ArticlePDF Available

Overt Attention in Contextual Cuing of Visual Search is Driven by the Attentional Set, but Not by the Predictiveness of Distractors

Article

Overt Attention in Contextual Cuing of Visual Search is Driven by the Attentional Set, but Not by the Predictiveness of Distractors

Abstract and Figures

Two experiments examined biases in selective attention during contextual cuing of visual search. When participants were instructed to search for a target of a particular colour, overt attention (as measured by the location of fixations) was biased strongly towards distractors presented in that same colour. However, when participants searched for targets that could be presented in one of two possible colours, overt attention was not biased between the different distractors, regardless of whether these distractors predicted the location of the target (repeating) or did not (randomly arranged). These data suggest that selective attention in visual search is guided only by the demands of the target detection task (the attentional set) and not by the predictive validity of the distractor elements.
No caption available
… 
No caption available
… 
No caption available
… 
Content may be subject to copyright.
Selective attention in contextual cuing Beesley, Hanafi, Vadillo, Shanks, & Livesey
1
Running Head: Selective attention in contextual cuing
Overt attention in contextual cuing of visual search is driven by the attentional set, but not by
the predictiveness of distractors
Tom Beesley1, Gunadi Hanafi1, Miguel A. Vadillo2,3 & David. R. Shanks4, & Evan J. Livesey5
1UNSW Australia, Sydney, Australia
2King’s College London, London, UK
3Universidad Autónoma de Madrid, Madrid, Spain
4University College London, London, UK
5University of Sydney, Sydney, Australia
Mailing address:
Dr Tom Beesley
School of Psychology
Matthews Building
UNSW Australia
Sydney, NSW
Australia, 2052
Tel: +61 (0)2 9385 3032
e-mail: t.beesley@unsw.edu.au
Selective attention in contextual cuing Beesley, Hanafi, Vadillo, Shanks, & Livesey
2
Abstract
Two experiments examined biases in selective attention during contextual cuing of visual search.
When participants were instructed to search for a target of a particular color, overt attention (as
measured by the location of fixations) was biased strongly towards distractors presented in that same
color. However, when participants searched for targets that could be presented in one of two possible
colors, overt attention was not biased between the different distractors, regardless of whether these
distractors predicted the location of the target (repeating) or did not (randomly arranged). These data
suggest that selective attention in visual search is guided only by the demands of the target detection
task (the attentional set) and not by the predictive validity of the distractor elements.
Selective attention in contextual cuing Beesley, Hanafi, Vadillo, Shanks, & Livesey
3
Humans and other animals are able to cope with the complexity of the surrounding environment by
filtering the incoming information, such that cognitive processes are directed only to stimuli of
primary importance. This is the role of selective attention, to determine the selective processing of
information both within and across different modalities (e.g., Broadbent, 1958; Evans & Craig, 1991;
Rock & Gutman, 1981). What exactly receives the focus of selective attention is determined by a
number of factors, commonly and broadly partitioned into bottom-up stimulus features and top-down
goal-directed processes (for reviews, see Awh, Belopolsky, & Theeuwes, 2012; Theeuwes, 2010).
This article is concerned with the latter and primarily how the goals of the agent, brought about
either by instruction or by experience with a task, come to determine the focus of attention.
The interdependency of the processes of learning and attention is demonstrated in many
common cognitive tasks. For example, when solving categorisation problems with exemplars
comprising multiple features, those features that are learnt to be most diagnostic are allocated
preferential attentional processing in the future (e.g., Rehder & Hoffman, 2005). Young infants will
tend to direct gaze towards the face of a parent over a stranger, presumably as a result of learning
about the rewarding properties of that stimulus. Even in simple and rapid visual detection tasks,
attention is automatically captured by a rewarding stimulus, even when this attentional capture is
counterproductive to the task demands (Le Pelley et al., 2015).
We focus here on the process of visual search, in which the cognitive task is to locate and
respond to one stimulus positioned within an array of many. The mechanisms of attentional selection
are well studied in this task. Triesman’s feature integration theory (Triesman & Gelade, 1980) has
provided the basis for modern theorising on the processes of visual search. Briefly, the model states
that features of the visual input are extracted in parallel by pre-attentive processing mechanisms,
with attention acting to guide the focus of further processing to enable appropriate feature binding.
Thus, in any complex visual search tasks that require the resolution of a conjunction of features (e.g.,
search for a red vertical line among blue vertical and red horizontal lines), the attentive process
Selective attention in contextual cuing Beesley, Hanafi, Vadillo, Shanks, & Livesey
4
moves the spotlight of attention serially from stimulus to stimulus until the target object is detected.
Such searches are relatively inefficient and determined by the number of distracting stimuli in the
configuration (the “set size”). In contrast, when searching for a single feature (e.g., a red target
among blue distractors) rather than a conjunction, the pre-attentive processing mechanisms receive a
unique hit and attentional resources can be allocated directly towards the unique object (search times
are not a function of the set size). This simple dichotomy of search into discrete pre-attentive and
attentive processes has been challenged in recent years by findings showing that the set size effect
varies for both conjunction and feature searches, suggesting that there are parallel processes
influencing conjunction search, and conversely that serial processes can play a role in feature search.
These data led Wolfe and colleagues to propose the Guided Search model (e.g., Wolfe, 1994; Wolfe,
Võ, Evans, & Greene, 2011), which suggests that the role of pre-attentive processing is to guide the
attentive process by restricting the range of to-be-searched objects to those that contain features
consistent with the target.
Visual search in the real world will engage not just attentional mechanisms, but also the
encoding and recall of memory for past search experiences. Indeed, many experiments to date have
shown that a stored representation for the configuration of the distractors can lead to a substantial
decrease in the time taken to locate and respond to the target. This “contextual cuing” effect (Chun &
Jiang, 1998) is thought to derive from a perceptual saving that results from the processing of fewer
distractors prior to the localisation of the target (although see Kunar, Flusberg, Horowitz, & Wolfe,
2007). This is perhaps best shown in data from eye-tracking studies of contextual cuing, which have
found that fewer fixations are made when searching repeating configurations compared to random
configurations (Harris, & Remington, 2017; Peterson & Kramer, 2001; Tseng & Li, 2004; Zhao et
al., 2012).
It has been demonstrated that selective attention plays an important role in contextual cuing.
Jiang and Chun (2001) presented participants with a visual search task in which configurations
Selective attention in contextual cuing Beesley, Hanafi, Vadillo, Shanks, & Livesey
5
comprised green and red stimuli. Importantly, the target (a T shape) was the same color on every
trial, say red (though this was counterbalanced across participants), and this ensured that the
distractors (L shapes) presented in that color would receive preferential attentional processing, even
though participants were not explicitly instructed about this regularity. An effective search strategy
in this task would therefore be to ignore all green items. Jiang and Chun presented repeated
configurations for which just the red items comprised the repeating configuration (while the green
stimuli were randomly arranged), while for other configurations the green (but not the red) stimuli
were repeated. Thus, the former repeating configurations contained useful information for detecting
the target presented in the attended color, while the latter repeating configurations only contained
useful information presented in the unattended color. This manipulation had a significant effect on
contextual cuing: learning was only observed for those configurations with repeating distractors
presented in the attended color, while no learning was observed for configurations with repeating
distractors presented in the unattended color. These experiments demonstrate the impact of top-down
control on the processing of distractors in the contextual cuing task. By fixing the color of the target
and therefore the “attentional set” that participants engage in the task, the possible search space is
narrowed to only those objects that constitute plausible targets (c.f. Guided Search; Wolfe, 1994).
This results in preferential processing of those stimuli, permitting associative learning to occur only
between these processed elements and the target position (for a recent discussion of associative
models of contextual cuing, see Beesley, Vadillo, Pearson, & Shanks, 2015, 2016).
The current article aims to address two questions that arise from Jiang and Chun’s (2001)
results. Firstly, to what extent is this modulation of the contextual cuing effect driven by a
preferential allocation of attention to the distractor stimuli of a particular color? It seems likely that
overt shifts of attention to relevant stimuli will occur, given it is well known that a standard
conjunction visual search of the type used in these studies results in a preferential allocation of
attention to those features of the configuration that are shared with the target (e.g., Motter & Belky,
Selective attention in contextual cuing Beesley, Hanafi, Vadillo, Shanks, & Livesey
6
1998). Our study attempts to confirm this hypothesis in the contextual cuing task. The second and
more important aim of this work is to examine to what extent these biases in top-down control of
attentional selection are driven only by the attentional set determined by the visual search task, or
whether the allocation of attention can also be driven by learning which elements of the
configurations are most useful for finding the target, as determined by their predictiveness.
A number of experiments in the human associative learning literature have demonstrated that
when a stimulus becomes a reliable predictor of events in the environment, that stimulus will receive
a biasing of attention towards it in the future (e.g., Beesley & Le Pelley, 2011; Beesley, Nguyen,
Pearson, & Le Pelley, 2015; Le Pelley, Beesley, & Griffiths, 2011; Le Pelley, Beesley, & Griffiths,
2014; Livesey, Harris & Harris, 2009; Mitchell, Griffiths, Seetoo, & Lovibond, 2012). These
attentional effects are thought to be elicited reflexively by the appearance of the stimuli (e.g., Le
Pelley, Pearson, Griffiths, & Beesley, 2015; Le Pelley, Vadillo & Luque, 2013; Luque, Vadillo, Le
Pelley, & Beesley, 2017) and such learned biases in processing have also been observed in implicit
learning tasks (Beesley & Le Pelley, 2010). These studies demonstrate a reciprocal relationship
between associative learning and attentional processing: as we learn about the usefulness of certain
stimuli in our environment (e.g., for predicting rewards), these stimuli come to be allocated greater
attentional processing; in turn this enhanced attentional processing will bias any future learning
episodes involving these stimuli (i.e., learning more about these stimuli compared to stimuli which
are not the focus on attention).
By mapping this to the contextual cuing task, we see that Jiang and Chun’s experiments test
the latter aspect of the relationship (attention modulates associative learning), but it is unclear
whether the former aspect (associative learning modulates attention) plays a role in contextual cuing.
The current experiments examined this question directly.
Selective attention in contextual cuing Beesley, Hanafi, Vadillo, Shanks, & Livesey
7
Experiment 1
In Experiment 1 we sought to establish that the effects ascribed to selective attentional processes in
Jiang and Chun’s (2001) study truly reflected a biasing of attention towards those stimuli that
contained relevant features, as determined by the demands of the visual search task. As we have
noted, given that effects of preferential eye movements of this kind have already been demonstrated
in visual search tasks, a measurable bias in eye movements was expected. Nevertheless, by
examining this in a similar procedure to that used by Jiang and Chun (2001), we were able to
establish that our procedure, measurements and analysis could reliably detect differences in
participants eye-movements, which would provide a suitable baseline for assessing biases in
attentional processing between the stimuli of our task.
The second aim was to examine whether such biases in attention could be driven by learning
to attend to relevant elements of the repeated configuration. Here we used a condition in which the
attentional set determined by the visual search task did not dictate a top-down biasing of attention to
one color over another, however the configuration of stimuli consisted of distractors that were
predictive of the target position (repeating) and distractors that were non-predictive of the target
position (random). Several tasks in our lab have established that participants can learn about such
semi-repeating configurations of context (e.g., Beesley & Shanks, 2012), and so the question of
interest was specifically whether such learning effects result in a biasing of attention to those
distractors that are predictive of the target.
Method
Participants
The experiment was approved by the local UNSW Sydney ethics committee. Sixty-eight
undergraduate psychology students from UNSW Australia participated in exchange for course credit.
This sample size yields 81% power to detect a moderate-to-large effect size of Cohen’s d = 0.7 in a
Selective attention in contextual cuing Beesley, Hanafi, Vadillo, Shanks, & Livesey
8
between groups contrast, and 98% power to detect effects of the same size in a within group contrast.
All participants had normal color vision and normal or corrected-to-normal visual acuity.
