vrije universiteit amsterdam · vrije universiteit. attention to emerging objects . academisch...
TRANSCRIPT
VU Research Portal
Attention to Emerging Objects
Schreij, D.B.B.
2012
document versionPublisher's PDF, also known as Version of record
Link to publication in VU Research Portal
citation for published version (APA)Schreij, D. B. B. (2012). Attention to Emerging Objects.
General rightsCopyright and moral rights for the publications made accessible in the public portal are retained by the authors and/or other copyright ownersand it is a condition of accessing publications that users recognise and abide by the legal requirements associated with these rights.
• Users may download and print one copy of any publication from the public portal for the purpose of private study or research. • You may not further distribute the material or use it for any profit-making activity or commercial gain • You may freely distribute the URL identifying the publication in the public portal ?
Take down policyIf you believe that this document breaches copyright please contact us providing details, and we will remove access to the work immediatelyand investigate your claim.
E-mail address:[email protected]
Download date: 25. Jul. 2021
VRIJE UNIVERSITEIT
Attention to Emerging Objects
ACADEMISCH PROEFSCHRIFT
ter verkrijging van de graad Doctor aan de Vrije Universiteit Amsterdam,
op gezag van de rector magnificus prof.dr. L.M. Bouter,
in het openbaar te verdedigen ten overstaan van de promotiecommissie
van de faculteit der Psychologie en Pedagogiek op vrijdag 13 januari 2012 om 11.45 uur
in de aula van de universiteit, De Boelelaan 1105
door
Daniël Bertus Bernard Schreij
geboren te Amsterdam
promotor: prof.dr. J.L. Theeuwes
copromotor: dr. C.N.L. Olivers
The studies presented in this dissertation were funded by VIDI grant 452‐06‐007 from the Netherlands Organization for Scientific Research (NWO) awarded to dr. Christian Olivers. Beoordelingscommissie Prof.dr. Ed Awh
Prof.dr. Pieter Roelfsema Prof.dr. Werner Schneider Prof.dr. Willem Verwey Dr. Artem Belopolsky Dr. Mark Nieuwenstein
Paranimfen: Paul van Klaveren, MSc Ronald Terpstra, MSc Voorblad: Fotografie: Sam A. Eftegarie, Ir. Handmodellen: Fong Lin, MSc Emma Hameleers, LLM Druk: Off Page, Amsterdam, the Netherlands
Table of contents Chapter 1: Attention and spatiotemporal object representations .................................................. 7
Chapter 2: Object representations maintain attentional control settings across space and time ............................................................................................................... 33
Chapter 3: Object-based attentional control settings depend more on the spatial than the temporal continuity of the object ....................................................... 45
Chapter 4: Object representations maintain attentional control settings for feature information ....................................................................................................... 59
Chapter 5: Abrupt onsets capture attention independent of top-down control settings ........................................................................................................................... 85
Chapter 6: Abrupt onsets capture attention independent of top-down control settings II: Additivity is no evidence for filtering ......................................................... 111
Chapter 7: Abrupt Irrelevant onsets cause inhibition of return regardless of attentional set .............................................................................................................. 137
Summary in Dutch / Nederlandse samenvatting ......................................................................... 149
References .................................................................................................................................... 155
Acknowledgements / Dankwoord ................................................................................................ 165
Curriculum Vitae ........................................................................................................................... 167
Author publications ...................................................................................................................... 168
Chapter 1 Attention and spatiotemporal object
representations
ATTENTION AND SPATIOTEMPORAL OBJECT REPRESENTATIONS
8
Introduction
Around 400 BC, the Greek philosopher Plato proposed his Theory of Forms, which states
that "Forms" (or "Ideas"), and not the material world of change known to us through
sensation, constitute the highest and most fundamental kind of reality. Almost three
millennia have passed since, but many contemporary philosophers and scientists still
agree with Plato’s view that our mind operates on representations of observed objects.
The brain can perform mental operations (e.g. assess an object’s function) on these
representations and uses them to construct a (spatial) model of its direct environment.
Such a model helps the brain to maintain a coherent experience of its surroundings and
also enables it to stay aware of objects of which it (temporarily) has no sensory input.
For instance, imagine that you see a coffee cup in front of you. If you completely turn
your gaze away from the cup, you will still be aware that the cup exists, despite the
absence of its sensory input. Object representations further aid the brain in keeping
track of objects that are temporarily occluded by other objects. For example, when you
witness a cat moving behind a tree, you will probably expect that same cat to reappear a
moment later (assuming it did not come to a halt behind the tree) and at the same time
know that it is not a different cat than the one you saw disappear earlier. In short, the
brain stays aware of objects by maintaining representations of them based on the
assumption that objects are always continuous in both space and time.
The main focus of this thesis lies on how this spatiotemporal continuity of an
object, or the lack thereof, affects the deployment of attention. The first part of the
presented research investigates if knowledge obtained during the first encounter with
an object influences the way it is attended to during subsequent encounters (Chapter 2
to 4). Think of the cup of coffee from before, which was standing on a table in front of
you. Would you expect to find this cup of coffee at the same spot of the table after you
temporarily lost sight of both objects? And what would your expectations be about the
location of the coffee cup when you are presented with a table that looks identical to the
one before, but of which you know it is actually a different table? We investigated if and
how the spatiotemporal continuity of an object affects the preservation of attentional
control settings that one has potentially established for internal properties or locations
of the object. The second part of this thesis addresses whether objects that suddenly
appear in the visual field involuntarily capture attention (Chapter 5 to 7). The sudden
appearance of an object can be seen as an event that violates spatiotemporal continuity
CHAPTER 1
9
and such an event is already known to attract the attention of an observer without his
intention (Jonides & Yantis, 1988; Yantis & Jonides, 1984). We investigated whether this
is also the case when a completely irrelevant newly appearing object has to compete for
attention with a cue that has task-relevant properties. Before discussing the main
research in detail, I first review concepts and background literature that are relevant to
the current work in the next sections of this introduction.
Object representations
Object correspondence through spatiotemporal continuity
For a coherent visual experience, it is important to keep track of objects in the direct
vicinity through space and time. Kahneman, Treisman and Gibbs (1992) were among the
first to provide evidence that episodic representations of objects assist in this process. In
their object preview paradigm, they presented the observer with two boxes located at
opposite sides of the screen. Initially, these boxes would each contain a letter that
disappeared after a short preview period, after which both boxes smoothly moved to
new locations. Once they arrived there, a letter reappeared in one of the boxes and
observers indicated if this letter was also present during the preview period. Responses
were faster when the letter matched one of the previewed ones than when it was a new
letter. More important however was that this repetition benefit was greater when the
matching letter also appeared in the same box (now at a different location), as compared
to the other box. Note that “sameness” here is determined by the motion trajectory of
the box – in other words, its spatiotemporal history – and does not refer to its
appearance (i.e. shape or other features). Kahneman, et al. (1992) proposed that we
create an episodic representation, or ‘object file’, for each object we observe. Object files
contain information about properties of their corresponding physical objects (e.g. color,
shape, or letter identity), and are preserved across space and time. When the
spatiotemporal history of the stimulus suggests that the same object is encountered, its
represented properties are readily available, and if they match the actual visible
properties, responses will be facilitated. In contrast, responses are slowed when the
visible object no longer matches the content of its associated object file.
It has been proposed that spatiotemporal continuity is in fact the most important
contribution to perceiving object constancy (Scholl, 2001, 2007), more so than other
properties such as object identity or surface features. Pylyshyn and colleagues
ATTENTION AND SPATIOTEMPORAL OBJECT REPRESENTATIONS
10
(Pylyshyn, 2000, 2001; Pylyshyn & Storm, 1988) for instance showed that observers are
able to track several identical objects (which thus could not be distinguished by surface
features) on basis of their spatiotemporal properties. Observers were presented with a
search field in which a large number of white dots were continuously moving in random
directions. At the beginning of a trial, one or more of these dots were briefly highlighted
and observers were instructed to track these for a few seconds. After this period, a
single dot was highlighted again and observers had to indicate whether this was one of
the dots they were supposed to track or not. Notably, observers were able to track up to
4 or 5 objects with high accuracy.
Furthermore, Mitroff and Alvarez (2007) used a variant of the object preview
paradigm to demonstrate that continuity in the surface features of an object (e.g. texture,
color or shape) from one instance to the other plays no significant role in same-object
perception, but the spatiotemporal continuity of the object does. When the
spatiotemporal constraints of an object were violated (e.g. when it jumped from one
location to another instead of gradually moving between them), no same-object benefits
were found, even when the object retained the same physical appearance. Vice versa, a
change in features did not affect same object benefits as long as spatiotemporal
consistency was preserved.
Spatiotemporal representations further enable us to keep track of objects that
temporarily disappear from vision through occlusion. This is nicely illustrated by the
tunnel effect (Burke, 1952; Michotte, 1963), which occurs when a moving object
disappears behind one end of an occluder (the ‘‘tunnel’’) and a different moving object
appears at the other end a moment later. When the second object emerges at about the
time and place that the first object should have emerged, assuming it underwent
continuous motion behind the occluder, people tend to perceive both objects as a single
instance. This effect has appeared to be very robust and as long as spatiotemporal
parameters are correct, people can tolerate large changes in color, shape or other
features to the second object (Flombaum & Scholl, 2006). Infants seem to be even less
susceptible to such changes than older children and adults, who will at a point infer that
the second object is different from the first when changes become too large (Bower,
1967; Bower, Broughto.J, & Moore, 1971; F. Xu & Carey, 1996, 2000; Y. D. Xu & Chun,
2006).
CHAPTER 1
11
Contents of object representations
A topic that has been less frequently investigated is which information about an object is
actually maintained with its representation. The object-preview paradigm of Kahneman
et al. (1992) suggests that a representation preserves information about the identity of a
response feature, enabling one to later respond faster to objects that remain consistent
with their object files. This has often been corroborated by other studies using variants
of the object preview paradigm (Kruschke & Fragassi, 1996; Mitroff, Scholl, & Wynn,
2004; Noles, Scholl, & Mitroff, 2005). However, most of these object-reviewing studies
have used displays in which a target object only contained a single response-related
feature. In fact, in many of these studies the previewed and target-defining features
were identical to the response features, such that same object benefits may have been
response-based rather than have a perceptual or attentional basis. Moreover, such
response-based effects have also been found for objects that were merely repeated on
basis of their features and were not spatiotemporally connected. Hommel and
colleagues for instance suggested that one integrates specific object features and
responses that co-occurred in a certain ‘time segment’ (usually a single trial) into an
episodic structure they coined an event file (Hommel, 1998, 2004, 2007; Hommel,
Musseler, Aschersleben, & Prinz, 2001). Keizer, Colzato and Hommel (2008) illustrated
this concept by showing that observers ‘integrate’ their responses to an object with the
direction they see it move. When observers were presented with a same object as in a
trial before that also moved in the same direction, there were response benefits.
However, if either the identity of the object or its moving direction was altered,
responses were slower compared to when both the object’s identity and its moving
direction were new. Keizer et al. therefore reasoned that when the current set of
features match those stored in an event file, previous response information is retrieved
from it and facilitates current responses. Conversely, the information from an event file
can interfere with responses if it only partially corresponds with the current situation.
This study thus shows that response information can be tied to specific objects of which
the correspondence is determined on the basis of feature context rather than
spatiotemporal continuity.
We are aware of only one previous study that looked at pure perceptual effects in
object representations. With a novel paradigm, Yi, Turke-Brown, Flombaum, Kim, Scholl
and Chun (2008) examined the effect of spatiotemporal continuity on the perseverance
ATTENTION AND SPATIOTEMPORAL OBJECT REPRESENTATIONS
12
of complex stimuli, in this case faces. They displayed a pillar on the left and on the right
side of the screen, which each served as an occluder. A trial consisted of two consecutive
events each involving the emergence of a face from behind a pillar. The second face
could either be the same as or different from the first, and it could appear from behind
the same or the other pillar. Using fMRI, Yi et al. found that brain activity in the right
fusiform face-area (FFA) decreased when the face had the same appearance in both
events, which was not that surprising as it is common for neurons to show such signs of
habituation after sustained stimulation. Importantly however, the habituation effect was
stronger when the same face had also re-emerged from the same pillar as it had just
disappeared behind. In other words, the neural coding of the face as one and the same
depended on whether its spatiotemporal history had been violated or not, suggesting
that face information is preserved with representations. This study largely inspired the
work described in the following section of this thesis, as it shows that an object
representation maintains more than simple feature or response information. We
addressed the question whether complex attentional control settings for an object are
also preserved.
Object representations and attentional control settings
The first part of this thesis focuses on the question if attentional selection settings that
have been established for an object can be stored with its spatiotemporal
representation. If we for instance see a platter with snacks and find bread at its top left
corner, do we intuitively expect to find the bread at this same location again the next
time we encounter this same platter? More technically, if a certain part or feature of an
object has been selected before, do these attentional selection settings then persist
across subsequent spatiotemporal changes that the object undergoes and affect
selection again later on?
As illustrated in Figure 1, our main paradigm consisted of two display objects
which were for the largest part hidden behind walls positioned at each edge of the
screen. On a trial, one of these displays would shift to the center of the screen revealing a
visual search array with multiple distractors and one target. This made selection
necessary as there was competition between multiple elements within the object. After
response, the display shifted back behind one of the walls which was not occupied by the
CHAPTER 1
13
other display object. Crucially, in the following trial either the same or the other display
would shift to the center of the screen.
Chapter 2 describes an experiment in which the potential same-object benefit on
selection was measured by repeating the target’s position over trials. As Maljkovic and
Nakayama (1996) demonstrated, observers unintentionally tend to start search at the
location which they previously found the target, leading to response benefits when the
target is indeed found there again the next time. If these selection settings are
furthermore connected to an object representation, observers should be even faster
when this target not only appears at the same location as on the previous trial, but also
on the same object. In other words, intertrial repetitions of the target location should
result in greater performance benefits when the spatiotemporal dynamics suggest that
the search display is the same object as the one which appeared the trial before, as it
would appear from its spatiotemporal trajectory.
The results indeed provided evidence that spatial selection settings for internal
object properties are stored with the object’s representation. Observers were found to
respond faster when the location of the target was repeated, but even more so when it
also appeared in the same object as in the preceding trial. This corroborates earlier
findings that repeating target locations facilitates search over trials (Maljkovic &
Nakayama, 1996) as do entirely repeated display configurations (Chun, 2000; Chun &
Figure 1: Example of the stimulus displays shown in a trial sequence of Chapter 2.
Until response
150 ms
150 ms
ATTENTION AND SPATIOTEMPORAL OBJECT REPRESENTATIONS
14
Jiang, 1998) and extend them by showing that such effects are partly object-bound.
Observers thus are strongly inclined to start search at the same spot as where they
previously found the target, especially when the current display object was
spatiotemporally connected to the previous one.
Chapter 3 looked further into the factor of spatiotemporal continuity itself,
which of course consists of the two separate components space and time. We examined if
same-object perception relies more heavily on only one of the two, or really requires
continuity in both dimensions to occur. This time, there was only one display present
which would shift to the opposite side of the screen on each trial. Centered in the middle
of the screen was a stationary narrow wall, which would temporarily and completely
occlude the display when it passed behind it. Importantly, the display could pass behind
this occluding wall in four possible manners: its motion trajectory was either continuous
(i.e. emerge from behind the occluder at the expected place and moment), temporally
discontinuous (emerge at the expected location, but a different moment), spatially
discontinuous (emerge at the expected moment, but a different location), or
discontinuous in both space and time. The benefits for a repeated target location were
found to be larger when the display object followed a spatially coherent trajectory,
regardless of whether there was a temporal discontinuity. This suggests that persistence
of attentional settings within an object is mostly determined by continuity in the spatial
dimension.
Chapter 4 further investigated if other attentional biases than for the target’s
location are preserved based on spatiotemporal object history. If so, similar object
benefits as for a repetition of the target’s location should be found for the repetition of
other internal object properties. The first experiment was identical to the experiment
described in Chapter 2, except for the addition of an occasional singleton color
distractor. If object representations preserve inhibitory tags for distracting elements in
an object, then one would expect a distractor to interfere less when it reappears in the
same object than in the other one. Distractor interference indeed was lessened when it
had been present the trial before but especially when it reappeared inside the same
object. Repetition of the distractor’s location however did not lead to any further
benefits, suggesting that attentional settings for a distractor are not maintained on the
basis of its spatial location.
CHAPTER 1
15
In a second experiment we examined whether selection settings for the target’s
features also survive occlusion of the object it is displayed on. On any trial, either or both
the target shape and color could be the same or different from the previous trial. A
change of the target’s features had a larger impact on RTs if the display object was
repeated from the previous trial, indicating that object representations can indeed hold
feature-based attentional settings. A mismatch between the observed target features
and those stored in the object representation consequently results in extra response
penalties. Notably, the effect was only apparent for a repetition of shape, but not for
color. As the target was defined by its shape and color was an irrelevant dimension, a
bias might exist toward maintaining only task-relevant information about the target.
In all previous experiments, the two display objects both had the exact same
appearance of a basic black slate. Still, participants were perfectly able to distinguish
them on the basis of their spatiotemporal history and link attentional control settings to
a specific representation. In a third experiment we investigated how much same-object
perception depends on, or can be disrupted by a change of, the exterior appearance of
the objects. The stimuli and task were similar to the first experiment, but this time the
search array was presented inside the display area of a mobile device, which could
either be a white IPod or a black Nokia N95. Between trials, when the display object was
completely occluded by one of the walls, this exterior could change causing the object to
have a completely different appearance when it re-emerged. As the specific object on
which the array was presented could also still switch from trial to trial, this experiment
pitted the spatiotemporal continuity of objects directly against their appearance to
determine which of these two factors forms the more important basis for same-object
perception. According to Flombaum and Scholl (2006) or Mitroff and Alvarez (2007) a
change to an object’s external features should not disrupt same-object effects as objects
are mainly individuated using their spatiotemporal properties. Alternatively, the
findings of Moore et al. (2010) or Moore and Enns (2004) would predict that a drastic
change of the object’s features would shatter same-object perception. We found
attentional selection to be only modulated by the object’s spatiotemporal history and to
remain unaffected by changes to or repetitions of its exterior. Apparently, the sameness
of an object is more predominantly designated by spatiotemporal continuity which thus
also as the only factor determines whether previously established attentional control
settings are retrieved.
ATTENTION AND SPATIOTEMPORAL OBJECT REPRESENTATIONS
16
Taken together, Chapters 2 to 4 show that an observer is able to maintain
attentional selection settings he has established for the internal properties of an object
with its spatiotemporal representation. Information of both the target’s location and its
defining features can survive with the object through occlusion, as long as its
spatiotemporal continuity suggests it to be a single instance. Whether an object is
perceived as the same seems to mostly depend on its continuity in the spatial dimension
and does further not rely on the object’s exterior features. A change to the appearance of
a spatiotemporally continuous object therefore did not have an influence on the reuse of
previously stored attentional control settings.
The above and other studies thus suggest that an object is mainly individuated
and tracked on the basis of its spatiotemporal continuity and not by its appearance.
Moore, Stephens and Hein (2010) reasoned that if features are irrelevant for the
maintenance of object representations, same object effects should still occur when
objects change color during smooth movement in a typical object-preview paradigm.
Instead, they found the opposite to be the case and argued that feature changes interfere
with same object perception, even when an object’s spatiotemporal continuity remains
intact. Moore et al. hence concluded that beside spatiotemporal factors, features also
play an important role in establishing and maintaining object correspondence. Other
studies supporting this notion reported that corrections of misdirected eye movements
to objects that are displaced during saccadic suppression are done on the basis of the
objects’ features (Hollingworth, Richard, & Luck, 2008; Richard, Luck, & Hollingworth,
2008), which suggests that visual working memory also maintains object
representations on the basis of features. Furthermore, a phenomenon known as change-
related persistence (Moore & Enns, 2004) shows that an abrupt change to a moving
stimulus can cause it to be seen as two separate objects: the original object and the
changed object. If the object underwent no changes in the same setting, it was perceived
as a single instance. These studies thus all show that disrupting the features of an object
can disrupt the representation of it as a single object, despite spatiotemporal
consistency, and therefore argue that the role of features in maintaining object
correspondence should not be underestimated.
There however is a possibility that not the changes per sé severed object
correspondence, but rather the abruptness with which they occurred. Abrupt luminance
or polarity changes have been shown to potentially signal the appearance of a new
CHAPTER 1
17
object (Rauschenberger, 2003b) which hence requires a new object representation to be
created (Kahneman, et al., 1992). This process likely requires the object to be evaluated
by attention first, which is hence automatically allocated to it upon its appearance
(Yantis & Jonides, 1984). Attentional capture by the appearance of new objects is the
topic of the second part of my thesis.
Capture of attention by new objects
The sudden appearance of a new object, often signaled by an abrupt onset of its features,
has shown to be an event that attracts the attention of an observer independent of his
goals and intentions (Posner, 1980; Todd & Van Gelder, 1979; Yantis, 1993). This
phenomenon is referred to as attentional capture. Before we continue, it is useful to first
give better description of attention and what is meant with attentional capture.
The world we live in provides such a vast amount of information that the
capacity-limited human brain would quickly become overwhelmed if it had to process
everything that came in through the senses. It therefore needs to make a selection from
its sensory input. This process is referred to as attentional selection. Only the objects or
locations we attend to are processed in more detail and reach our conscious experience
or awareness (Simons & Chabris, 1999). The locus of attentional focus in the
environment frequently changes to obtain a more complete representation of it. When
the properties of the stimulus features in the environment determine what attention is
focused on, orienting is said to be stimulus-driven, or ‘bottom-up’. This exogenous
attention is very fast, crude and not under our voluntary control (Broadbent, 1958;
Treisman & Gelade, 1980; Treisman & Sato, 1990). Top-down (or endogenous) orienting
on the other hand refers to the goal-driven way in which one directs his attention.
Contrary to bottom-up allocation, top-down allocation is “voluntary”, under control of
the observer and led by his goals, beliefs or intentions (Yantis, 1993). Top-down
attention engages and disengages slower from locations than its bottom-up counterpart
(Theeuwes, Kramer, & Atchley, 2000) and operates serially, as opposed to the presumed
parallel nature of bottom-up processing (Egeth & Yantis, 1997).
It is generally agreed that whether or not an object gets represented in visual
cortex depends on the collaborative outcome of both processes (Cave & Wolfe, 1990;
Wolfe, 1994). Brain areas located later in the visual stream are shown to affect the
processing of information by areas earlier in the stream via re-entrant processing (Di
ATTENTION AND SPATIOTEMPORAL OBJECT REPRESENTATIONS
18
Lollo, Enns, & Rensink, 2000), but before these higher areas can do so, they first have to
receive information from the earlier areas. Therefore, the initial evaluation of a scene is
thought to be stimulus-driven, and at this stage only the most salient objects in the
visual field win the competition for representation (Donk & van Zoest, 2008; Theeuwes,
2010), which consequently also results to the suppression of other, less salient, objects.
Endogenous processing kicks in shortly after, during which the goals and intentions, also
called the attentional control settings (ACS), of the observer come into play. Top-down
biases can subsequently increase the strength of relevant, but less salient stimuli, if
salience caused an irrelevant object to win the competition (Desimone & Duncan, 1995;
Reynolds & Desimone, 2003; Tsotsos et al., 1995). In those cases in which an object is so
salient that it is represented at the expense of other stimuli, regardless of the volitional
goals of the observer, one speaks of attentional capture.
