The Role of Context in Volitional Control of Feature-Based Attention

Abstract

Visual selection can be biased toward nonspatial feature values such as color, but there is continued debate about whether this bias is subject to volitional control or whether it is an automatic bias toward recently seen target features (selection history). Although some studies have tried to separate these 2 sources of selection bias, mixed findings have not offered a clear resolution. The present work offers a possible explanation of conflicting findings by showing that the context in which a trial is presented can determine whether volitional control is observed. We used a cueing task that enabled independent assessments of the effects of color repetitions and current selection goals. When the target was presented among distractors with multiple colors (heterogeneous blocks), Experiment 1 revealed clear goal-driven selection effects, but these effects were eliminated when the target was a color singleton (pop-out blocks). When heterogeneous and pop-out displays were mixed within a block (Experiment 2), however, goal-driven selection was observed with both types of displays. In Experiment 3, this pattern was replicated using an encoding-limited task that included brief displays and an A' measure of performance. Thus, goal-driven selection of nonspatial features is potentiated in contexts where there is strong competition with distractors. Selection history has powerful effects, but we find clear evidence that observers can exert volitional control over feature-based attention. (PsycINFO Database Record

Full text

Feature-based attention 1
The role of context in volitional control of feature-based attention
Artem V. Belopolsky1 & Edward Awh2
VU University Amsterdam1
University of Oregon, Eugene2
Journal of Experimental Psychology: HPP
(In Press)
©American Psychological Association
http://dx.doi.org/10.1037/xhp0000135
This article may not exactly replicate the final version published in the APA journal. It is not the copy of
record.
Address for correspondence:
Artem Belopolsky
Dept. of Cognitive Psychology
VU University Amsterdam
Van der Boechorststraat 1
1081 BT Amsterdam
The Netherlands
Phone: +31 20 598 8714
Fax: +31 20 598 8971
Email: A.Belopolskiy@vu.nl

Feature-based attention 2
Abstract
Visual selection can be biased towards non-spatial feature values such as color, but there is continued
debate about whether this bias is subject to volitional control or whether it is an automatic bias towards
recently seen target features (selection history). Although some studies have tried to separate these two
sources of selection bias, mixed findings have not offered a clear resolution. The present work offers a
possible explanation of conflicting findings by showing that the context in which a trial is presented can
determine whether volitional control is observed. We used a cueing task that enabled independent
assessments of the effects of color repetitions and current selection goals. When the target was
presented amongst distractors with multiple colors (heterogeneous blocks), Experiment 1 revealed clear
goal-driven selection effects, but these effects were eliminated when the target was a color singleton
(pop-out blocks). When heterogeneous and pop-out displays were mixed within a block (Experiment 2),
however, goal-driven selection was observed with both types of displays. In Experiment 3 this pattern was
replicated using an encoding-limited task that included brief displays and an A’ measure of performance.
Thus, goal-driven selection of non-spatial features is potentiated in contexts where there is strong
competition with distractors. Selection history has powerful effects, but we find clear evidence that
observers can exert volitional control over feature-based attention.
Keywords: feature-based attention, goal-driven control, top-down control, biased competition

Feature-based attention 3
The role of context in volitional control of feature-based attention
Since the typical visual scene contains far more information than an observer has the capacity to
process at once, selective attention is crucial for directing limited processing resources towards the most
relevant aspects of the environment. Thus, a central research question in psychology has been to
understand how selective attention is controlled. The most prominent models of attentional control
propose that attention can be allocated in a voluntary fashion (consistent with the observer’s current
goals) or in an automatic fashion that is determined by the physical properties of the stimulus display
(i.e. “top-down vs. bottom-up” control, Corbetta & Shulman, 2002; Egeth & Yantis, 1997; Jonides, 1981;
Posner, Davidson, & Snyder, 1980; Theeuwes, Olivers, & Belopolsky, 2010; c.f. Awh, Belopolsky, &
Theeuwes, 2012). In line with this framework, a consensus has emerged that observers can direct spatial
attention to specific locations “at will” (e.g., Eriksen & Hoffman, 1973; Moran & Desimone, 1985; Posner
et al., 1980) or spatial attention can be “captured” at specific locations regardless of the observer’s will by
salient events such as an abrupt onset (Jonides, 1981; Yantis & Jonides, 1990). In either case, target
discrimination at attended locations is faster and more accurate than at unattended locations (Eriksen &
Hoffman, 1973; Jonides, 1981; Posner et al., 1980; Yantis & Jonides, 1990).
In line with the findings from the spatial attention literature, many influential theories of visual
selection also describe control processes that direct attention towards non-spatial features such as color,
orientation or motion. For example, Guided Search (Wolfe, Cave, & Franzel, 1989) suggests that
observers can prepare to search for a target with a certain feature (e.g. red) or dimension (e.g. color), by
placing a ‘higher weight on the output of one channel than on others’ (Wolfe, 2007, p.105). Similarly,
according to the Dimensional Weighting Account (Found & Müller, 1996) “preattentive saliency
computations may be biased by top-down signals reflecting expectations of particular stimulus attributes”
(Müller et al., 2010, p.118). This in turn biases the processing in favor of the elements sharing the target
feature or dimension in a subsequently presented display. In support of this idea it has been
demonstrated that observers are more efficient in finding the target among distractors, when they
possess the advance knowledge of the target feature (Müller, Reimann, & Krummenacher, 2003), and
they have difficulty ignoring distractors that share features with the target (Folk, Remington, &
Jonhnston, 1992). Similarly, neuroimaging studies have demonstrated that attending to color or motion
results in increase in baseline activity in the corresponding areas of early visual cortex (Chawla, Rees, &
Friston, 1999; Saenz, Buracas, & Boynton, 2002), spreading across the whole visual field, including the
empty areas (Serences & Boynton, 2007). In addition, recent work (Zhang & Luck, 2008) has shown that a
distractor that shares a color with the target elicits a larger visually-evoked neural response even when it
is presented at an unattended location.