Participants were randomly allocated to one of two between-subject conditions: instructed or
learning. The experiment was approved by the ethics committee of the School of Psychology,
UNSW Australia, and all participants gave informed consent.
Materials and Apparatus
Participants were tested individually in a quiet room with a standard desktop computer and a
58.4 cm widescreen eye tracking monitor (TX-300, Tobii Technology, Danderyd, Sweden) which
samples eye gaze at 300 Hz. Participants sat at an average viewing distance of 59 cm (SD = 2.8cm),
using a chin rest to maintain a fixed position. The eye tracker was calibrated using a five-point
procedure at the start of the experiment. Stimulus presentation was controlled by MATLAB using
the Psychophysics Toolbox extensions (Brainard, 1997; Kleiner, Brainard & Pelli, 2007; Pelli,
1997). Responses to the target stimulus were made by pressing the ‘m’ or ‘z’ key on a standard
keyboard.
Distractor stimuli were an ‘L’ shape (rotated 0°, 90°, 180°, or 270°) while the target stimulus
was a ‘T’ shape (rotated at either 90° or 270°). Stimuli were arranged in a square grid of 144 evenly
spaced cells (12 x 12) which was positioned centrally on the screen and was 240 mm (23°) square.
The grid itself was invisible to participants. The fixation cross (displayed centrally before each trial)
was 11 mm (1.1°) square. The stimuli were 13 mm (1.3°) square. The background of the screen was
grey (RGB: .6, .6, .6) and the stimuli were presented in either red (RGB: .7, .13, .13) or blue (RGB:
.25, .41, .88). There was a small offset in the vertical line of the ‘L’ distractors, which increased the
similarity between the ‘L’ distractor and the target ‘T’, making the search task more difficult
(Duncan & Humphreys, 1989).
Design
Selective attention in contextual cuing Beesley, Hanafi, Vadillo, Shanks, & Livesey
9
Experiment 1 employed a 2 x 2 x 4 mixed-model design, with a between-subjects factor of task
instruction (instructed vs. learning) and within-subject factors of configuration (repeated vs. random)
and epoch (1 to 4). The two between-subject conditions experienced both repeated and random
configurations over four epochs of 120 trials. All search configurations contained 16 distractors and
one target stimulus, with equal numbers of red and blue distractors in each configuration. Four
“repeated” search configurations were trained, each of which contained a subset of distractors for
which the position and orientation was maintained across multiple presentations (termed predictive).
These repeated search configurations contained eight such predictive distractors presented in one
color (for example, red) which were intermixed with a set of eight distractors that were placed and
orientated randomly on each trial in the alternative color (blue). For the instructed condition, the
color of the target was always the same as the color of the predictive distractors within the repeated
configurations (red). For the learning condition, the target was presented in one color on half the
trials and in the alternative color on the remainder of trials. For the purely “random” configurations,
all 16 distractors were randomly arranged on each trial with an equal number of red and blue
distractors. For both the instructed and learning conditions the targets in purely random
configurations were colored in the same manner as for repeating configurations. Table 1 shows the
design of the experiment; a schematic illustrating the differences between the conditions is also
presented in Figure 1.
Two red and two blue distractors were placed in each quadrant of the screen. Eight target
locations were used, with one from each quadrant assigned to the repeated configurations and one
from each quadrant assigned to the random configurations. These eight target positions were chosen
at random from one of five locations within each quadrant that were approximately equidistant from
the centre of the screen. Distractors could not appear in these target locations.
The four repeated configurations were presented 60 times each (15 times in each epoch) across
the course of the experiment. The same number of random trials was used, resulting in 480 trials in
Selective attention in contextual cuing Beesley, Hanafi, Vadillo, Shanks, & Livesey
10
total. The same repeated configuration or the same target position could not occur on consecutive
trials. Target orientation was determined randomly but an equal number of presentations of each
orientation was maintained within each epoch.
Procedure
Participants were seated with a chin rest adjusted according to the participants’ height. The
eye-tracker was then calibrated and participants received instructions about the nature of the search
task: in the instructed condition, participants were told that the target would always be in one color
(i.e., they were instructed to attend to one color), while in the learning condition, participants were
told that the target could be in either color (i.e., they may learn to attend more to one color on the
basis of the predictiveness of the distractors, but were not instructed to do so). An example of a
search trial was presented and participants were shown the two correct responses for the two possible
orientations of targets.
Each trial commenced with a fixation cross presented in the centre of the screen for 1000 ms,
which was then replaced immediately by the search configuration. Participants searched for the
target stimulus and responded with a left or right response depending on its orientation. RTs were
recorded from the onset of the search configuration. Following a valid response (z or m) the
configuration was removed from the screen. The response-stimulus interval (hereafter RSI) was 1000
ms. If participants made an incorrect response to the target orientation, “ERROR!” appeared in the
centre of the screen for 2000 ms, prior to the RSI. A rest-break of 20 seconds was given every 120
trials (splitting the experiment into 4 equal parts). Trials started automatically after these breaks.
Measuring the distribution of attention from fixations
For our analysis of how attention was distributed across the different distractors, we took
each fixation that was made during the task and calculated two metrics of distractor processing. The
first was the number of distractors present within the “attentional spotlight” region surrounding the
Selective attention in contextual cuing Beesley, Hanafi, Vadillo, Shanks, & Livesey
11
centre point of each fixation. We defined the width of this spotlight as 300 pixels (7.7°) in diameter,
but very similar ordinal results were observed for analyses conducted with smaller spotlights of (at
least) 100 pixels (2.6°) in diameter. A distractor was deemed to be “attended” if the centre point of
that distractor fell within the area of the attentional spotlight (i.e., the Euclidian distance from the
fixation to the distractor was less than the radius). The number of distractors attended was summed
across all fixations for a given trial, with the metric reflecting the mean number attended on each
trial. The second metric was simply the average distance of the nearest distractor of each type to the
position of each fixation. If attention is biased towards one type of distractor over another, we would
expect more distractors to fall within the attentional spotlight and for the nearest distractor of that
type to be closer to the points of fixation. We had no a priori reason to anticipate different patterns of
results from these two metrics, but we include both to provide a more comprehensive examination of
the attentional effects.
Results
Three participants in the instructed condition and two in the learning condition produced accuracy
that was below 90% and were removed from the final analysis. Accuracy of responses for the
remaining sample was high in both the instructed (N = 31; 97.8%; standard error of the mean, SE =
0.4) and learning conditions (N = 32; 97.9%; SE = 0.3). Data from trials on which an inaccurate
response was made or the reaction time was 2.5 standard deviations or more from the participant
mean (2.8%; SD = 0.55) did not contribute to the analyses.
For each trial, the percentage of missing samples resulting from tracking errors (e.g., due to
blinks) was calculated, and the data from the eye with the lowest proportion of missing samples were
used for that trial. Missing data that spanned a gap of no more than 75 milliseconds were replaced by
interpolating between the data immediately preceding and following the gap. The average proportion
of missing samples following this interpolation procedure was 2.2% (SD = 2.4). Fixations were
Selective attention in contextual cuing Beesley, Hanafi, Vadillo, Shanks, & Livesey
12
determined by a displacement method (Salvucci & Goldberg, 2000). The range of values of both the
vertical and horizontal coordinates of the gaze data were analysed in 150-ms windows. If neither
coordinate deviated beyond a range of 75 pixels (1.9°), then the analysed window was deemed a
fixation. Fixation length was determined by extending this window until a displacement exceeded
this threshold. Fixation position was determined by the mean horizontal and vertical pixel values
across the fixation sample. Trials without any detected fixations did not contribute to the analysis.
This led to an exclusion of 48% of the data for one participant in the learning condition and the data
from this participant were therefore not included in any analyses; for the remaining participants,
1.3% (SD = 2.33) of trials on average were removed.
Figure 2 shows the average RT (panel A) and average number of fixations per trial (panel B).
Reaction times decrease across the course of the experiment and are shorter for repeated compared to
random configurations, demonstrating the typical contextual cuing effect. While the pattern of data
looks similar in the instructed and learning conditions, RTs are longer in the learning condition and
the contextual cuing effect seems to be weaker. The pattern in the number of fixations per trial is
remarkably similar to the RT data.
The RT data were subjected to a mixed model ANOVA with within-subjects factors of
configuration (repeated vs. random) and epoch (1-4), and a between-subjects factor of condition
(instructed vs. learning). This revealed a main effect of configuration, F(1,60) = 49.38, ηp2 = .45, p <
.001, reflecting a mean contextual cuing effect (RT for random configurations minus RT for repeated
configurations) of 163 ms (SD = 191). There was also a main effect of epoch, F(3,180) = 76.03, ηp2 =
.56, p < .001, reflecting a decline of RT across blocks, as well as a main effect of condition, F(1,60)
= 50.40, ηp2 = .46, p < .001, indicating that responses were faster in the instructed (Mean = 2023 ms;
SD = 504) than in the learning condition (Mean = 2911 ms; SD = 480). There was an interaction
between configuration and epoch, F(3,180) = 3.85, ηp2 = .06, p = .011, suggesting that the contextual
cuing effect increased in magnitude across epochs. The configuration by condition interaction was
Selective attention in contextual cuing Beesley, Hanafi, Vadillo, Shanks, & Livesey
13
significant, F(1,60) = 6.10, ηp2 = .09, p = .016, indicating that the contextual cuing effect was
stronger in the instructed condition (221 ms; SD = 170) than the learning condition (106 ms; SD =
195). The epoch by condition interaction was also significant, F(3,180) = 7.55, ηp2 = .11, p < .001,
suggesting that improvements in RT across epoch were greater for the learning condition compared
to the instructed condition. The three-way interaction was not significant, F(3,180) = 1.13, p = .34.
To examine whether contextual cuing was present for each condition, the data were separately
subjected to a Repeated Measures ANOVA with factors of configuration and epoch. In each
condition there were main effects of configuration and epoch, Fs ≥ 9.13, ps ≤ .005. However, the
interaction effect was significant only in the instructed condition, F(3,90) = 5.19, ηp2 = .15, p = .002,
and not in the learning condition, F(3,90) = 1.54, p = .211.
The fixation data were subjected to an identical overall ANOVA, which revealed a main
effect of configuration, F(1,60) = 50.57, ηp2 = .46, p < .001, demonstrating a contextual cuing effect,
with a saving of 0.58 fixations (SD = 0.66), on average, for repeated compared to random
configurations. There was also a main effect of epoch, F(3,180) = 86.26, ηp2 = .59, p < .001,
reflecting a decline in the number of fixations across epochs, as well as a main effect of condition,
F(1,60) = 73.16, ηp2 = .55, p < .001, indicating that fewer fixations were made in the instructed
condition (5.5; SD = 1.2) than in the learning condition (8.5; SD = 1.6). There was an interaction
between configuration and epoch, F(3,180) = 3.15, ηp2 = .05, p = .026, revealing that the contextual
cuing effect increased in magnitude across epochs. Unlike in the RT data, the configuration by
condition interaction was not significant, F(1,60) = 4.20, ηp2 = .06, p = .057. The epoch by condition
interaction was also significant, F(3,180) = 6.89, ηp2 = .10, p < .001, suggesting that the decrease in
the number of fixations across epochs was greater for the learning condition compared to the
instructed condition. The three-way interaction was not significant, F < 1. To examine whether
contextual cuing was present for each condition for the fixation data, the data were separately
subjected to a Repeated Measures ANOVA with factors of configuration and epoch. In each
Selective attention in contextual cuing Beesley, Hanafi, Vadillo, Shanks, & Livesey
14
condition there were main effects of configuration and epoch, Fs ≥ 10.94, ps ≤ .002. However, the
interaction effect was significant in the instructed condition, F(3,90) = 3.52, ηp2 = .11, p = .018, but
not in the learning condition, F(3,90) = 1.44, ηp2 = .05, p = .24.