The capture of attention is often operationalized as speeded search performance
when an otherwise non-predictive stimulus happens to be the target of a visual search.
In other words, a stimulus is said to capture attention when it is searched with priority,
even when it is irrelevant to the task. Jonides and Yantis (1988; see also Yantis &
Jonides, 1984; Yantis & Jonides, 1990) demonstrated that onsets enjoy such
prioritization with an experiment in which observers were instructed to find a pre-
specified letter among a set of other letters. Some letters were initially masked by figure-
eight placeholders that were present from the start of the trial. These letters, which
constituted the no-onset stimuli, were revealed by removing the placeholders’ line
segments that camouflaged their identity (a method of presentation first introduced by
Todd & Van Gelder, 1979). Simultaneously with the revelation of the no-onset letters,
one other letter appeared at a previously blank location in the visual field and
constituted the onset stimulus. Importantly, the onset stimulus was the target only at
chance level and hence gave the observers no reason to start search at the onsets as they
were not predictive of the target’s location. When the target was one of the no-onset
letters, search times increased linearly with the number of items present in the display
(also called set size). For targets appearing at the onset location however, search times
remained constant regardless of the set size. The search slopes (response time as a
function of number of items present in the display) for onset targets were found to be
shallow, which is considered to be a hallmark for attentional capture. Theeuwes (1994b)
further found that abrupt onsets interfere with search when observers are explicitly
CHAPTER 1
19
instructed to look for a color singleton and thus an onset is not part of their attentional
set. In this same situation, onsets are even shown to capture one’s gaze (also called
oculomotor capture; Theeuwes, Kramer, Hahn, & Irwin, 1998).
Possible explanations for attentional capture by onsets
So what makes an onset so special that it receives attentional priority? It has been
suggested that this tendency is a result of evolutionary development. As onsets signal
the appearance of a new entity that could be a dangerous predator, it is best for
organisms to attend to it as quickly as possible to take appropriate action. In other
words, onsets signal potentially behaviorally urgent events (Franconeri & Simons,
2003).
Others have proposed that it is the luminance increment accompanying an onset
which triggers the shift of attention. The magnocellular visual pathway of the brain is
known to be sensitive to high temporal frequencies. This neurological system is thought
to be very influential in signaling the location to which attention should be directed
(Breitmeyer & Ganz, 1976). The luminance increment of an onset might activate visual
pathways sensitive to high temporal frequencies, which are then responsible for
directing attention to the onset. In support of this notion, Franconeri, Hollingworth and
Simons (2005) demonstrated that removing the luminance increase accompanying an
onset makes it lose capturing power. Likewise, Enns, Austen, Di Lollo, Rauschenberger
and Yantis (2001) showed that certain types of luminance contrast changes to old
objects are already sufficient to capture attention (albeit to a lesser degree).
A profound alternative explanation was given by Kahneman and Treisman
(1984), who proposed that it is not the physiological properties of an onset which
attract attention, but rather that the onset signals the presence of a novel perceptual
object for which a new ‘object file’ needs to be created. Since this process requires
attention to evaluate the object first (although see Kahneman, Treisman, & Burkell,
1983), attention is automatically allocated to any new object which appears in the visual
field. Yantis and Hillstrom (1994) indeed demonstrated that a new object was still
prioritized if its appearance was not accompanied by luminance transient. The presence
of a new perceptual object seemed enough to produce the flat search slopes typical for
attentional capture when the onset was the target. In addition, Enns, et al. (2001)
showed that search for targets indicated by luminance changes was far less effective
ATTENTION AND SPATIOTEMPORAL OBJECT REPRESENTATIONS
20
than targets signaled by the sudden appearance of the new object. They found that
search slopes were steeper for targets designated by luminance increments, in contrast
to the search slopes of new object targets, which remained zero. Furthermore,
Franconeri, Simons and Junge (2004) showed that a new object still captured attention
when it appeared during a period of saccadic suppression. Even though observers thus
not saw the appearance (and hence a luminance increase) of the object itself, the sudden
presence of the object still elicited an involuntary shift of attention. This was further
corroborated by Brockmole and Henderson (2005) who found that observers also made
significantly more eye-movements to new objects that were inconspicuously added
during presentations of natural scenes.
Are abrupt onsets unique in their ability to capture attention?
Feature singletons also capture attention
Some have argued that involuntary shifts of attention can also be elicited by so-called
feature singletons, which are objects that possess unique properties in relation to the
other objects surrounding them, such as a distinct color or shape. In his famous study,
Theeuwes (1992) demonstrated that the presence of a color singleton interferes with
search when another item is the target. He presented observers with a circular array of
circles within which observers had to locate a unique diamond shape and respond to the
orientation of a line inside it. Whenever one of the circles carried a unique color,
response times were higher than when all items were colored the same. Theeuwes
inferred that the color singleton must have captured attention, thereby delaying
selection of and response to the target. Indeed, the diamond target could also be
regarded as a singleton itself because of its unique shape, but a singleton in the color
dimension is simply regarded as being more salient and therefore demands higher
priority in attentional processing (Theeuwes, 1992). Even after extensive practice and
explicit instructions to ignore the color singleton, search performance remained poor
when the distractor singleton was present, suggesting that observers simply were
unable to ignore a singleton and prevent their attention to be captured by it. This finding
has often been corroborated by many other studies (Kim & Cave, 1999; Kumada, 1999;
Lu & Zhou, 2005; Nothdurft, 1993, 2006; Pashler, 1988; Theeuwes, 1995b, 2010)
CHAPTER 1
21
Only onsets can capture attention independent of attention set
Yantis and Egeth (1999) however argued that in most studies where search was
disrupted by a singleton distractor, it was in one or another way task relevant. Either the
target was a singleton itself, or the locations of the singleton distractor and target
coincided at higher than chance level, giving observers an incentive to initially attend
the distractor because of its predictive ability. Singleton distractors were often, as they
called it, nominally irrelevant. Yantis and Egeth (1999) stated that the many cases in
which a singleton did influence search, one could better speak of a top-down
“attentional misguidance” by the singleton than of stimulus-driven attentional capture.
According to Yantis (1993) one can only speak of true stimulus-driven attentional
capture, when the investigated distractor stimulus property is independent of either the
defining or the reported attribute of the target (notions first introduced by Duncan,
1985). The defining property designates the specific feature that defines the target and
the observer has to look for. For instance, if one has to look for a red letter among
distractors with a different color, the defining attribute is the color red. The reported
attribute of the target forms the basis for response. A reported attribute could for
instance be the identity of a target letter. Yantis and Egeth (1999) henceforth made sure
that these criteria of Yantis (1993) were met in their experiments, in which observers
had to indicate the presence of a vertical among randomly oriented line segments and
one line could have a unique color. The singleton distractor and target only coincided at
chance level and the singleton shared neither a reported nor a defining attribute with
the target, as the target was defined by its orientation and the singleton was defined in
the color dimension. The resulting response times showed no benefits when the target
also was the color singleton, nor did the presence of a singleton distractor inflict a cost.
Yantis and Egeth (1999) hence argued that if a highly salient feature singleton is not part
of the observer's attentional control setting, that singleton does not necessarily control
the deployment of attention. Franconeri and Simons (2003) further demonstrated that
onsets are one of few dynamic events that can produce shallow search slopes, and that
color singleton targets are unable to do so if color is unpredictive in an exact same task
environment. This and other studies thus suggest that new-object onsets are truly
unique in their capability to elicit purely stimulus-driven shifts of attention (Enns, et al.,
2001; Franconeri & Simons, 2003; Jonides & Yantis, 1988).
ATTENTION AND SPATIOTEMPORAL OBJECT REPRESENTATIONS
22
One may then wonder, what top-down set in Theeuwes (1992) paradigm allowed
the singleton distractor to capture attention. Bacon and Egeth (1994) proposed that
observers chose to adopt a singleton detection mode, in which they strategically use the
singleton status of the target to guide search. As a consequence however, any item which
possesses unique features is bound to capture attention, thus also singleton distractors.
Conversely, when the target is more difficult to find because it possesses no salient
properties, observers have to fall back on feature detection mode, in which attention
traverses locations that match the task-relevant visual features (e.g. a red element or a
diagonal bar) in a serial fashion. Indeed, when Bacon and Egeth (1994) increased the
variety of shapes in Theeuwes’ (1992) search task, thereby revoking the target’s
singleton status, a color singleton distractor did not interfere with search anymore (but
see Theeuwes, 2004). It is however safe to assume that in most studies showing capture
by onsets (i.e. Jonides & Yantis, 1988; Yantis & Hillstrom, 1994) observers were always
operating in feature detection mode, as the target was defined by a specific letter
identity which possessed no unique properties compared to the other letters present.
Still, onsets were prioritized in search, providing further evidence that attentional
capture by onsets occurs regardless of observers’ search strategies.
Others have proposed that objects can only capture attention if they fall within
the attentional window (Belopolsky & Theeuwes, 2010; Belopolsky, Zwaan, Theeuwes, &
Kramer, 2007; Theeuwes, 1994a, 2010). When search requires attention to be focused
on a small region, the attention window is said to be narrow and attentional capture is
often precluded because salient distractors are likely to fall outside the window. A small
attentional window is thus in a sense similar to feature detection mode, as both assume
effortful search leaves little resources for salient distractors to capture attention.
Conversely, if the task permits or induces attention to be distributed across the visual
field, the attentional window is said to be wide, granting irrelevant salient elements the
opportunity to capture attention. This distributed state is thus in a way comparable to
singleton detection mode, in which selection is guided by unique features across the
visual field. Contrary to the search modes of Bacon and Egeth (1994) however, it has
been shown that attentional capture by onsets can indeed be prevented when attention
is highly focused on a specific location, albeit an (upcoming) target has to appear there
with high spatial certainty (Theeuwes, 1991b; Yantis & Jonides, 1990). In this sense, the
CHAPTER 1
23
modulation of the attentional window size seems to be the only top-down measure one
has against attentional capture by abrupt onsets (Theeuwes, 2010).
Capture of attention by onsets is always contingent on attention set
Some however argue that any form of attentional capture is always completely
contingent on attentional control settings, even capture by onsets (Folk & Remington,
1998; Folk, Remington, & Johnston, 1992). In this view, attentional allocation is fully
under top-down control and objects or events will never capture attention if the
observer is not actively looking for them or has no intrinsic reasons to attend them. To
this end, Folk et al. (1992) devised a pre-cue paradigm in which observers had to
respond to the identity of a target letter (X or =) which was either designated by a
unique red color or by an onset. The stimulus displays used in this paradigm are
depicted in Figure 2. The target letter could appear in one of four placeholder boxes that
were positioned on the corners of an imaginary diamond shape. In the color singleton
condition, the red target appeared together with three white letters which filled the
remaining boxes (hence giving it a singleton status) and in the onset condition the target
appeared as the only white character. 150 ms before presentation of the target display,
one of the boxes was briefly cued. Importantly, this cue could also be defined either by
Figure 2: Example of a trial sequence and the stimulus displays used in the Folk et al. (1992, 1998) pre-cue paradigm. A target, which could be defined as an onset or color singleton, was preceded by a cue of the same or the other dimension. It was found that a cue only had effect when it was of the same dimension as the target. The cue is hence assumed to only capture attention if it contains a task-relevant property. In reality the background was black, black lines where white and the gray elements were red.
ATTENTION AND SPATIOTEMPORAL OBJECT REPRESENTATIONS
24
an onset, in which case only one of the boxes was surrounded by four white dots, or by a
color singleton, in which case one box was surrounded by four red and the other boxes
by four white dots. The cue and target could equally likely appear at any of the four
possible locations and only coincided at chance level. Participants were informed about
the cue’s uninformativeness and were advised that it would be best to ignore it. The
critical finding was that when the dimensions of the cue and target matched and when
they appeared at the same location, there was a considerable response benefit, which
was absent when their dimensions differed; even when the cue was an onset which
according to most studies should have captured attention. Thus, although participants
knew the cue was uninformative, they were not able to ignore it when it possessed a
property that also defined the target. This led Folk et al. (1992) to conclude that a
stimulus is only able to capture attention when it has properties that are contained in
one’s attention set. They called this phenomenon contingent attentional capture (CAC).
How could it be then, that so many studies did find attentional capture by
irrelevant abrupt onsets even when observers were not looking for them? Folk and
Remington (1998) reasoned that in most studies onsets were always task-relevant, in
the sense that the presentation of the target display as a whole was often accompanied
by an onset (or luminance increase), which thus can be regarded as a signal that search
can start. Henceforth, people likely establish a default attention set for onsets, in the
absence of any other clear top-down control settings. Gibson and Kelsey (1998) made a
similar argument that observers’ attention is prone to be captured by any change in the
display (or display wide features) that might indicate the appearance of the target.
When the appearance of the target display in their study was for instance signaled by an
onset and color change, this latter dimension also captured attention, while it did not do
so if an onset was the only signaling attribute. Gibson and Kelsey (1998) hence warned
that many studies claiming to find pure stimulus-driven attention capture might be
confounded by goal-directed processes. However, there have been studies that
eliminated such transients which warn for the target’s appearance and still found
stimulus-driven attention capture. As discussed before, Franconeri et al. (2004) made
sure an observer never directly saw the display change, by presenting the target display
during a period of saccadic suppression. They instructed participants to briefly make an
eye movement to a point below the screen, and at the peak level of this saccade the
target display appeared. Even though participants thus never witnessed the display
CHAPTER 1
25
change, irrelevant salient items still possessed the power to capture attention. In their
second experiment, Franconeri et al. (2004) presented the target display to participants
without telling them what to search for, so that they could not form an attentional set.
After presenting the whole display, participants were told the target’s identity by means
of a voice prompt (see also Belopolsky, Theeuwes, & Kramer, 2005). The revelation of
the target was thus never signaled by a display change, giving observers no incentive to
adopt a default attention set for dynamic events. Even so, when one item in the array
underwent a large contrast change, it still received attentional priority.
Rather than being the result of the strategies of an observer, it can be argued that
the contradicting findings of contingent capture on the one hand and stimulus-driven
capture on the other are attributable to critical differences in experimental design. For
one, whereas onsets mostly appear simultaneously with the target display in studies
supporting stimulus-driven capture, in a typical contingent capture paradigm the onset
cue precedes the target display with at least 150ms. It has been shown that attention can
quickly disengage from a location to which it made an exogenous shift (Kim & Cave,
1999; Lamy, Tsal, & Egeth, 2003; Theeuwes, et al., 2000). This gives attention ample
opportunity to have disengaged from the cued location once the target appears,
especially when the cue is defined by color, making the onset an irrelevant property. On
the other hand, when the target is also defined by an onset, disengagement from an
onset cue naturally would be more difficult as the onset property has become task
relevant and attention shifts might be partly under control of the slower top-down
component. In this case, attention is thus more likely to linger at the cued location,
resulting in the cue validity effect common to the contingent capture paradigm.
Theeuwes (1991a) indeed showed that interference effects dramatically increase when
distractors and targets changed roles from trial to trial. Due to the large uncertainty this
brings about, substantial top-down processing is needed to determine that a distractor
which captured attention is not the target, causing the disengagement of attention to
take longer.
Filtering costs
Folk and Remington (1998) however warned that the simultaneous presentation of a
target and distractor may lead to a non-spatial filtering operation, of which the effects
can easily be confused with those of attentional capture. The filtering account was first
introduced by Kahneman, et al. (1983), who showed that response times increase
ATTENTION AND SPATIOTEMPORAL OBJECT REPRESENTATIONS
26
linearly with the number of items that are concurrently presented with a target, even
when these do not resemble the target at all (e.g. a random dot pattern when the target
is a letter). Kahneman et al. claimed this happens because the appearance of new objects
necessitates the creation of new object files, which process competes with that of the
reallocation of attention and thereby delays it. According to this line of reasoning,
attention thus goes directly to a target, but when it appears together with other items,
attention takes more time to do so. One might quickly notice that both the filtering and
stimulus-driven capture accounts attribute the RT costs incurred by new objects to the
creation of corresponding object files. The critical difference however is that the filtering
account assumes attention is never allocated to the object in doing so (and hence is a
non-spatial process), while the stimulus-driven capture account assumes that a
representation for an object cannot be created without attention evaluating it first,
which is why the object captures attention in the first place (and hence is a spatial
process). Both accounts thus differ vastly in the assumed underlying processes that are
involved in the processing of new objects, but make similar phenomenal predictions
concerning RTs, making it difficult to dissociate them empirically or determine mutual
falsification. Folk and Remington (1998) stated that attentional capture can only be
dissociated from filtering costs (as these are very fleeting according to them) by
separating the presentation of the distractor and target in time, which is exactly what
they did in their contingent capture paradigm. They claimed that many studies finding
response delays while simultaneously presenting a target and distractor actually
demonstrate the occurrence of filtering costs rather than attentional capture.
As can be seen, controversies still remain regarding the effects on attention
caused by the sudden appearance of new objects and how these effects are modulated
by top-down control. This thesis contributes new insights on this matter by directly
pitting stimulus-driven attention capture by new-object onsets against contingent
capture. Importantly, we did this within the CAC task devised by Folk et al. (1992) so
that any results could not be attributed to the use of different paradigms.
Onsets capture attention independent of control settings
The second part of the research presented in this thesis is dedicated to further explore
the ability of abrupt onsets to capture attention when observers are strongly induced to
adopt an attention set for a different feature dimension. To this end, we utilized the
CHAPTER 1
27
contingent capture pre-cue paradigm (Folk & Remington, 1998; Folk, et al., 1992), but
made some important changes. First, both the cue and target were always presented as a
color singleton. Observers had to find a unique red character (again an “X” or “=”) among
three similar white characters and hence were induced to have a strong attention set for
the color red throughout each experiment, making the onset dimension completely
irrelevant. The cue thus also consisted as one set of four red dots again, among other
sets of white dots surrounding the other boxes. Second, the onset distractor did not
appear as a cue surrounding one of the boxes briefly before the target was presented,
but concurrently and as a new object (a placeholder box containing a character)
occupying a previously empty space. This assured that attention did not have the time to
disengage from the onset’s location before the target appeared, which might have been
the case in the classic contingent capture studies of Folk and colleagues, in which any
capture effects might have been absorbed by the inter-stimulus interval (ISI) between
cue and target. Thus, by combining the specifics of well-established contingent capture
and onset capture paradigms, we investigated how irrelevant onsets affected the
allocation of attention, when it is under strict top-down control.
In the experiments presented in Chapter 5, the onset distractor was always
present on half of the trials. Experiment 1 and 2 both demonstrated that the onset
distractor interfered equally in both invalid and valid cue conditions. Experiment 2
controlled for competition by other display changes that could have lessened the
capture effect by the onset, but its results were no different from Experiment 1. Both
Experiment 1 and 2 however left the possibility that the RT costs inflicted by an onset
distractor were the result of a non-spatial filtering operation rather than attentional
capture. Experiment 3 therefore applied the identity intrusion method (Theeuwes,
1995b) to further dissociate these two options. Basically, this method relies on the
assumption that the information at attended locations always gets processed at a high
(semantic) level. Henceforth, if the onset distractor captured attention, it can be
expected that the identity of its character is processed. One may then assume this
character will interfere more with response to the target, if their identities are
incongruent than when they are the same. Since the filtering account assumes that
attention does not visit the locations at which a new object appears, it would predict that
the identity of the distractor would have no effect on response times. These findings
favored the stimulus-driven capture account: after changing the onset distractor
ATTENTION AND SPATIOTEMPORAL OBJECT REPRESENTATIONS
28
character to carry one of the possible target identities (X or =, instead of just a neutral
“O” as in the first two experiments), we indeed found that RT costs were higher when
target and distractor identity were incongruent than when the distractor congruent.
Finally, Experiment 4 controlled if the onset did not capture attention because
participants were searching in singleton detection mode. When each no-onset distractor
character was given a unique color (instead of all being white) and observers were thus
forced to adopt a feature search mode to find the red target, the onset still captured
attention.
Folk and colleagues (Folk & Remington, 1998; Folk, Remington, & Wu, 2009)
nevertheless proposed an alternative explanation for the identity intrusion effect that
we found, namely that parallel processing of the target and distractor was the source of
interference. Lavie (1995) demonstrated that displays which pose a low perceptual load
leave observers resources to process all items in the visual field in parallel, during which
items whose identities compete with that of the target cause interference which is
reflected in prolonged response times. Importantly, parallel processing is spatially non-
specific, which implies that (spatial) attention does not need to visit the locations of the
items for their identities to be processed. In contrast, if the complexity of a display
increases (e.g. due to the addition of extra distractors) search will claim more resources
which prevents parallel search and thereby eliminates the interference effects.
According to Folk et al. (2009), the displays used in our Experiment 3 posed a low
perceptual load, because the color target and onset distractor both were very salient and
therefore ‘popped out’. Henceforth, they claimed that the onset interfered because it was
processed in parallel with the target and not because it captured spatial attention.
However, this perceptual load account was recently challenged by Tsal and
Benoni (2010a, 2010b, 2010c) who showed that the degree of interference by certain
distractors rather depends on the dilution of the display. Dilution is dissociable from
perceptual load and refers to the mere presence of different neutral letters whose
features are visually similar to those of the distractor. To demonstrate this concept, Tsal
and Benoni (2010c) made participants search for a pre-specified letter in the presence
of a highly conspicuous distractor letter. Low-diluted displays only contained these two
letters and therefore simultaneously posed a low perceptual load. Conversely, in high-
diluted displays more distractor letters were present, which either resembled the target
(causing the perceptual load to be high) or possessed a different color (giving the target
CHAPTER 1
29
a unique color, resulting in a low load). The important finding was that a conspicuous
incongruent distractor letter interfered when the display dilution was low, but did not
do so anymore in highly diluted displays, regardless of perceptual load. From a similar
perspective, one could argue that the displays used in Experiment 3 constituted a low
perceptual load as the target and distractor both were very salient, but were diluted by
the presence of the other no-onset distractors. Since the dilution account would in this
case predict that the identity of the onset distractor does not modulate interference, our
finding favors the assumption that the onset distractor captured spatial attention.
The research described in Chapter 6 was conducted in response to Folk et al.
(2009), who argued that filtering costs were still the best explanation for the effects
elicited by the onset distractor in Chapter 5. If the effects of onset presence and color-
based cuing both originate from a same process of spatial capture, these two factors
would have been expected to show an under-additive relationship (Sternberg, 1969).
After all, if capture by the onset nullified any effects of the cue, attention should always
move directly from onset to target and not be affected by cue validity. Filtering and
contingent capture are more compatible in this sense, because attention is expected to
only go to the item that matches the attentional set for color and not go to the location of
the abrupt onset, which hence causes a delay only because it triggers a non-specific
filtering operation. Because capture and filtering operations are presumed to take place
during independent stages of processing, additive effects of color-based cuing and onset
interference would be expected. As pointed out by Folk et al. (2009), this was exactly
what was found in the experiments described in Chapter 5: The interference caused by
the onset presence was equally strong for trials with valid or invalid color cues.
Therefore, if one accepts that the color cue captures spatial attention, one can only infer
that the onset cannot do so as well.
Experiment 1 of Chapter 6 investigated if such a pattern of under-additivity is a
useful diagnostic to determine if two consecutive events separated shortly in time both
captured attention. Assuming it is, then the underadditive relationship should also be
apparent for two events that capture attention contingent on attention set. We took the
Folk and Remington (1998) paradigm and briefly made one of the distractor boxes red
between presentation of the first color cue and the target. This distractor cue was
always invalid. In other words, the displays contained a red cue (valid or invalid), then
potentially a red distractor (invalid), and finally, the target display with a red target.