Feature-based attention 4
Nevertheless, despite the diverse array of evidence showing that selection can be biased towards
non-spatial features, the extant data do not yet make a strong case for goal-driven control over these
visual biases. Many past studies are ambiguous because they relied on blocked designs in which the
relevant non-spatial feature (e.g., the target-defining color) was held constant across multiple trials (e.g.,
Folk et al., 1992; Wolfe, Butcher, Lee, & Hyle, 2003). This design allows for inter-trial priming effects in
which the repetition of a target feature yields long-lasting benefits in both the speed and accuracy of
responses. Moreover, these selection history effects have been shown to occur automatically, regardless
of whether the observer is aware of the repetition or whether their current goals have shifted to a
different feature value (e.g., Maljkovic & Nakayama, 1994). Thus, when selection history is confounded
with the putative effects of goal-driven selection, there is no clear evidence that goal-driven selection has
had an effect (Awh et al., 2012; Belopolsky, Schreij, & Theeuwes, 2010; Theeuwes, 2013).
This point was clearly illustrated in a study by Theeuwes and Van der Burg (2007) that directly
compared the effect of spatial and non-spatial precues in a visual search task. To allow separation of goal-
driven selection and selection history, Theeuwes and Van der Burg varied the specific location or color
that was cued on a trial-by-trial basis. In addition, to prevent priming from the physical presentation of
the cues, the locations and color precues were communicated with words that did not require the
physical presence of the cued feature value. Finally, all targets in this study were color singletons, because
they reasoned that such “pop-out” targets (combined with brief, masked displays and a behavioral
measure of perceptual sensitivity) would provide a cleaner index of early stages of visual processing that
occur during the first feedforward sweep of visual activity (Nothdurft, Gallant, & van Essen, 1999;
Treisman, 1988). The results showed a striking contrast between the efficacy of the spatial and non-
spatial cues. Clear evidence for goal-driven selection – independent of selection history effects – was
observed with the spatial precues, but no such effect was observed with the color precues. Thus,
Theeuwes and Van der Burg (2007) concluded that while observers could exert goal-driven control over
spatial attention, only automatic priming of specific feature values was possible with non-spatial cues.
Indeed, when the color word cues were replaced with physical cues that contained the cued feature
value, Theeuwes and Van der Burg (2007) observed reliable benefits of the color cues; critically, this effect
was observed regardless of whether the cue was predictive or not, suggesting that it was not connected
with the volitional selection goals of the observer.
To summarize, despite clear evidence for goal-driven selection in the spatial domain, the role of
volitional control in the selection of non-spatial features is less certain. Clear evidence for goal-driven
selection of non-spatial features requires an experimental design that can disentangle the automatic
biases that are caused by selection history and the biases in visual selection that are subject to volitional
control. This motivates an experimental design that includes trial-by-trial variations in the cued feature
value (so that it is possible to test whether current goals bias visual selection in the absence of repetition

Feature-based attention 5
priming), and that employs abstract cues rather than physical presentations of the cued feature value. To
date, a handful of studies have fulfilled these criteria (Leonard & Egeth, 2008; e.g., Mortier, Theeuwes, &
Starreveld, 2005; Müller & Krummenacher, 2006; Müller et al., 2003; Theeuwes, Reimann, & Mortier,
2006; Theeuwes & Van der Burg, 2007; Zehetleitner, Krummenacher, Geyer, Hegenloh, & Müller, 2011),
but the findings have been mixed. While some studies have seen evidence of goal-driven selection that
cannot be explained by selection history (Müller & Krummenacher, 2006; Müller et al., 2003; Zehetleitner
et al., 2011), the size of the effects (about 10 ms faster for valid than for neutral trials) has often been
modest. In a recent study large cueing effects were observed, but only for the smaller display sizes
(Leonard & Egeth, 2008). One concern with this result is that with only a few elements in the display,
there is a large uncertainty about the target feature, which could imply the decisional and not attentional
origin of the cueing effects (Meeter & Olivers, 2006). Moreover, various studies that used almost identical
experimental designs did not find reliable benefits of the non-spatial cues (Mortier et al., 2005; Theeuwes
et al., 2006; Theeuwes & Van der Burg, 2007). Thus, our goal in the present work was to examine the
boundary conditions of goal-driven selection with non-spatial cues, with an eye towards reconciling some
of the conflicting findings in the literature.
One characteristic of all of the prior work noted above is that the target stimuli were color
singletons within an array of homogenous distractors. Because pop-out stimuli are highly salient even
when they are not relevant to the current task (e.g., Maljkovic & Nakayama, 1994), one might question
whether or not observers in these studies were strongly motivated to use the color cues. Both Theeuwes
et al. (2006) and Muller and Krummenacher (2006) attempted to address this concern by including filler
trials on which participants had to report the cue or compliance with the cue; participants were extremely
accurate in reporting the cue or indicated high compliance. Nevertheless, we hypothesized that increased
competition between the target stimulus and the surrounding distractors might help to elicit more robust
evidence of goal-driven selection. Such an interaction between the effects of visual selection and the
degree of distractor interference is predicted by the biased competition theory of selective attention
(Desimone & Duncan, 1995). According to this framework, attention enhances the signal-to-noise ratio for
relevant stimuli by biasing competitive interactions between targets and distractors. In line with this
claim, many studies have shown that the effects of visual selection are indeed enhanced when there is
substantial distractor interference in a target display (Awh, Matsukura, & Serences, 2003; Awh, Sgarlata,
& Kliestik, 2005; Dosher & Lu, 2000; Kastner, De Weerd, Desimone, & Ungerleider, 1998; Shiu & Pashler,
1994).
All of these considerations led us to employ a task similar to that used by Theeuwes and Van der
Burg (2007), except that we also included an active manipulation of the strength of distractor interference
in the target display. Observers searched for either a blue or an orange target. Before the search display
participants received a word cue indicating the likely target color. We included two types of search trials.

Feature-based attention 6
In the pop-out condition the target was a singleton, presented among homogeneously colored distractors.
In the heterogeneous condition, the target was presented amongst an array of distractors that each had a
unique color thereby amplifying the amount of distractor interference. Motivated by the biased
competition perspective, we reasoned that the latter condition would provide a clear incentive to engage
in goal-driven selection based on the cue, and would magnify the performance benefit observed for
validly cued stimuli. To anticipate our findings, in Experiment 1 where the pop-out and heterogeneous
conditions were blocked, cueing effects were either absent or minimized in the pop-out condition
depending on the analytic approach. By contrast, when the pop-out and heterogeneous conditions were
intermixed within blocks in Experiment 2 we found more robust evidence of goal-driven selection in the
pop-out condition. These findings suggest that the pop-out target displays used in past studies may not
have been conducive to robust goal-driven selection effects. Finally, Experiment 3 replicated these
empirical patterns using masked displays and with accuracy as a dependent measure. This suggests that
the benefits of goal-driven selection influenced the perceptual encoding of the target stimuli rather than
just efficiency of post-perceptual decision or response processes.
Experiment 1
Participants received a word cue indicating the upcoming target color (orange or blue) with 80%
validity. The word cue did not share any features with the upcoming target; thus, this cue eliminated the
possibility of low-level priming from the physical presentation of the relevant hue. Participants received
ample time to use the cue and bias their visual selection. The pop-out and heterogeneous conditions
were blocked. This design allowed us to examine whether there were goal-driven selection effects that
were not contingent on selection history.
Methods
Participants
Twenty naïve participants (7 females, mean age of 24, age range 20-32 years) from VU University
Amsterdam with normal or corrected to normal vision participated in the experiment.
Apparatus & Stimuli
Experiment was programmed in E-Prime (Schneider, Eschman, & Zuccolotto, 2002). Stimuli were
presented on a 22-inch Samsung Syncmaster 2233RZ monitor at 120 Hz.
A trial started with the presentation of a white instructional cue (“ORANGE” or “BLUE”, Courier
new; Font size 18) at the center of a black display for 2000 ms (see Figure 1). It was followed by a display
with a fixation dot (0.5°) which after 300 ms was replaced by a search display. The search display