Figure 3 shows data pertaining to the distribution of fixations to the different distractor types
for repeated and random configurations across the two conditions. Recall that each configuration
contained two different sets of distractors, with each set appearing in a distinct color. For “repeated
configurations” half of the distractors (those in one color) were predictive of the target location,
while the other half of the distractors were randomly arranged and therefore nonpredictive. For
entirely random configurations, we continue to demarcate these into two sets referred to as
“predictive” and “nonpredictive” (although both sets of distractors are nonpredictive of the target
location) since these different sets of distractors were colored in a manner that corresponded to the
two sets of distractors in repeated configurations. Thus the fixation data in random configurations
provides a baseline for overt attention towards stimulus features in the absence of any predictive
information. It is clear that for the instructed condition, attention was biased towards the distractors
that were in the same color as the target (P). This effect was consistent across configurations with
repeating elements and those that were entirely random. However, in the learning condition there
was no clear attentional bias to either set of distractors, in either the configurations with repeating
elements or those that were entirely random.
These data were assessed with a mixed-model ANOVA (for each metric) with within-subject
factors of configuration (repeating vs. random) and distractor type (P vs. NP) and a between-subjects
factor of condition (instructed vs. learning). For the attentional spotlight metric, this revealed a main
effect of configuration, F(1,60) = 40.81, ηp2 = .41, p < .001, with a greater number of distractors
falling within the attentional spotlight of the fixations on trials with random configurations compared
to trials with repeated configurations (mirroring the earlier results of more fixations overall for
random configurations). There was a main effect of distractor type, F(1,60) = 200.75, ηp2 = .77, p <
Selective attention in contextual cuing Beesley, Hanafi, Vadillo, Shanks, & Livesey
15
.001, indicating that predictive distractors were more likely to fall within the attentional spotlight (on
average 7.1 per trial; SD = 1.8) than non-predictive distractors (5.9, SD = 2.5). There was also a main
effect of condition, F(1,60) = 92.57, ηp2 = .61, p < .001, with a greater number of fixations on
distractors overall in the learning condition (again mirroring the finding of a greater number of
fixations made and longer RTs in this condition). There was a significant configuration by condition
interaction, F(1,60) = 4.72, ηp2 = .07, p = .034, revealing that the difference in the processing of
repeated and random configurations was driven primarily by the data from the instructed condition.
There was a significant distractor type by condition interaction, F(1,60) = 189.50, ηp2 = .77, p < .001,
indicating that the attentional bias towards predictive distractors in the instructed condition (on
average 2.3 more predictive distractors attended; SD = 0.14) was greater than that in the learning
condition (0.0; SD = 0.1). The configuration by distractor type interaction was not significant,
F(1,60) = 2.34, ηp2 = .04, p = .13. The three-way interaction was significant, F(1,60) = 5.26, ηp2 = .08,
p = .025. This three-way interaction appears to result from a difference in the magnitude of
attentional bias towards P distractors over NP distractors between repeated and random
configurations across the two conditions. In the instructed condition, the attentional bias towards P
distractors over NP distractors was actually greater in random configurations compared to that in
repeated configurations (2.5 vs 2.1), t(30) = 3.40, d = .61, p = .002. In the learning condition, there
was no difference in the size of the attentional bias across repeated and random configurations (0.1
vs 0.0), t < 1.
An identical ANOVA on the distance metric found no effect of configuration, F < 1, but did
find an effect of distractor type, F(1,60) = 296.13, ηp2 = .83, p < .001, and condition, F(1,60) = 12.90,
ηp2 = .18, p = .001. The distractor type by condition interaction was significant, F(1,60) = 267.99, ηp2
= .82, p < .001, indicating that the difference in the distance of P and NP distractors to the point of
fixation was greater in the instructed condition (a difference of 1.4°; SD = 0.08°) than in the learning
condition (0.04°; SD = 0.03°). No other interaction effects were significant, Fs ≤ 1.05, ps ≥ .31.
Selective attention in contextual cuing Beesley, Hanafi, Vadillo, Shanks, & Livesey
16
These data suggest that attention was biased towards P distractors and away from NP
distractors in the instructed condition, but not in the learning condition. To test the null hypothesis
that the contextual cuing effect for repeated configurations does not result in a bias of attention in the
learning condition, the P and NP data for repeated configurations in this condition were subjected to
Bayesian paired t-tests2 separately for the two metrics. These revealed BF01 = 4.8 and BF01 = 4.0 for
the attentional spotlight and distance metrics, respectively. The conclusions were the same (BF01
2.9) when the data were analysed over epochs 3 and 4, or epoch 4 alone (i.e., where we would expect
contextual cuing to be at its strongest). There is therefore evidence to suggest that the contextual
cuing effect in the learning condition does not result in an attentional bias.
Discussion
Experiment 1 examined whether contextual cuing results in attentional biases to distractor
stimuli that differ in terms of their surface feature similarity to the target stimulus (i.e., color) and in
their ability to predict the position of the target. It was observed that when the distractors shared
stimulus properties that were relevant to the search task (i.e., they were presented in the same color
as the target), and participants were instructed about this regularity, then attention was biased
towards the processing of those distractors. However, this attentional bias was present for both
repeated and random configurations, indicating that it did not differ as a function of the relevance of
the specific set of distractors (i.e., with respect to locating the target). In other words, distractors that
were predictive of the target position were not favored over those that were non-predictive. In fact,
the attentional bias to distractors presented in the target color was larger in random configurations
compared to repeated configurations in the instructed condition. This is the opposite of the result that
would be predicted if participants had developed a bias towards predictive distractors and is possibly
the result of there being fewer fixations for repeated configurations overall (i.e., a floor effect may
have reduced the observed attentional bias). When the color of the distractors was rendered irrelevant
to the search task by making the target appear in each color with equal frequency (the learning
Selective attention in contextual cuing Beesley, Hanafi, Vadillo, Shanks, & Livesey
17
condition), no attentional bias was observed to the different distractor types. Overall, these results
suggest that the learning of predictive information in the configurations does not lead to changes in
the allocation of overt attention to that information during contextual cuing.
It is noteworthy that contextual cuing was observed in both the instructed and the learning
condition. That is, even in the learning condition, which showed no attentional bias towards
predictive distractors, a contextual cuing effect was observed. Thus, an attentional bias towards
predictive distractors is not a necessary condition for contextual cuing. However, it does appear that
the selective processing of distractors, as brought about by the attentional set determined by the
visual search task in the instructed condition, led to an enhancement in the contextual cuing that
occurred, in so much that the contextual cuing effect was greater in this condition compared to that
observed in the learning condition.
One possible explanation as to why a bias in selective attention failed to develop in the
learning condition could be that paying attention to the predictive distractors might provide a
disadvantage in the search task. That is, on half of all trials, the target in repeated configurations
appeared in the opposing color to that of the predictive distractors. In contrast, in the instructed
condition, there was no disadvantage to focusing attention on the predictive distractors, since the
target was always presented in the same color as these distractors. Consequently, Experiment 2
examined this possibility by attempting to minimise this disadvantageous aspect of selective
attention in the learning group.
It is also worth noting that the clear demonstration of a contextual cuing effect in the fixation
data (Figure 2B) lends support to the attentional guidance theory of contextual cuing. Chun and Jiang
(1998) suggested that contextual cuing is driven by the guidance of attention towards the target
location in the repeated configurations. Other researchers (Kunar, Flusberg, Horowitz, & Wolfe,
2007) have proposed that the contextual cuing effect might be driven instead by response selection
Selective attention in contextual cuing Beesley, Hanafi, Vadillo, Shanks, & Livesey
18
mechanisms. This non-attentional account denies that attention arrives at the target location earlier in
repeated configurations; rather, it proposes that repeated configurations permit a faster response to
the target by lowering the response threshold. However, given that our data showed that fewer
fixations were required in the repeated configurations than in random configurations and that eye-
gaze has been shown to be tightly coupled with attention (Deubel & Schneider, 1996), our data seem
to be incompatible with such a non-attentional account of contextual cuing (see also Harris, &
Remington, 2017; Peterson & Kramer, 2001; Tseng & Li, 2004; Zhao et al., 2012).
Experiment 2
The data from the learning condition of Experiment 1 suggest that contextual cuing can occur in the
absence of an attentional bias towards predictive distractors. However, we have suggested that for
the learning condition, shifting overt attention towards the predictive distractors may have been at
odds with conducting an efficient search for the target, since for half of the trials in this condition,
the target was not presented in the same color as the predictive distractors. It is possible that this
potential detrimental effect on target search may have hindered the development of a strong
attentional bias to predictive distractors in the learning condition.
We addressed this issue in Experiment 2 by training a modification of the learning condition
here termed the “split” condition – in which the distractors were red and blue, but in repeated
configurations the color of the predictive distractors matched that of the target. For example, if a
participant was trained with repeating configurations that contained predictive red distractors, then
the target was always presented in red in these configurations (and the target was always presented in
blue for random configurations). Thus, attending to the predictive distractors would now be
beneficial in terms of the primary task of visual search, at least for these trials. Consistent with this
prediction, Geyer, Shi and Muller (2010) have shown that the relationship between the target and
distractor colors plays a crucial role in contextual cuing. Geyer et al. trained participants with
Selective attention in contextual cuing Beesley, Hanafi, Vadillo, Shanks, & Livesey
19
configurations containing both predictive and nonpredictive distractors (as in the current conditions).
Stronger contextual cuing effects were observed when those predictive elements were trained in the
same color as the target, compared to when they were colored differently from the target. This result
demonstrates the importance of perceptual features in contextual cuing, and on the basis of these
findings we would predict greater contextual cuing in the split condition compared to the learning
condition of Experiment 1.
Recall that in Experiment 1, despite the absence of any effect on overt attention in the
learning group, a contextual cueing effect was still observed for this condition. It is possible that
grouping the predictive elements together by color is in some way beneficial for the encoding of the
configuration in memory. We sought to examine this in Experiment 2 in two ways. Firstly, we
compared learning in the split condition to that in a “mono” condition in which all of the stimuli (all
of the distractors and the target) were presented in one color. This mono condition was otherwise
identical to the split condition: for repeating configurations, half of the distractors were predictive of
the target position and half were randomly arranged. Secondly, in a final test phase of the
experiment, we presented the configurations of the split condition in both the training configuration
of colors and the reverse of these colors. That is, for “switched” trials, all of the blue elements (e.g.,
the predictive distractors and the target) were now presented in red, while the red elements (e.g., the
random distractors) were presented in blue. Should color grouping information play an important
role in the encoding or retrieval of stored representations in contextual cuing, we might expect that
this disruption of color would have a significant impact on performance in these switch trials. The
mono condition also provides a further test of the process of selective attention in contextual cuing,
in the sense that it provides a means to observe whether selective attention to predictive distractors
occurs in the absence of any biases that may be brought about as a result of the attentional set that is
determined by the visual search task.
Selective attention in contextual cuing Beesley, Hanafi, Vadillo, Shanks, & Livesey
20
Method
Participants
The experiment was approved by the local UNSW Sydney ethics committee. Sixty-two
undergraduate psychology students from The University of New South Wales participated in the
experiment in exchange for course credit. This sample size yields 77% power to detect a moderate-
to-large effect size of d = 0.7 in a between groups contrast, and 96% power to detect effects of the
same size in a within group contrast. All participants had normal color vision and normal or
corrected-to-normal visual acuity. Participants were randomly allocated to one of the two between-
subject conditions (split vs. mono). The experiment was approved by the ethics committee of the
School of Psychology, UNSW Australia, and all participants gave informed consent.
Design
The first phase of Experiment 2 employed a 2 x 2 x 3 mixed-model design. The between-
subject factor was color condition (split vs. mono) and the within-subject factors were configuration
(repeated vs. random) and epoch number (1 to 3). The search array in the ‘split’ group consisted of 8
red and 8 blue distractors, with a red target presented for all repeated configurations and a blue target
presented for all random configurations (see Table 2). In contrast, the search array in the mono’
group consisted of 16 red distractors and 1 red target. The training phase lasted for 3 epochs of 120
trials. All colors were counterbalanced across participants. This phase of the experiment continued in
the mono condition for one additional epoch of 120 trials. For comparison between the two between-
subject conditions, the trials from the 4th epoch in the mono condition are not analysed but are shown
in Figure 4.