ATTENTION AND SPATIOTEMPORAL OBJECT REPRESENTATIONS
30
According to the contingent capture theory, the new red distractor should capture
attention away from the initially cued location, since red is what the observers are
looking for, and capture by red is moreover what explains the original cuing effect. If
such capture by a new distractor did indeed erase or reduce all prior cuing effects, the
red distractor should attenuate the benefits of a valid color cue, resulting in under-
additivity. If, on the other hand, there is still residual activation of the first cue strong
enough to affect the reorienting toward or the identification of the target, we may again
observe additivity between cue validity and the presence of the red distractor. The latter
was found to be the case, showing that even two events which must have captured
attention following the contingent capture hypothesis did not show underadditive
effects. This makes underadditivity an unreliable diagnostic for attentional capture.
Clearly, neither an onset distractor nor a contingent distractor was powerful
enough to erase the effects of the prior cueing effect. Would there exist a process at all
which is able to do so? It was hypothesized that only an event which invokes a process
involving a strong stimulus-driven attention shift, like an onset, and at the same time
possesses a task-relevant feature (making disengagement from its location harder) is
able to hold attention sufficiently long for the earlier cueing effect to dissipate.
Experiment 2 thus reintroduced the onset distractor with the goal to find out if the
‘capturing power’ of this distractor increased if it also carried task-relevant features.
Instead of having its normal white bounding box, the onset possessed a red bounding
box each other block. With such a distractor, underadditivity was indeed found between
the cueing and onset effect, suggesting that some events are able to overcome prior
cueing effects, given that these events possess characteristics that elicit both stimulus-
driven and goal-driven attention shifts.
Experiment 3 finally investigated whether responses would benefit when the
onset occasionally was the target itself. If present, the onset was a distractor the
majority of cases and the target was among the no-onset characters. However,
incidentally the red target letter was located inside the onset, albeit at less than chance
level, giving observers no incentive to purposefully start search at the onset.
Nevertheless, responses to onset targets were of the same magnitude as those for a
validly cued no-onset target, indicating that the onset captured attention.
Finally, Chapter 7 provided further evidence that spatial attention had really
been allocated to the location of the onset target described in the previous chapter. A
CHAPTER 1
31
phenomenon which is often used as a spatial marker for attention is Inhibition of Return
(IOR). IOR is best described as a slowing in the allocation of attention to a previously
visited location when the time between first and subsequent attentional shift exceeds
300 ms, only when the first shift of attention to this location was involuntary or
exogenous (Klein, 2000; Posner & Cohen, 1984). IOR is therefore considered a hallmark
for stimulus-driven capture as it is rarely found after voluntary orientation to a location
(although see Berlucchi, 2006). We thus added a condition in which the ISI between cue
and target display was 1000ms instead of 150ms. In the long ISI condition, the onset, if
present, still appeared after 150 ms, but its identity remained hidden until the revelation
of the target display. If attention was initially captured by the onset but shortly after
disengaged from its location, people should be slower if the target appeared at the onset
location a moment later after all. If the onset on the other hand was suppressed by a
non-spatial filtering process, such signs of IOR are not expected as the filtering account
claims attention did not go to the onset upon its appearance. While the onset target as
before enjoyed a response benefit in the short ISI conditions, responses were indeed
prolonged when the ISI was extended, indicative of IOR. As filtering would not have
predicted such a pattern, the results once again flavored stimulus-driven attention
capture by the onset.
Conclusion
In the first part of this thesis, we showed that an observer maintains attentional
selection settings for the internal properties of an object with its spatiotemporal
representation. These selection settings involve information about the location and
features of a target and influence how attention treats this object the next time it is
encountered, as long as its spatiotemporal continuity indicates that it is a single
instance. Whether an object is perceived as the same, seems to depend more on its
continuity in the spatial than the temporal dimension. It does furthermore not rely on
the object’s exterior features, as a change to the appearance of a spatiotemporally
continuous object did not eliminate the reuse of previously stored attentional control
settings.
The second part of this thesis provided evidence that suddenly appearing objects
enjoy a special status for attentional allocation as their spatiotemporally discontinuous
nature demands priority, even when observers are not set to look for them, or are
ATTENTION AND SPATIOTEMPORAL OBJECT REPRESENTATIONS
32
actively looking for something else. The only top-down control we thus seem to be able
to exert on attentional capture by onsets is adjusting the size of the attentional window,
as onsets have been shown to only capture attention if they fall within this area of focus.
Chapter 2
Object representations maintain attentional control settings across space
and time
Schreij, D. & Olivers C.N.L (2009) Object representations maintain attentional control settings across space and time
Cognition, 113(1), 111-116
OBJECT-BASED ATTENTIONAL CONTROL SETTINGS
34
Abstract
Previous research has revealed that we create and maintain mental representations for
perceived objects on the basis of their spatiotemporal continuity. An important question
is what type of information can be maintained within these so-called object files. We
provide evidence that object files retain specific attentional control settings for items
presented inside the object, even when it disappears from vision. The objects were
entire visual search displays consisting of multiple items moving into and out of view. It
was demonstrated that search was speeded when the search target position was
repeated from trial to trial, but especially so when spatiotemporal continuity suggested
that the entire display was the same object. We conclude that complete spatial
attentional biases can be stored in an object file.
CHAPTER 2
35
Introduction
Visual attention is the mechanism by which we select relevant information from a rich
visual world. Evidence so far indicates that attention can be directed not only to specific
locations (Posner, Snyder, & Davidson, 1980) or features (e.g. Wolfe, 1994), but also to
entire objects (see Scholl (2005), for a review). In most of the classic studies on object-
based selection, objects were presented abruptly on the screen and, either after a set
time or after a response, disappeared instantly. Researchers then assume that each trial
provides an independent performance measurement of the condition at hand. In the real
world, objects often behave quite differently. Rather than covering a well-defined
instance in time, objects typically appear and disappear gradually (e.g. by occlusion,
moving in and out of the periphery, or looming). In other words, objects have a history
across space and time. The present study investigates the effect of this spatiotemporal
history of an object on the attentional selection settings applied to that object. More
specifically, it addresses the question as to whether an object representation includes
the way in which attention has previously treated that object, and whether this
information is then preserved across space and time, even when an object temporarily
disappears from vision.
Theorists have argued that for an object representation to be preserved across
space and time, some type of indexing or tokenization of objects is necessary (e.g.
Pylyshyn, 2001), A token is an episodic representation that allows the observer to refer
to “that object, then and there” and thus to track it across space and time (Pylyshyn &
Storm, 1988). Kahneman, Treisman and Burkell (1983) referred to such temporary
episodic representations as object files. Evidence for object files comes from the object
reviewing paradigm (Kahneman, et al., 1992), in which observers typically are
presented with a preview of two objects, each containing a letter. The letters disappear
and the objects then move to a new position, after which one of the initial letters re-
emerges in either one of the objects. The task is to identify the letter. Identification is
speeded when the target letter emerges in the same object as it did before, even though
this object has now changed position. This effect was referred to as “the object-specific
preview benefit”, and we refer to it here as the same object benefit.
An important question is what information about an object is maintained across
space and time. In other words, what properties are bound to an object file or index?
Pylyshyn and co-workers (Pylyshyn, 2001) have argued that indexes are primitive and
OBJECT-BASED ATTENTIONAL CONTROL SETTINGS
36
preconceptual, and should thus contain little information on specific object features, as
long as spatiotemporal continuity is preserved (see also Mitroff & Alvarez, 2007).
However, Flombaum and Scholl (2006) demonstrated that, when a moving object is
briefly occluded by another display element, a change to the object’s color or shape is
better detected when the object moves in a spatiotemporally coherent fashion,
suggesting a link between spatiotemporal and object feature representations. In
addition, the same object benefit in the classic object file studies imply that information
about the identity of the response feature (the letter inside the object) must have been
preserved across the object’s translation (Gordon & Irwin, 1996, 2000; Kahneman, et al.,
1992).
In the present study we investigate the possibility that entire attentional
selection settings as applied to a specific object are maintained in its mental
representation. Imagine the case when only a part of an object is relevant to an observer,
for example the door handle when approaching one’s car, or one’s favorite flavor when
presented with a tray of different tea bags. If a certain part or feature of an object has
been selected before, do these selection settings then survive with the object across
subsequent spatiotemporal changes, and affect attention later on? So far, this question
has been largely unexplored. Object reviewing studies have typically used displays with
only a single response-related feature inside the specified target object. This means that
there was no competition for selection in these displays, and hence no way of testing
whether specific selection settings were preserved with the object. In fact, in many
object reviewing studies the previewed and target-defining features were identical to
the response features, such that same object benefits may have been response-based
rather than perceptual in nature (although this was certainly not the case for all object
file studies, see Kruschke & Fragassi, 1996; Noles, et al., 2005).
In our experiment, the objects of interest consisted of entire visual search
displays containing more than one item, but only one target. Thus, there was
competition between multiple elements within the object, and hence the need for
selection (of the search target). Figure 3 illustrates the procedure. Within a trial, one of
two search displays gradually emerged from behind one of four walls on each side of the
screen, and, after response, moved back to a random unoccupied location behind one of
the walls. Crucially, the search display could emerge from the same side as it
disappeared to on the previous trial and would thus constitute the same object in terms
CHAPTER 2
37
of spatiotemporal continuity, or it could come from a different side and would thus
constitute a different object. A potential same object benefit on selection was then
measured through repeating the target position. Following the object reviewing logic, if
observers selected a certain location within a search display on trial n, they may be
inclined to select the same location again on trial n+1, especially so if the search display
is perceived as being the same object as before. If so, intertrial repetitions of the target
location should result in greater performance benefits when the spatiotemporal
dynamics suggest that the search display is the same object as that appeared the trial
before.
The rationale for this manipulation was inspired by Yi, Turk-Browne, Flombaum,
Scholl and Chun (2008). They asked participants to respond to particular faces while
measuring fMRI activity in the fusiform face area (FFA). The faces appeared from pillars
on either side of the screen. They found that if the same face was repeated from one trial
to the next, FFA activity was reduced, in line with general habituation effects. The
Until Response
Figure 3: Example of stimulus display for a typical trial in Experiment 1. For printing purposes, these images were converted to black and white, but the walls were brick-colored and the search elements were green or red. A typical trial started with a screen with both search displays hidden behind the walls for 1000 ms. Then, over a time course of 150 ms, one of the search displays slid to the center of the screen, and with this exposed the search array. Participants were to report the identity of the letter in the diamond shape. When the participant had given a response, the search display shifted back behind one of the unoccupied walls (within another 150 ms).
150 ms
150 ms
OBJECT-BASED ATTENTIONAL CONTROL SETTINGS
38
important finding was that this reduction was greater when the repeated face appeared
from behind the pillar where it had disappeared on the previous trial. In other words,
FFA activity was modulated by the spatiotemporal history of the face, in addition to its
identity. Whereas that study looked at identity processing as a function of
spatiotemporal object history, here we looked at spatial selection as a function of such
object history.
Experiment
Participants searched for a green diamond inside an array of green circles (as in
Theeuwes, 1992) and responded to the identity of the letter inside it (N or M). By
independently varying the response feature from the target-defining feature, we could
decouple potential response-based benefits from spatial selection-based benefits. The
two main manipulations involved the repetition of the target location, and the
spatiotemporal dynamics of the entire visual search display, both from trial to trial. We
expected that repeating the target location would result in faster response times relative
to a location change, but especially so when the dynamics of the display suggested that
observers were dealing with the same object. For this purpose, after each trial, the
display moved behind one of four walls. On the next trial, the display could then emerge
from behind the same wall (same object) or a different wall (different object).
Previous studies have found evidence for the binding of object features to a
specific motion direction (Keizer, et al., 2008). However, these experiments did not
assess the effects of object binding on selection settings. Furthermore, we were not
interested in the binding of such settings to motion direction per se (which could be
regarded as just another object feature), but to a pure object-type representation that is
maintained across space and time. Thus, in our experiment, whether or not a display
constituted the same object (in spatiotemporal terms) was uncorrelated with whether it
moved in the same direction as it did on the previous trial, by not making objects
consistently return to their starting location. If benefits occur independently of an
object’s motion direction, they can unambiguously be ascribed to the object’s
spatiotemporal characteristics.
CHAPTER 2
39
METHOD
Participants
Sixteen students from the Vrije Universiteit of Amsterdam participated and received
course credit or money in return. They were between 18 and 29 years of age (average
22), reported normal or corrected-to-normal vision and no color blindness.
Stimuli and apparatus
The experiment was run in a dimly lit cubicle, on a PC with a 19” CRT screen (1024 x 768
resolution, 120 Hz), viewed from 75 cm. Stimulus presentation and response recording
were done in E-prime 1.2 (Psychological Software Tools, 2003). Images of a wall were
positioned on each side of the display (covering 7.4 deg visual angle on the left and right
sides, 2.5 deg at the top and bottom), revealing a central square grey background area
(CIE(.289,.316), 4.9 Cd/m2). Each of two square panels (11.0 deg visual angle, black
background, white border, 39.0 Cd/m2) containing the search arrays was initially placed
behind one of the walls, revealing only a single edge. To enhance the “objectness” of the
panels, a thin grey shadow was drawn behind it at the right and bottom sides, 3.2 Cd/m2.
The search elements were positioned on an imaginary circle with a radius of 14.2 deg
visual angle, with at the center a white fixation cross. The distractors were green
(CIE(.280,.623), 7.2 Cd/m2) circles with a diameter of 3.1 deg. The target was a green
diamond with a diameter of 3.7 deg to equate for area. All shapes contained a grey “M”
or “N”.
Design and Procedure
The main factors were: 1) Display Identity (3 levels: the search array appeared on the
same object, which had the same motion direction before search as on the previous trial;
the search array appeared on the same object, which had a different motion direction
than on the previous trial, or it appeared on a different object altogether). 2) Target
Location (target position was the same as or different from the previous trial). 3)
Response (again same as or different from previous trial). All factors were randomly
mixed within blocks. Participants were instructed to look for the diamond and to
respond as fast as possible to the letter inside it by pressing N or M on the keyboard,
while making as few errors and eye movements as possible. They practiced 96 trials, and
then completed 10 blocks of 96 trials each. After each block, the participant received RT
OBJECT-BASED ATTENTIONAL CONTROL SETTINGS
40
and accuracy scores, followed by a break. The experiment lasted approximately 45
minutes.
On each trial, one of the two search displays slid to the middle of the screen from
behind a wall within 150 ms. To prevent possible benefits from previewing the target
location while the display was moving, the display only contained distractor shapes until
it reached the center of the screen. One of the distractors then changed into a target and
RT measurements started. Given the speed of the motion and the abruptness of the halt,
these changes were not consciously noticeable. Participants reported whether the
diamond shape contained an “M” or an “N” character by pressing the corresponding key
on the keyboard. The display stayed on until response and then slid back behind any of
the three unoccupied walls, again within 150 ms. This could be the wall it had emerged
from, or a different wall. Correct and incorrect responses were followed by a high and a
low sound respectively. A new trial then started after 1000 ms.
Results
Trials of which RTs deviated more than 2.5 SDs from the mean (3% of the total) were
removed. Erroneous responses (5%) were analyzed separately. The means of the
remaining RTs were subjected to a 3-way repeated-measures ANOVA with Display
Identity (same object from same side; same object from different side or different
object), Target Location (same or different) and Response (same or different) compared
to previous trial as factors.
Participants were significantly slower when the location of the target had
changed, Target Location, F(1,15) = 106.48, p < .001. They were also slower when the
required response changed, Response, F(1,15) = 7.97, p < .01. There was no main effect
of Display Identity (F < 1, p = .348).
Important for the present investigation, the interaction between Display Identity
and Target Location was significant, F(1,15) = 5.13, p < .05. As shown by Figure 4a,
observers benefited from a repeated target location, but especially so when the search
array appeared on the same object. Figure 4d shows the same object benefits underlying
this interaction. For repeated displays, there was a significant same object benefit when
the same object had the same motion direction [t(15) = 2.10, p < .05] and when it had a
different motion direction [t(15) = 2.28, p < .05] as on the previous trial (with no
difference between these conditions, t(15) = 0.80, p = .434). Thus, there was a clear
CHAPTER 2
41
object-based benefit, regardless of the object’s departure side and motion direction.
When the search display configuration changed over trials, there were no significant
same object costs or benefits.
D)
A) B) C)
620
640
660
680
700
720
740
760
780
Mea
n C
orre
ct R
T (m
s)
same display, same sidesame display, different sidedifferent display
0%
4%
8%
12%
same differentTarget Location
Erro
r
620
640
660
680
700
720
740
760
780
Mea
n C
orre
ct R
T (m
s)
same display, same sidesame display, different sidedifferent display
0%
4%
8%
12%
same different
Response
Erro
r
620
640
660
680
700
720
740
760
780
Mea
n C
orre
ct R
T (m
s)
Same responseDifferent Response
0%
4%
8%
12%
same different
Target Location
Erro
r
-15
-10
-5
0
5
10
15
20
25
30
same different Target location
Sam
e O
bjec
t Ben
efit
(ms)
same sidedifferent side
Figure 4: Results of the Experiment. Mean response times and error rates for A) same and different target locations, as a function of display identity: the search display could have the same start side as in a previous trial, or a different start side, or be a different object altogether. B) same and different responses, as a function of target location (same or different as previous trial). C) display identity as a function of response (same or different as on previous trial). Panel D) shows the same object benefits underlying the interaction of panel A, i.e. the RT differences between same (with both types of start side) and different search display, for the different feature changes in display configurations. The error bars display the standard errors of the mean same object benefit.
OBJECT-BASED ATTENTIONAL CONTROL SETTINGS
42
There also was an interaction between Target Location and Response, F(1,12) =
40.92, p < .001 (Figure 4b). Participants were particularly fast when both the display
configuration and the response feature repeated, as compared to when only one or none
of these properties repeated. There was no interaction between Display Identity and
Response Feature, F(1,15) = 2.30, p = .123 (Figure 4c).
The error pattern followed that of the RTs and there was no evidence for a
speed/accuracy trade-off. The only significant effect was the interaction of Target
Location and Response, F(1,15) = 32,73, p < .001. Relatively more errors were made
when the location changed but the response stayed the same.
Discussion
The results support the hypothesis that spatial selection settings for internal object
properties are stored in the object’s mental representation. Repeating the target location
led to greater benefits when these repetitions occurred within the same object, than
when they occurred within different objects – as defined by the spatiotemporal
trajectory of moving back and forth from behind an occluder. This indicates that
observers have a stronger tendency to look at the same spot as where they previously
found the target, when the current display object was spatiotemporally linked to the
previous object.
The results also indicate that the repetition benefits are object-specific (i.e. tied to
the spatiotemporal history of the display object) rather than feature-specific (i.e. tied to
a single motion direction). Same object benefits occurred both for displays that had
returned to their old position (and hence came from the same direction) and for displays
that had not returned to their old position (and thus re-emerged from a new direction).
The fact that we retain location-specific control settings for an object is
consistent with findings of Kristjansson, Mackeben and Nakayama (2001). They asked
observers to search for a target, which was positioned within a cue consisting of two
horizontal lines. Participants were slower when the position of the target within the cue
changed between trials. If we regard the cue as an object within which the target could
appear, these findings corroborate ours, in that search performance deteriorated when
the new spatial relationship between the target and cue object was inconsistent with the
one participants had learned on the previous trial.
CHAPTER 2
43
Studies by Tipper and colleagues also suggest the linking of attentional control
settings to object files (Tipper, Brehaut, & Driver, 1990; Tipper, Grison, & Kessler, 2003;
Tipper, Weaver, Jerreat, & Burak, 1994). These studies showed that inhibition of
distractor objects (as measured through negative priming and inhibition of return)
survives across spatiotemporal changes of these objects. This at least suggests that
object files can include an inhibitory tag. However, note that in the Tipper studies only a
single, response-related feature was used inside the target object, and hence it may have
been the response associated with the object that may have been inhibited, and not
selection settings per se, as applied to information contained within the object file.
It has been demonstrated before that target location repetitions facilitate search
from one trial to the next (Maljkovic & Nakayama, 1996). In addition, repetition of the
entire spatial configuration of a display also facilitates search (Chun, 2000; Chun & Jiang,
1998). In the current study we extend these findings by showing that such repetition
effects are partly object-bound. Search was speeded when target location was repeated,
but more so when presented on an object that could be spatiotemporally linked to the
search display of the previous trial. The interaction between target location and
response repetition we found here is also found more often in the intertrial priming
literature (Hommel, et al., 2001; Huang, Holcombe, & Pashler, 2004). A repetition of both
factors typically yields fastest performance, whereas a repetition of either property may
lead to even slower performance than when both change. In any case, none of the
response effects here interacted with display identity suggesting that the specific
responses are not linked to the object. Our results show that the attentional selection
settings prior to response selection, in contrast, are linked to the object representation.
Thus, using the objects’ spatiotemporal consistency, we can keep its mental
representation active when it disappears from vision and retrieve useful selection
information from this representation when it reappears.
Acknowledgments
This work was supported by VIDI grant 452-06-007 from the Netherlands Organization
for Scientific Research (NWO) awarded to CNLO. We thank the reviewers for their very
helpful comments and suggestions.
Chapter 3 3: Object-based attentional control settings
depend more on the spatial than the temporal continuity of the object
Schreij, D. & Olivers C.N.L (submitted) Object-based attentional control settings depend more on the spatial than the temporal continuity of
the object
OBACS DEPEND MORE ON SPATIAL THAN TEMPORAL CONTINUITY
46
Abstract
It has been shown that specific attentional control settings for an object are preserved
by binding them to the object’s spatiotemporal history. The current study investigates if
the persistence of these attentional control settings is more constrained by spatial or by
temporal factors. Observers searched a display for a target shape among multiple
distractors. The target location could repeat or alternate from trial to trial. This visual
search display was integrated in a moving object that would then pass behind a
stationary occluder before the next visual search display appeared. The occluded motion
trajectory was either continuous (i.e. emerge from behind the occluder at the expected
location and moment), temporally discontinuous (emerge at the expected location, but a
different moment), spatially discontinuous (emerge at the expected moment, but a
different location), or discontinuous in both space and time. Search target repetition
benefits were larger when the display object followed a spatially coherent trajectory,
regardless of whether there was a temporal discontinuity. We conclude that persistence
of attentional settings within an object is mostly determined by continuity in the spatial
dimension.
CHAPTER 3
47
Introduction
For a coherent visual experience, it is important to keep track of objects through space
and time. Kahneman, Treisman and Gibbs (1992) provided evidence that episodic
representations of objects assist in this process. In their classic object preview paradigm
they presented participants with two boxes, each containing a unique letter. They
allowed participants to preview the display for about a second, after which the letters
disappeared and both boxes started to move to new locations. Once both objects had
arrived at their destination, a single letter reappeared in one of them, and participants
had to respond to its identity. Responses were found to be faster when the target letter
matched one of the previewed ones than when it was a new letter. Importantly, this
repetition benefit was greater when the matching letter appeared in the same box (now
at a different location), as compared to the other box. Note that “sameness” here is
determined by the motion trajectory of the box – in other words, its spatiotemporal
history. Kahneman, et al. (1992) proposed that we create an episodic representation, or
‘object file’, for an object we observe. Object files contain information about properties
of their corresponding physical objects (e.g. color, shape, or letter identity), and are
preserved across space and time. When the spatiotemporal history of the stimulus
suggests that the same object is reencountered, its represented properties are readily
available, and if they match the actual visible properties, responses are facilitated. In
contrast, responses are slowed when the visible object no longer matches the content of
its associated object file.