Feature-based attention 7
consisted of six outline circles (2.8° in diameter) positioned on an imaginary circle with a radius of 6.7°
(clock positions 1, 3, 5, 7, 9 and 11). One of the circles was the target and could be either orange (CIE:
.475, .467; 16.5 cd/m2) or blue CIE: .153 .106; 6.3 cd/m2). In the pop-out condition the target circle was
unique and all distractor circles were green (CIE: .196, .701; 15.9 cd/m2). In the heterogeneous condition
the target was not unique since the distractor circles were all different colors (red (CIE: .574, .390; 14.0
cd/ m2), green (CIE: .196, .701; 15.9 cd/m2), purple (CIE: .302, .219; 9.0 cd/m2, yellow (CIE: .39, .54; 96.0
cd/m2 and pink (CIE: .409, .393; 43.6 cd m2). Inside each circle there was a light gray line segment (CIE:
.256, .440, 30.8 cd/m2). In the distractor circles, the line segments were tilted 22.5° to either side of the
horizontal or vertical plane. The target circle always contained either horizontal or vertical line segment.
The search display was presented until response was made or until 2000 ms have elapsed. The inter-trial
interval was 1500 ms.
Design & Procedure
Each observer was seated 70 cm from a computer screen, with head positioned on a chinrest. Participants
were instructed to search for either an orange or a blue circle among distractor circles and to determine
the orientation of a line segment inside of it. On every trial the color of the target was equally likely to be
either orange or blue, while its position on the imaginary circle and the line orientation inside of it was
chosen randomly. Participants were asked to respond quickly and accurately by pressing the “z” key when
the line segment inside the target circle was oriented vertically and the “m” key when it was oriented
horizontally. Participants were informed that the word cue preceding the search display indicated the
target color with 80% validity. The pop-out and heterogeneous conditions were presented in separate
blocks, while target color and the presentation of valid (80% of trials) or invalid (20% of trials) color cue
were mixed randomly within blocks. Half of the participants started with the pop-out condition (10 blocks
of 40 trials) and the other half started with the heterogeneous condition (10 blocks of 40 trials). Each
search condition was preceded by two respective practice blocks (40 trials each). In practice blocks
participants received feedback about correctness of their response after each trial (a word “correct” or
“incorrect” in the middle of the screen). After each block, participants received feedback about their
average reaction time and accuracy.

Feature-based attention 8
Figure 1. Time-course of a typical trial in Experiments 1 and 2. Participants had to search for either an
orange (thick line) or a blue (dashed line) circle and determine the orientation of the line inside it.
Before the search display they were provided with a word cue, which indicated the likely color of the
target with 80% validity. Top: an example of a trial in the pop-out condition. Bottom: an example of a
trial in the heterogeneous condition. In Experiment 1 pop-out and heterogeneous conditions were
blocked, while in Experiment 2 they were mixed.

Feature-based attention 9
Results and Discussion
Error rates
Overall, participants made very few errors (3%). The analysis on error rates showed no significant effects
or interaction.
Reaction times
Trials in which participants responded faster than 150 ms or slower than 1500 ms were excluded
from further analysis. This led to a loss of 1.8% of the trials.
A within-subject analysis of variance (ANOVA) with search condition (pop-out vs. heterogeneous)
and cue validity (valid vs. invalid) revealed a main effect of search condition (F(1,19) = 129.33, p<.001),
indicating that as expected it took longer to find the target in the heterogeneous condition (787 ms) than
in the pop-out condition (610 ms). The cue validity was also significant (F(1,19) = 33.76, p<.001),
indicating that participants were faster in finding the target when the word cue indicated the upcoming
target correctly. The search condition by cue validity interaction was significant (F(1,19) = 34.02, p<.001),
indicating that the cueing effect was larger in the heterogeneous condition (142 ms) than in the pop-out
condition (14 ms). Importantly, planned comparisons revealed that a significant cueing effect was present
in both conditions (heterogeneous condition, t(1, 19) = 5.90, p<.001; pop-out condition t(1, 19) = 3.04,
p<.01). These cueing effects, however, do not yet demonstrate goal-driven selection of the cued color,
because the analysis did not correct for the expected effects of selection history. Instead, these effects
could have been caused by the automatic benefits of responding to a target color that is the same as the
target color in the preceding trial. To examine this possibility, we performed an inter-trial analysis.
Inter-trial effects
There are two ways to examine the inter-trial effects in this task. On the one hand, trials can be
grouped based on whether or not the color cued in a trial is the same as the color of the most recently
presented target. This approach allows us to examine whether having an immediate experience with
selecting a specific color facilitates the voluntary selection of the same color on the next trial. In other
words, selecting the color “blue” on one trial should facilitate the usage of the cue word “BLUE” in guiding
attention on the next trial. This is similar to the logic that was previously used to parse out the inter-trial
and goal-driven influences in control of visual attention (Belopolsky et al., 2010; Folk & Remington, 2008).
On the other hand, trials could be grouped based on whether the color of the current target matches the
color of the most recent target, regardless of the cued color. This approach enables a focus on the
stimulus-driven consequences of repeating a target color. Throughout this paper, we report the results of
both analytic approaches, and discuss their distinct strengths and weaknesses.
Previous target-current cue analysis. The results of this inter-trial analysis are presented in Figure
2. A within-subject analysis of variance (ANOVA) with search condition (pop-out vs. heterogeneous), cue