The split condition received a final epoch of 120 trials (the 4th epoch) which we term the
switch phase”. Here the colors of all the stimuli (both sets of distractors and the target) in the
repeated configurations were switched (as described above). Since we presented both repeated (non-
Selective attention in contextual cuing Beesley, Hanafi, Vadillo, Shanks, & Livesey
21
switched) and switched trials equally often in this phase, this meant there were both repeated
configurations with blue targets and repeated configurations with red targets. Therefore, for the trials
that presented random configurations, the target was presented in red for one half of the trials and in
blue for the other half. This ensured that there was an equal distribution of red and blue targets
throughout the experiment in this condition. Repeated (non-switched), switched and random
configurations were presented in an intermixed manner during this phase (40 trials each).
Materials & Procedure
Experiment 2 used the same materials and procedure as Experiment 1, with the addition of
the switched configurations in the fourth epoch of the split condition. For the split condition, the
transition from the training phase to the switch phase occurred seamlessly without any signal to the
participant.
Results
Three participants in the split condition and four in the mono condition produced accuracy rates that
were below 90% and were therefore removed from the final analysis. For the remaining participants,
accuracy of responding was high in both the split (N = 28; 98.7%; SE = 0.2) and mono (N = 27;
97.8%; SE = 0.4) conditions. Data from trials on which an inaccurate response was made or on
which the reaction time was 2.5 standard deviations or more from the participant mean (3.0%; SD =
0.6) did not contribute to the analyses. The processing of eye gaze data into fixations was conducted
in an identical manner to Experiment 1. The average proportion of missing samples following the
interpolation procedure was 2.0% (SD = 2.8). Two participants (one in each condition) had more
than 15% of trials without any eye-gaze data and were not included in the final analysis; on average,
0.8% (SD = 1.5) of trials did not contain a single fixation and were therefore not analysed.
Figure 4 shows RTs (A) and number of fixations (B) for repeated and random configurations
for the split and mono conditions. The data show a contextual cuing effect in both conditions, but
Selective attention in contextual cuing Beesley, Hanafi, Vadillo, Shanks, & Livesey
22
this appears stronger in the split condition compared to the mono condition. This difference in the
extent of contextual cuing is also observed in the fixation data.
The RT data were analysed with a mixed-model ANOVA with within-subject factors of
configuration (repeated vs. random) and epoch (1-3) and between subject-factor of condition (split
vs. mono). This revealed a main effect of configuration, F(1,51) = 24.18, ηp2 = .32, p < .001,
reflecting a mean contextual cuing effect of 228 ms (SD = 350). There was also a general decrease in
RTs as revealed by a main effect of epoch, F(2,102) = 154.90, ηp2 = .75, p < .001. There was no main
effect of condition, F < 1. There was a significant interaction between configuration and epoch,
F(2,102) = 4.61, ηp2 = .08, p = .012, suggesting that contextual cuing increased with training, and
also a significant interaction between configuration and condition, F(1,51) = 5.97, ηp2 = .11, p = .018,
indicating that contextual cuing was stronger in the split condition (338 ms; SD = 383) than in the
mono condition (114 ms; SD = 275). The epoch by condition interaction was not significant,
F(2,102) = 2.88, p = .061, nor the three-way interaction, F(2,102) = 1.57, p = .21.
To assess the contextual cuing effect in each condition, two repeated-measures ANOVAs
were conducted with within-subject factors of configuration and epoch. In the split condition this
revealed main effects of configuration, F(1,26) = 21.03, ηp2 = .45, p < .001, and epoch, F(2,52) =
129.02, ηp2 = .83, p < .001, and a significant interaction effect, F(2,52) = 6.37, ηp2 = .20, p = .003. In
the mono condition there was a significant main effect of configuration, F(1,25) = 4.45, ηp2 = .15, p =
.045, a significant main effect of epoch, F(2,50) = 46.98, ηp2 = .65, p < .001, but no interaction
between these factors, F < 13.
The fixation data were analysed with an identical overall ANOVA, which revealed a main
effect of configuration, F(1,51) = 23.66, ηp2 = .32, p < .001, demonstrating a contextual cuing effect,
with a saving of 0.79 fixations (SD = 1.20), on average, for repeated configurations compared to
random configurations. There was also a main effect of epoch, F(2,102) = 151.63, ηp2 = .75, p < .001,
Selective attention in contextual cuing Beesley, Hanafi, Vadillo, Shanks, & Livesey
23
but no main effect of condition, F < 1. There was a significant interaction between configuration and
epoch, F(2,102) = 6.94, ηp2 = .12, p = .001, and also importantly between configuration and
condition, F(1,51) = 4.50, ηp2 = .08, p = .039, which confirms in the fixation data that the contextual
cuing effect was larger in the split condition (a saving of 1.11 fixations; SD = 1.32) compared to the
mono condition (a saving of 0.45 fixations; SD = 0.97). The remaining interaction effects were not
significant, Fs ≤ 2.12, ps ≥ .13. We also assessed whether the contextual cuing effect was present in
the fixation data in each condition. In the split condition this revealed main effects of configuration,
F(1,26) = 19.31, ηp2 = .43, p < .001, and epoch, F(2,52) = 97.85, ηp2 = .79, p < .001, and a significant
interaction effect, F(2,52) = 8.58, ηp2 = .25, p = .001. In the mono condition there was a significant
main effect of configuration, F(1,25) = 5.27, ηp2 = .17, p = .030, a significant main effect of epoch,
F(2,50) = 58.33, ηp2 = .70, p < .001, but no interaction between these factors, F < 1.
Figure 5 shows the attentional spotlight and distance metrics, as described in Experiment 1,
for the data from epochs 1-3. As was the case for the learning condition of Experiment 1, there was
very little evidence of differential distractor processing in either the split or the mono condition. The
attentional spotlight data were subjected to a mixed-model ANOVA with within-subjects factors of
configuration (repeated vs. random) and distractor type (predictive vs. non-predictive) and between-
subjects factor of condition (split vs. mono). This revealed a main effect of configuration, F(1,51) =
30.29, ηp2 = .37, p < .001, which mirrors the finding from the main fixation analysis, that a greater
number of distractors were fixated for random configurations than for repeated configurations. The
main effects of distractor type and condition were not significant, Fs < 1, and none of the interaction
effects were significant, Fs ≤ 2.86, ps ≥ .10.
The data for the distance metric were subjected to the same analysis process. There were no
main effects nor any significant two-way interaction effects, Fs ≤ 1.44, ps ≥ .24, however the three-
way interaction was significant, F(1,51) = 4.06, ηp2 = .07, p = .049. It is noteworthy that for repeated
configurations in the split condition, fixations were (numerically at least) closer to predictive
Selective attention in contextual cuing Beesley, Hanafi, Vadillo, Shanks, & Livesey
24
compared to non-predictive distractors, while the reverse is true for the mono condition. A two-way
ANOVA on the data from the split condition found no main effects, Fs 1, but did find a significant
interaction effect, F(1,26) = 4.89, ηp2 = .16, p = .036. However, the difference in distractor distance
was not significant in either the repeated or random configurations, ts ≤ 1.63, ps ≥ .11. A two-way
ANOVA on the data from the mono condition found no main effects nor an interaction effect, Fs < 1.
Figure 6 shows RTs and numbers of fixations to different configurations in the “test” phase
for the split condition. In both RTs and fixations, it is clear that a contextual cuing effect is observed
on trials in which the trained arrangement of colors was used, as well as on trials in which those
colors were switched. Paired samples t-tests supported these conclusions with significant differences
in RTs for trained vs. random, t(26) = 2.53, d = .49, p = .018, switched vs. random, t(26) = 3.96, d =
.76, p = .001, but not for trained vs. switched trial types, t < 1. A similar pattern of results was
observed for the fixation data: trained vs. random, t(26) = 2.60, d = .50, p = .015; switched vs.
random, t(26) = 4.12, d = .79, p < .001; trained vs. switched, t < 1.
Discussion
In Experiment 2, all participants experienced repeated configurations that contained both
distractors that were predictive and distractors that were non-predictive of the target position. We
observed contextual cuing effects in both of these conditions. However, as was observed for the
learning condition in Experiment 1, significant contextual cuing effects (present in both RT and
fixation data) did not occur as a result of an attentional bias to predictive distractors over non-
predictive distractors. While we observed a significant three-way interaction in the fixation distance
metric, our follow up analyses, as well as the lack of any such effects in the attentional spotlight
metric, suggests that we should be cautious about interpreting these effects as related to a specific
attentional bias towards predictive distractors. We can therefore conclude that the ability to detect the
target at a more rapid rate (and after fewer fixations) when presented with a repeated configuration is
Selective attention in contextual cuing Beesley, Hanafi, Vadillo, Shanks, & Livesey
25
not a result of selectively fixating on the predictive information within that configuration. An
exploratory analysis of the eye-gaze data is presented in the General Discussion to examine in more
detail the time-course and efficiency of search through repeated and random configurations.
Experiment 2 also examined the role of color segmentation in contextual cuing. Participants
were either trained with predictive and non-predictive distractors in different colors (split condition)
or with all distractors in the same color (mono). We found that the cuing effect was stronger when
the predictive and non-predictive distractors were segmented by color (split) compared to when they
were presented in the same color (mono). This suggests that in our task, perceptual segmentation
leads to a stronger trace of the repeated configuration of distractors in memory, perhaps by
facilitating the formation of a configural representation of these similar elements (Beesley, et al.,
2015, 2016). This finding is at odds with the results of a series of experiments presented by Conci
and von Mühlenen (2011), in which they examined how perceptual segmentation of the distractors
modulates contextual cuing. When the search task ensured the target features were not preferentially
attended, Conci and von Mühlenen found that the contextual cuing effect was significantly weaker
when the configuration was segmented by perceptual features (compared to cuing for homogenous
configurations). This is inconsistent with our finding of enhanced cuing for segmented
configurations, but may be explained by the different amounts of predictive context in the two
designs. In our experiments, repeated configurations contained both predictive and nonpredictive
distractors, while in Conci and von Mühlenen’s experiments the repeated configurations were
entirely predictive. Therefore perceptual segmentation may be beneficial when it constrains the
processing of information to a relevant set (and avoids processing irrelevant information), but is a
hindrance when processing is necessary across information presented in two distinct perceptual
features. Indeed, similar benefits of segmentation have been observed for partially predictive
configurations (Geyer, Shi, & Müller, 2010).
Selective attention in contextual cuing Beesley, Hanafi, Vadillo, Shanks, & Livesey
26
In a final phase of the experiment, for those in the split condition, the colors of the distractors
in these repeating configurations were switched in a final phase. We found that this had no
detrimental effect on the size of the contextual cuing effect that was observed. One interpretation of
these data is that the representation of the repeating configuration exists in a form that is independent
of the surface level features of the distractor elements (e.g., distractor colors). Taken together with
the more substantial cuing effect in the split condition compared to the mono condition, these data
may suggest that color information acts to modulate the encoding of repeated configurations in visual
search but the subsequent behavior elicited by these configurations is driven by the spatial
configuration and not the surface features.
Statistical comparisons of the data from Experiments 1 and 2
As a means to elucidate the variables affecting contextual cuing and attentional processing, we
provide a statistical analysis of the four between-subjects conditions conducted across the two
experiments. Since the two experiments were drawn from different samples, and the data were
collected in different time periods, the conclusions should be taken as tentative at best; full
experimental control of these variables is necessary to be conclusive about their importance.