Later work has indicated that the information that can be tied to an object file
ranges from simple features, such as color and shape, to more complex representations
such as faces and abstract concepts (Flombaum & Scholl, 2006; Gordon & Irwin, 1996,
2000; Kruschke & Fragassi, 1996; Noles, et al., 2005; Yi, et al., 2008). Recently, Schreij
and Olivers (2009; submitted) have provided evidence that attentional control settings
are also preserved across spatiotemporally coherent instances of objects. They devised a
paradigm in which two visual search displays were hidden behind walls flanking the
screen. On each trial, one of the displays would move to the middle of the screen.
Participants were instructed to find a diamond-shaped target among circle-shaped
distractors, and respond to the letter printed inside (cf. Theeuwes, 1992). After
response, the display would move back behind one of the walls. There were two
important manipulations: First, from one trial to the next, either the same display object
OBACS DEPEND MORE ON SPATIAL THAN TEMPORAL CONTINUITY
48
would re-emerge again from behind the wall it had just disappeared behind, or the other
display object would now slide to the middle of the screen. Second, within the emerged
display object, the target was either positioned at the same location as on the previous
trial, or at a different location. Previous work has shown that observers show a
preference for the location where they previously found the target (Maljkovic &
Nakayama, 1996). The hypothesis was that if such spatial selection biases are at least
partially stored with a representation of the entire display object, then this information
may subsequently be retrieved when the spatiotemporal dynamics of the object suggest
that the same search display has re-appeared (i.e. by emerging from the wall the
previous display had disappeared behind), and thus lead to further benefits. This was
indeed found: Search on repeated target trials was especially fast when the display re-
emerged from its last-known position. This provides evidence that relevant target
location information is stored with the spatiotemporal representation of an object.
Follow-up work has indicated that not only target location, but also target feature
information is stored with the object (Schreij & Olivers, submitted).
The present study investigates which aspect of the stimulus is the most important
for the preservation of attentional control settings. In earlier work, Mitroff and Alvarez
(2007) had already demonstrated that continuity in surface features of the object (e.g.
texture, color or shape) from one instance to the other does not necessarily result in a
same object benefit. Instead, the perception of an object as one and the same appears to
depend primarily on its spatiotemporal continuity, as the same object benefit
disappeared whenever the displacement of an object violated natural spatiotemporal
constraints. Likewise, Schreij and Olivers (submitted) found visual search benefits for
spatiotemporally coherent displays regardless of a change in identity of the display
objects – that is, whether they were part of a black phone or a white music player from
instance to instance. The importance of spatiotemporal continuity is also nicely
illustrated by the tunnel illusion (Burke, 1952), which occurs when a moving object
disappears behind an occluder (the ‘‘tunnel’’), followed by the appearance of a different
moving object at the other end. When the second object emerges at about the time and
place that the first object would be expected to emerge, people tend to perceive both
objects as a single instance that underwent continuous motion behind the occluder, in
some cases even when the object changes its appearance. This effect has previously been
used to show that both spatial and temporal continuity play a role in the preservation of
CHAPTER 3
49
object features such as shape and color (Flombaum, Kundey, Santos, & Scholl, 2004;
Flombaum & Scholl, 2006; we will return to this in the discussion). Here we use a variant
of the tunnel paradigm to investigate if spatial continuity, temporal continuity, or both
are necessary for the preservation of attentional control settings across two instances of
an object.
Examples of the stimulus displays are shown in Figure 5. In a visual search
display, participants had to locate a diamond shaped target among circle distractors and
identify the letter (M or N) inside it. The display was positioned at one of four quadrants
of the screen, in the center of which there was a brick wall serving as an occluder. After a
response (i.e. between trials) the entire search display object moved to the other side of
the screen, passing behind the brick wall in the process. When the search display re-
emerged from the other side of the wall, the target could have the same or a different
location than before the object disappeared. Target location repetitions were expected
to result in faster response times (Maljkovic & Nakayama, 1996). The crucial
manipulation was that the displacement of the display object to the other side of the
occluding object was not always spatiotemporally coherent and could breach spatial and
temporal constraints. Concretely, the search display could 1) move to its destination
smoothly and uninterrupted (continuous), 2) pause its movement for a full second while
occluded by the wall (temporally discontinuous), 3) re-emerge from a different vertical
position than when it disappeared (spatially discontinuous), 4) or be subject to both
these disruptions concurrently (spatiotemporally discontinuous).
On the basis of our previous findings (Schreij & Olivers, 2009), we expected to
find a same object benefit for the continuous condition, when both spatial and temporal
continuity are preserved, relative to the spatiotemporal discontinuous condition. That is,
search target repetition effects were expected to be larger when the display object was
seen as one and the same as the previous search display, then when it was seen as
different. Our interest lies in which conditions of discontinuity are able to eliminate this
benefit. If an object only needs to be continuous over time, a spatial discontinuity should
not eliminate same object benefits. Alternatively, if the object only needs to be spatially
continuous, a temporal discontinuity should not eliminate the same object benefit. The
final possibility is that the maintenance of an object representation requires continuity
in both space and time, in which case neither the spatial or temporal discontinuity
conditions should show benefits for a repeated target location. The results suggest that
OBACS DEPEND MORE ON SPATIAL THAN TEMPORAL CONTINUITY
50
1000 ms
Temporally discontinuous
Spatially discontinuous
Spatiotemporally discontinuous
Figure 5: Illustration of the stimulus displays in a typical trial. In the Continuous condition, the search display moved behind the wall to the other side of the screen in one smooth movement. In the Temporally discontinuous condition, the display stopped moving for 1 second when completely occluded by the wall. In the Spatially discontinuous condition, the display moved smoothly to the other side of the screen again, but had swapped its position along the vertical axis (from bottom to top of the screen or vice versa) once it re-emerged from the other side of the occluder. In the Spatiotemporally discontinuous condition, it paused for a second while occluded, and had swapped vertical position once re-emerged. In the real stimulus displays, the walls had a slight red hue and the background was gray.
1000 ms
Continuous
CHAPTER 3
51
object spatial continuity is in principle sufficient to generate a same object benefit for
attentional control settings.
Experiment
METHOD
Participants
Fourteen students (7 male) from the VU University of Amsterdam participated in
exchange for money or course credits. All were between 17 and 25 years of age and
reported having normal or corrected-to-normal vision and no color blindness.
Apparatus
The experiment was run on a HP Compaq with a 2.6 GHz Pentium 4 processor and 512
MB of RAM. The stimuli were presented on a 19” Iiyama Vision Master Pro 454 CRT
screen with loudspeakers, with a refresh rate of 120 Hz and with a resolution of 1024 x
768 pixels. The “M” and “N” keys on a normal keyboard were used to register the
responses. Stimulus presentation and response recording were done in E-prime 1.2
(Psychological Software Tools, 2003). The experiment was executed in a dimly lit and
soundproof room, in which participants were seated at a distance of approximately 75
cm from the screen.
Stimuli
An image of a wall (7.4o visual angle wide) was positioned in the middle of the display,
stretching from the top to the bottom of the screen. The wall was drawn on an evenly
colored grey background (CIE(.289,.316), 4.9 Cd/m2). A square area containing the
search array was placed next to the wall, in one of four quadrants of the display. This
search display had a black background and a white border, CIE(.282,.310), 39.0 Cd/m2,
with 0.07o width. The diagonal of the square search display was 5.2o. To generate an
impression of depth in the display and to enhance the perception of the search display
as a real object, a thin shadow was drawn behind it at the right and bottom sides,
CIE(.0272,.330), 3.2 Cd/m2. The search elements were positioned on an imaginary circle
with a radius of 2.3o visual angle. A white fixation cross was located at the center of this
circle. The individual search elements consisted of circles and diamonds, with a visual
angle of 0.5o and 0.6o respectively. The difference in visual angle between a circle and a
OBACS DEPEND MORE ON SPATIAL THAN TEMPORAL CONTINUITY
52
diamond shape was necessary, because this adjustment equalized the surface sizes of
these two shapes. The shapes were all colored green, CIE(.280,.623), 7.2 Cd/m2, except
for the distractor, if present, which was colored red, CIE(.619,.355), 9.0 Cd/m2. The
shapes contained either an “M” or “N” character, of a grey color identical to that of the
main background. The display moved from its starting position to the center of the
screen (behind the wall) in 300 ms, and another 300 ms to move from behind the wall to
its new position. A continuous displacement from origin to destination would thus take
600 ms.
Design and Procedure The main factors of interest were: 1) Target Location (the target position was the same
as, or different from, the previous trial). 2) Response (again same as, or different from,
previous trial). 3) Spatiotemporal continuity. The display would either a) move behind
the wall with constant speed and vector (continuous), b) stop for a second when behind
the wall (temporally discontinuous), c) reappear from behind the wall on the opposite
vertical half of the screen than it disappeared behind (spatially discontinuous), or d) be
subject to both latter events at the same time (spatiotemporally discontinuous).
Together, this resulted in a 2x2x4 design.
Participants were tested in a one hour session. Before the experiment started,
oral and written instructions were given to familiarize them with the task. They were
asked to look for the diamond target, while ignoring all other items, and to respond as
fast as possible while making as few errors and eye movements as possible (they did not
have to keep fixation at a central point). Participants were to report whether the
diamond shape contained an “M” or a “N” character by pressing the corresponding key
on the keyboard. Correct and incorrect responses were followed by a short high
frequency tone and a slightly longer low frequency tone respectively. Participants were
first presented with a practice block containing 80 trials. After completion, participants
were requested to call the experimenter to check their scores. The main experiment
consisted of 8 blocks of 112 trials each. In each block there were an equal number of
combinations for each of the factors. After each block, the participant received RT and
accuracy scores followed by a short break.
At the beginning of a trial, the search display was located at one of the four
possible locations next to the wall. After response, the display would move behind the
wall to a new location at the other side of the wall. When moving, the search display only
CHAPTER 3
53
contained green circles (with N or M inside each of them), as to not give away the
display arrangement during the motion. It deserves mentioning that given the speed of
the motion, and the abruptness of the halt, these changes were not consciously
noticeable if one was not instructed about their occurrence beforehand. As soon as the
display came to a halt on the opposite side of the wall, one of the items changed into the
diamond target again and response time measurements were started for a next trial.
Results The analyses focused on response times (RTs). Erroneous responses (4.6% of the trials)
and trials on which RTs were deviating more than 2.5 SD from the mean for each cell
(another 2.1%) were removed from the dataset. The means of the remaining RT data
were submitted to an ANOVA with Target Location (same, different), Response (same,
different) and Spatiotemporal Continuity (continuous, temporally discontinuous,
spatially discontinuous, spatiotemporally discontinuous) as factors.
There was a significant main effect of Target Location, F(1,13) = 77.88, p < .001.
As expected, responses were faster when the target location was repeated between
trials. There also was a significant main effect of Response, F(1,13) = 6.69, p < .05,
reflecting faster responses when the response feature (M or N) was repeated.
Figure 6: Results of the experiment. A) Mean correct RTs for the various Spatiotemporal Continuity conditions as a function of repeated or alternated target locations. B) Same object effects for a continuously moving display and displays that suffered a temporal or spatial discontinuity in their displacement. The spatiotemporal discontinuous condition was taken as a baseline and was substracted from RTs in the other conditions to obtained the depicted values. Error bars depict one standard error of the mean effect.
B)
A)
OBACS DEPEND MORE ON SPATIAL THAN TEMPORAL CONTINUITY
54
Importantly, the Target Location x Spatiotemporal Continuity interaction was
significant, F(3,39) = 5.97, p < .01. This interaction is depicted Figure 6a. The effects of
target location repetition were larger for displays that had followed either a fully
continuous or a spatially continuous trajectory, than those that had followed a spatially
disrupted or both temporally and spatially disrupted trajectory. Figure 6b depicts this
same interaction, but now as “same object effects” (SOE) relative to the spatiotemporally
discontinuous condition, which was taken as a baseline. Thus, the SOE is the additional
location repetition benefit found for spatially and/or temporally coherent objects. As
can be seen, continuous objects resulted in a same object effect of 22 ms, which was
significant at F(1,13) = 17.62, p < .001. The temporal discontinuity did not destroy this
effect, still resulting in a significant 17 ms same object effect, F(1,13) = 6.27, p < .05,
which did not differ significantly from the continuous condition, F<1, p> .5. However, the
spatial discontinuity reduced the same object effect to nil, F < 1, p > .5, a reduction that
was significant when compared to the continuous and temporally discontinuous
conditions, F(1,13) = 17.62, p < .001, and F(1,13) = 10.51, p < .01 respectively. Thus,
when the display was subject to a spatial or spatiotemporal discontinuity, the effect
elicited by the repetition or alternation of the target’s location was attenuated compared
to the continuous and temporal discontinuous conditions. Finally, the Target Location x
Response interaction was significant, F(1,13) = 19.86, p < .001. Responses were speeded
when both the target location and the response feature were repeated. No other
interactions reached significance (Fs < 1.29, ps > .29).
An ANOVA on the error rates with the same factors revealed no significant effects.
Table 1 shows the error rates as a function of target location and spatiotemporal
continuity.
Average error rates SearchField GapEvent different same
None 5% 4% Spatial 5% 5% Spatiotemporal 5% 4% Temporal 4% 5%
Table 1: The error rates as a function of target location and spatiotemporal continuity.
CHAPTER 3
55
Discussion
Schreij and Olivers (2009; submitted) have shown that observers maintain attentional
control settings for relevant locations or features inside an object. When the same object
is later reencountered (for instance after disocclusion) these attentional control settings
are more readily available to guide attention. The current study shows that the retrieval
of these attentional control settings is hampered when a display object has been subject
to a spatial or spatiotemporal discontinuity while moving behind an occluder. If the
display object reappeared from behind a different part of the wall than its previous
trajectory would lead to expect (spatial discontinuity) or additionally paused for a
moment when entirely occluded by the wall (i.e. spatiotemporal discontinuity), the
beneficial effect of a repeated target location was attenuated compared to when the
display moved in a spatiotemporally continuous fashion. A temporal discontinuity in the
object’s movement also did not disrupt the maintenance of attentional control settings.
When the display stopped behind the occluder for a moment but continued along the
same trajectory as before it disappeared, the additional beneficial effects of target
repetition were preserved. In other words, there was a same object benefit. The
persistence of attentional control settings for an object thus appears to be more
constrained by the spatial than by the temporal characteristics of the object’s dynamics.
Response selection itself (i.e. whether an N or M had to be pressed) was not
affected by the spatiotemporal continuity of the display object, corroborating the
findings of Schreij and Olivers (2009). This indicates that only spatial selection settings
are maintained with an object representation, and response selection is not. However,
response repetition effects did in turn interact with location repetition effects, which
suggests a cascade of stages. First, the target is spatially selected within the object with
help of information stored in the spatiotemporal object representation, and then the
response is selected, which is in turn influenced by the spatial selection process.
Our findings are largely in line with an earlier study of Flombaum and Scholl
(2006), who deployed a change detection task to demonstrate that when the occlusion
and disocclusion of an object occurs in a spatiotemporally incoherent fashion, the
disoccluding object tends to be perceived as a different instance. Flombaum and Scholl
presented participants with various oscillating shapes that were temporarily occluded
by other, stationary, objects along their trajectory. During occlusion a moving object
could change shape or color and participants had to detect this change once the object
OBACS DEPEND MORE ON SPATIAL THAN TEMPORAL CONTINUITY
56
reappeared. Detection rates were higher when an object moved behind an occluder in a
spatiotemporal coherent way, than when it paused for a moment behind the occluder, or
reappeared from an unexpected part of the occluder. Consistent with object-file theory,
Flombaum and Scholl reasoned that in order to efficiently compare the current state of
an object with its previous state, it has to be perceived as the same object in terms of its
spatiotemporal history. However, their finding that a temporal discontinuity also breaks
the object file is at odds with our present findings showing that attentional settings are
preserved across such discontinuities. This is not just a fluke of the present study. Note
that in our previous study (Schreij & Olivers, 2009), there was also a temporal
discontinuity between trials: There a search display would disappear behind a wall on
one trial, and after a pause of one second the next trial would start and a display would
appear from either the same wall or a different wall. This suggests that observers can
bridge a temporal gap as long as there is sufficient other evidence (in this case spatial)
that one is dealing with the same object.
Why then did Flombaum and Scholl (2006) find that temporal discontinuities
affected object persistence? One possible explanation is that in their study the temporal
consistency of the objects was not the only thing that changed. As the task was a change
detection task, the objects regularly changed features form one appearance to the next.
It could be that object persistence is mediated by featural consistency of the object when
it is subject to a temporal discontinuity. That is, briefly pausing objects will be regarded
as one and the same, unless they also change feature, in which case there is sufficient
evidence for a new object. This notion is supported by a study of Flombaum, Kundey,
Santos and Scholl (2004), who used a foraging task to explore the tunnel effect in rhesus
monkeys. The monkeys watched as a lemon rolled down a ramp containing two occluder
screens. When the lemon came to rest behind the first occluder (located halfway across
the ramp), a kiwi started rolling from the other end of the screen and eventually became
occluded behind the second screen at the bottom of the ramp. If the kiwi emerged at the
moment the lemon should have, had it rolled on, then the monkeys only searched for
food behind the second screen – as if the lemon had just transformed into a kiwi. In
contrast, when the disappearance of the lemon and the emergence of the kiwi was
interrupted by a brief pause (i.e. a temporal discontinuity), most monkeys searched for
food behind both occluders, apparently perceiving two distinct objects. In other words,
they must have inferred (on the basis of the featural difference in combination with the
CHAPTER 3
57
temporal gap) that the lemon must have remained behind the first occluder. Importantly
however, when a lemon disappeared behind the first occluder, but a lemon also re-
appeared at the other end, a temporal gap had no such effect: The monkeys then only
searched for the lemon behind the last occluder the majority of times. This demonstrates
that a temporal gap does not need to disrupt object persistence, given that the occluding
and disoccluding objects are featurally consistent. In the current study, the features of
the search display object also did not change during occlusion, possibly explaining why
its representation survived a temporal discontinuity.
In conclusion, the current study shows that attentional control settings are best
preserved when an object passes behind an occluder in a spatiotemporal coherent
fashion. Whether one perceives an object as the same or different, and hence whether
one is able to efficiently retrieve previously established attentional control settings from
the object’s representation or not, seems to largely depend on the object’s spatial
continuity and to a lesser degree on its continuity in the temporal dimension.
Acknowledgments
This work was supported by VIDI grant 452-06-007 from the Netherlands Organization
for Scientific Research (NWO) awarded to CNLO. We thank the reviewers for their very
helpful comments and suggestions.
Chapter 4 Object representations maintain
attentional control settings for feature information
Schreij, D. & Olivers C.N.L (under revision). Object representations maintain attentional control settings for feature information.
OBACS FOR FEATURE INFORMATION
60
Abstract
For stable perception, we maintain mental representations of objects across space and
time. An important question remains what information is stored in or linked to such a
representation. Recently, we reported evidence for the preservation of spatial
attentional control settings across instances of the same object, as defined by
spatiotemporal history. Here we further investigate which attentional settings are
stored with a spatiotemporal object representation. Observers conducted visual search
for a target among multiple distractors on a search display, in which the target location
or feature (here shape) could repeat from trial to trial. Importantly, the entire visual
search display was part of an object that could move in and out of view. Responses were
speeded when the target property repeated, but especially when the motion suggested
that the same object had emerged. We show that this same object benefit is tied to task-
relevant target information, and not to irrelevant target or distractor information.
Additionally, we show that the maintenance of these attentional control settings is not
affected by a change in object identity, but is specific to an object’s spatiotemporal
history.
CHAPTER 4
61
Introduction
To help keep a stable percept of the world, it is proposed that the visual system
maintains mental representations of objects across space and time. Classical empirical
support for this idea was provided by Kahneman, Treisman and Gibbs (1992; see also
Treisman & Kahneman, 1983) with their object preview paradigm. Participants were
initially shown two boxes each containing a unique letter. Participants previewed the
display for about a second, after which the letters disappeared. Both boxes then moved
to new locations. Once both objects had arrived at their destination, a single letter
reappeared in one of them, and participants had to respond to its identity. As would be
expected, responses were considerably faster when the target letter matched one of the
previewed ones than when it was a whole new letter. The important result was that this
benefit was greater when the matching letter appeared in the same box (now at a
different location), as compared to the other box. This same object benefit led
Kahneman, et al. (1992) to propose that we create an episodic representation, which
they termed an ‘object file’, for each object we observe. Object files contain information
about the properties of the corresponding physical objects (e.g. location, color, shape, or
letter identity). When the same object is then encountered again, its represented
properties are readily available, and when they match the actual visible properties, a
rapid response is achieved. In contrast, responses are slowed when there is a mismatch
between the visible object and its object file.
It has been proposed that it is especially spatiotemporal continuity which
determined whether an object is experienced as one and the same (Scholl, 2001).
Consistent with this, Mitroff and Alvarez (2007) used a variant of the object preview
paradigm, and demonstrated that continuity in the surface features of the object (e.g.
texture, color or shape) from one instance to the other plays no significant role in the
same object benefit, but the spatiotemporal continuity of the object does. When
spatiotemporal constraints of an object were violated (e.g. when it jumped from one
location to another instead of gradually moving between them), no same object benefits
were found, even when the object retained the same physical appearance. Vice versa, a
change in features did not alter the same object benefit as long as spatiotemporal
consistency was preserved. Spatiotemporal continuity enables us to keep track of an
object even when it temporarily disappears from vision (for instance through occlusion;
Flombaum, Scholl, & Pylyshyn, 2008; Pratt & Sekuler, 2001). In other words, an object
OBACS FOR FEATURE INFORMATION
62
does not cease to exist in our experience if we cannot see it for a moment and we will be
aware that the object is ‘there’ as long as it moves behind an occluder in
spatiotemporally coherent way. Flombaum and Scholl (2006) illustrated this with a
study in which participants were presented with various shapes that were moving back
and forth across the screen, during which these shapes were temporarily occluded by
other, stationary, objects. During occlusion a moving object could change shape or color
and participants had to detect this change once the object reappeared. Detection rates
were higher when an object moved behind an occluder in a spatiotemporal coherent
way, than when it paused for a moment behind the occluder (temporal gap), or
reappeared from an unexpected part of the occluder (spatial gap). Consistent with
object-file theory, Flombaum and Scholl reasoned that, in order to efficiently compare
the current state of an object with its previous state, it has to be perceived as the same
object in terms of its spatiotemporal history.
In a comparable study, Yi, Turk-Browne, Flombaum, Kim, Scholl and Chun (2008)
used fMRI to investigate the effect of spatiotemporal history on physically identical or
distinct faces. They displayed two pillars, one on the left, and one on the right. These
served as occluders for the faces. Each trial consisted of a sequence of two events, each
involving the emergence of a face from behind a pillar. The second face could be the
same as the first or different, and it could appear from behind the same or a different
pillar. Yi et al. found that brain activity in the right fusiform face-area (FFA) was reduced
when the identity of the face was the same from one event to the next. This is expected
on the basis of habituation of neurons in response to sustained stimulation. Importantly,
this habituation was stronger when the same face had also re-emerged from the same
pillar as where it had just disappeared. In other words, the neural coding the face as one
and the same depended on whether its spatiotemporal history had been violated or not.