Feature-based attention 10
matching the previous target (match vs. mismatch) and cue validity (valid vs. invalid) revealed that the
cueing effect was much larger on the trials on which the word cue matched the previous target feature
(cue match by cue validity interaction (F(1,19) = 49.33, p<.001), and this effect was much larger for the
heterogeneous trials than for the pop-out trials (three-way interaction, F(1,19) = 25.99, p<.001). There
was no main effect of cue match or cue match by search condition interaction (both Fs <1). Planned
comparisons showed that the cueing effect was significant in all conditions, except when the cue did not
match the previous target feature on the pop-out trials. For the heterogeneous condition the cueing
effect was 215 ms when the cue matched the previous target feature (t(19) = 8.94, p<.001) and 68 ms
when it mismatched (t(19) = 2.33, p<.05). For the pop-out condition the cueing effect was 29 ms when the
cue matched the previous target feature (t(19) = 4.42, p<.001) and -4 ms when it mismatched (t(19) =
0.67, p=.5).
The inter-trial analysis suggests that inter-trial priming has a profound effect on biasing feature
selection in both heterogeneous and pop-out conditions. The apparent cueing effect diminished from 215
ms to 68 ms in the heterogeneous condition, and from 29 ms to -4 ms in the pop-out condition. Thus, this
analysis suggests that only the heterogeneous condition produced reliable evidence of goal-driven
selection in Experiment 1

Feature-based attention 11
Figure 2. Mean reaction times (RTs) in the pop-out and heterogeneous conditions as a function of cue
validity and whether the word cue matched the color of the target on the previous trial in Experiment
1. The error bars represent standard error of the mean for within-subject designs normalized for the
cue validity factor (Loftus & Masson, 1994). The inset illustrates the conditions plotted on the graph,
the search target was either orange (thick line) or blue (dashed line).
One possible concern with this analysis is that the magnitude of the cueing effects co-varied with
overall response speed. For example, Zehetleitner et al. (2011) found larger cueing effects for targets of
relatively low salience. They argued that since the evidence accumulation rate is higher for salient
targets, then the effect of top-down dimension cueing on further rate increase is limited (e.g., a ceiling
effect). To partial out the large differences in response speed between the two search conditions we
performed the inter-trial analysis on normalized data. The data was normalized per search condition
relative to the respective mean and standard deviation (z-score). All main effects and interactions were
preserved. The cueing effect was much larger on the trials on which the word cue matched the previous
target feature (cue match by cue validity interaction (F(1,19) = 48.46, p<.001), and this effect was much
larger for the heterogeneous trials than for the pop-out trials (three-way interaction, F(1,19) = 17.76,

Feature-based attention 12
p<.001). There was no main effect of cue match or cue match by search condition interaction (both Fs <1).
Note that the normalization procedure does not change the planned comparisons for each condition. This
analysis suggests that the overall response speed did not determine the pattern of results.
To reiterate, the preceding analysis grouped trials based on whether the current cue word
matched or mismatched the color of the most recent target. One concern for this analysis is that
repetition priming effects were imbalanced in the valid and invalid conditions. For example, in the
mismatch condition (see the inset in Figure 2), the target color repeats for the invalid cues, but not for the
valid cues. Therefore, target repetition priming would have enhanced performance in the invalid (but not
the valid condition) yielding a smaller validity effect. Thus, we next present an analysis that groups trials
based on whether the target color repeated or not.
Previous target-current target analysis. Here, the data were analyzed by grouping trials based on
whether the color of the prior target matched the color of the current target (regardless of the currently
cued color). This analysis can assess whether repetitions of the target color interact with the cue validity
(e.g., Leonard & Egeth, 2008; Weidner & Müller, 2013). A within-subject analysis of variance (ANOVA)
with search condition (pop-out vs. heterogeneous), target repetition (same target vs. different target) and
cue validity (valid vs. invalid) revealed a significant main effect of target repetition (F(1,19) = 49.33,
p<.001), as well as significant search condition by target repetition interaction (F(1,19) = 25.99, p<.001)
and search condition by cue validity interaction (F(1,19) = 32.35, p<.001). Importantly, there was no
interaction between target repetition and cue validity and no three-way interaction between search
condition, target repetition and cue validity (all F<1). This analysis suggests that both target repetition
priming and goal-driven control have independent effects on visual selection. One caveat for this
conclusion, however, is that the previous target-current target analysis suffers from a confound between
cue validity and the match between current cue and past target colors. For example, in the different
target condition the currently cued color differs from the color of the most recent target in the valid trials
but not the invalid trials. Selection of the cued color may be stronger in the invalid condition, such that
more time is needed to find targets in an uncued color. Such an effect would slow RT in the invalid
condition, thereby leading to an overestimate of cueing effects in the different target condition.
To summarize, both trial-by-trial analyses indicate that selection history has a strong influence on
performance in this task, although the precise consequences for cueing effects depend upon whether
history effects are defined based on the current target or the currently cued color. These differences
notwithstanding, both analytic approaches support two clear conclusions. First, cueing effects are larger
in the heterogeneous condition than in the pop-out condition, dovetailing with past studies showing that
visual selection effects are amplified in the displays that contain strong distractor interference (Awh et al.,
2003, 2005; Dosher & Lu, 2000; Kastner et al., 1998; Shiu & Pashler, 1994). Second, this study provides
clear evidence of goal-driven selection of color that cannot be explained by selection history.

Feature-based attention 13
Experiment 2
Depending on the analytic approach, Experiment 1 showed that cueing effects were either
absent or much smaller in the pop-out condition than in the heterogeneous condition. The goal of
Experiment 2 was to examine whether this was due to an experimental context in which target
localization could be efficiently guided without color selection. Although previous studies showed that
participants actively process color cues like the ones in the present study (Müller & Krummenacher, 2006;
Theeuwes et al., 2006), this does not establish that they actively use it to guide their attention. To
examine this issue we intermixed the pop-out and heterogeneous trials within each block. Because
observers did not know which type of target array would be presented, we reasoned that they would be
more likely to engage in goal-driven selection even during pop-out trials.
Methods
Participants
Twenty- eight naïve participants (8 females, mean age of 23, age range 19-30 years) from VU
University Amsterdam with normal or corrected to normal vision participated in the experiment. One
participant’s data was discarded from analysis because of extremely high error rate (23%).
Stimuli, Design & Procedure
The experiment was exactly the same as Experiment 1, except that the pop-out and
heterogeneous conditions were now randomly mixed within blocks. That is participants were equally
likely to get a pop-out or a heterogeneous trial. Participants completed 20 blocks (40 trials each) of
experimental trials, preceded by 2 blocks of practice (also 40 trials each).
Results and Discussion
Error rates
Overall, participants made very few errors (4%). The analysis on error rates showed no significant effects
or interaction, only the cue validity approached significance (F(1,26) = 3.94 p=.06), with slightly fewer
errors being made when the cue was valid (3.8% vs. 4.7% for valid and invalid conditions, respectively).
Reaction times
Trials in which participants responded faster than 150 ms or slower than 1500 ms were excluded
from further analysis. This led to a loss of 1.6% of the trials.
A within-subject analysis of variance (ANOVA) with search condition (pop-out vs. heterogeneous)
and cue validity (valid vs. invalid) revealed a main effect of search condition (F(1,26) = 341.81, p<.001),