Nevertheless, it is notable that across experiments there were substantial differences in the size of the
contextual cuing effects. Of particular note is that the split condition in Experiment 2 showed a
comparable contextual cuing effect to that of the instructed condition in Experiment 1, while the
remaining two conditions showed comparably small contextual cuing effects. To simplify the
analysis we took the data from the first 3 epochs of trials (trials 1-360) and computed a contextual
cuing score for each condition of interest by subtracting the reaction time for repeated configurations
from that for random configurations. We observed that the cuing effect for the Split condition in
Experiment 2 (338 ms, SE = 74 ms) was somewhat similar in size to that of the Instructed condition
in Experiment 1 (208 ms, SE = 32 ms), though there was little evidence in favor of the equivalence
of these cuing effects, BF01 = 1.15. In contrast, the cuing effect in the Split condition of Experiment
Selective attention in contextual cuing Beesley, Hanafi, Vadillo, Shanks, & Livesey
27
2 was stronger than that observed for the Learning condition in Experiment 1 (72 ms, SE = 37 ms),
and there was considerable evidence in support of this difference, BF10 = 23.46. A similar pattern of
results was observed in the fixation data (Split vs. Instructed, BF01 = 1.60; Split vs. Learning, BF10 =
3.78).
These comparisons appear to suggest that the Split condition showed a stronger contextual
cuing effect than the Learning condition. The only difference between the two conditions was that in
the Split condition, the target was always presented in the same color as the predictive distractors,
while in the Learning condition the target could be in either color (across trials). This suggests that
the coincidence of the distractor and target colors promotes the learning of associations, a finding
which is consistent with data presented by Geyer, Shi, and Müller (2010). Note that despite these
suggested differences in the size of the cuing effects in the Split and Learning conditions, the profile
of eye-movements was identical in the two conditions: in neither condition did we observe greater
processing of predictive over non-predictive distractors.
General Discussion
The four conditions across these two experiments paint a very consistent pattern. Selective attention
(fixations) towards particular sets of distractors was prioritised according to the attentional set
determined by the nature of the search task, but showed no sensitivity to the predictiveness of
distractors. When participants were explicitly instructed to search for a target that always appeared in
a single color and to ignore distractors of a different color (Experiment 1: instructed condition), overt
attention was strongly biased towards distractors appearing in the same color as the target. The eye
gaze analyses of the instructed group of Experiment 1 show this bias clearly, regardless of whether
those distractors were predictive (on repeated configuration trials) or nonpredictive (on random
configuration trials). In contrast, when the task was to search for a target that could appear in one of
two colors, but distractors in just one color were predictive on repeated trials, participants did not
Selective attention in contextual cuing Beesley, Hanafi, Vadillo, Shanks, & Livesey
28
fixate the predictive distractors any more than the non-predictive distractors in either a configuration-
specific manner (i.e., on repeated trials only) or a feature-general manner (i.e., on all trials). This was
the case in the learning condition of Experiment 1, when the predictive distractors predicted the
location but not the color of the target, and in the split condition of Experiment 2, when the
predictive distractors conveyed information about both the location and color of the target.
Furthermore, when only a single color was used for all distractors and targets in the mono condition
of Experiment 2, thus removing uncertainty about the color of the target without producing the
strong task-driven biases in selective attention seen in Experiment 1, there was again no bias towards
fixating predictive compared to non-predictive distractors.
These results clearly suggest that selection biases in contextual cuing are driven by the top-
down demands of the search task and not the learned predictiveness of the distractors within repeated
configurations. To provide further support for this conclusion, the data from the selective attention
metrics were combined for the three conditions which appeared to show no bias in selective attention
(epochs 1 to 4 in the learning and mono conditions, and epochs 1 to 3 in the split condition). These
data were subjected to Bayesian paired samples t-tests to compare attention to predictive and non-
predictive distractors for repeated configuration trials. This revealed that for both the attentional
spotlight, BF01 = 7.96, and the distance metric, BF01 = 5.00, there was considerable support for the
null hypothesis that there was no effect of selective attentional processing.
Given the evidence that there is no bias in overt attention towards predictive distractors in
repeated configurations, it is especially noteworthy that this predictive information clearly did
improve the efficiency of target detection during search in all of the conditions tested. In every
condition, participants required fewer fixations to locate the target on repeated configuration trials
than on random configuration trials. Thus, despite their inherent usefulness for improving visual
search, predictive distractors appear to receive no bias in selective processing. One explanation for
this somewhat paradoxical finding might be that the demonstration of equivalent fixation profiles to
Selective attention in contextual cuing Beesley, Hanafi, Vadillo, Shanks, & Livesey
29
repeated and random distractors does not necessarily equate to equivalent processing of the two sets
of distractors. In other words, more weight may be placed on the processing of predictive distractors
without this leading to a higher probability of those distractors being fixated. Indeed, learning in the
absence of overt attentional processing has been suggested by Jiang and Leung (2005), who gave
participants a task similar to that used by Jiang and Chun (2001), in which repeated configurations
contained predictive distractors in both “attended” and “unattended” colors (as determined by the
color of the target). In a later stage, Jiang and Leung reversed the assignment of colors for the
distractors (as in the switch phase of Experiment 2) and found that once the unattended distractors
became the focus of attention, contextual cuing was observed. This result suggests that learning
about predictive elements of the search configuration may occur even when those elements are not
the focus of overt processing. Indeed, contextual cuing can occur even in the presence of attention
demanding stimuli such as color singletons (Conci & von Mühlenen, 2009; Harris & Remington,
2017). These findings are consistent with our own, which demonstrate that strong contextual cuing
effects occur without a strong biasing of attention to the relevant predictive content.
There remains a question as to why repeated configurations have fewer fixations: does the
fixation profile of repeating configurations tell us anything about the efficiency of search in
contextual cuing? To examine this further, we subjected the fixation data from the learning (epochs 1
to 4), split (epochs 1 to 3), and mono conditions (epochs 1 to 4) to an exploratory analysis. Figure 7
shows the data grouped by the number of fixations within a trial. Figure 7A shows the mean pixel
distance from the target of each fixation within these different trial types. Firstly, it is quite clear that
across the trials of different length (3 fixations to 10 fixations), the pattern of data is remarkably
similar between the repeated and random search configurations. Secondly, in line with the previous
analyses conducted by Tseng and Li (2004), our data suggest that visual search performance in these
tasks has two characteristic components: an inefficient search process in which attention moves in a
non-productive manner, failing to move towards the target and consistently away from the target
Selective attention in contextual cuing Beesley, Hanafi, Vadillo, Shanks, & Livesey
30
location in trials with more fixations, followed by an efficient search process in which consistent
monotonic increments are made towards the target position (the final 3-4 fixations). The latter
efficient search process is extremely consistent across the different trial types, suggesting that once
this phase is reached, the trial terminates in a similar manner across paths of different lengths. In
contrast, the trials differ in the length of the first, inefficient search process.
Figure 7B shows the proportion of trials containing a given number of fixations for the
repeated and random configurations. As the data clearly show, the benefit of search through repeated
configurations is driven by a greater number of trials having fewer fixations, particularly in the range
of 1-5 fixations. Conversely, there is a consistent pattern of repeated configurations being less likely
to have trials with more fixations (6-20). To provide some statistical basis to these claims, we
subjected the data to a two-way Repeated Measures ANOVA with factor of configuration (repeated
vs. random) and fixations (1-20), which importantly revealed a significant interaction between the
factors, F(19,1577) = 7.73, ηp2 = .09, p < .001, suggesting that the pattern in the proportion of trials
with different numbers of fixations differed between the repeated and random configurations. An
analysis of the data on trials with fewer than 6 fixations revealed that there was a greater proportion
of these trials for repeated compared to random configurations, t(83) = 5.52, d = .60, p < .001 (a
complementary analysis of the RT data is presented in the Appendix).
Thus, these data suggest that for repeated configurations, the inefficient period of visual
search is more likely to cease at an earlier stage in the search process and there will therefore be an
earlier transition into the efficient search process (over the final 3 or 4 fixations, as suggested by
Figure 7A). For each participant, we calculated the maximum distance from the target of all the
fixations in a trial, and calculated the mean of these maximal distances for repeated and random
configurations. This revealed that search tends not to stray as far from the target location for repeated
configurations (maximum distance = 17.7°, SE = 0.12) compared to search through random
configurations (maximum pixel distance = 18.0°, SE = 0.11), t(83) = 2.63, d = .29, p = .010.
Selective attention in contextual cuing Beesley, Hanafi, Vadillo, Shanks, & Livesey
31
Overall, our data suggest that participants search directly for the perceptual features that
define the target (as demonstrated in the instructed condition), and not for the cues that might inform
them as to where the target is located. In doing so, they fixate distractors that occur in the repeated
configurations, which provide information about where the target is located, but this sampling is not
strategic and thus participants do not prioritise selection of these predictive distractors over non-
predictive distractors in future encounters with the same repeated configuration.
Search through repeated configurations appears to progress, at least initially, in a somewhat
random fashion, with fixations evenly distributed towards predictive and non-predictive distractors.
Once some beneficial predictive information has been encountered (i.e., when a match is made with
a stored representation in memory), this leads to a termination in the inefficient search at an earlier
time point (Figure 7), and also to a reduction in the maximal distance that fixations drift from the
target location.
Why would the predictive distractors in contextual cuing fail to receive prioritised attention
when there is evidence of biases in attention on the basis of learned predictiveness in a wide range of
other tasks? There are other visual cognition paradigms that exhibit oculumotor biases towards
predictive distractors (e.g., Le Pelley et al., 2015) and other learning tasks that exhibit learned
predictiveness biases towards multiple spatially discrete predictive and nonpredictive features (e.g.,
Livesey & McLaren, 2007). Thus the procedure we employed in the contextual cuing task in which
predictive and nonpredictive information were spatially distributed and simultaneously presented
would be expected to lead to such biases.
Although it may seem somewhat counterintuitive, our contention is that it may be optimal in
tasks like contextual cuing for attention to be strongly controlled by a search for the defining features
of the target, even at the expense of attending to predictive cues. On a typical contextual cuing trial,
the target is presented simultaneously with the predictive distractors, and search times are typically
Selective attention in contextual cuing Beesley, Hanafi, Vadillo, Shanks, & Livesey
32
relatively brief. Repeated configurations also possess reasonably high similarity with random
configurations, meaning that the learned components of a repeated array may not be immediately
apprehended. These properties of the task are important because they may limit the usefulness of
selecting predictive distractors for further attention relative to simply investing attention in those
perceptual features that define the target itself. The predictive information that repeated distractors
convey essentially directs the participant to look to another location to find the target. Thus by the
time the specific components of a repeated configuration are processed, biasing attention more
strongly towards them may serve little purpose and in fact may actually hinder rapid target selection
(if predictive distractors capture attention, they may prevent attention from reaching the target
quickly). Indeed, the benefits that are provided by intentionally encoding the predictive content of
the displays are minimal (a saving in the order of 200 ms or 1 fixation in the conditions examined
here).
That is not to say that predictive information is neglected in the encoding of the display; our
data quite clearly show that predictive information was learnt and had significant control of behavior
in terms of producing shorter search times. Like all implicit learning tasks, the learning that occurs in
the contextual cuing task might be best considered incidental, in the sense that it is arguably
superfluous to the explicit demands of the task (i.e., to respond accurately to the target orientation).
Our data certainly support this notion that participants learn incidentally and do not strategically
encode greater amounts of predictive information in the displays as they search; if anything, it would
appear that search is somewhat random until the point at which predictive information is
encountered, and then transitions rapidly to an efficient search process which results in target
detection within 3 or 4 fixations. It is possible that the learning of predictive information occurs
somewhat covertly (that is, not determined exclusively by the spatial location of the fixations), either
by an initial (perhaps rapid) global processing of the whole scene, or sequentially through the non-
strategic search process we have identified.
Selective attention in contextual cuing Beesley, Hanafi, Vadillo, Shanks, & Livesey
33
This account suggests that an attentional bias to predictive distractors fails to emerge in
contextual cuing because it is not optimal for the individual to develop one. This account thus
implies that learned predictiveness effects are strategic and initiated voluntarily, which is a
description that may not characterise the general attentional consequences of learned predictiveness
particularly well (for a review see Le Pelley et al., 2016). An alternative explanation is simply that in
visual search, selective attention is so strongly dominated by searching for the defining physical
properties of the target that biases to other features (voluntary or otherwise) are extremely difficult to
observe. Biasing stimulus selection towards the unique features of the target may be the single most
important factor in determining good performance in visual search, and this differs from most
learned predictiveness tasks, where attending to the specific features of the predicted outcome can
occur sometime after attending to the predictive cues. This may be why the reciprocal relationship
between predictive learning and attention that is observed in many other experimental contexts does
not appear to hold in this one.