Many studies have thus attempted to assess how object representations are
referenced and maintained, but few have examined what kind of information is actually
stored within, or bound to, the representation. With this purpose, Schreij and Olivers
(2009) investigated if attentional control settings are maintained with object
representations. They devised a paradigm in which two visual search displays were
hidden behind walls flanking the screen on all sides. On each trial, one of the displays
would move to the middle of the screen. Participants were instructed to find a diamond-
shaped target among circle-shaped distractors, and respond to the letter printed inside
CHAPTER 4
63
(cf. Theeuwes, 1992). After response, the display would move back behind one of the
unoccupied walls. There were two important manipulations: First, from one trial to the
next, either the same display object would re-emerge again from behind the wall it had
just disappeared behind, or the other display object would now slide to the middle of the
screen. Second, within the emerged display object, the target was either positioned at
the same location as on the previous trial, or at a different location. Previous work has
shown that when observers have selected the location of a target, they are likely to
select that location again on the next trial (Maljkovic & Nakayama, 1996). The
hypothesis was that if such spatial selection biases are at least partially stored with the
spatiotemporal representation of the entire display object, then this information is
subsequently retrieved when the same display object appears again from behind the
same wall, leading to further benefits. This was indeed found: Search on repeated target
trials was especially fast when the display re-emerged from its last-known position.
Thus, the display object’s spatiotemporal history affected the way in which locations
within the object were selected. The study of Schreij and Olivers (2009) shows that
relevant target location information is stored with the spatiotemporal representation of
an object. One of the goals of the current study was to investigate if information less
relevant to the task is also stored within a representation, such as the presence of an
irrelevant salient object. To this end, we replicate Schreij and Olivers (2009) in
Experiment 1 but in one condition added a color-defined singleton distractor to the
search array.
Second, we ask if besides the target location, relevant target feature information
is also stored with an object’s spatiotemporal representation. In Experiment 2, the target
shape, the target color, or both could repeat from trial to trial, but only target shape was
relevant. The target location was never repeated. As with spatial biases, the question
was whether the bias for target features would depend on the display object being
spatiotemporally the same or different.
Third, we investigated if the attentional control settings which are maintained for
target locations are affected by a change in the physical appearance of the display object,
especially when the spatiotemporal continuity of the object defines it as one and the
same. Instead of having two simple square displays as objects, in Experiment 3, we
presented the search arrays in the display area of either a cell phone or a digital music
OBACS FOR FEATURE INFORMATION
64
player. The type of device could change from trial to trial, independent of the
spatiotemporal trajectory.
Experiment 1
This experiment sought to replicate Schreij and Olivers (2009) in showing an object-
based bias towards the target location. Figure 7 shows an example display. As in Schreij
and Olivers (2009), participants searched for a diamond shape and responded to the
letter N or M inside it. From trial to trial, the target location could be repeated, or it
could change. In addition, the search display appeared from behind either one of two
walls positioned at the left or right side of the screen. Schreij and Olivers (2009) actually
used walls at all four sides of the screen in order to dissociate movement direction and
object history. Since this made no difference, and to simplify the experimental design,
we used only two walls in this and subsequent experiments. As in our previous study,
we predicted a same object benefit for target location repetitions. Response times
should be speeded when the target location repeats, but especially so when it appears to
150 ms
150 ms
Figure 7: Example of stimulus display for a typical distractor present trial in Experiment 1. For printing purposes, these images were converted to black and white, but the walls were brick-colored and the search elements were green or red. A typical trial started with a screen with both search search displays hidden behind the walls for 1000 ms. Then, over a time course of 150 ms, one of the search displays slid to the center of the screen, and with this exposed the search array. Participants were to report the identity of the letter in the diamond shape. When the participant had given a response, the search display shifted back behind the wall it departed from (within another 150 ms).
Until response
CHAPTER 4
65
be the same display object, as when it re-emerges from the side on which it disappeared
on the previous trial.
The novel addition was that on certain trials one of the circles would carry a
salient deviating color, and thus form a singleton distractor. Theeuwes (1992) has
shown that such a color distractor interferes with search, resulting in increased RTs.
Here we assessed if this distractor interference effect is also modulated by the
spatiotemporal history of the search display object. Just like biases towards the target
location are bound to the display object, biases away from the distractor may also be
bound to it. If so, distractor interference should be reduced for search array repetitions
when the same spatiotemporal object is reencountered. If on the other hand only
(relevant) target information is saved, and (irrelevant) distractor information is not, the
effect of a distractor should be unaffected by the spatiotemporal continuity of the object
on which it is presented.
METHOD
Participants
Twelve students from the VU University Amsterdam participated and received course
credit or money in return. All of them were between 19 and 29 years of age (average
21), reported normal or corrected-to-normal vision and no color blindness.
Apparatus
The experiment was run on a HP Compaq with a 2.6 GHz Pentium 4 processor and 512
MB of RAM. The stimuli were presented on a 19” Iiyama Vision Master Pro 454 CRT
screen with loudspeakers, with a refresh rate of 120 Hz and with a resolution of 1024 x
768 pixels. The “M” and “N” keys on a normal keyboard were used to register the
responses of the participants. Stimulus presentation and response recording were done
in E-prime 1.2 (Psychological Software Tools, 2003). The experiment was executed in a
dimly lit and soundproof room, in which participants were seated at a distance of
approximately 75 cm from the screen.
Stimuli
Images of a wall (7.36o visual angle wide) were positioned on the left and the right side
of the display, stretching from the top to the bottom of the screen. Behind these walls
there was an evenly colored grey background (CIE(.289,.316), 4.9 Cd/m2). Two square
OBACS FOR FEATURE INFORMATION
66
areas containing the search arrays were placed behind the walls on either side. The
contents of these search displays would be occluded by the walls and thus not visible to
the participant, until one of them would slide to the middle of the screen. The search
displays had a black background and a white border, CIE(.282,.310), 39.0 Cd/m2, with
0.07o width. The size of the search display was 512x384 pixels. To generate an
impression of depth in the display and to enhance the perception of the search display
as a real object, a thin shadow was drawn behind it at the right and bottom sides,
CIE(.0272,.330), 3.2 Cd/m2. The search elements were positioned on an imaginary circle
with a radius of 14.24o visual angle. A white fixation cross was located at the center of
this circle. The individual search elements consisted of circles and diamonds, with a
visual angle of 3.07o and 3.68o respectively. The difference in visual angle between a
circle and a diamond shape was necessary, because this adjustment equalized the
surface sizes of these two shapes. The shapes were all colored green, CIE(.280,.623), 7.2
Cd/m2, except for the distractor, if present, which was colored red, CIE(.619,.355), 9.0
Cd/m2. The shapes contained either an “M” of “N” character, of a grey color identical to
that of the main background.
Design and Procedure
The main factors of interest were: 1) Spatiotemporal Object (the search array appeared
on the same object as the previous trial, or on a different object). 2) Target Location
(target and distractor positions were the same as, or different from, the previous trial).
3) Response (again same as, or different from, previous trial). 4) Distractor: An
irrelevant color singleton could be absent or present in the display (replacing one of the
standard distractors). Together, this resulted in a 2x2x2x2 design. So for example, on
trial n, the search display could come from the other side of the display than on trial n-1
(and therefore constitute a different object), but the target could be in the same location
as on trial n-1 and have the same response feature as on trial n-1 or any other
combination of these factors. Distractor: All factors were randomly mixed within blocks,
except distractor presence, which was blocked.
Participants were tested in a half hour session. Before the experiment started,
oral and written instructions were given to familiarize them with the task. They were
asked to look for the diamond target, while ignoring all other items, and to respond as
fast as possible while making as few errors and eye movements as possible. To keep
CHAPTER 4
67
them motivated, participants were asked to write down their average response time and
accuracy score after each block. They were first presented with one singleton distractor
absent and one singleton distractor present practice block (with the order
counterbalanced across participants), each containing 80 trials. After completion of the
practice blocks participants were requested to call the experimenter to check their
scores. Practice was repeated until scores were reasonable (no less than 85% correct
and average RT < 1000 ms). The main experiment consisted of 10 blocks of 80 trials
each, with distractor present and distractor absent blocks alternating in
counterbalanced order (across participants). In each block there were an equal number
of combinations of same and different conditions for each of the remaining factors. After
each block the participant received RT and accuracy scores, followed by a short break.
At the beginning of a trial only the two walls were visible, each having a search display
hidden behind them of which only the left- or right-most edge was visible (for right- and
left-sided displays respectively). After 1000 ms, one of the search displays would take
150 ms to slide to the middle of the screen from behind a wall, revealing the search
display. Until it arrived at the center of the display, the search display only contained
green circles (with N or M inside each of them), as to not give away the display
arrangement during the motion. As soon as it came to a halt, one of the items then
changed into a diamond target, another into a red singleton distractor (in the distractor
present condition). It deserves mentioning that given the speed of the motion, and the
abruptness of the halt, these changes were not consciously noticeable if one was not
instructed about their occurrence beforehand. It still looked like a smoothly entering
display.
When the search display arrived at the middle of the screen, response time
measurements started and participants were to report whether the diamond shape
contained an “M” or an “N” character by pressing the corresponding key on the
keyboard. The display stayed on until response and then slid back behind the wall that it
came from (again within 150 ms). Correct and incorrect responses were followed by a
short high frequency tone and a slightly longer low frequency tone respectively.
Results
Figure 8 depicts the results of Experiment 1. The RT data were trimmed from 2.5 SD
from the mean (2.6% of trials) and erroneous responses were removed (another 4.8%).
OBACS FOR FEATURE INFORMATION
68
The means of the remaining RTs were subjected to a 4-way repeated-measures ANOVA
with Spatiotemporal Object (same or different compared to previous trial), Target
Location (same or different), Response (same or different), and Distractor (absent,
present) as factors. When a new display object was presented, responses were overall
slower (717 ms) than when the old one appeared again (707 ms); Spatiotemporal
Figure 8: Results of Experiment 1. Mean response times and error rates for A) same and different search displays, as a function of object type (same or different as on previous trial). B) same and different search displays, as a function of response feature (same or different as on previous trial). C) same and different responses, as a function of object type (same ore different as previous trial). Panel D) shows the same object benefits underlying the interaction of panel A, i.e. the RT differences between ‘different object’ and ‘same object’ conditions, for the different feature changes in display configurations. The error bars show the standard error.
B) C)
D)
A)
CHAPTER 4
69
Object, F(1,11) = 11.07, p = .007. Participants were also significantly slower when the
target location differed from the previous trial, compared to same previous target
locations (761 vs. 662 ms); Target Location, F(1,11) = 31.92, p < .001. A change in
Response caused no significant difference. The main effect of Distractor was significant
F(1,11) = 5.57, p = .038. Responses took longer when there was a color distractor
present (721 ms) than when it was absent (702 ms).
As shown in Figure 8a, repeated display configurations led to improved
performance, but especially so when they appeared on the same object. This was
confirmed by a significant Spatiotemporal Object by Target Location interaction, F(1,11)
= 13.83, p = .003. Figure 8d depicts this same object benefit in terms of the difference in
RT between the same and different object conditions. Separate t-tests showed a reliable
same object benefit for repeated display configurations, t(11) = 4.43, p < .001, and no
such benefits for changing configurations.
The interaction between Target Location and Response was also significant,
F(1,11) = 28.762, p < .001 (Figure 8b). When both factors repeated, participants were
faster than when either one or none repeated. There also was a trend towards an
interaction between Distractor and Target Location, F(1,11) = 4.01, p = .070. The
presence of a distractor resulted in greater RT costs when the display configuration
changed, than when it repeated. There was no interaction between Spatiotemporal
Object and Response (F< 0.5, p > .5), nor between Spatiotemporal Object and Distractor,
or Spatiotemporal Object, Target Location, and Distractor (F<0.5, p >.5 and F<0.9, p > .3
respectively).
The only significant effect on the errors was an interaction between Target
Location and Response, F(1,11) = 23.09, p = .001. Participants made more errors when
either the response feature or the display configuration changed, than when both
remained the same or when both changed together.
Discussion
A clear same object benefit occurred for selecting target locations, confirming the results
of Schreij and Olivers (2009). Observers benefitted when the target location repeated
from one trial to the next, but especially so when the display object re-appeared from
the side it had just disappeared to. The presence of a color distractor slowed down
search overall, in line with previous findings in the literature (Theeuwes, 1992).
OBACS FOR FEATURE INFORMATION
70
However, this cost did not interact with the spatiotemporal properties of the display
object. Such an interaction would be expected if selection settings pertaining to the
distractor were stored with the object representation. For example, if the distractor
becomes subject to location-based or feature-based inhibition, such inhibitory settings
may then be expected to carry over to the next search display when it appears on the
same object. Apparently the control settings for the target were bound to the
spatiotemporal object representation but the settings for the distractor were not.
Apparently, the object file stores what is relevant and not what is irrelevant to the task.
Experiment 2
Experiment 1 suggests that objects carry an attentional bias that affects selection when
the same object emerges again. This bias was spatial in nature, as the same object
benefit was measured through the repetition of the target location. The main goal of the
present experiment was to see if similar object benefits would also occur for the
repetition of target features instead of location. The second goal was to see if such
object-based biases are again limited to relevant features, or could also include
irrelevant features.
Participants searched for a unique shape, which could be a diamond among
circles, or vice versa. Furthermore, the color of the search elements could also change
between red and green. So on any trial, either or both the target shape and color could
be the same or different from the previous trial (and the same would go for the singleton
distractor). This allowed for feature-based priming effects to occur. Importantly, the
target and distractor locations always changed from trial to trial, so there were no
opportunities for location-based effects. If object representations can also carry feature-
based attentional settings, we would expect a same object benefit for repeated target
features. For example, if observers found a diamond-shaped target on trial n, they would
be biased towards a diamond-shaped target again on trial n + 1 if the search display is
the same object in terms of spatiotemporal history.
As the target is defined by its shape and not by its color, information about the
target color is not relevant to the task. If, as the results of Experiment 1 suggest, task-
irrelevant information is not stored with the object, then the effect of a color change on
search would not be modulated by the spatiotemporal continuity of the display object.
Furthermore, we again introduced an irrelevant, color-defined distractor. On the basis of
CHAPTER 4
71
previous experiments with feature changes between trials (Hickey, Olivers, Meeter, &
Theeuwes, 2011; Pinto, Olivers, & Theeuwes, 2005), we expected this distractor to
strongly interfere with search. However, since it was again irrelevant, it may not stick
with the spatiotemporal representation of the display object.
METHOD
Participants
Fifteen students from the VU University Amsterdam (aged 18-21 years) participated for
course credits or money. They reported normal or corrected-to-normal vision and no
color blindness.
Apparatus, Stimuli, Design, and Procedure
The method was the same as in Experiment 1, except for the following changes.
Participants were instructed to look for the unique shape, which could be a diamond
among circles, or the other way around. Items could also change color from trial to trial,
between red and green. If the target was red, the singleton distractor (if present) would
be green, and vice versa. The location of the target and distractor never remained the
same and changed on every trial. Thus, instead of Target Location, we manipulated a
new factor, Target Feature, with four levels: “no change”, “color change”, “shape change”,
and “both change”, as considered relative to the previous trial. This factor was presented
randomly mixed with the previous factors Spatiotemporal Object (same, different), and
Response (same, different) in 14 counterbalanced distractor present and distractor
absent blocks of 96 trials each, resulting in a 2x2x2x4 design.
Results
The RT data were trimmed, removing 2.6% of the total number of trials, and erroneous
responses were omitted, removing another 9.7%. The remaining RTs are depicted in
Figure 9 and were submitted to an ANOVA with Target Feature (no change, color change,
shape change, both change), Spatiotemporal Object (same, different), Distractor (absent,
present) and Response (same, different) as factors. Overall, participants were faster
when there was a switch of object (791 vs. 802 ms), Spatiotemporal Object, F(1,14) =
9.82, p = .007. They were slower when the target-defining feature changed (834 vs. 796
ms), Target Feature, F(3,42) = 42.43, p < .001. They were also slower in responding
when a distractor was present (853 ms, vs. 740 ms when absent), Distractor, F(1,14) =
OBACS FOR FEATURE INFORMATION
72
Figure 9: Results of Experiment 2. Mean response times and error rates for A) same and different search displays, as a function of object type (same or different as on previous trial). B) same and different search displays, as a function of response feature (same or different as on previous trial). C) same and different responses, as a function of object type (same ore different as previous trial). Panel D) shows the same object benefits underlying the interaction of panel A, i.e. the RT differences between ‘different object’ and ‘same object’ conditions, for the different feature changes in display configurations. The error bars show the standard error, for each individual “same vs. different object” contrast.
C)
B)
D)
A)
CHAPTER 4
73
57.65, p < .001. The interaction of most importance here, between Spatiotemporal
Object and Target Feature, was significant, F(3,42) = 3.88, p = .021. This interaction is
shown in Figure 9a, and further clarified in Figure 9d, which shows the relative same
object benefits (i.e. the difference in RT between same object and different object trials).
Note that rather than there being a same object benefit with repeated feature
configurations, there was now a marked same object cost when the feature
configuration changed over trials. Planned comparisons confirmed that when the target
shape changed, or shape and color changed together, there were significant same object
costs, t(14) = 3.06, p < .05 and t(14) = 3.33, p < .05 respectively. There were no costs or
benefits in the condition in which the target changed color only, or did not change at all
(ts < 1.2, ps > .2).
There also was a significant interaction between Distractor and Target Feature,
F(3,42) = 7.03, p = .001. The distractor was more disruptive when the target shape, or
the target shape and the color changed. The interaction between Response and Target
Feature was also significant, F(3,42) = 15.95, p < .001. When the target changed shape,
or both shape and color, participants were faster when the response feature changed
too. When the target did not change, or changed only color, same responses were faster.
The presence of a distractor even enlarged this effect, resulting in a three-way
interaction, F(3,42) = 4.08, p = .019. There was no Response x Spatiotemporal Object (p
> .3 ), nor a Spatiotemporal Object x Distractor (p > .1) or a Spatiotemporal Object x
Target Feature x Distractor interaction (p > .4). As shown in Figure 9c, RTs remained
similar for each combination of Response and Spatiotemporal Object.
An ANOVA on the error rates with the same factors revealed that participants
made significantly more errors when a distractor was present than when it was absent,
F(1,14) = 57.65, p < .001. Participants also made more errors when the object changed
over trials, Spatiotemporal Object, F(1,14) = 8.50, p = .011, when the Target Feature
changed, F(3,42) = 19.51, p < .001, or when Response changed, F(1,14) = 10.57, p = .006.
The interaction between Response and Target Feature was significant, F(3,42) = 15.90, p
< .001. More erroneous responses were made when the display configuration changed
but the response feature was repeated. None of these findings jeopardizes the
conclusions based on the RT data.
OBACS FOR FEATURE INFORMATION
74
Discussion
The results provide evidence for the idea that feature-based attentional control settings
can be stored in an object file. The change of search display features between trials had a
differential effect on RTs depending on the spatiotemporal history of the entire display
object. However, where Experiment 1 found a same object benefit for repeated
configurations, here the object-based effect was expressed as a same object cost when
the feature configuration changed from one trial to the next. Participants performed
worse when the spatiotemporal trajectory suggested that they were looking at the same
display as on the previous trial, but in fact the feature configuration had changed.
Apparently, the reappearance of an object (in spatiotemporal terms) creates the
expectation of a certain target definition (here shape). When this expectation is violated,
attentional settings need to be reconfigured, resulting in a selection delay. A possible
reason for finding same object costs with changing configurations rather than same
object benefits with same configurations is that although the features of the search
elements did not necessarily change over trials, the locations of the target and distractor
did (on every trial). Any location-based expectations (such as found in Experiment 1)
may thus have worked against the feature-based benefit.
When merely the color changed, there were no particular same object effects.
This finding once again supports that it is mainly a bias towards task-relevant features
which is stored with the object representation. Because the target was defined by its
shape and not by its color, information about the target’s color was irrelevant for the
task and henceforth apparently not kept with the representation. This conclusion is
further corroborated by the lack of any object-based modulations of the singleton
distractor effect. Even though the distractor by itself caused considerable slowing, this
did not depend on object history.
Finally, a change of response interacted with a change of target feature, indicating
that participants linked specific responses to specific features of the target. Such feature-
response bindings have been found before (Hommel, et al., 2001) and will be discussed
further in the General Discussion.
Experiment 3
Experiments 1 and 2 demonstrated that selection settings for relevant properties of the
search array, like target features and locations, are linked to an object representation.
CHAPTER 4
75
The current experiment investigates if a change in the physical appearance of this object
weakens or severs this link. Will attentional control settings for the target still be
retrieved for an object that is the same in terms of spatiotemporal continuity, but has
changed physical appearance during occlusion? For this purpose, we presented the
Figure 10: Illustration of a stimulus display in Experiment 3. The search arrays were now presented inside the screen area of a portable device, which was either a white Apple Ipod, or black Nokia N95.
OBACS FOR FEATURE INFORMATION
76
search array inside the display area of two distinct mobile devices, as illustrated in
Figure 10. When one of these mobile devices moved to the center of the screen, it could
change its appearance when it was temporarily occluded by the wall. This manipulation
allows us to determine if attentional control settings are linked more strongly to
spatiotemporal continuity, continuity in terms of object identity, or both. Earlier work
by Mitroff and Alvarez (2007) has indicated that the persistence of an object
representation depends on the spatiotemporal continuity rather than on the physical
appearance of the object. We may therefore expect the same to happen for the
attentional control settings.
In order to keep the number of experimental manipulations within limits, we
chose to leave out the singleton distractor in this experiment, since its presence did not
interact with the spatiotemporal representation of an object in both previous
experiments.
METHOD
Participants
Eleven students from the VU University Amsterdam (aged 18-24 years, 3 male)
participated for course credits or money. They reported normal or corrected-to-normal
vision and no color blindness. One participant reported suffering from severe fatigue
during the experiment. As this was confirmed by her data (large standard errors and
mean RT beyond 3 SD from the mean of the remainder of the group), she was removed
from the dataset.
Apparatus
The apparatus remained unchanged from the previous experiments.
Stimuli
The search display was now presented as the display of a mobile device (see Figure 10),
which could either be a white Apple iPod or a black Nokia N95. To make these devices fit
on the screen, the location of the search display had to be slightly shifted upwards
compared to the central position used in Experiments 1 and 2. A further small change
was that whereas in the previous experiments the objects stuck out a little from the
inner side of the occluding wall, in the present experiment they stuck out from the outer
CHAPTER 4
77
side. This way, an object could change appearance while it moved behind the wall before
re-emerging on its path towards the middle of the screen.
Design, and Procedure
The experiment was similar to Experiment 1. The factor Singleton Distractor was
dropped. No singleton distractor was present in the display. Instead, Object Identity was
introduced as a factor, with the object identity either being the same or different as
compared to the previous trial. The factors of this experiment thus consisted of Object
Identity, Spatiotemporal Object, Response and Target Location, all with the levels same
and different as on the previous trial.