Feature-based attention 14
indicating that as expected it took much longer to find the target in the heterogeneous condition (763
ms) than in the pop-out condition (604 ms). The cue validity was also significant (F(1,26) = 97.0, p<.001),
indicating that participants were faster in finding the target when the word cue indicated the upcoming
target correctly. The search condition by cue validity interaction was also significant (F(1,26) = 75.01,
p<.001), indicating that the cueing effect was larger in the heterogeneous condition (142 ms) than in the
pop-out condition (31 ms). Importantly, planned comparisons revealed that a significant cueing effect was
present in both the heterogeneous condition (t(1, 26) = 9.79, p<.001) and in the pop-out condition (t(1,
26) = 6.08, p<.01). Just as in Experiment 1, however, these cueing effects cannot be clearly interpreted
until the effects of selection history are examined. Thus, we again performed analysis conditional on the
trial-by-trial match between the prior target and either the current cue or current target.
Inter-trial effects
Previous target-current cue analysis. The results are presented in Figure 3. A within-subject
analysis of variance (ANOVA) with search condition (pop-out vs. heterogeneous), cue matching the
previous target (match vs. mismatch) and cue validity (valid vs. invalid) revealed that the cueing effect was
much larger on the trials on which the word cue matched the previous target feature (cue match by cue
validity interaction (F(1,26) = 96.38, p<.001), and this effect was larger for the heterogeneous trials than
for the pop-out trials (three-way interaction, F(1,26) = 53.52, p<.001). Unlike Experiment 1 there was a
main effect of cue match (F(1,26) = 5.37, p<.05), indicating that RTs were faster when the word cue
matched (648 ms) the target feature on the previous trial relative to when it mismatched it (667 ms). The
cue match by search condition interaction was not significant (F <1). Planned comparisons showed that
the cueing effect was significant in all conditions. For the heterogeneous condition the cueing effect was
217 ms when the cue matched the previous target feature (t(26) = 14.07, p<.001) and 63 ms when it
mismatched (t(26) = 3.63, p<.005). For the pop-out condition the cueing effect was 47 ms when the cue
matched the previous target feature (t(26) = 7.36, p<.001) and 13 ms when it mismatched (t(26) = 2.25,
p<.05). Thus, although there were strong inter-trial priming effects, Experiment 2 also provided reliable
evidence of goal-driven selection in both the heterogeneous and pop-out conditions. That is, goal-driven
selection was clearly present when we controlled for target-cue priming. These findings suggest that a
higher probability of distractor interference elicited more robust goal-driven selection effects with the
pop-out displays, a result that falls in line with past work examining the impact of distractor probability on
visual selection (Awh et al., 2003, 2005).

Feature-based attention 15
Figure 3. Mean reaction times (RTs) in the pop-out and heterogeneous conditions as a function of cue
validity and whether the word cue matched the color of the target on the previous trial in Experiment
2. The error bars represent standard error of the mean for within-subject designs normalized for the
cue validity factor (Loftus & Masson, 1994). The inset illustrates the conditions plotted on the graph,
the search target was either orange (thick line) or blue (dashed line).
Previous target-current target analysis. A within-subject analysis of variance (ANOVA) with
search condition (pop-out vs. heterogeneous), target repetition (same target vs. different target) and cue
validity (valid vs. invalid) revealed a significant main effect of target repetition (F(1,26) = 96.38, p<.001), as
well as significant search condition by target repetition interaction (F(1,26) = 53.52, p<.001) and search
condition by cue validity interaction (F(1,26) = 75.87, p<.001). Importantly, there was a significant
interaction between target repetition and cue validity (F(1,26) = 5.37, p<.05). The three-way interaction
between search condition, target repetition and cue validity was not significant (F<1). This analysis
suggests that the influences of target repetition priming and goal-driven control on visual selection are
not independent. Specifically, for both search conditions the cueing effects were larger when the target
did not repeat (25 ms vs 35 ms and 130 ms vs 150 ms, respectively for the pop-out and heterogeneous

Feature-based attention 16
search conditions). This seems to be primarily driven by the trials on which the target feature did not
repeat and the word cue was invalid. Note that in this case, the word cue matched the target on the
previous trial (see the inset of Figure 3). Thus, when participants were cued to attend the same color as
the target that had just been processed in the prior trial, strong selection of the cued color may have
extended the time required to find a target in the uncued color. Qualitatively similar effects consistent
with this account were observed in both the pop-out and heterogeneous search conditions.
To summarize, both analytic approaches provide clear evidence for goal-driven selection of
colors that was stronger in the heterogeneous than in the pop-out condition. Moreover, in the pop-out
condition, these selection effects were enhanced relative to those observed in Experiment 1, suggesting
that a higher probability of distractor interference potentiated this form of non-spatial selection.
Experiment 3
Experiment 2 demonstrated that goal-driven selection effects can be reliably observed even with
pop-out targets when there is a strong need to resolve interference between targets and distractors. The
stage of processing affected by goal-driven selection, however, is somewhat ambiguous. Faster RTs in the
valid condition could reflect either changes in the initial perceptual encoding of the stimulus, or changes
in the efficiency of post-perceptual response or decision processes. To further examine whether goal-
driven selection of colors affected the initial perception of the target, Experiment 3 employed brief
stimulus displays and backward masks. A’ was the primary dependent variable in this unspeeded task, and
discrimination difficulty was staircased within-subjects. As in Experiment 2 the pop-out and
heterogeneous trials were intermixed.
Methods
Participants
Twenty-six participants (7 females, mean age of 23, age range 19-30 years) from VU University
Amsterdam with normal or corrected to normal vision participated in the experiment. Two participants’
data were discarded from analysis because in more than 20% of the trials they did not produce any
response. Due to a network malfunction the experiment ended early for two participants. They had only
slightly fewer number of trials (553 and 580 out of 600).
Stimuli, Design & Procedure
The experiment was very similar to Experiment 1, except that it was modified for measuring A’.
The search display was presented for 100 ms and backward masked (Figure 4). The mask consisted of 20
randomly generated gray line segments randomly positioned within each circle. The mask display was