While our data suggest that attentional selection is controlled exclusively by the attentional
set determined by the search task, the data from Experiment 2 reveal that the perceptual features of
the distractors and their relationship with the target features did have a significant effect on the
contextual cuing that developed. In Experiment 2, the cuing effect (in both RTs and the number of
fixations) was greater in the split condition compared to that in the mono condition. That is, while
grouping the predictive and non-predictive distractors by color did not lead to an overt attentional
bias towards those predictive distractors, it had a clear effect on the strength of the memory
representation that formed for those configurations.
Furthermore, the data from our final test phase suggest that this representation may well be
stored in a form independent from (or at least insensitive to) the color information in which it has
been presented: when the colors of the predictive and non-predictive distractors were switched in this
final phase, intact cuing was observed in both RTs and fixations. It would seem therefore that the
Selective attention in contextual cuing Beesley, Hanafi, Vadillo, Shanks, & Livesey
34
benefit of segregating the predictive and nonpredictive information occurs during the initial encoding
of the configuration, rather than the recall of that information from memory. While further
experimental evidence will be needed to support these conclusions, these findings may well have
important implications for the manner in which surface feature information is realized in formal
models of contextual cuing (e.g., Brady & Chun, 2007; Beesley et al., 2015, 2016).
In conclusion, the data from these experiments illuminates the interaction between associative
learning and overt attentional processes in the contextual cuing task. Overt attentional biases are
driven by physical properties of the stimuli governed by the target search template and are not
preferentially directed by the associative strength of distractors contained within repeating
configurations. Furthermore, in line with the findings of Tseng and Li (2004), our data suggest that
the contextual cuing effect occurs when a stored representation is activated following an
unsystematic (i.e., random) search process, which leads to an early termination of inefficient search,
and an earlier engagement of efficient search.
Selective attention in contextual cuing Beesley, Hanafi, Vadillo, Shanks, & Livesey
35
Footnotes
1. A mixed-model ANOVA with the same factors was conducted on the accuracy data of
Experiment 1 and found only a main effect of epoch, F(3,180) = 6.59, ηp2 = .10, p < .001,
with accuracy increasing across epochs from 96.6% in epoch 1 to 98.3% in epoch 4. No other
main effects or interactions were significant, Fs ≤ 2.16, ps ≥ .10.
2. Bayesian statistics were conducted in JASP (Version 0.8.0.0) with the default Cauchy prior
width of 0.707.
3. A mixed-model ANOVA with the same factors was conducted on the accuracy data of
Experiment 2 and found only a main effect of epoch, F(2,102) = 4.96, ηp2 = .09, p = .01, with
accuracy increasing across epochs from 97.5% in epoch 1 to 98.4% in epoch 3. No other
main effects or interactions were significant, Fs ≤ 2.53, ps ≥ .09.
Selective attention in contextual cuing Beesley, Hanafi, Vadillo, Shanks, & Livesey
36
Appendix
The RT data for epochs 1-3 from the learning, split, and mono conditions were subjected to further
analysis in order to evaluate the observed contextual cuing effect across the range of RTs (Figure 8).
In Figure 8A, RTs are split into 10 bins by each decile of reaction time for both the repeated and
random configurations. At each decile, the reaction times were faster in the repeated than the random
configurations, ts(83) ≥ 2.70, ds ≥ .15, ps ≤ .01. As in the fixation data presented in Figure 7B, a
similar pattern of data emerged in the reaction time data, with a greater percentage of trials having
faster reaction times in the repeated configurations compared to the random configurations. An
ANOVA with factors of configuration and bin found a significant interaction, F(21,1743) = 5.52, ηp2
= .06, p < .001, supporting this conclusion. These analyses are consistent with those presented by
Johnson, Woodman, Braun, and Luck (2007).
Selective attention in contextual cuing Beesley, Hanafi, Vadillo, Shanks, & Livesey
37
Selective attention in contextual cuing Beesley, Hanafi, Vadillo, Shanks, & Livesey
38
Author note
This work was supported by Grant ES/J007196/1 from the Economic and Social Research Council, awarded
to David R. Shanks and Tom Beesley. Miguel A. Vadillo was supported by Grant 2016-T1/SOC-1395 from
Programa de Atracción de Talento Investigador, Comunidad de Madrid. This research formed the basis of an
Honours dissertation by Gunadi Hanafi. Correspondence concerning this article should be addressed to Tom
Beesley, School of Psychology, UNSW Australia, Sydney, New South Wales, Australia, 2052. Email:
t.beesley@unsw.edu.au.
Selective attention in contextual cuing Beesley, Hanafi, Vadillo, Shanks, & Livesey
39
Awh, E., Belopolsky, A. V., & Theeuwes, J. (2012). Top-down versus bottom-up attentional
control: A failed theoretical dichotomy. Trends in Cognitive Sciences, 16, 437443.
Beesley, T., & Le Pelley, M. E. (2010). The effect of predictive history on the learning of
sub-sequence contingencies. The Quarterly Journal of Experimental Psychology, 63, 108135.
Beesley, T., & Le Pelley, M. E. (2011). The influence of blocking on overt attention and
associability in human learning. Journal of Experimental Psychology: Animal Behavior
Processes, 37, 114-120.
Beesley, T., Nguyen, K. P., Pearson, D., & Le Pelley, M. E. (2015). Uncertainty and
predictiveness determine attention to cues during human associative learning. Quarterly Journal of
Experimental Psychology, 68, 21752199.
Beesley, T., & Shanks, D. R. (2012). Investigating cue-competition in contextual cuing of
visual search. Journal of Experimental Psychology: Learning, Memory & Cognition, 38, 709725.
Beesley, T., Vadillo, M. A., Pearson, D., & Shanks, D. R. (2015). Pre-exposure of repeated
search configurations facilitates subsequent contextual cuing of visual search. Journal of
Experimental Psychology: Learning, Memory, and Cognition, 41, 346-362.
Beesley, T., Vadillo, M. A., Pearson, D., & Shanks, D. R. (2016). Configural learning in
contextual cuing of visual search. Journal of Experimental Psychology: Human Perception and
Performance, 42, 1173-1185.
Brady, T. F., & Chun, M. M. (2007). Spatial constraints on learning in visual search:
modeling contextual cuing. Journal of Experimental Psychology: Human Perception and
Performance, 33, 798-815.
Brainard, D. H. (1997). The psychophysics toolbox. Spatial Vision, 10, 433436.
Broadbent, D. E. (1958). Perception and communication. Oxford: Oxford University Press.
Selective attention in contextual cuing Beesley, Hanafi, Vadillo, Shanks, & Livesey
40
Chun, M. M., & Jiang, Y. (1998). Contextual cueing: Implicit learning and memory of visual
context guides spatial attention. Cognitive Psychology, 36, 28-71.
Conci, M., & von Mühlenen, A. (2009). Region segmentation and contextual cuing in visual
search. Attention, Perception, & Psychophysics, 71, 151424.
Conci, M., & Mühlenen, A. Von. (2011). Limitations of perceptual segmentation on
contextual cueing in visual search. Visual Cognition, 19, 203-233.
Deubel, H., & Schneider, W. X. (1996). Saccade target selection and object recognition:
Evidence for a common attentional mechanism. Vision research, 36, 1827-1837.
Evans, P. M., & Craig, J. C. (1991). Tactile attention and the perception of moving tactile
stimuli. Perception and Psychophysics, 49, 355-364.
Geyer, T., Shi, Z., & Müller, H. J. (2010). Contextual cueing in multiconjunction visual
search is dependent on color- and configuration-based intertrial contingencies. Journal of
Experimental Psychology. Human Perception and Performance, 36, 515532.
Harris, A. M. and R. W. Remington (2017). Contextual cueing improves attentional
guidance, even when guidance is supposedly optimal. Journal of Experimental Psychology: Human
Perception and Performance, 43, 926-940.
Jiang, Y., & Chun, M. M. (2001). Selective attention modulates implicit learning. The
Quarterly Journal of Experimental Psychology, 54, 1105-1124.
Johnson, J. S., Woodman, G. F., Braun, E. L., & Luck, S. J. (2007). Implicit memory
influences the allocation of attention in visual cortex. Psychonomic Bulletin & Review, 14, 834839.
Kleiner, M., Brainard, D. H., & Pelli, D. G. (2007). What’s new in Psychtoolbox-3?
Perception, 36, ECVP Abstract Supplement.
Selective attention in contextual cuing Beesley, Hanafi, Vadillo, Shanks, & Livesey
41
Kunar, M. A., Flusberg, S., Horowitz, T. S., & Wolfe, J. M. (2007). Does contextual cuing
guide the deployment of attention? Journal of Experimental Psychology: Human Perception and
Performance, 33, 816.
Le Pelley, M. E., Beesley, T., & Griffiths, O. (2011). Overt attention and predictiveness in
human contingency learning. Journal of Experimental Psychology: Animal Behavior Processes, 37,
220-229.
Le Pelley, M. E., Beesley, T., & Griffiths, O. (2014). Relative salience versus relative
validity: Cue salience influences blocking in human associative learning. Journal of Experimental
Psychology: Animal Learning and Cognition, 40, 116132.
Le Pelley, M. E., Mitchell, C., Beesley, T. G., George, D.N. & Wills, A. J. (2016). Attention
and associative learning in humans: An integrative review. Psychological Bulletin, 142, 1111-1140.
Le Pelley, M. E., Pearson, D., Griffiths, O., & Beesley, T. (2015). When goals conflict with
values: Counterproductive attentional and oculomotor capture by reward-related stimuli. Journal of
Experimental Psychology: General, 144, 158171.
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, 18881900.
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.
Livesey, E. J., & McLaren, I. P. L. (2007). Elemental associability changes in human
discrimination learning. Journal of Experimental Psychology: Animal Behavior Processes, 33, 148.
Selective attention in contextual cuing Beesley, Hanafi, Vadillo, Shanks, & Livesey
42
Luque, D., Vadillo, M. A., Le Pelley, M. E., & Beesley, T. (in press). Prediction and
uncertainty: Examining controlled and automatic components of learned attentional biases. Quarterly
Journal of Experimental Psychology.
Mitchell, C. J., Griffiths, O., Seetoo, J., & Lovibond, P. F. (2012). Attentional mechanisms in
learned predictiveness. Journal of Experimental Psychology: Animal Behavior Processes, 38, 191
202.
Motter, B. C., & Belky, E. J. (1998). The guidance of eye movements during active visual
search. Vision research, 38, 1805-1815.
Pelli, D. G. (1997). The VideoToolbox software for visual psychophysics: Transforming
numbers into movies. Spatial Vision, 10, 437442.
Rehder, B., & Hoffman, A. B. (2005). Eyetracking and selective attention in category
learning. Cognitive Psychology, 51, 141.
Rock, I., & Gutman, D. (1981). The effect of inattention on form perception. Journal of
Experimental Psychology: Human Perception and Performance, 7, 275285.
Salvucci, D. D., & Goldberg, J. H. (2000). Identifying fixations and saccades in eye-tracking
protocols. Proceedings of the symposium on eye tracking research & applications, 71-78.
Theeuwes, J. (2010). Top-down and bottom-up control of visual selection. Acta
Psychologica, 135, 77-99.
Treisman, A. M., & Gelade, G. (1980). A feature-integration theory of attention. Cognitive
Psychology, 12, 97-136.
Tseng, Y. C., & Li, C. S. R. (2004). Oculomotor correlates of context-guided learning in
visual search. Perception & Psychophysics, 66, 1363-1378.