Results
Erroneous trials (6%) and trials deviating more than 2.5 SD from the mean (another
2.3%) were removed from the dataset. The results are shown in Figure 11 and Figure
12. An ANOVA was performed with Spatiotemporal Object (same, different), Object
Identity (same, different), Target Location (same, different) and Response (same,
different) as factors. The main effects of Target Location and Response were found to be
significant, F(1,9) = 67.47, p < .001 and F(1,9) = 10.20, p < .05 respectively. Participants
were faster when the target location or response feature repeated from trial to trial.
Importantly, the Target Location x Spatiotemporal Object interaction was significant
again, F(1,9) = 9.20, p < .05. Participants responded faster to a target location repetition
when the spatiotemporal object was repeated too. The Target Location x Response
interaction was also significant, F(1,9) = 8.88, p < .05. Participants responded faster
when both Target Location and Response were unaltered from the previous trial. There
was no interaction between Spatiotemporal Object and Object Identity, F(1,9) < 0.3, p >
.5, nor between Target Location and Object Identity, F(1,9) < 0.02, p > .88, nor between
Object Identity, Spatiotemporal Object, and Target Location, F(1,9) < 0.26, p > .62.
However, there was a significant three-way interaction between Spatiotemporal Object,
Object Identity and Response, F(1,9) = 5.57, p < .05 (see Figure 12). Response
repetitions led to faster RTs than a response change, but only if both object history and
object identity were consistent with the display being the same object. If either of these
factors indicated a different object, no such response repetition benefits were found. If
anything, RTs were then faster to response changes.
OBACS FOR FEATURE INFORMATION
78
A) B)
C)
Figure 11: Results of Experiment 3 concerning the effects of Target Location on Response and Object. Mean response times and error rates for a repeated or changed target location that is A) presented on the same or a different spatiotemporal object and B) contains a same or different response feature. Panel C) displays the same object benefits for repeated or different target locations. Error bars depict the standard error.
CHAPTER 4
79
A)
C
Different object
B) Same object
Figure 12: Results of Experiment 3 concerning the effects of object appearance. Mean response times and error rates for a repeated or changed response that is presented on the same or a different object when A) the object retains the same physical appearance or B) the object changes appearance in between trials. Panel C) displays the same object benefits for repeated or different responses, depending on a repetition or change in object appearance. Error bars depict the standard error.
OBACS FOR FEATURE INFORMATION
80
A similar analysis on the error rates only revealed main effects of Object Identity and
Response, F(1,9) = 5,32, p < .05 and F(1,9) = 7.47, p < .05 respectively. Participants made
more errors when the object appearance or the response was different from the
previous trial.
Discussion
The current results corroborate the findings of Experiment 1. There was once again a
clear same object benefit on target selection. The important new result is that this
benefit was not affected by a change in object appearance, suggesting that the object
representation that is used to store selection settings is predominantly determined by
spatiotemporal continuity rather than by identity. This is consistent with earlier work
showing that the persistence of an object’s perceptual representation is dependent on
the spatiotemporal continuity rather than the surface features of the object (cf. Mitroff &
Alvarez, 2007).
This is not to say that object identity had no effect at all. The three-way
interaction with the spatiotemporal object history and response shows that object
identity information is bound to response information. This is consistent with earlier
work showing episodic binding of object features and responses (Hommel, 1998; Keizer,
et al., 2008). We add that this response selection is not only modulated by object
identity, but also by the spatiotemporal history of the object, as we found earlier in
Experiment 1. In conclusion, although object appearance does not appear to guide our
spatial attention, it appears to affect the way in which we respond to it, consistent with
separate perceptual and response selection mechanisms.
General Discussion
To make sense of the outside world, the visual system carves it up into separate objects.
An important carrier of object representations appears to be spatiotemporal continuity.
This makes sense, since objects tend to be stable across space and time. This has been
confirmed by previous studies showing same object benefits (or different object costs)
when objects were defined in terms of spatiotemporal history (Egly, Driver, & Rafal,
1994; Flombaum & Scholl, 2006; Kahneman, et al., 1992; Mitroff, Scholl, & Noles, 2007;
Muller & Kleinschmidt, 2003; Scholl, 2001). Where these previous studies focused on the
link between the spatiotemporal history of an object and the responses that should be
made to that object, the current study explored the link between the spatiotemporal
CHAPTER 4
81
history of an object and the attentional selection of perceptual properties of that object.
To this end, objects consisted of entire visual search displays, with the target location or
target-defining feature being decoupled from the response feature. We report evidence
for the following:
1) Experiments 1 and 3 showed that seeing the same object (in terms of
spatiotemporal history) retrieves spatial selection settings for that object.
Repeating the target location in the visual search task was especially beneficial
when the search array was presented on the same display object that had just
disappeared behind a wall. This replicates Schreij and Olivers (2009), and
confirms the conclusion that target location settings are stored with an object.
2) Experiment 2 showed an important new finding: Seeing the same object retrieves
selection settings for target features, in this case shape. When the target was
presented on the same object but had a different shape than in the previous trial,
responses were slowed. This indicates that relevant information on feature
selection is stored with the object representation, as a mismatch of the observed
target feature and the one stored in the representation causes interference.
3) Both Experiment 1 and 2 showed that although target information is stored with
an object, distractor information is not. A salient singleton distractor was
introduced in the displays, and although it interfered overall with search, such
interference effects did not vary with the spatiotemporal history of the object.
This was the case regardless of whether same object benefits were spatial
(Experiment 1) or feature-based (Experiment 2) in nature, and regardless of
whether distractor interference was weaker (Experiment 1) or stronger
(Experiment 2). This suggests that relevant information is stored with an object,
but irrelevant information is not.
4) The same conclusion can be reached on the basis of the fact that, in Experiment 2,
only the target-defining feature (shape) was linked with the object history,
whereas an irrelevant feature (color) was not. This is unlikely due to color being
a weaker feature than shape: If anything, previous studies using the same visual
search displays have shown that color is a stronger selection property than shape
(Theeuwes, 1992). Additionally, in the present experiments a color-defined
singleton distractor was sufficiently powerful to interfere with search for a shape.
OBACS FOR FEATURE INFORMATION
82
5) Finally, Experiment 3 showed that the preference for previous target locations is
tied to the spatiotemporal history of an object, but not to its identity. Location
repetition effects were not modulated by object identity, nor were the effects of
spatiotemporal history weakened by a discordant change in object identity. The
dissociation between object definitions based on spatiotemporal continuity and
object identity is consistent with a number of other studies (Kahneman, et al.,
1992; Mitroff & Alvarez, 2007; Mitroff, et al., 2007; Scholl, 2001) which show that
it is mainly spatiotemporal consistency that determines what is and what is not
the same object, and not its visual appearance.
In all, the results show that both location-based and feature-based selection settings are
stored with an object, that the most important definition of an object here is one in
terms of spatiotemporal continuity rather than identity, and that what is stored is
relevant target information instead of information about irrelevant features or
distractors.
Effects on response selection
The main result in Experiment 1 and 3 was that spatiotemporal history affected spatial
selection within the object, as measured by target location repetition. In turn, the target
location repetition interacted with response, such that a repeated target location further
facilitated a repeated response. Response selection per se did not interact with
spatiotemporal history, although in Experiment 3 response selection depended on
overall object appearance, both in terms of identity and spatiotemporal history. In all,
this suggests a cascade of processes first retrieving target location information on the
basis of object history, followed by response selection on the basis of the target location.
Such interactions with response selection fit well within the theory of event
coding (Hommel, 1998; Hommel, et al., 2001). This theory states that task-relevant
stimulus and response features are integrated into an episodic structure called an event
file (comparable to Logan’s instance theory (Logan, 1988) . These event files are nicely
illustrated by Experiment 4 of Keizer, Colzato and Hommel (2008). They presented
participants with two sequential images of faces or houses (S1 and S2), which moved in
a certain diagonal direction. Regardless of its content, S1 had to be responded to with a
pre-defined left or right response (as indicated by an arrow prior to the image).
CHAPTER 4
83
However, S2 had to be responded to on the basis of its motion direction, again with a left
or right key press. On each trial, S2 could be the same as or different from S1 in terms of
content (house or face) and motion direction (and thus also the required response) It
was found that response times were longer when one feature was repeated while the
other was alternated, compared to complete repetition or alternation of both features.
Keizer et al. reasoned that we integrate the given response with the observed
combination of features into an event file. When both movement direction and
face/house features were then repeated, the previous response could quickly be
retrieved and executed from the event file. However, when one feature was repeated
and the other was alternated, the stimulus was partly In conflict with its event file,
resulting in response penalties. In the last case, in which both features were alternated,
there was neither a match nor conflict with an event file and thus no effect on responses.
In the present study too, the spatiotemporal properties as well as identity of the display
object may be integrated with the given response, thus facilitating the same response on
the next trial, when the same combination appears again.
The Keizer et al. (2008) study also provides an alternative explanation for our
object-based effects. So far we have suggested that selection settings are bound to
objects, as defined by their spatiotemporal continuity. Alternatively, these settings may
be tied to motion direction per se (regardless of it being the same object or not). Note
that in the present set up, the re-appearance of a previous object meant that it came
from the same side as on the previous trial, and thus had the same motion direction.
However, using very similar displays as in the current study, Schreij and Olivers (2009)
have dissociated movement direction from spatiotemporal object continuity by enabling
an object to move to and from four different sides of the screen. Therefore, an object did
not need to move back to where it had just come from and could thus re-emerge from a
different direction on the next trial. Schreij and Olivers found a similar same object
benefit regardless of whether the direction of movement was the same, suggesting that
it is not motion direction, but object representation that drives the effect. A further
argument against binding at the level of motion direction is that Keizer et al. found a link
between object identity and motion direction, whereas in Experiment 3 we failed to find
such a link, suggesting a potential dissociation between motion binding and object
binding.
OBACS FOR FEATURE INFORMATION
84
In conclusion, task-relevant selection settings, like for target locations and
features are maintained in the spatiotemporal representation that we create for an
object. Information that is not directly relevant for the task, like the presence of a
distractor, or a task-irrelevant target feature, is most likely not stored in this
representation. Although an object’s physical appearance does not affect spatial
selection of target information inside that object, it does influence response selection
given that the object is spatiotemporally continuous. This suggests that we possibly
store information inside an object representation at different levels, which each
influence different stages of attention and response selection the next time we
encounter and interact with the same object.
Chapter 5 Abrupt onsets capture attention
independent of top-down control settings
Schreij, D., Owens, C. & Theeuwes J. (2008) Abrupt onsets capture attention independent of top-down control settings
Perception & Psychophysics, 70 (2), 208-218
ATTENTIONAL CAPTURE BY ABRUPT ONSETS INDEPENDENT OF ACS
86
Abstract
Previous research using a spatial cueing paradigm in which a distractor cue preceded
the target has shown that new objects presented with abrupt onset only capture
attention when observers are set to look for them (Folk, et al., 1992). In the present
study we used the same spatial cueing paradigm and demonstrated that even when
observers have an attentional set for a color singleton or a specific color feature, an
irrelevant new object presented with abrupt onset interfered with search. We also show
that the identity of the abrupt onset distractor affects responses to the target, indicating
that at some point spatial attention was allocated to the abrupt onset. We conclude that
abrupt onsets or new objects over-ride top-down set for color. Abrupt onsets or new
objects appear to capture attention independently of top-down control settings.
CHAPTER 5
87
Introduction
One of the most fundamental questions is whether we are able to control what we select
from our environment. Overt or covert selection may either be controlled by the
properties of the stimulus field, or by intentions, goals and beliefs of the observer (see
recent reviews, e.g. Burnham, 2007; Rauschenberger, 2003a; Ruz & Lupianez, 2002;
Theeuwes & Godijn, 2001). When we intentionally select only those objects and events
needed for our current tasks, selection is said to occur in a voluntary, goal-directed
manner. When, irrespective of our goals and beliefs, specific properties present in the
visual field determine what we select, this selection is said to occur in an involuntary,
stimulus-driven manner. These two mechanisms of selection have been referred to as
bottom-up and top-down attentional control (Eriksen & Hoffman, 1972; Hoffman, 1979;
Posner, 1980; Theeuwes, 1991a; Yantis & Jonides, 1984). When objects or events receive
priority of processing, independent of the observer's goals and beliefs, one refers to this
as attentional capture when such events or objects only capture our attention (e.g.
Yantis, 1996). When such events trigger an exogenous saccade to the location of the
object or event, one refers to this as oculomotor capture (Theeuwes, et al., 1998).
Even though the controversy whether salient, static singletons capture attention
in a purely bottom-up way continues (for recent discussions see Leber & Egeth, 2006;
Theeuwes, 2004; Theeuwes & van der Burg, 2008), there appears to be less controversy
about attentional capture by suddenly appearing new objects or abrupt onsets. The
finding that abrupt onsets might capture our attention dates back to the early research
of Eriksen and Hoffman (1972) and Jonides (1981), which showed that participants’
attention was automatically drawn to an exogenous cue. Subsequent research by Todd
and Van Gelder (1979) showed that onset stimuli were detected faster than their no-
onset counterparts in tasks requiring rapid eye movement responses. As the task
demands were made more complex, Todd and van Gelder (1979) observed that the
advantage for onset stimuli increased with the complexity of decisions that had to be
made by participants. Yantis and Jonides (1984) demonstrated that peripheral cues
captured attention because of their abrupt onset. In their experiments, participants had
to search for a specific target letter embedded in an array of two or four non-target
letters. While participants searched for the target letter, a new letter suddenly appeared
in an empty location.
ATTENTIONAL CAPTURE BY ABRUPT ONSETS INDEPENDENT OF ACS
88
Following these demonstrations of the special status of onsets on attentional
capture, Theeuwes and colleagues (Theeuwes, et al., 1998; Theeuwes, Kramer, Hahn,
Irwin, & Zelinsky, 1999) showed that abrupt onsets also have the ability to capture the
eyes. In this so called oculomotor capture paradigm, participants were instructed to
make a saccadic eye movement towards the only gray element in the display. On some
trials, an irrelevant new object presented with abrupt onset was added to the display.
Participants knew the onset was irrelevant and also knew that they had to ignore it. The
condition in which a to-be-ignored onset was presented somewhere in the visual field
was compared to a control condition in which no sudden-onsets were added to the
display. The results showed that when no new object was added to the display,
observers made saccades that generally went directly to the uniquely colored circle.
However, in those trials in which a new object was added to the display, in about 30 to
40% of the trials the eye went in the direction of the abrupt onset. Moreover, in a
subsequent eye movement study Theeuwes and colleagues (Theeuwes, De Vries, &
Godjin, 2003) showed that under the very same circumstances irrelevant salient static
singletons (such as a uniquely color element) only captured attention but not the eyes.
Therefore transient singletons seem to have a different effect than static singleton
confirming the special role of abrupt onsets.
All these studies demonstrate the special status of abrupt onsets in capturing
attention. The reason for this special status in capturing attention may be because
onsets are accompanied by luminance transient (e.g. Jonides & Yantis, 1988; Theeuwes,
1990, 1994b, 1995a; Yantis & Jonides, 1984) or because they represent a new object
(Abrams, Davoli, & Suszko, 2007; Yantis & Hillstrom, 1994). Regardless of the
underlying mechanism, it is generally agreed that onsets have the ability to capture
attention in an exogenous way.
Whether attentional capture by onsets is truly exogenous has been challenged by
the Involuntary Contingent Orienting Hypothesis of Folk, Remington and Johnston (1992).
According to this hypothesis whether or not an object captures attention is completely
dependent on attentional control settings. Participants are able to compose a certain
attentional set, which contains the dimensions or features of the target the participant
has to look for in a task (also called task set). Only elements in the visual field that
possess the properties that match the information in the attentional set, will capture
attention. This holds for static stimuli, as well as dynamic events like onsets or motion.
CHAPTER 5
89
To provide evidence for this hypothesis, Folk et al. (1992) conducted an experiment in
which participants were induced to adopt an attention set for a certain stimulus
property, like an onset or a color. In their paradigm, the presentation of a search display,
which contained a stimulus dimension participants had to look for, was preceded by a
cue which gave an incorrect or correct (henceforth called valid or invalid) indication of
the location where the target would appear. This cue could be the same or a different
dimension as the target element. For example, a color search display could either be
preceded by an onset cue or by a color cue. The critical finding here was that only when
the cue was of the same dimension as the target element, a considerable validity effect of
the cue was found. When the cue was from a different dimension as the target, it did not
affect the response times to the target, regardless of its validity.
The contingent capture hypothesis of Folk et al. (1992) is consistent with notions
put forward by Bacon and Egeth (1994) regarding top-down set for specific search
modes (i.e., feature search vs. singleton detection mode) and with notions suggested by
Yantis and Egeth (1999) regarding an top-down attentional set for singletons. In this
respect, the contingent capture hypothesis of Folk et al. (1992) may account for findings
of attentional capture by static singletons (Hickey, McDonald, & Theeuwes, 2006;
Theeuwes, 1991b, 1992; Theeuwes, et al., 2000). However, the predictions of the
involuntary contingent orienting hypothesis regarding abrupt onsets seem to be
inconsistent with earlier findings which have shown that abrupt onsets are unique in the
ability to capture attention without an attentional set (Jonides & Yantis, 1988; Yantis &
Egeth, 1999).
The discussion whether it is transient luminance or the appearance of a new
object that causes abrupt onsets to capture attention (e.g. Yantis & Hillstrom, 1994) led
to discussions regarding the original Folk et al. (1992) studies. The onset in Folk et al.’s
experiments consisted of the appearance of a character inside a boundary box, instead of
the presentation of an object a previously empty location. This could be a violation of the
requirement that the onset has to be a new perceptual object to capture attention.
Instead of being regarded as the appearance of a new perceptual object, these onsets
could just be perceived as a property change of a previous present object (the bounding
box), which by itself does not always capture attention (Jonides & Yantis, 1988). To
address this issue, Folk and Remington (1999) conducted a series of new experiments
using new perceptual objects in combination with their usual pre-cueing paradigm.
ATTENTIONAL CAPTURE BY ABRUPT ONSETS INDEPENDENT OF ACS
90
Consistent with earlier findings, they found that onsets presented in empty locations
(i.e., so called new objects) only captured attention when participants were set for
onsets but not for color. In line with their hypothesis, capture was fully contingent on
the attentional control settings.
However, there is one important difference between the experimental paradigms
favoring capture by abrupt onsets and salient singletons (Theeuwes, 1992, 1994b;
Yantis & Jonides, 1984) and the pre-cueing paradigm of Folk and colleagues supporting
the contingent orienting hypothesis. Experiments using the first paradigm, presented
the target and distracting element simultaneously exactly at the moment participants
needed to start searching. However, in the classic pre-cueing paradigm of Folk et al.
(1992) the distracting element (the cue) preceded the search display by 150 ms. In other
words; participants had to ignore a “cue” that preceded the search display. As argued by
Theeuwes et al. (2000) it is possible that the delay between cue and search display was
long enough to overcome attentional capture by the irrelevant cue (see also Theeuwes,
1994b). In other words, disengagement of attention from the cue may have been
relatively fast when the cue and target did not share the same defining properties (e.g.,
the cue was red and the target was an onset), while disengagement from the cue may
have been relatively slow in cases where the cue and target share the same defining
properties (e.g., both were red). Such a mechanism could explain why there are RT costs
when the cue and target have the same defining characteristics and no costs when cue
and target are different. In this view, the contingent capture hypothesis can explain why
it may be easier to disengage attention from a particular location when an element
presented at that location is not in line with the top-down control settings. However, this
does not imply that there is no capture of attention by the irrelevant cue singleton; it
simply indicates that after a certain time participants are able to exert top-down control
over the erroneous capture of attention by the irrelevant singleton.
In line with this explanation, Theeuwes and colleagues (2000) provided strong
evidence for the claim that once attention is captured by an irrelevant singleton it only
takes a very brief time to disengage attention from that location. Theeuwes et al. (2000)
used a visual search task similar to that of Theeuwes (1992) in which participants
searched for a shape singleton (a single gray diamond among 8 gray circles). Prior to the
presentation of the target display a color singleton was presented at different SOAs (50,
100, 150, 200, 250 and 300 ms). Theeuwes et al. showed that for conditions in which
CHAPTER 5
91
target and distractor were presented in close temporal proximity (< 100 ms), the
distractor interfered with search, suggesting that there was not enough time to
overcome attentional capture. However, when the singleton distractor was presented a
considerable time (SOAs 150 to 300 ms) before the presentation of the singleton target,
the distractor no longer interfered with search suggesting that participants were able to
overcome capture by the irrelevant singleton
The current research was intended to further explore the ability of abrupt onsets
to capture attention, while using the classic pre-cueing paradigm of Folk et al. (Folk &
Remington, 1999; Folk, et al., 1992; Folk, Remington, & Wright, 1994). A search display
was preceded by a cue display of the same dimension as the target. Participants were set
for color, since they had to search for a color singleton throughout the whole
experiment. In some trials an abrupt onset (i.e., a new perceptual object) was presented
at a random and empty location. However, unlike Folk and Remington (1999) we
presented the abrupt onset not during the cue display but simultaneously with the
search display, as is typically done in traditional visual search experiments investigating
the role irrelevant distractors (e.g. S. E. Christ & Abrams, 2006; Theeuwes, 1994b,
1995a; Yantis & Johnson, 1990; Yantis & Jonides, 1990). As noted, by presenting the
target and the onset distractor simultaneously, the data will reveal any potential capture
effect of the onset distractor because unlike in the spatial cueing paradigms of Folk and
Remington (1998, 1999) there is no time to recover from capture (see Theeuwes, et al.,
2000).
Experiment 1
The current experiment used the Folk et al. (1992) pre-cuing paradigm. We created two
experimental conditions. One condition was regarded as the “no-onset” condition and
was the same as the color cue, color target condition of Folk et al. (1992). The
participant’s goal was to find a red character among white characters which appeared
inside placeholder boxes at four possible locations in the visual field. Before the search
display appeared, a pre-cue that had the same color as the target was presented at any of
the four potential target locations. Since the cue was a color singleton and participants
were set for color, it was expected that the cue would capture attention and result in a
significant difference in response times depending on the validity of the cue, just as in
Folk et al. (1992).
ATTENTIONAL CAPTURE BY ABRUPT ONSETS INDEPENDENT OF ACS
92
In the other condition, the “onset” condition, a boundary box containing a white
character appeared at a random empty location between two of the already present
boundary boxes. This extra character could be regarded as an abrupt onset of a new
perceptual object. The Contingent Involuntary Orienting Hypothesis of Folk et al. (Folk &
Remington, 1999; Folk, et al., 1992; Folk, et al., 1994) would predict that when
participants are set to search for color, the sudden appearance of an abrupt onset should
have no effect on performance.
To ensure that difference in response times could be attributed to attentional
capture and not by “attentional misguidance”, it was made sure that the appearance of
the new object would meet the criteria set by Yantis (1993) for stimulus-driven
attentional capture. This means that the distracting element should not share a defining
or reporting property (Duncan, 1985) with the target character. In this case, the defining
property was the red color of the target character, which the onset did not share since it
was white. The reporting property in this task was character identity, so it was made
sure that the identity of the onset distractor never had the identity of a target character.