Feature-based attention 17
presented for 1500 ms. The inter-trial interval was 500 ms. Participants completed 15 blocks of 40 trials
each, preceded by 2 practice blocks (also 40 trials each). As in Experiment 2 participants were equally
likely to receive a pop-out or a heterogeneous trial. The orientation of the line segment within the target
circle was also equally likely to be either horizontal or vertical. All conditions were mixed randomly within
each block. Half of the participants responded by pressing the “z” key when the line segment inside the
target circle was oriented vertically and the “m” key when it was oriented horizontally. The response
mapping was reversed for the other half of the participants. It was stressed that participants should be as
accurate as possible. The difficulty of the task was adjusted online by using a procedure similar to
Theeuwes & van der Burg (2007). Specifically, the length of the target line was adjusted to ensure that the
performance was kept around 75% correct. Each 10 trials we calculated the accuracy and increased the
length of the line by 0.05° if the accuracy dropped below 75% and decreased the length by 0.05° of the
line if the accuracy exceeded 75%.
Figure 4. Time-course of a typical heterogeneous trial in Experiment 3. Participants had to search for
either an orange (thick line) or a blue (dashed line) circle and determine the orientation of the line
inside it. The search displays were presented for 100 ms and quickly masked. The pop-out trials looked
exactly the same except that the distractor circles were all green. The staircase was used to keep
performance around 75 %.
Results and Discussion
Trials on which participants did not make any response were not included in the analysis. This
resulted in rejection of 0.9% of all trials. The data was analyzed using A’ as a dependent measure. A’ is a
nonparametric analogue of the d’ statistics (Stanislaw & Todorov, 1999). A’ ranges from .5, which
indicates that signal cannot be distinguished from noise, to 1, which corresponds to perfect performance.

Feature-based attention 18
We used A’ and not d’ because in the inter-trial analysis some participants did not make any false alarms
(FA) in some conditions1.
A within-subject analysis of variance (ANOVA) with search condition (pop-out vs. heterogeneous)
and cue validity (valid vs. invalid) revealed a main effect of search condition (F(1,23) = 92.56, p<.001),
indicating that as expected participants performed worse in the heterogeneous condition (0.71) than in
the pop-out condition (0.80). The cue validity was also significant (F(1,23) = 45.62, p<.001), indicating that
participants performed better in reporting the target when the word cue indicated the upcoming target
correctly. The search condition by cue validity interaction was also significant (F(1,23) = 9.67, p<.01),
indicating that the cueing effect was significantly larger in the heterogeneous condition (0.17) than in the
pop-out condition (0.09). Importantly, planned comparisons revealed that cueing effect was significant in
both the heterogeneous (t(1, 23) = 7.45, p<.001) and the pop-out condition (t(1, 23) = 3.63, p<.005). To
test whether these cueing effects included a goal-driven component, we examined the inter-trial effects.
Inter-trial effects
Previous target-current cue analysis. The results of this analysis are presented in Figure 5. They
showed that the cueing effect was much larger on the trials on which the word cue matched the previous
target feature (cue match by cue validity interaction (F(1,23) = 7.27, p<.05), and this effect was larger for
the heterogeneous trials than for the pop-out trials (three-way interaction, F(1,23) = 7.09, p<.05). The cue
match by search condition was also significant F(1,23) = 6.62, p<.05), suggesting that cue matching the
previous target feature resulted in better performance on the pop-out trials, but not on the
heterogeneous trials. Planned comparisons showed that the cueing effect was significant in all conditions.
For the heterogeneous condition the cueing effect was 0.26 when the cue matched the previous target
feature (t(23) = 8.77, p<.001) and 0.10 when it mismatched (t(23) = 2.46, p<.05). For the pop-out
condition the cueing effect was 0.10 when the cue matched the previous target feature (t(23) = 3.70,
p<.005) and 0.07 when it mismatched (t(23) = 2.44, p<.05).
Previous target-current target analysis. A within-subject analysis of variance (ANOVA) with
search condition (pop-out vs. heterogeneous), target repetition (same target vs. different target) and cue
validity (valid vs. invalid) revealed a significant main effect of target repetition (F(1,23) = 7.27, p<.05), as
well as significant search condition by target repetition interaction (F(1,23) = 7.09, p<.05) and search
condition by cue validity interaction (F(1,23) = 12.82, p<.005). There was a marginally significant
interaction between target repetition and cue validity (F(1,23) = 3.17, p=.09) and a significant three-way
interaction between search condition, target repetition and cue validity (F(1,23) = 6.62, p<.05). This
analysis suggests that the influences of target repetition priming and goal-driven control on visual
selection are not independent. Specifically, for the heterogeneous search condition the cueing effects
were larger when the target did not repeat (0.11 vs. 0.24, for same and different, respectively). As in

Feature-based attention 19
Experiment 2, this seems to be primarily driven by performance in the invalid trials when the cued color
matched the color of the prior target. Our hypothesis is that selection of the cued color was potentiated
by the match with the recently processed target color, yielding a greater cost when the current target did
not match the selected color. Unlike Experiment 2, this was not the case for the pop-out search condition
(0.1 vs. 0.07, for same and different target conditions, respectively).
To summarize, Experiment 3 replicated and extended the conclusions of Experiment 2. We
observed goal-driven selection of colors in both the pop-out and heterogeneous conditions. Because this
was an encoding-limited task and we employed an A’ measure of performance, these data suggest that
goal-driven selection enhanced the initial perception of targets in the cued color.
Figure 5. Mean A’ in the pop-out and heterogeneous conditions as a function of cue validity and
whether the word cue matched the color of the target on the previous trial in Experiment 3. The error
bars represent standard error of the mean for within-subject designs normalized for the cue validity
factor (Loftus & Masson, 1994). The inset illustrates the conditions plotted on the graph, the search
target was either orange (thick line) or blue (dashed line).