Selective attention in contextual cuing Beesley, Hanafi, Vadillo, Shanks, & Livesey
43
Wolfe, J. M. (1994). Guided search 2.0 a revised model of visual search. Psychonomic
Bulletin & Review, 1, 202-238.
Wolfe, J. M., Võ, M. L. H., Evans, K. K., & Greene, M. R. (2011). Visual search in scenes
involves selective and nonselective pathways. Trends in Cognitive Sciences, 15, 7784.
Zhao, G., Liu, Q., Jiao, J., Zhou, P., Li, H., & Sun, H.-J. (2012). Dual-state modulation of the
contextual cueing effect: Evidence from eye movement recordings. Journal of Vision, 12, 1-13.
Selective attention in contextual cuing Beesley, Hanafi, Vadillo, Shanks, & Livesey
44
Table 1.
Condition
Target color (% of trials)
Instructed (Repeated)
Red (100%)
Instructed (Random)
Red (100%)
Learning (Repeated)
Red (50%) or blue (50%)
Learning (Random)
Red (50%) or blue (50%)
Note: The manipulations of predictive and randomised distractors, their colors, and the colors of the
targets for Experiment 1. The colors of the stimuli were counterbalanced, such that in the instructed
condition half of the participants searched for a blue target with blue predictive distractors, while in
the learning condition half the participants were trained with blue predictive distractors and red
random distractors.
Selective attention in contextual cuing Beesley, Hanafi, Vadillo, Shanks, & Livesey
45
Table 2.
Condition
Target color (% of trials)
Split (Repeated)
Red (100%)
Split (Random)
Blue (100%)
Mono (Repeated)
Red (100%)
Mono (Random)
Red (100%)
Note: The manipulations of predictive and randomised distractors, their colors and the colors of the
targets for Experiment 2. The colors of the stimuli were counterbalanced, such that in the split
condition half of the participants searched for a blue target with blue predictive distractors.
Selective attention in contextual cuing Beesley, Hanafi, Vadillo, Shanks, & Livesey
46
Figure 1. A schematic of the different conditions in Experiments 1 and 2. All configurations actually
contained 16 distractors and the different sets of distractors were evenly distributed across the screen.
Distractors presented in boxes reflect those that repeated across trials and therefore were predictive
of the target location. For simplicity, the same configurations are presented in all four conditions, but
note that both repeated and random configurations were generated randomly for each participant in
each condition.
Selective attention in contextual cuing Beesley, Hanafi, Vadillo, Shanks, & Livesey
47
Figure 2. Reaction time (A) and number of fixations (B) to repeated and random configurations in
the instructed and learning conditions of Experiment 1. Error bars show standard error of the mean.
Selective attention in contextual cuing Beesley, Hanafi, Vadillo, Shanks, & Livesey
48
Figure 3. Attention to distractor types P (predictive) and NP” (nonpredictive; see text for
description) in repeated and random configurations for the instructed and learning conditions of
Experiment 1. A: mean number of distractors falling within the “attentional spotlight” of all fixations
of a trial. B: mean distance of the nearest distractor of each type to each fixation. Error bars show
standard error of the mean.
Selective attention in contextual cuing Beesley, Hanafi, Vadillo, Shanks, & Livesey
49
Figure 4. Reaction time (A) and number of fixations (B) to repeated and random configurations in
the split and mono conditions of Experiment 2. Error bars show standard error of the mean.
Selective attention in contextual cuing Beesley, Hanafi, Vadillo, Shanks, & Livesey
50
Figure 5. Attention to distractor types “P” and “NP” (see text for description) in repeated and
random configurations for the split and mono conditions of Experiment 2. A: mean number of
distractors falling within the “attentional spotlight” of all fixations of a trial. B: mean distance of the
nearest distractor of each type to each fixation. Error bars show standard error of the mean.
Selective attention in contextual cuing Beesley, Hanafi, Vadillo, Shanks, & Livesey
51
Figure 6. Detailed analysis of the reaction time (left) and number of fixations (right) in epoch 4 of
the split condition in Experiment 2. Trained and switched configurations were different presentations
of the same set of repeated configurations, with the labels reflecting whether the distractors were
presented in the same color arrangement as presented in epochs 1-3 (trained) or in the reverse
arrangement of colors (switched). Error bars show standard error of the mean.
Selective attention in contextual cuing Beesley, Hanafi, Vadillo, Shanks, & Livesey
52
Figure 7. Fixation data for correct responses from the learning, split and mono conditions, grouped
according to the number of fixations per trial and separately for repeated and random configurations.
A: mean pixel distance from the target of each fixation in turn (left to right); for clarity of
presentation the data for random configurations are offset horizontally from those for repeated
configurations. B: percentage of trials containing a given number of fixations (error bars show
standard error of the mean).
Selective attention in contextual cuing Beesley, Hanafi, Vadillo, Shanks, & Livesey
53
Figure 8. Reaction time data for correct responses from the learning, split, and mono conditions. A:
data presented separately by each decile of reaction time and for repeated and random
configurations. B: percentage of trials within a given reaction time bracket (200 ms bin) for repeated
and random configurations.
... Moreover, using a similar analysis to the one proposed by Beesley et al. (2018), we compared the mean distance (degrees of visual angle) between fixations to target locations. This technique allowed us to precisely measure the proximity of fixations to the target across all trials. ...
... Figure 8, Panel B shows the percentage of trials with different numbers of fixations taken to find the target for predictive and random configurations, respectively. Combining Panels A and B in Figure 8, it can be inferred that the benefit of contextual learning of predictive configurations is driven by a greater number of trials having fewer fixations, particularly in the range of 3-5 fixations (see also Beesley et al., 2018). ...
... function of epochs confirmed the iMAP results; fixations tended to be closer to the target for predictive configurations compared to random configurations. Adopting an analytic technique fromBeesley et al. (2018), we also grouped the fixation data by the number of fixations within a trial. Analyzing data in this way presented a dynamic analysis of fixations across the course of the trials. ...
Article
Full-text available
The contextual cueing effect (CCE) refers to the learned association between predictive configuration and target location, speeding up response times for targets. Previous studies have examined the underlying processes (initial perceptual process, attentional guidance, and response selection) of CCE but have not reached a general consensus on their contributions to CCE. In the present study, we used eye tracking to address this question by analyzing the oculomotor correlates of context-guided learning in visual search and eliminating indefinite response factors during response priming. The results show that both attentional guidance and response selection contribute to contextual learning.
... There has been rather little focus on the question of whether a sequential prediction is necessary for the establishment of a learned attentional bias, or whether mere association will suffice. In a visual search study reported by Beesley et al. (2018), participants were required to make a response about the orientation of a target stimulus that was presented simultaneously with an array of distractors. In one condition, a configuration of relevant distractors provided information about the location of the target. ...
Article
Learning permits even relatively uninteresting stimuli to capture attention if they are established as predictors of important outcomes. Associative theories explain this "learned predictiveness" effect by positing that attention is a function of the relative strength of the association between stimuli and outcomes. In three experiments we show that this explanation is incomplete: learned overt visual-attention is not a function of the relative strength of the association between stimuli and an outcome. In three experiments, human participants were exposed to triplets of stimuli that comprised (a) a target (that defined correct responding), (b) a stimulus that was perfectly correlated with the presentation of the target, and (c) a stimulus that was uncorrelated with the presentation of the target. Participants' knowledge of the associative relationship between the correlated or uncorrelated stimuli and the target was always good. However, eye-tracking revealed that an attentional bias toward the correlated stimulus only developed when it and target-relevant responding preceded the target stimulus. We propose a framework in which attentional changes are modulated during learning as a function the relative strength of the association between stimuli and the task-relevant response, rather than an association between stimuli and the task-relevant outcome. (PsycInfo Database Record (c) 2022 APA, all rights reserved).
... An important factor determining which parts of repeated contexts are learned in contextual cueing is attention (e.g., Jiang & Leung, 2005; see also Beesley, Hanafi, Vadillo, Shanks, & Livesey, 2018;Jiang & Chun, 2001;. In their contextual cueing task, Jiang and Leung (2005) composed the search display of black and white distractors; the target was constantly white throughout the experiment (black for half of the participants; the following sentences apply to white targets). ...
Article
Full-text available
In contextual cueing tasks, participants can use a repeating local context to learn to detect the target, yet most contextual cueing studies have relied on repeating global context properties. We examined whether observers can use local context repetitions in a similar manner as they use global context repetitions. In addition, we examined how reward-predicting context features modulate the use of local and global contexts. Participants searched through contexts in which either the entire context configuration or only a local context around the target repeated, intermixed with novel contexts. Half of the context items appeared in a color signaling either low or high reward. We found that local context repetitions led to comparable benefits in response times and fixation count as global context repetitions did. Surprisingly, reward magnitude did not affect performance in local nor in global contexts. The results suggest that a local chunk of distractors can be used for context learning and attention guidance in a similar manner as the global context configuration. We suggest that the proportion of repeated and novel context trials is crucial for context learning and that our combination of locally and globally repeating contexts provided an environment that facilitated learning in both context types because it allowed predicting the target location from the context in most of the trials.
... This can be tested by recording the eye movements. According to this, the gaze may have shifted to the co-actor's attended context in the joint group while participants in the single group would have moved their gaze mostly to the attended context (Beesley et al., 2018). ...
Article
Full-text available
Regarding the effects of joint action on visual memory, previous research has focused on the memory of a single object that a participant and their co-actor attended together (i.e., a shared situation), while the literature on memory has demonstrated that spatial regularity composed of multiple objects can also be learned. We aimed to examine whether the visuospatial regularity of the co-actor's attended objects could be strongly encoded. We repeatedly presented the same configuration of two targets and two sets of distractors in different colors (i.e., blue and red) to participants. In Experiment 1, pairs of participants simultaneously searched for the same target in the joint group while individual participants searched for the target alone in the single group. As a result, greater facilitation in reaction time was observed in earlier epochs in the joint group, reinforced by the learning of visuospatial regularity, compared to the single group. Experiment 2 examined whether the co-actor's attended context could be strongly encoded although two persons simultaneously searched for different targets (i.e., parallel situation) such that one searched for the blue target and the other for the red target. The results showed no evidence regarding participants' learning visuospatial regularity of the co-actor's attended objects, indicating that co-actor's learning information cannot be shared in this situation. This study revealed that facilitation of visuospatial learning in joint action would require two individuals to attend to the same objects when they perform the task.
... Hence, when the pattern of distractors in which a target is embedded indicates where the target will be located, responding is reliably faster. As with other spatial cuing effects, the savings observed on repeated trials are most likely the result of several learning mechanisms, including the more efficient direction of attention toward the target location (e.g., Beesley et al., 2018;Tseng & Li, 2004), and changes in response thresholds based on the familiarity of the repeated configuration (Kunar et al., 2007;Sewell et al., 2018). Among these processes and consistent with a prediction-based facilitation account, there is evidence that learning improves the efficiency with which the target is processed once it is located, or at least the rate at which evidence for the decision about the target is accumulated (Sewell et al., 2018). ...
Article
Full-text available
One of the mechanisms proposed to underpin perceptual learning is the reduction in salience of predicted stimuli. This reduction is held to affect the representation of (conditioned) stimuli before they have been associated with motivationally meaningful consequences but may also affect (unconditioned) stimuli that automatically elicit responding. The purpose of this article is to review past findings and present new evidence of phenomena across a range of domains that are consistent with the idea that responses automatically triggered by stimulating events will be reduced by prediction. We argue that prediction-based attenuation may serve several adaptive functions. (PsycInfo Database Record (c) 2021 APA, all rights reserved).
... For instance, guiding attention incidentally towards the relocated target facilitated the updating of the previously acquired contextual associations. It thus appears that flexible contextual cueing in fact depends on the attentional engagement towards relevant target-context associations (Jiang & Chun, 2001;Beesley, Hanafi, Vadillo, Shanks, & Livesey, 2018). Conversely, being distracted from the search context reduces the size of contextual cueing (Conci & von Mühlenen, 2009). ...
