/BF03211656
Theeuwes, J. (1994). Endogenous and exogenous control of visual selection.
Perception, 23, 429–440. http://dx.doi.org/10.1068/p230429
Theeuwes, J., & Belopolsky, A. V. (2012). Reward grabs the eye: Oculomotor
capture by rewarding stimuli. Vision Research, 74, 80–85.
http://dx.doi.org/10.1016/j.visres.2012.07.024
Theeuwes, J., Kramer, A. F., Hahn, S., & Irwin, D. E. (1998). Our eyes do
not always go where we want them to go: Capture of the eyes by new
objects. Psychological Science, 9, 379–385. http://dx.doi.org/10.1111/
1467-9280.00071
Theeuwes, J., Kramer, A. F., Hahn, S., Irwin, D. E., & Zelinsky, G. J.
(1999). Influence of attentional capture on oculomotor control. Journal
of Experimental Psychology: Human Perception and Performance, 25,
1595–1608. http://dx.doi.org/10.1037/0096-1523.25.6.1595
Tomie, A., Grimes, K. L., & Pohorecky, L. A. (2008). Behavioral characteristics
and neurobiological substrates shared by Pavlovian signtracking
and drug abuse. Brain Research Reviews, 58, 121–135. http://
dx.doi.org/10.1016/j.brainresrev.2007.12.003
Uslaner, J. M., Acerbo, M. J., Jones, S. A., & Robinson, T. E. (2006). The
attribution of incentive salience to a stimulus that signals an intravenous
injection of cocaine. Behavioural Brain Research, 169, 320–324. http://
dx.doi.org/10.1016/j.bbr.2006.02.001
Waters, A. J., Shiffman, S., Sayette, M. A., Paty, J. A., Gwaltney, C. J., &
Balabanis, M. H. (2003). Attentional bias predicts outcome in smoking
cessation. Health Psychology, 22, 378–387. http://dx.doi.org/10.1037/
0278-6133.22.4.378
Yantis, S. (2000). Goal-directed and stimulus-driven determinants of attentional
control. In S. Monsell & J. Driver (Eds.), Attention and
performance XVIII (pp. 73–103). Cambridge, MA: MIT Press.
Received April 13, 2014
Revision received September 24, 2014
Accepted October 7, 2014
This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
REWARD AND ATTENTION 171