Feature-based attention 20
General Discussion
The present results demonstrate that the context in which visual search is performed has a clear
influence on whether evidence of goal-driven control over feature-based visual selection will be observed.
When pop-out and heterogeneous conditions were blocked in Experiment 1, volitional control over
selection of target features was clear in the heterogeneous condition, but either absent or much smaller
in the pop-out condition. The findings in the pop-out condition are consistent with the previous findings in
the literature that argue against the possibility of goal-driven attention to features (e.g., Theeuwes, 2013).
However, when in Experiments 2 and 3 the two search conditions were intermixed within the same
blocks, robust evidence for volitional control was observed in the pop-out condition. Thus, a task context
where there is a high probability of distractor interference appears to potentiate the goal-driven selection
of non-spatial features. Finally, Experiment 3 replicated and extended this empirical pattern with a task
that employed data-limited displays and an analysis of perceptual sensitivity, showing that attention
affected the initial perceptual encoding of the stimulus rather than the efficiency of post-perceptual
response or decision processes.
As noted above, many earlier demonstrations of voluntary feature-based selection relied on
block designs and thus confounded voluntary orienting with automatic effects of recent selection history
(e.g., Folk et al., 1992; Wolfe et al., 2003). The present results indeed show that inter-trial priming of
previous target features plays an important role in feature-based selection (Kristjánsson, Wang, &
Nakayama, 2002), though the nature of this effect depended on how selection history was defined. When
we focused on the correspondence between the color of the current cue and the color of the most recent
target, we found that cueing effects were approximately four times larger when those colors matched
than when they did not. This analysis, however, was affected by differential effects of target repetitions in
the valid and invalid conditions of the study. When we focused on the correspondence between the color
of the previous target and the color of the current target (regardless of what color was cued), we
observed larger cueing effects when those colors did not match in two of the three studies (Experiments 2
and 3). This finding echoes the results of a study by Weidner & Müller (2013) who used displays similar to
those in our heterogeneous condition. The previous target-current target analysis also had a limitation,
however, because cue validity was confounded with whether the voluntarily selected color (i.e. cued
color) matched the most recent target color.
Nevertheless, although both analytic approaches have their limitations, three primary
conclusions were clearly supported by both approaches. First, cueing effects were much larger in the
heterogeneous condition where there was strong distractor interference. This is consistent with many
studies that have shown enhanced visual selection with increased competition from distractors (Awh et
al., 2003, 2005; Dosher & Lu, 2000; Kastner et al., 1998; Shiu & Pashler, 1994). Second, goal-driven

Feature-based attention 21
selection effects were clear even when we accounted for selection history effects, in line with models that
argue for volitional control over feature-based attention (e.g., Müller et al., 2010). And finally, the
strength of goal-driven orienting effects was augmented in an experimental context where there was a
high probability of distractor interference. This context effect dovetails with past studies that have shown
enhanced spatial selection when the probability of interference is high (Awh et al., 2003, 2005), and it is
consistent with the claim that excluding interference is one of the primary functions of visual selective
attention (Desimone and Duncan, 1995; Dosher and Lu, 2000; Kastner et al., 1998; Shiu and Pashler,
1994). Moreover, our findings offer an explanation for why there have been mixed findings in past studies
of non-spatial selection that have employed displays that contain relatively low levels of distractor
interference (e.g., Leonard & Egeth, 2008; Mortier et al., 2005; Müller & Krummenacher, 2006; Müller et
al., 2003; Theeuwes et al., 2006; Theeuwes & Van der Burg, 2007; Zehetleitner et al., 2011). Thus, these
findings may help to establish the boundary conditions for procedures that will produce consistent
evidence for goal-driven selection of non-spatial features.
To summarize, we provide clear evidence that volitional control over feature-based selection can
enhance processing during the initial stages of stimulus encoding. We show that the context determines
the degree to which the observers use cue information to bias visual selection. This insight regarding the
boundary conditions for observing these goal-driven effects may help to reconcile the conflicting findings
that have been reported in past studies of feature-based selection.

Feature-based attention 22
Notes
1. We also re-analyzed the data using accuracy as the dependent measure, since in the forced-
choice tasks it is typically considered to be a measure of sensitivity unaffected by response bias
(Stanislaw & Todorov, 1999). The results were statistically identical.

Feature-based attention 23
Table 1
Mean Percentage Hits, mean percentage False alarms (FA) and mean A’ as a function of search
condition, match between the word cue and the previous target feature and cue validity in Experiment 3
Pop-out trials
Heterogeneous trials
Cue match to previous
target
Cue mismatch to
previous target
Cue match to previous
target
Cue mismatch to
previous target
valid
invalid
valid
invalid
valid
invalid
valid
invalid
Hits (%)
75.4
(16.8)
64.5
(19.0)
72.3
(14.8)
66.7
(17.0)
68.3
(13.9)
47.2
(21.3)
66.9
(13.0)
61.4
(17.2)
FA (%)
18.0
(9.3)
28.0
(15.9)
20.5
(8.7)
27.1
(16.6)
23.9
(9.9)
42.7
(18.1)
26.6
(8.0)
34.8
(20.1)
A’
0.85
(0.13)
0.75
(0.13)
0.83
(0.10)
0.76
(0.16)
0.80
(0.09)
0.54
(0.16)
0.78
(0.08)
0.69
(0.18)
Standard deviations are shown in parentheses