Since in the current task the onset character would always be an ‘O’ and the target
character would either be an ‘X’ or an ‘=’, one can argue that this demand was met as
Cue display Fixation display Fixation display
Search display
1000-1400 ms 50ms 100 ms
Until response (or 2000ms)
No Onset
Onset
Figure 13: The sequence of events for a typical trial. First a fixation display was shown for 500ms, after which the central fixation cross was turned off for 50 ms. Then the fixation display was shown again for a random period of 1000, 1100, 1200, 1300 or 1400 ms. The cue display was presented for 50 ms. After an ISI of 100ms, the search display was presented for 2000 ms or until the participant responds. This is an example of an invalid trial, since the location of the red dots in the cue display and the location of the red character in the search display differ.
CHAPTER 5
93
well, since the distractor identity was not among the target identities the participants
were to respond to.
METHOD
Participants
Fourteen first-year students from the University of Sydney, School of Psychology
participated in the study, in exchange for course credit. The participants ranged in age
from 18 to 25 and all reported normal or corrected-to-normal visual acuity and color
vision.
Apparatus
The stimuli were presented on a 15” TFT screen with a Dell OptiPlex GX520, containing
an Intel Pentium IV 3 GHz and 512 MB of internal memory. The experiment was created
and run with E-prime 1.1 (SP3). The slides consisted of BMP images and had a resolution
of 640x480 pixels.
Stimuli
There were three basic types of displays: a fixation display, a cue display and a search
display, all of which had a black background. The fixation display consisted of a light
white fixation cross at the center of the screen, surrounded by four light gray,
RGB(167,169,172), placeholder boxes measuring a width of 2.6o visual angle, using an
approximate distance to the screen of 40 cm. The four boxes were positioned above,
below, to the left and to the right of the fixation cross, along a virtual circle with a
diameter of 20o visual angle, with the fixation cross as the center.
The cue display consisted of the same elements as the fixation display, with the
addition of four dots, with a diameter of 1.4o visual angle, positioned along the outside of
the center of each rib of all the placeholder boxes. One set of these dots surrounding one
of the placeholder boxes had a red color, (RGB (236,43,39); luminance 62.2 cd/m2) and
indicated the cued location. All the dots surrounding the other boxes had a bright white
color, RGB (255,255,255). In the search display an “X” (Myriad Roman, 21pt) or “=”
(Myriad Roman bold, 22pt) were placed in each of the boxes. There were always two ‘X’s
and two ‘=’ present.
ATTENTIONAL CAPTURE BY ABRUPT ONSETS INDEPENDENT OF ACS
94
Three of the characters inside the boxes were bright white and one was red,
which designated the target character. The search display could contain all possible
combinations of characters in the four boxes of which one character always was red.
In the onset condition one extra character “O” with a bright white color and
placed inside a light grey bounding box, would appear in the search display, located
between two other boxes on the virtual circle on which all other boxes were placed.
Examples of these display types, along with their order of appearance are presented in
Figure 13.
Design
There were two within-subject conditions. In the onset condition an additional object
(an abrupt onset) was added to the display. In the no-onset condition, everything was
the same except that no new object was added to the display. In both conditions, the cue
display could give a correct or incorrect indication of the location where the target
character might appear, but was correct at chance level. In 25% of the trials the red dots
surrounded the box where the target character would appear, which was considered a
valid cue, and in the other 75% of the trials it was invalid and the red dots surrounded a
box other that the target.
In the onset condition an extra distractor character was presented
simultaneously with the search display. This character appeared inside a bounding box,
identical to the other four boxes present on the display, in a previously empty location.
The extra onset character could appear between any of two other already present
placeholder boxes and appeared equally often in each of the four possible locations,
throughout the experiment.
The onset and no-onset condition were presented in 6 mixed blocks of 80 trials
each. 50% of the trials had no-onset search displays and the other 50% consisted of
onset search displays. These two types of search displays were randomly mixed within a
block.
Procedure
Participants were tested in a one hour session. Before the experiment started, oral
instructions were given to familiarize them with the task to be performed. It was
stressed that they keep using both hands for pressing the two different buttons on the
keyboard and to not move their eyes away from fixation during a trial, because this
CHAPTER 5
95
could impair their performance. The experiment commenced with the presentation of
general instruction slides that explained the main course of the experiment.
Participants were told that the abrupt onset was irrelevant to the task and would
never contain a target character. Finally, the participants were stressed to react as fast
as possible without making too many mistakes. After these general instruction slides,
the participant began with a block of 40 practice trials. They were prompted to press the
space bar when they were ready to begin with the real trials, after which they were
presented with 6 blocks of 80 trials. At the end of each block, participants were advised
to take a rest and were forced to wait for 30 seconds, before they could press a key to
begin with the next block of trials.
Trials began with the presentation of the fixation display for 500 ms. Then, the
fixation cross blinked off and on for 50 ms, to notify the participant of the start of a trial.
The fixation display remained on screen for a period randomly chosen from a set of
1,000, 1,100, 1,200, 1,300 or 1,400 ms, to eliminate any effects of expectancy. After this
fore period, a cue display was presented for 50 ms, after which the fixation display was
shown again for 100 ms. This served as an inter-stimulus interval (ISI) after which the
search display was presented until the participant responded or, when no response was
detected, for 2,000 ms. When no response was detected the trial was counted as an
error. Throughout a trial, the four place holder boxes were constantly visible.
After a response was given, a distinctive sound was played for a correct or
incorrect response. If the response was incorrect, the experiment paused for 10 seconds,
showing a counter counting from 10 to 0 seconds, to let the participants regain their
focus. If a response was not made before 2000 ms, it was registered as an error and the
participant had to wait 10 seconds before the trial procedure continued. Following the
response of a participant, there was an intertrial interval of 500ms, before the fixation
cross blinked again to indicate the start of the next trial.
Results
All response times (RTs) above 1000 ms (which is approximately 4 SD from the mean)
were regarded as errors and removed from the data set, as were incorrect responses.
This led to a loss of only 5.2% of the trials. Figure 14 shows the participants’ mean RT
and error percentages in the SOA and cue validity conditions. The individual mean RTs
ATTENTIONAL CAPTURE BY ABRUPT ONSETS INDEPENDENT OF ACS
96
were submitted to a repeated measures analysis of variance (ANOVA) with onset
presence (onset or no onset) and cue validity (valid or invalid) as factors.
There was a significant main effect of the presence of the sudden-onset, F(1,13) =
7.915, p<.05, such that participants were slower in their response to the target when an
onset was present. In addition, cue validity was highly significant F(1,13) = 276.852, p <
.001, replicating the traditional Folk et al (1992) effect demonstrating that a cue that
shares the feature properties of the target captures spatial attention. It is important to
note that there was no significant interaction between the validity of the cue and the
presence of an onset F(1,13) = 1.329, p = .270.Error rates for each condition were well
below 10%. Participants made significantly more errors in the invalid than in the valid
cue condition, F(1,13) = 15.443, p=.002, indicating that an invalid cue not only made
participants respond slower, but also less accurate. There were no differences in error
rates between the onset and no onset conditions, F(1,13) =.240.
Figure 14: Experiment 1: Mean Response time for validly and invalidly cued positions with and without the presence of an abrupt onset. Error bars represent the standard error of the mean difference scores between onset vs. no-onset for each validity condition.
CHAPTER 5
97
Discussion
The present results replicate the one of main findings of Folk et al. (1992): when
participants have an attentional set for a color, a “to-be-ignored” cue that has the same
color as the target captures attention. Even though the cue was not informative about
the location of the upcoming target singleton, participants were not able to ignore it.
This is a replication of the classic finding of Folk et al. (1992) and signifies the notion
that attentional capture is (or at least can be) contingent on top-down settings which are
established ‘off-line’ on the basis of current attentional goals. According to the
‘contingent capture’ model, only stimuli that match the top-down control settings will
capture attention; stimuli that do not match the top-down settings should be ignored.
Even though it is clear that participants were set for a color singleton, the
presence of an abrupt onset slowed responding. Indeed, regardless of whether the color
cue was valid or invalid, in the onset present condition RT was slowed by about 10 ms.
One way to explain this slowing is that attention was captured by the abrupt onset, an
interpretation that is consistent with findings from many previous studies and various
paradigms showing the ability of onsets in capturing spatial attention (see e.g.
Belopolsky, et al., 2005; S.E. Christ & Abrams, 2008; Donk & van Zoest, 2008; Gellatly,
Cole, & Blurton, 1999; R. W. Remington, Johnston, & Yantis, 1992; Theeuwes, 1990,
1994b; Yantis & Jonides, 1984). If the distraction effect caused by the abrupt onset is
indeed due to attentional capture, then one has to conclude that this finding is
inconsistent with the contingent capture hypothesis of Folk et al. (1992) because, as we
show in this experiment participants were set for color and therefore onsets should not
capture attention.
Experiment 2
The first experiment showed that even when the classic Folk et al. (1992) spatial cueing
paradigm is used, there is an effect of the appearance of an onset distractor when
presented at an empty location in the search display. This finding is consistent with
other studies using different types of paradigms demonstrating the extent to which
abrupt onsets can capture attention (S.E. Christ & Abrams, 2008; R. W. Remington, et al.,
1992; Theeuwes, 1994b, 1995a).
Even though the present study confirms the notion that onsets capture attention
in a purely exogenous way, one could argue that in the present experiment participants
ATTENTIONAL CAPTURE BY ABRUPT ONSETS INDEPENDENT OF ACS
98
were set to look for onsets. Indeed, the onset was presented simultaneously with the
target elements (the “X” and “=”) inside the placeholder boxes, making it possible that
the abrupt onset captured attention because participants were set to look for it. For
example, Gibson and Kelsey (1998) argued that the onset of a new object (an abrupt
onset) may capture attention because observers are prepared for the abrupt onset of the
entire display (see also Burnham, 2007). Participants may adopt such a set for display-
wide features, because the abrupt onset of the search display typically signals the
presence of the target in a very general sense. In other words, it is feasible that in
addition to an attentional set for color, participants also adapted a default set for abrupt
onsets because the abrupt onset of the elements inside the placeholder boxes signalled
the presence of the target.
In order to address this concern, in Experiment 2 the pre-masking elements were
placed inside the placeholder boxes. They consisted of an overlapping ‘X’ and ‘=’ and ‘|’
to hide the identities of the characters to appear. When the identities of the elements to
search needed to be revealed, the irrelevant line segments were removed in analogy to
figure-eight pre-masking characters as used in Yantis & Jonides (1984).
METHOD
Participants
Thirteen first year students from the University of Sydney participated in this
experiment in exchange for course credit. The participants ranged in age from 18 to 24
and all reported normal or corrected-to-normal visual acuity and color vision. None of
the participants had participated in the previous experiment of this study.
Experiment 1 Experiment 2
Figure 15: Examples of the fixation displays used in Experiment 1 and Experiment 2
CHAPTER 5
99
Apparatus and Stimuli
The apparatus and stimuli were similar to Experiment 1, with the exception of white
pre-masking characters (62.2 cd/m2), which are placed inside the boundary boxes.
Figure 15 provides an example. When the search display was presented, the extra line
segments hiding the characters on the search display were removed, making the
characters to search visible.
Design and Procedure
The design and procedure was identical to those of experiment 1. Instead of appearing
at the empty location inside a placeholder box, the target character was now revealed by
changing the color of one of the pre-masking characters and, at the same time, removing
the line segments that hide it.
Results
Erroneous responses were removed, as were responses above 1000 ms which led to a
loss of 5.2 % of the trials. Figure 16 shows the participants’ mean RT and error
percentages in the SOA and cue validity conditions. As in Experiment 1, there was a
main effect of onset presence, F(1,12) = 28.410, p < .001, indicating that the presence of
an onset slowed search for the color singleton. Also, the main effect of cue was highly
significant, F(1,12) = 78.470, p < .001. Consistent with Experiment 1, the interaction
between cue validity and onset presence was not reliable F(1,12) = 0.170, p = .687.
Error rates were well below 10%. Again, participants made significantly more
errors in the invalid relative to the valid cue condition, F(1,12) = 21.429, p=.001. There
were no differences in error rates for the onset and no-onset condition, F(1,12) =.302.
Discussion
The current results are basically identical to those of Experiment 1. The presence of an
onset distractor resulted in longer response latencies relative to a condition in which
there was no onset present. The results indicate that whether or not the search elements
placed inside the placeholder boxes were presented with onsets (as in Experiment 1) or
with offsets (as in Experiment 2) had no effect on the impact of the abrupt onsets. It
appears that in the current experiments, in addition to an attentional set for color,
participants did not adopt a general default set for onsets as is advocated by the display-
wide visual feature notion of Gibson and Kelsey (1998).
ATTENTIONAL CAPTURE BY ABRUPT ONSETS INDEPENDENT OF ACS
100
Experiment 3
Experiment 1 and 2 clearly show that the presence of an abrupt onset slows responding
to target. Even though participants were set for color, the onset had an effect on their
performance. Even as previous studies (Belopolsky, et al., 2005; S.E. Christ & Abrams,
2008; Enns, et al., 2001; Gellatly, et al., 1999; Theeuwes, 1990, 1994b; Yantis & Jonides,
1984) suggested that onsets capture attention, one still could argue that the 15 to 25 ms
cost caused by the onsets has nothing to with attentional capture. Indeed, Folk and
Remington (1998) offered an alternative explanation for increases in response times in
conditions in which a distractor was present. They suggested that the increase in search
time caused by the irrelevant singleton is due to what they call "filtering costs" a notion
first introduced by Kahneman, Treisman and Burkell (1983). In the current experiments
the presence of the abrupt onset may have slowed the deployment of attention to the
target item by requiring an effortful and time-consuming filtering operation. According
to this line of reasoning, attention goes directly to the uniquely colored item; and simply
Figure 16: Experiment 2: Mean Response time for validly and invalidly cued positions with and without the presence of an abrupt onset. Error bars represent the standard error of the mean difference scores between onset vs. no-onset for each validity condition.
CHAPTER 5
101
because the onset is present, directing attention to the uniquely colored item takes more
time than when no onset is present. The filtering cost explanation is compatible with the
contingent capture hypothesis because spatial attention only goes to the item (the
uniquely color item) that matches the attentional set for color; it assumes that spatial
attention does not go the location of the abrupt onset.
To determine whether the performance costs due to the onset are the result of
erroneous attentional capture or filtering costs, we employed the so-called “identity
intrusion technique” first introduced by Theeuwes (1995b) and Theeuwes and Burger
(1998). Instead of presenting a neutral character “O” inside the abrupt onsets, in
Experiment 3, the element inside the abrupt onset was either compatible or
incompatible to the response to the target. The underlying notion is that if attention is
allocated to the location of the abrupt onset distractor, its identity will be processed (e.g.
Kramer & Jacobson, 1991). Given this assumption, if attention is captured by the abrupt
onset, then a compatibility effect should be found, with longer RTs when the element
inside the distractor is incompatible with the target than when it is compatible. If the
abrupt onset does not capture attention and spatial attention is never allocated to the
location of the onset, then one does not expect a compatibility effect whatsoever.
METHOD
Participants
Twenty-one first-year students from the University of Sydney, School of Psychology
participated in the study, in exchange for course credit. The participants ranged in age
from 18 to 25 and all reported normal or corrected-to-normal visual acuity and color
vision. None of the participants had participated in any of the previous experiments.
Apparatus and Stimuli
The apparatus and stimuli were identical to Experiment 2 with the exception of the
onsetting element. Instead of an “O” character inside the onset placeholder, an “X” or “=”
would appear with the same font and color properties as all the other characters.
Design and Procedure
The design and procedure were basically the same as in the previous experiments
except that the compatibility of the element inside the onset distractor was manipulated.
When an onset distractor was present, in half of the trials the item inside the onset
ATTENTIONAL CAPTURE BY ABRUPT ONSETS INDEPENDENT OF ACS
102
distractor was compatible with the response to the target; in the other half it was
incompatible. There were 6 blocks of 80 trials.
Results
Erroneous trials and trials with responses above 1000 ms were removed from the data,
which resulted in a total loss of 6.7% of the trials. Figure 17 presents the results. To
investigate whether the overall presence of a distractor had an effect, the individual
mean RTs were submitted to an analysis of variance, with distractor presence (onset vs.
no onset), cue validity (valid vs. invalid) as factors. The main effect of distractor
presence was significant, F(1,20) = 22.977, p < .001, indicating the presence of the onset
Figure 17: Experiment 3: Mean Response time for validly and invalidly cued positions with and without an abrupt onset. In the compatible condition the character inside the abrupt onset is compatible with the response to the target; in the incompatible condition it is incompatible with the response to the target. Error bars represent the standard error as recommended by Masson and Loftus (2003).
CHAPTER 5
103
slowed search. Again, the cue validity was significant, F(1,20) = 75.774, p < .001. As in
the previous experiments, there was no significant interaction between the cue
condition and presence of the onset distractor, F(2,40) = 0.321, p =.649,
Another ANOVA was performed on the individual mean RTs for the distractor
conditions, with compatibility (compatible vs. incompatible) and cue condition (valid vs.
invalid) as factors. The main effect of compatibility was significant, F(1,20) = 5.222, p <
0.05, indicating the identity of the element inside the onset distractor affected the speed
with which participants responded to the target. When the element inside the onset
distractor was compatible participant responded faster (729 ms) then when it was
incompatible (759 ms). The interaction between cue validity and compatibility failed to
reach significance, F(1,20) = 0.929, p = .347.
All error rates were well below 10%. Again error rates were higher in the invalid
condition than in the valid condition F(1,20) = 15.977, p<.001. The presence of the onset
had no significant effect on errors, F(1,20) = .186 nor did the compatibility of the onset
distractor with the target F(1,20) = 1.463.
Discussion
The current experiment shows a small but clear compatibility effect, suggesting that
attention was allocated to the location of the onset distractor. The effect size of
compatibility is comparable to that reported by Theeuwes (1995b). On the basis of this
finding, one has to conclude that even when participants are set to look for a color
singleton, irrelevant abrupt onsets can capture attention. Note that the effect of the
onset is not modulated by the validity of the cue suggesting that even when attention is
directed towards the location where the target item is going to appear, the onset may
pull attention away towards the location of the onset. The observed compatibility effect
indicates that the onset does not merely cause some type of non-spatial filtering costs
but shows that the onset truly pulls attention to its location.
Experiment 4
The results of Exp. 1-3 show that when participants have a clear attentional set for color,
an irrelevant abrupt onset captures attention. Even though this clearly suggests that
onsets capture attention independent of a top-down set for color, one still may rescue
the contingent capture hypothesis of Folk et al. (1992) by assuming that in our
Experiment 1-3, participants engaged in what has been called singleton detection mode
ATTENTIONAL CAPTURE BY ABRUPT ONSETS INDEPENDENT OF ACS
104
(Bacon & Egeth, 1994; see also Lamy & Egeth, 2003). The idea behind this is that
participants can choose to search in a particular search mode. When participants engage
the singleton detection mode, they choose to direct attention to the location having the
largest feature contrast. In this mode, the most salient singleton will capture attention
regardless of whether it is the target or not. If, however, participants engage what is
called a feature detection mode they choose to direct their attention to a particular
feature (e.g., the color red) instead of to any singleton. According to Bacon and Egeth
(1994) in this mode "goal directed selection of a specific known featural singleton identity
may override stimulus-driven capture by salient singletons" (p.493). Bacon and Egeth
(1994) suggested that when observers 'choose' a feature search mode, attentional
capture by irrelevant singletons is eliminated. The notion that choosing a search
strategy allows attentional control suggests that attentional capture is under top-down
control (but see Theeuwes, 2004 who criticized the circularity of these search concept).
If we apply this type of reasoning to the current experiments, it is possible that
participants searched in singleton detection mode, allowing the onset singleton to
capture attention. Because participants were always looking for a singleton (the only red
element among gray elements), it is possible that the other singleton (i.e., the onset)
captured attention. To test whether the onset captures attention even when participants
were engaged in a feature search mode, we changed the display such that participants
no longer could search for a singleton. Instead of a red target among gray non-target
elements, participants had to search for one particular color (e.g., red) among elements
which each had a different color (e.g., green, yellow and blue). This way, participants
were forced to use the feature search mode. The question was whether if even in this
set-up the onset would capture attention.
METHOD
Participants
Eight students from the Vrije University Amsterdam were paid for participation. They
ranged in age from 18 to 24 and all reported normal or corrected-to-normal visual
acuity and color vision.
CHAPTER 5
105
Apparatus and Stimuli
The apparatus and stimuli were similar to Experiment 2, with the exception that the
non-target characters were colored instead of white. The colors red, green, yellow and
blue (all matched for luminance at 29 cd/m2) were randomly assigned to each of the
four characters.
Design and Procedure
Participants were instructed to look for one particular color throughout the whole
experiment. Participants were balanced across the four different colors such that two
participants consistently searched for red, two for green, two for yellow and two for
Figure 18: Experiment 4: Mean Response time for validly and invalidly cued positions with and without the presence of an abrupt onset in a task in which participants search for a particular color feature. Error bars represent the standard error of the mean difference scores between onset vs. no-onset for each validity condition.
ATTENTIONAL CAPTURE BY ABRUPT ONSETS INDEPENDENT OF ACS
106
blue. The color of the cue in the cue display matched that of the color a particular
participant was looking for.
Results
Erroneous trials and trials with response above the 1000 ms were removed from the
data set, which resulted in a loss of 8% of the total trials. Figure 18 displays the
participants’ mean RT and error percentages in the onset presence and cue validity
conditions. There was a main effect of onset presence, F(1,7) = 6.575, p = .037, indicating
that the presence of an onset slowed search for the color singleton. Also, the main effect
of cue was significant, F(1,7) = 11.731, p = .011. The interaction between cue validity
and onset presence was not reliable, F(1,7) = 2.121, p = .189.
Error rates were well below 10%. Participants made significantly more errors in
the invalid than in the valid cue condition, F(1,7) = 7.881, p=.026. There were no
differences in error rates for the onset and no-onset condition, F(1,7) =.127, p=.732.
Discussion
The current results are basically identical to those of the previous three experiments.
Even when participants are forced to search for a particular color feature, the irrelevant
abrupt onset captured attention. The present findings suggest that regardless whether
participants are set to search for a unique color singleton (the singleton detection mode
as in our Exp. 2) or a specific color feature (the feature search mode as in our Exp. 4), an
abrupt onset captures attention and interferes with search.
General Discussion
The present results are clear. In conditions in which participants have a clear attentional
set for color regardless whether they are looking for a singleton or for a specific color,
they cannot prevent attentional capture by an irrelevant abrupt onset. The results are
consistent with the Contingent Capture Hypothesis in showing that an attentional set for
color results in strong capture by a color cue. However, according to Contingent Capture
hypothesis of Folk et al. (Folk & Annett, 1994; Folk & Remington, 1999; Folk, et al., 1992)
an attentional set for color should have prevented attentional capture by abrupt onsets
because Contingent Capture hypothesis assumes that capture is fully dependent on
attentional control settings.
CHAPTER 5
107
The pattern of results obtained in the present study indicates that the effect of
the attentional set on capture is very strong, generating cueing effects of about 40 ms.
The distracting effect of the onset is relatively small (about 8 to 10 ms) and appears to
be additive with the cueing effects. This pattern of results implies that the distracting
effect of the onset rides on top of the contingent capture effect of the color cue,
suggesting that independent of whether attention is allocated to a valid or an invalid
condition, the onset captures spatial attention for a very brief time before a response to
the target can be emitted.