Article
Full-text available
Visual search for a target is faster when the spatial layout of nontarget items is repeatedly encountered, illustrating that learned contextual invariances can improve attentional selection (contextual cueing). This type of contextual learning is usually relatively efficient, but relocating the target to an unexpected location (within otherwise unchanged layouts) typically abolishes contextual cueing. Here, we explored whether bottom-up attentional guidance can mediate the efficient contextual adaptation after the change. Two experiments presented an initial learning phase, followed by a subsequent relocation phase that introduced target location changes. This location change was accompanied by transient attention-guiding signals that either up-modulated the changed target location (Experiment 1), or which provided an inhibitory tag to down-modulate the initial target location (Experiment 2). The results from these two experiments showed reliable contextual cueing both before and after the target location change. By contrast, an additional control experiment (Experiment 3) that did not present any attention-guiding signals together with the changed target showed no reliable cueing in the relocation phase, thus replicating previous findings. This pattern of results suggests that attentional guidance (by transient stimulus-driven facilitatory and inhibitory signals) enhances the flexibility of long-term contextual learning.
... The search advantage for repeated patterns seems to be largely driven by an improvement in attentional guidance towards the usual location of the target. For instance, participants make fewer fixations in repeated search displays than in new displays (Beesley et al., 2018;Peterson & Kramer, 2001). Similarly, search slopes (i.e., the average amount of time needed to find the target relative to the number of elements in the search display) are usually shallower for repeated than for new search displays (Chun & Jiang, 1998), suggesting that visual search is more efficient in the former. ...
Article
Full-text available
It is usually easier to find objects in a visual scene as we gain familiarity with it. Two decades of research on contextual cuing of visual search show that repeated exposure to a search display can facilitate the detection of targets that appear at predictable locations in that display. Typical accounts for this effect attribute an essential role to learned associations between the target and other stimuli in the search display. These associations improve visual search either by driving attention towards the usual location of the target or by facilitating its recognition. Contrary to this view, we show that a robust contextual cuing effect can also be observed when repeated search displays do not allow the location of the target to be predicted. These results suggest that, in addition to the mechanisms already explored by previous research, participants learn to ignore the locations usually occupied by distractors, which in turn facilitates the detection of targets even when they appear in unpredictable locations.
... Thus, neither focusing on the set of colored items (cf. Beesley, Hanafi, Vadillo, Shanks, & Livesey, 2018;Jiang & Leung, 2005) nor a general attentional weighting of the color dimension (cf. Krummenacher & Müller, 2012) would have been helpful for increasing task performance in novel contexts. ...
Article
Recent work on contextual cueing has shown that reward can facilitate context learning, e.g., when salient but task-irrelevant context features signal reward magnitude. Whether task-relevant context features yield a similar facilitating effect is unclear. Also, emergence and persistence of context learning for contexts associated with different reward magnitudes remains unclear. The present study investigated whether reward increases the speed with which context learning emerges, resulting in an earlier but asymptotically similar contextual cueing effect, or whether reward persistently increases context learning, visible as a larger contextual cueing effect on an asymptotical level. Reward was associated to the predominant orientation of the L-distractors, and the number of context repetitions was increased considerably. Results showed contextual cueing, i.e., faster responses and fewer fixations in repeated compared to novel contexts for all reward magnitudes. Moreover, a high reward led to a more pronounced contextual cueing effect. We developed a model-based approach to explicitly assess the non-linear decline and asymptotic level of the response time curves and to quantify how they were altered by reward. A hierarchical Bayesian parameter estimation revealed that reward decreased the asymptotic level of the repeated contexts’ response times. Our results therefore show that reward leads to a persistent advantage in contextual cueing rather than to earlier but asymptotically similar context learning.
... Studies which examined eye movements in Contextual Cueing seem to support this interpretation. Observers make fewer fixations in repeated compared to novel displays (Beesley, Hanafi, Vadillo, Shanks, & Livesey, 2018;Tseng & Li, 2004). These fewer fixations were in fact associated with longer fixation durations, suggesting that context repetition aids in the planning of eye movement decisions rather than the speed at which items are processed (Zang, Jia, & Müller, 2015). ...
Article
When the spatial configuration of a search display is presented repeatedly, response times to finding the target within that configuration are shorter compared to completely novel configurations, even though observers do not have explicit recognition of the repetition. This phenomenon is known as Contextual Cueing and selective attention is thought to be necessary for the effect. Previous research has suggested that repetition of the context of unattended items does not appear to improve performance; only repetition of attended items does. It has been proposed that this occurs because unattended items are pre-attentively filtered and thus do not contribute to performance. Here we demonstrate that so-called “unattended” items do contribute to performance, just not to contextual cueing. We approach this question from the perspective of the parallel processing of the scene that unfolds at the start of each item and that has been recently modelled by the Target Contrast Signal Theory. We show that the processing time per item during parallel evaluation of the scene is not affected by context repetition, suggesting that the locations of the items rejected in this stage are not integrated into the memory representation underlying contextual cueing. Other alternatives are also discussed.
Article
Despite clear, finite limits on the availability of human attention, limited effort has been directed toward understanding how user attention is allocated in contexts such as digital application environments where consumers frequently confront an abundance of competing alternatives. This is surprising given the growing significance of application platforms, ecosystems, and other application environments that offer seemingly endless choice. In these types of environments the value delivered to users and the success of application developers is often highly dependent on the attention that users allocate to an application after they have adopted it. There is, therefore, considerable value in efforts to better understand what drives the post-adoption allocation of consumer attention toward specific applications. As such, we draw on attention theory to propose a model of attention allocation and then test this model on a dataset that includes over 2200 hedonic desktop applications and more than 300 users. Analysis using fractional regression indicates that user attention is most likely to be allocated to applications that have been used both recently and persistently as well as by applications that are popular and current. Finally, our results highlight the negative impact that a large application portfolio has on the allocation of user attention to individual applications.
Article
Full-text available
Prior research has suggested that attention is determined by exploiting what is known about the most valid predictors of outcomes and exploring those stimuli that are associated with the greatest degree of uncertainty about subsequent events. Previous studies of human contingency learning have revealed evidence for one or other of these processes, but differences in the designs and procedures of these studies make it difficult to pinpoint the crucial determinant of whether attentional exploitation or exploration will dominate. Here we present two studies in which we systematically manipulated both the predictiveness of cues and uncertainty regarding the outcomes with which they were associated. This allowed us to demonstrate, for the first time, evidence of both attentional exploration and exploitation within the same experiment. Moreover, while the effect of predictiveness persisted to influence the rate of novel learning about the same cues in a second stage, the effect of uncertainty did not. This suggests that attentional exploration is more sensitive to a change of context than is exploitation. The pattern of data is simulated with a hybrid attentional model.
Article
Full-text available
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. (PsycINFO Database Record (c) 2014 APA, all rights reserved).
Article
Full-text available
Contextual cuing is the enhancement of visual search when the configuration of distractors has been experienced previously. It has been suggested that contextual cuing relies on associative learning between the distractor locations and the target position. Four experiments examined the effect of pre-exposing configurations of consistent distractors on subsequent contextual cuing. The findings demonstrate a facilitation of subsequent cuing for pre-exposed configurations compared to novel configurations that have not been pre-exposed. This facilitation suggests that learning of repeated visual search patterns involves acquisition of not just distractor-target associations but also associations between distractors within the search context, an effect that is not captured by the Brady and Chun (2007) connectionist model of contextual cuing. We propose a new connectionist model of contextual cuing that learns associations between repeated distractor stimuli, enabling it to predict an effect of pre-exposure on contextual cuing. (PsycINFO Database Record (c) 2014 APA, all rights reserved).
Article
Full-text available
Two studies of human contingency learning investigated the influence of stimulus salience on the cue competition effect of blocking. These studies demonstrated that blocking (defined as a difference in responding to blocked and control cues) was greater for target cues that had high "semantic salience" than those of lower salience. Moreover participants showed weaker responding to high salience blocked cues than low salience blocked cues, but a corresponding difference was not observed for control cues. These findings suggest that the influence of relative salience on associative learning depends on the relative validity of the cues in question. Use of eye tracking in Experiment 2 demonstrated that participants' overt attention to cues was also influenced by both relative salience and relative validity. We describe three associative learning models, based on the attentional theory proposed by Mackintosh (1975), that are able to account for our key findings. (PsycINFO Database Record (c) 2013 APA, all rights reserved).
Article
Visual search through previously encountered contexts typically produces reduced reaction times compared with search through novel contexts. This contextual cueing benefit is well established, but there is debate regarding its underlying mechanisms. Eye-tracking studies have consistently shown reduced number of fixations with repetition, supporting improvements in attentional guidance as the source of contextual cueing. However, contextual cueing benefits have been shown in conditions in which attentional guidance should already be optimal-namely, when attention is captured to the target location by an abrupt onset, or under pop-out conditions. These results have been used to argue for a response-related account of contextual cueing. Here, we combine eye tracking with response time to examine the mechanisms behind contextual cueing in spatially cued and pop-out conditions. Three experiments find consistent response time benefits with repetition, which appear to be driven almost entirely by a reduction in number of fixations, supporting improved attentional guidance as the mechanism behind contextual cueing. No differences were observed in the time between fixating the target and responding-our proxy for response related processes. Furthermore, the correlation between contextual cueing magnitude and the reduction in number of fixations on repeated contexts approaches 1. These results argue strongly that attentional guidance is facilitated by familiar search contexts, even when guidance is near-optimal. (PsycINFO Database Record
Article
This article presents a comprehensive survey of research concerning interactions between associative learning and attention in humans. Four main findings are described. First, attention is biased toward stimuli that predict their consequences reliably (). This finding is consistent with the approach taken by Mackintosh (1975) in his attentional model of associative learning in nonhuman animals. Second, the strength of this attentional bias is modulated by the value of the outcome (). That is, predictors of high-value outcomes receive especially high levels of attention. Third, the related but opposing idea that may result in increased attention to stimuli (Pearce & Hall, 1980), receives less support. This suggests that hybrid models of associative learning, incorporating the mechanisms of both the Mackintosh and Pearce-Hall theories, may not be required to explain data from human participants. Rather, a simpler model, in which attention to stimuli is determined by how strongly they are associated with significant outcomes, goes a long way to account for the data on human attentional learning. The last main finding, and an exciting area for future research and theorizing, is that and modulate both deliberate attentional focus, and more automatic attentional capture. The automatic influence of learning on attention does not appear to fit the traditional view of attention as being either or . Rather, it suggests a new kind of "derived" attention. (PsycINFO Database Record
Article
It has been suggested that attention is guided by two factors that operate during associative learning: a predictiveness principle, by which attention is allocated to the best predictors of outcomes, and an uncertainty principle, by which attention is allocated to learn about the less-known features of the environment. Recent studies have shown that predictiveness-driven attention can operate rapidly and in an automatic way to exploit known relationships. The corresponding characteristics of uncertainty-driven attention, on the other hand, remain unexplored. In two experiments we examined whether both predictiveness and uncertainty modulate attentional processing in an adaptation of the dot probe task. This task provides a measure of automatic orientation to cues during associative learning. The stimulus-onset-asynchrony of the probe display was manipulated in order to explore temporal characteristics of predictiveness- and uncertainty-driven attentional effects. Results showed that the predictive status of cues determined selective attention, with faster attentional capture to predictive than to nonpredictive cues. In contrast, the level of uncertainty slowed down responses to the probe regardless of the predictive status of the cues. Both predictiveness- and uncertainty-driven attentional effects were very rapid (at 250 ms from cue onset) and automatically activated.
Article
Two experiments explore the role of configural representations in contextual cuing of visual search. Repeating patterns of distractors (contexts) were trained incidentally as predictive of the target location. Training these repeating contexts in consistent configurations led to stronger contextual cuing than when contexts were trained in inconsistent configurations. Computational simulations with an elemental associative learning model of contextual cuing demonstrated that purely elemental representations could not account for the results. However, a configural model of associative learning was able to simulate the ordinal pattern of data.