Feature-based attention 24
References
Awh, E., Belopolsky, A. V., & Theeuwes, J. (2012). Top-down versus bottom-up attentional control: a failed
theoretical dichotomy. Trends in Cognitive Sciences, 16(8), 437–443.
Awh, E., Matsukura, M., & Serences, J. T. (2003). Top-down control over biased competition during covert
spatial orienting. Journal of Experimental Psychology: Human Perception and Performance, 29(1),
52.
Awh, E., Sgarlata, A. M., & Kliestik, J. (2005). Resolving visual interference during covert spatial orienting:
online attentional control through static records of prior visual experience. Journal of
Experimental Psychology: General, 134(2), 192.
Belopolsky, A. V., Schreij, D., & Theeuwes, J. (2010). What is top-down about contingent capture?
Attention, Perception, & Psychophysics, 72(2), 326.
Chawla, D., Rees, G., & Friston, K. J. (1999). The physiological basis of attentional modulation in
extrastriate visual areas. Nature Neuroscience, 2, 671–676.
Corbetta, M., & Shulman, G. L. (2002). Control of goal-directed and stimulus-driven attention in the brain.
Nature Neuroscience, 3, 201– 215.
Desimone, R., & Duncan, J. (1995). Neural mechanisms of selective visual attention. Annu. Rev. Neurosci.,
18, 193–222.
Dosher, B. A., & Lu, Z.-L. (2000). Noise exclusion in spatial attention. Psychol. Sci., 11, 139–146.
Egeth, H. E., & Yantis, S. (1997). Visual Attention: Control, Representation, and Time Course. Annu. Rev.
Psychol., 48, 269–297.
Eriksen, C. W., & Hoffman, J. E. (1973). The extent of processing of noise elements during selective
encoding from visual displays. Perception & Psychophysics, 14, 155–160.
Folk, C. L., & Remington, R. W. (2008). Bottom-up priming of top-down attentional control settings. Visual
Cognition, 16(2-3), 215–231.
Folk, C. L., Remington, R. W., & Jonhnston, J. C. (1992). Involuntary covert orienting is contingent on
attentional control settings. J. Exp. Psyc.: Hum. Perc. & Perf., 18(4), 1030–1044.
Found, A., & Müller, H. J. (1996). Searching for unknown feature targets on more than one dimension:
Investigating a “dimension-weighting” account. Perc. & Psyc., 58, 88–101.
Jonides, J. (1981). Voluntary versus Automatic Control over the Mind’s Eye’s Movement. In J.B. Long &
A.D. Baddeley (Eds). Attention and Performance (Vol. IX, pp. 187–203). Hillsdale, N.J.: Lawrence
Erlbaum Associates.
Kastner, S., De Weerd, P., Desimone, R., & Ungerleider, L. G. (1998). Mechanisms of directed attention in
the human extrastriate cortex as revealed by functional MRI. Science, 282, 108–111.
Kristjánsson, Á., Wang, D., & Nakayama, K. (2002). The role of priming in conjunctive visual search.
Cognition, 85, 37–52.
Leonard, C. J., & Egeth, H. E. (2008). Attentional guidance in singleton search: An examination of top-
down, bottom-up, and intertrial factors. Visual Cognition, 16(8), 1078–1091.
Loftus, G. R., & Masson, M. E. J. (1994). Using confidence intervals in within-subject designs. Psycon. Bull.
& Rev., 1, 476–490.
Maljkovic, V., & Nakayama, K. (1994). Priming of pop-out: I. Role of features. Memory & Cognition, 22(6),
657–672.
Meeter, M., & Olivers, C. N. L. (2006). Intertrial priming stemming from ambiguity: A new account of
priming in visual search. Visual Cognition, 13, 202–221.
Moran, J., & Desimone, R. (1985). Selective Attention Gates Visual Processing in the Extrastriate Cortex.
Sci., 229, 782–784.
Mortier, K., Theeuwes, J., & Starreveld, P. (2005). Response selection modulates visual search within and
across dimensions. J Exp Psychol Hum Percept Perform, 31(3), 542–57.
Müller, H. J., & Krummenacher, J. (2006). Locus of dimension weighting: Preattentive or postselective?
Visual Cognition, 14(4-8), 490–513.
Müller, H. J., Reimann, B., & Krummenacher, J. (2003). Visual search for singleton feature targets across
dimensions: Stimulus and expectancy-driven effects in dimensional weighing. Journal of
Experimental Psychology: Human Perception and Performance, 29, 1021–1035.

Feature-based attention 25
Müller, H. J., Töllner, T., Zehetleitner, M., Geyer, T., Rangelov, D., & Krummenacher, J. (2010). Dimension-
based attention modulates feed-forward visual processing. Acta Psychologica, 135(2), 117–122.
Nothdurft, H.-C., Gallant, J. L., & van Essen, D. C. (1999). Response modulation by texture surround in
primate area V1: correlates of “popout” under anesthesia. Visual Neuroscience, 16(01), 15–34.
Posner, M. I., Davidson, B. J., & Snyder, C. R. R. (1980). Attention and the detection of signals. J. Exp. Psyc.:
Hum. Perc. & Perf., 109, 160–174.
Saenz, M., Buracas, G. T., & Boynton, G. M. (2002). Global effects of feature-based attention in human
visual cortex. Nature Neuroscience, 5(7), 631–632.
Schneider, W., Eschman, A., & Zuccolotto, A. (2002). E-Prime: User’s guide. Psychology Software
Incorporated.
Serences, J. T., & Boynton, G. M. (2007). Feature-based attentional modulations in the absence of direct
visual stimulation. Neuron, 55(2), 301–312.
Shiu, L., & Pashler, H. (1994). Negligible effect of spatial precuing on identification of single digits. Journal
of Experimental Psychology: Human Perception and Performance, 20(5), 1037.
Stanislaw, H., & Todorov, N. (1999). Calculation of signal detection theory measures. Behavior Research
Methods, Instruments, & Computers, 31(1), 137–149.
Theeuwes, J. (2013). Feature-based attention: it is all bottom-up priming. Philosophical Transactions of
the Royal Society B: Biological Sciences, 368(1628), 20130055.
Theeuwes, J., Olivers, C. N. ., & Belopolsky, A. (2010). Stimulus-driven capture and contingent capture.
Wiley Interdisciplinary Reviews: Cognitive Science, 1(6), 872–881.
Theeuwes, J., Reimann, B., & Mortier, K. (2006). Visual search for featural singletons: No top-down
modulation, only bottom-up priming. Visual Cognition, 14(4-8), 466–489.
Theeuwes, J., & Van der Burg, E. (2007). The role of spatial and nonspatial information in visual selection.
Journal of Experimental Psychology-Human Perception and Performance, 33(6), 1335–1351.
Treisman, A. M. (1988). Features and Objects: The Fourteenth Bartlett Memorial Lecture. QUART J EXP
PSYCH, 40A, 201–237.
Weidner, R., & Müller, H. J. (2013). Dimensional weighting in cross-dimensional singleton conjunction
search. Journal of Vision, 13(3), 25.
Wolfe, J. M. (2007). Guided search 4.0: Current Progress with a model of visual search. In W. Gray (Ed.)
Integrated models of cognitive systems (pp. 99–119). New York: Oxford University Press.
Wolfe, J. M., Butcher, S. J., Lee, C., & Hyle, M. (2003). Changing Your Mind: On the Contributions of Top-
Down and Bottom-Up Guidance in Visual Search for Feature Singeltons. Journal of Experimental
Psychology, 29(2), 483–502.
Wolfe, J. M., Cave, K. R., & Franzel, S. L. (1989). Guided Search: An Alternative to the Feature Integration
Model for Visual Search. J. Exp. Psyc.: Hum. Perc. & Perf., 15, 419–433.
Yantis, S., & Jonides, J. (1990). Abrupt Visual Onsets and Selective Attention: Voluntary Versus Automatic
Allocation. J. Exp. Psyc.: Hum. Perc. & Perf., 16, 121–134.
Zehetleitner, M., Krummenacher, J., Geyer, T., Hegenloh, M., & Müller, H. J. (2011). Dimension intertrial
and cueing effects in localization: support for pre-attentively weighted one-route models of
saliency. Attention, Perception, & Psychophysics, 73(2), 349–363.
Zhang, W., & Luck, S. J. (2008). Feature-based attention modulates feedforward visual processing. Nature
Neuroscience, 12(1), 24–25.