The current findings are inconsistent with those of Folk and Remington (1999)
who conducted an experiment very similar to the current one. In their experiment they
also had conditions in which participants were set for a unique color and they showed
that a new object presented with abrupt onset had no effect when participants were set
for color. Indeed, the conclusion of this study was that the appearance of a new object
could not override a top-down set of color. Even though on the face of it, these studies
are very similar, there is one important difference between the current study and that of
Folk and Remington (1999). In the current study the irrelevant onset was presented
simultaneously with the search display, while in Folk and Remington (1999) the onset
was presented during the cue display, i.e., the onsetting new object was presented
before the presentation of the search display. In Folk and Remington’s Experiments 1
and 2 the SOA between cue and search display was 150 ms, a SOA identical to those used
in the original Folk et al. (1992) study. As argued, recovery from attentional capture can
be relatively fast (see Theeuwes, et al., 2000). Therefore, it is possible that in Folk and
Remington’s experiments participants had their attention captured by the abrupt onset
but were able to quickly disengage attention when they realized that the new object was
not a uniquely colored item. Data show that 150 ms SOA between cue and search display
certainly provides enough time to recover from capture (see Kim & Cave, 1999;
Theeuwes, et al., 2000). To address this issue, in Folk and Remington’s Experiment 3, the
SOA was reduced to 50ms. The reasoning was that if the abrupt onsets captures
attention then this should become visible when the search and cue display are presented
relatively close in time (i.e., within 50 ms). Even though this manipulation did not
change the overall pattern of results, Folk and Remington (1999) indicate that there was
a small distracting effect of the onset when participants were set for color. Indeed,
Figure 4 of Folk and Remington (1999) seems to suggest that the onset caused a
ATTENTIONAL CAPTURE BY ABRUPT ONSETS INDEPENDENT OF ACS
108
distraction effect of about 8 ms. As noted by Folk and Remington this effect could very
well indicate the “tail effect” of the recovery from capture. In our experiments the SOA
was basically “zero” because the cue and search display were presented simultaneously.
Obviously, in those conditions, the distracting effect of the onset (effect size of 10 to 20
ms) does become reliable as we demonstrated in all four experiments.
Folk and Remington (1999) also employed the identity intrusion technique (as
we used in our Experiment 3) to show in another way that the onset did not capture
attention. In their Experiment 4, they placed a character inside the abrupt onset that was
either compatible or incompatible with the response to the target. The abrupt onset with
the character was presented for 50 ms during the cue display, immediately masked for
50 ms followed by the search display after another 50 ms. In the critical condition, in
which participants were set for color, Folk and Remington did not find a compatibility
effect suggesting that the onset did not capture spatial attention. Note that we used the
very same technique in our Experiment 3 and showed that there was a small, yet reliable
compatibility effect. It is not immediately clear why we found a compatibility effect and
Folk and Remington (1998) did not, but it is feasible that when the onset is not relevant,
a very brief presentation of the interfering character (i.e., 50 ms) inside the onset
followed by a mask as was employed by Folk and Remington (1999) may not be long
enough to allow enough processing to cause a compatibility effect.
Alternatively, Folk and Remington (1998) argued that compatibility effects as
reported in previous studies (Theeuwes, 1995b; Theeuwes & Burger, 1998) may not
reflect the allocation of spatial attention but maybe the result of parallel processing of
target and distractor information. Even though this is a possibility, it still would imply
that at some point, either in serial or in parallel, attentional resources were allocated to
the location of the abrupt onset. Indeed, it seems to be unlikely that Folk and
Remington’s criticism of the identity intrusion technique implies that identity
information at the location of the distractor is processed without attention. Moreover,
Folk and Remington (1999) used this very same identity intrusion technique to argue
that the absence of a compatibility effect indicates that spatial attention did not go to the
abrupt onset or new object.
Even though the present study shows that abrupt onset capture attention even
when participants have a top-down set for color, this does not imply that onset capture
is never under top-down control. For example, Theeuwes (1991a) showed that when
CHAPTER 5
109
the upcoming target position was cued in advance by a 100% valid cue, an abrupt onset
presented elsewhere in the visual field ceased to capture attention (see also Juola,
Koshino, & Warner, 1995; Yantis & Jonides, 1990). In addition, Martin-Emerson and
Kramer (1997) showed that the capture effect of the abrupt onset is reduced with an
increasing number of no-onset elements in the display (see also Miller, 1989; von
Muhlenen, Rempel, & Enns, 2005). Also, Boot, Brockmole and Simons (2005) showed
that capture by onsets is eliminated when participants have to execute a concurrent
auditory task, suggesting that onset capture may dependent on the resources available.
The present study shows that onsets capture attention regardless whether
participants look for a color singleton (i.e., the singleton detection mode) or look for a
specific color (i.e., the feature search mode). Irrespective of these search modes and
consistent with the contingent capture hypothesis, the matching color cue captured
attention resulting in a large spatial cueing effect. At the same time, regardless of the
top-down search modes, the irrelevant onset captured attention. The discovery that the
search modes had no significant effect in our study adds to the growing number of
studies that supports the notion that these search modes may not be a very useful
distinction (Lamy & Egeth, 2003; Theeuwes, 2004; Theeuwes & van der Burg, 2008).
In conclusion, the present study demonstrates that even when participants have
a clear attentional set for color, an irrelevant abrupt onset or new object captures
attention. In other words, the appearance of the new object overrides a top-down set for
color regardless whether participants are engaged in a feature search or a singleton
detection mode. Since the abrupt onset or new object was always irrelevant for the task
and was presented at an empty location that never contained a target we argue that this
attentional capture is genuinely exogenous.
Author Note
This work was based on the master’s thesis of D.S. We thank Jim Brockmole, Angus
Gellatly, and Michael Proulx for excellent comments on an early version of the
manuscript.
Chapter 6 Abrupt onsets capture attention
independent of top-down control settings II: Additivity is no evidence for filtering
Schreij, D., Theeuwes, J. & Olivers C.N.L (2010). Abrupt onsets capture attention independent of top-down control settings II:
Additivity is no evidence for filtering. Attention, Perception & Psychophysics 72(3), 672-682
ATTENTIONAL CAPTURE BY ABRUPT ONSETS INDEPENDENT OF ACS II
112
Abstract
Is attentional capture contingent on top-down control settings or involuntarily driven by
salient stimuli? Supporting the stimulus-driven attentional capture view, Schreij, Owens
and Theeuwes (2008) found that an onset distractor caused a response delay in spite of
participants having adopted an attentional set for a color feature. However, Folk,
Remington and Wu (2009) claimed that this delay reflects separate, non-spatial filtering
costs instead, because the onset effects were additive with color-based capture effects,
and capture should have caused underadditivity. The current Experiment 1 shows that
contingent capture caused by additional color cues is also additive, just like the onset
effect. This makes additivity a dubious diagnostic with regard to spatial capture.
Experiment 2 demonstrates that it is possible to obtain underadditivity when attention-
demanding distractors have sufficient capturing power. Experiment 3 shows that the
abrupt onset interference turns into benefits when the locations of the onset and the
target coincide. Together, these results argue in favor of stimulus-driven attentional
capture by abrupt onsets.
CHAPTER 6
113
Introduction
A frequently debated issue in attention research is which underlying cognitive
mechanisms are fundamental for the ability of objects to capture attention. Can
attentional capture be driven by bottom-up, stimulus-related factors, or is capture
always contingent on what the observer is looking for (i.e. his or her top-down
attentional set)? The latter stance has been advocated by Folk and Remington and
colleagues on the basis of spatial cueing studies in which observers looked, for example,
for a red target among white distractors (Folk & Remington, 1998; Folk, et al., 1992).
Prior to the target display, a cue appeared which unpredictably indicated the target
position (valid cue), or one of the distractor positions (invalid cue). Even though the
cues were uninformative, valid cues resulted in faster response times (RTs), but only
when the cue matched the target-defining property – that is, when it was also red. When
the cue was instead defined by a single abrupt onset, there was no cueing effect. This
suggests that the cue captures attention in a spatial manner, but only when it carries a
feature that is relevant to the goals of the observer. In other words, attentional capture
is contingent on the top-down attentional set (for a different account of contingent
capture results see Belopolsky, Schreij, & Theeuwes, 2010).
Recently, Schreij, Owens and Theeuwes (2008) reported evidence that appears to
be inconsistent with the contingent capture hypothesis. In a task that was very similar to
the spatial cueing paradigm of Folk et al. (1992), Schreij et al. (2008) found that even
when observers had an attentional set for color, the presence of an irrelevant abrupt
onset nevertheless slowed participants in finding the target. As in the Folk et al.
paradigm, participants searched for a red target, and an attentional set for redness was
indeed demonstrated by a strong validity effect of the matching red pre-cue. At the same
time, the presence of an abrupt onset in the target display interfered with responding,
regardless of whether the red cue was valid or invalid. Consistent with earlier claims
suggesting that onsets capture attention in a stimulus-driven manner (Belopolsky, et al.,
2005; S.E. Christ & Abrams, 2008; Enns, et al., 2001; Gellatly, 1999; R. Remington,
Johnston, & Yantis, 1986; Theeuwes, 1990, 1994b; Yantis & Jonides, 1984), Schreij et al.
(2008) concluded that even when observers adopt a clear top down set for color, they
cannot prevent attentional capture by the onset. According to the contingent capture
hypothesis, the presence of the onset should have had no effect on responding because it
was completely irrelevant to the task.
ATTENTIONAL CAPTURE BY ABRUPT ONSETS INDEPENDENT OF ACS II
114
However, Folk, Remington, and Wu (2009) have questioned whether the
interference caused by the abrupt onset distractor was due to a bottom-up shift of
attention to the location of the onset. In line with earlier claims of Folk and Remington
(1998), they argued that the interference was due to non-spatial filtering costs, a notion
first introduced by Kahneman, Treisman and Burkell (1983). According to the filtering
explanation, irrelevant new objects that appear simultaneously with the target compete
for attention and need to be filtered out. This filtering operation slows RTs to the target.
In other words, an irrelevant object can cause a delay in the deployment of attention to a
relevant object, without itself invoking a shift of attention. Specifically, in the Schreij et
al. (2008) study, the onset would be competing with the target, thus causing filtering
costs. The filtering explanation is compatible with the contingent capture hypothesis
because attention only goes to the item that matches the attentional set for color; it
assumes that attention does not go to the location of the abrupt onset, which only causes
a non-specific filtering cost. Because capture and filtering operations are presumed to
take place during independent stages of processing, additive effects of color-based
cueing and onset interference would be expected (Sternberg, 1969) As pointed out by
Folk et al. (2009), this was exactly what was found by Schreij et al. (2008): the
interference caused by the onset presence was equally strong for trials with valid as for
trials with invalid color cues. Therefore, if one accepts that the color cue captures spatial
attention, then the onset cannot do so too. According to Folk et al. (2009), if the onset
presence and color-based cueing would both operate on the same process of spatial
capture, one would expect these two factors to show an underadditive relationship
instead. After all, if one assumes that capture by the onset made the cue obsolete, then
attention should move directly from onset to target, regardless of cue validity. Thus,
according to Folk et al. (2009), the elimination of attentional capture is rather
instantaneous, in that the subsequent appearance of another salient object “should
effectively eliminate the [earlier] effects” (p. 309).
We however assume that the local activation of an object does not have to be
instantaneous. Activity takes some time to build up, and may take even more time to
dissipate. Especially when a distractor object looks like a target object, resource
allocation may be more sustained, making it difficult to disengage from this object (see
Theeuwes, et al., 2000). If one accepts that an object’s capturing power can be sustained
for a while (especially when it looks task-relevant) then the additivity of color-based
CHAPTER 6
115
cueing and onset interference is not so strange from an attentional capture perspective:
The abrupt onset may briefly capture attention away from the red cue, but because the
onset is not relevant, whereas the red cue was, attention rapidly returns to the latter. An
alternative solution is that attention first lingers for quite some time at the cued location
even after the cue has disappeared and the target display is already present. Thus
attention can already gather some evidence about whatever object is present at that
location (which may be a target). This process is then disrupted (but not re-set) because
attention is captured away by the abrupt onset. When attention in turn disengages from
the abrupt onset, the color cues may indeed already be forgotten, as Folk et al. suggest,
and attention moves directly back to the target. However, in the case of a valid cue, a
large part of the evidence on which response to make might have already been
accumulated by then, resulting in a speeded response (in other words a cueing effect
that is additive with an onset effect).
This paper investigates the claim that two objects that subsequently capture
attention should always show an underadditive relationship. If Folk et al. (2009) are
correct and capture by a subsequent object always nullifies any previous capture effects,
such underadditivity should also be found when both consecutive capturing objects are
contingent upon the observer’s attentional set. However, if additivity is found in this
case as well, the criterion of underadditivity cannot be regarded as a suitable diagnostic
for or against the occurrence of attentional capture.
Using further adaptations of the Folk et al. (1992) paradigm, Experiment 1
demonstrates that interference by a distractor is also additive with the cueing effect
when the distractor carries the task-relevant feature and thus, according to the theory of
contingent attention capture, is assumed to capture attention. Experiment 2 shows that
a distractor is actually able to eliminate the effects of a pre-cue completely, resulting in
an underadditive relation, when it both has a strong bottom-up signal and is contingent
on the participant’s attentional set. As these experiments show, the occurrence of
additivity is not a reliable diagnostic for or against attentional capture, but depends on
the relative strength of the distractor. Experiment 3 shows that the additional abrupt
onset yielded considerable RT benefits rather than costs when its location incidentally
coincided with the target. This too indicates that the new abrupt onset element actually
attracted attention to its location rather than resulting in general filtering costs.
ATTENTIONAL CAPTURE BY ABRUPT ONSETS INDEPENDENT OF ACS II
116
Figure 19: An illustration of sequence of events for a trial of experiment 1. First a fixation display is shown for 500ms, after which the central fixation cross is turned off for 50ms. After this, the fixation display is shown again for a period of 1000ms. Then a cue display will appear for 50 ms and after an ISI of 150ms in which the fixation display is show again, the target display will be presented for 2000 ms or until the subject responds. When there was a distractor cue, it was presented for 50 ms between the cue and target display and was preceded and succeeded by a 50 ms fixation display. This is an example of a valid trial, since the location of the red balls in the cue display and the location of the red letter in the target display are the same. In reality the background was black, black lines where white and the gray elements were red.
CHAPTER 6
117
Experiment 1
This experiment investigated whether a pattern of additivity can serve as a useful
diagnostic against attentional capture. We took the Schreij et al. (2008) version of the
Folk et al. paradigm, but we left out the onset distractor and instead briefly made one of
the other distractor boxes red. This red distractor was always invalid. Thus, the displays
contained a red cue (valid or invalid), then potentially followed by a red distractor
(invalid) and afterwards the target display. The contingent capture theory states that the
new red distractor should capture attention away from the initially cued location. After
all, red is what the observers are looking for, and capture by red is what explains the
original cueing effect. If such capture by a new distractor would indeed erase or reduce
all prior cueing effects, the red distractor should here attenuate the benefits of a valid
color cue, resulting in underadditivity. If, on the other hand, there is still residual
activation of the first cue, strong enough to affect the reorienting towards, or the
identification of the target, then we may again observe additivity between cue validity
and the presence of the red distractor.
To further support the claim that both cues captured attention, it was
investigated if cued distractor identities interfered with responses to the target,
following the identity intrusion method introduced by Theeuwes (1996; see also
Theeuwes & Burger, 1998). The assumption is that if attention is shifted to the location
of the cue, the identity of the object at that position will be preferentially processed (e.g.
Kramer & Jacobson, 1991). If the distractor identity is compatible with the target
identity, performance may benefit, relative to when distractor and target are
incompatible. This compatibility effect would further strengthen the claim that the
distractor captured spatial attention (See Folk, Leber, & Egeth, 2002 for a similar
argument).
METHOD
Participants
Ten students, between 18 and 27 (average 21) years old, of the Vrije Universiteit of
Amsterdam participated in this experiment in return for money or course credits. All
reported no color blindness and normal or corrected-to-normal vision.
Apparatus & stimuli
ATTENTIONAL CAPTURE BY ABRUPT ONSETS INDEPENDENT OF ACS II
118
The experiment was run on a HP Compaq with a 2.6 GHz Pentium 4 processor and 512
MB of RAM. The stimuli were presented on a 19” Iiyama Vision Master Pro 454 CRT
screen with loudspeakers, with a refresh rate of 120 Hz and with a resolution of 1024 x
768 pixels. The “X” and “M” keys on a normal keyboard were used to register the
participants’ responses. Stimulus presentation and response recording were done in E-
prime 1.2 (Psychological Software Tools, 2003). The experiment was executed in a dimly
lit and soundproof room, in which participants were seated at a distance of
approximately 75 cm from the screen. All displays had a uniform black background. The
fixation display consisted of a bright white (CIE(0.286, 0.311), 59,50 cd/m2) fixation
cross at the center of the screen, surrounded by four light gray (CIE(0.285, 0.306),
luminance 28.28 cd/m2) placeholder boxes measuring a width of 3.4o visual angle. The
four boxes were positioned above, below, to the left and to the right of the fixation cross,
along a virtual circle with a radius of 6.6o visual angle, with the fixation cross as the
center. The cue display consisted of the same elements as the fixation display, with the
addition of four dots, with a diameter of 0.5o visual angle, positioned along the outside of
the center of each rib of all the placeholder boxes. One set of these dots surrounding one
of the placeholder boxes had a red color, (CIE(0.621, 0.345), 10.43 cd/m2) and indicated
the cued location. All the dots surrounding the other boxes had a bright white color,
(CIE(0.286, 0.311), 59,50 cd/m2). Until the presentation of the target screen, each of the
boxes contained a bright white figure, which consisted of overlapping “X”,”|” and “=”
symbols. When the target display was presented, the irrelevant line segments were
removed, revealing an “X” (Myriad Roman, 21pt) or an “=” (Myriad Roman bold, 22pt)
inside each of the boxes. At this same moment the color of the target character turned
from white into red. There were always two ‘X’s and two ‘=’s present. To serve as a
distractor cue, one of the placeholder boxes briefly flashed from white to red (CIE(0.621,
0.345), 10.43 cd/m2)
Design & Procedure
There were two important factors. The first was the validity of the first color cue, which
could be valid or invalid. The second was the presence of a red distractor (absent or
present). This resulted into a 2 x 2 factorial design. The red distractor was always invalid
and would never appear on a cued location, while the first color cue was valid in only
25% of the trails and invalid for the rest. The first cue validity was varied within blocks
CHAPTER 6
119
and the presence of the second cue was varied between blocks. There were 9 blocks of
80 trials, of which the first block was a practice block.
Participants were tested in a 30 minute session. Before the experiment started,
oral instructions were given to familiarize them with the task. Participants were told to
keep an index finger on each of the two response buttons and to not move their eyes
away from fixation during a trial, because this would impair their performance. The
target character appeared in one of the present boxes on the display, it was equally often
an X or a =, randomly mixed within blocks. At the end of each block, participants were
advised to take a rest and were forced to wait for at least 30 seconds, before they could
continue.
Trials began with the presentation of the fixation display for 1000 ms, after
which the fixation cross blinked off and on for 100 ms, to notify the participant of the
start of a trial. The fixation display then remained on the screen for another 100 ms after
which a cue display was presented for 50 ms. If a distractor cue was present, it appeared
50 ms after the first cue disappeared, with a duration of 50 ms. Afterwards, the fixation
display was again presented for 50 ms before the target display appeared. If the second
cue was absent, then the first cue would simply be followed up by a fixation display for
150 ms, after which the search display was presented until the participant responded
(with a maximum of 2,000 ms). The participant was instructed to look for the red item,
and press “X” when it was an X, or press ”M” when it was an =. A distinctive sound was
played for a correct or incorrect response. If the response was incorrect, the experiment
paused for 5 seconds to let the participants regain their focus. There was an inter-trial
interval of 500ms.
Results and discussion
Incorrect responses were removed from the dataset, resulting in a loss of 3.3% of the
trials, as were responses with RTs below and above 2.5 SD from the mean (another 2%).
The remaining data is depicted in Figure 20 and was first submitted to an ANOVA with
Red Distractor (absent, present) and Cue Validity (valid, invalid) as factors.
There was a significant main effect of Cue Validity, F(1,9) = 19.77, p < .001.
Participants were slower after an invalid cue than after a valid cue. The presence of the
red distractor slowed the responses by 18 ms, Red Distractor, F(1,9) = 13.76 , p < .001.
There was no interaction between Distractor Cue and Cue Validity, F(1,9) < 0.5, p > .5,
ATTENTIONAL CAPTURE BY ABRUPT ONSETS INDEPENDENT OF ACS II
120
indicating additive effects of distractor presence and cue validity. The same analysis of
the error pattern revealed no significant effects, F(1,9) = 0.882, p = .372.
Another analysis investigated possible compatibility effects between the target
character and the characters located at the color cue and the red distractor. When there
was no red distractor present and the cue was invalid, the character at the cued location
showed a reliable compatibility effect with the target, t(9) = 2.61, p < .05. An
incompatible character made participants respond more slowly than a compatible
character (by 14 ms). When a distractor cue was present, this compatibility effect
disappeared, t(9) = 1.59, p = .147. However, in this case there was a significant
compatibility effect with the character at the red distractor location, t(9) = 2.87, p < .05
(by 14 ms). Thus, the identity of the character at the distractor cue interfered with the
response to the target.
500
510
520
530
540
550
Mea
n C
orre
ct R
T (m
s)
Distractor cue absentDistractor cue present
0%
1%
2%
3%
4%
5%
6%
invalid validCue validity
Erro
r
Figure 20: Results of Experiment 1. The RTs and error percentages are shown as a function of cue validity, in the situations where a second red distractor was absent or present.
CHAPTER 6
121
An analysis on the errors with similar factors yielded no surprises. Error rates
were equal among all conditions and levels.
The contingent capture account leaves no other option than assuming that the
red distractor captured attention, since a) its color matched the participant’s attention
set for the task and b) it resulted in increased RTs. If one denies that the red distractor
captures attention, one would run the risk of also having to deny that the red cue
captures attention – something which obviously goes against the claims of contingent
capture. Even so, we find almost perfect additivity between the red cue and the red
distractor. An invalid cue added about 20 ms to response times, and so did the presence
of a red distractor, regardless of cue validity.
The costs inflicted by the red distractor in this experiment were generally of
greater magnitude than the costs inflicted by an onset distractor in Schreij et al. (2008).
Most likely this is because the red distractor contained a task relevant feature, which
may have resulted in more trouble disengaging attention from the distractor. The
observation that the cue induced compatibility effects also corroborate the claim that
the distractor successfully captured attention away from the first cue, since there is only
a compatibility effect of the first cue when the distractor cue is absent and, when
present, the distractor cue yields its own compatibility effect. Yet despite these
somewhat stronger distractor effects, the relationship with the cueing effect remained
additive.
It is also unlikely that the costs generated by the distractor cue are due to
filtering, since according to the filter account, filtering costs generally only accompany
the appearance of a new perceptual object, and the distractor cue was actually formed
by a feature change to an old, already present object. In addition, this change occurred
100 ms before the target appearance, while filtering costs are claimed to be usually only
manifested with simultaneous appearance as the target (Folk & Remington, 1998;
Kahneman, et al., 1983).
To conclude, it appears that the onset effects as found in Schreij et al. (2008) are
very similar to contingent capture effects, in terms of their additivity with earlier
capture effects. It seems that contingent capture is partly a sustained phenomenon of
which lingering effects are not easily terminated by subsequent events. It is possible that
such sustained effects are due to the seemingly task-relevant nature of the capturing
event, which may trigger slower but longer lasting top-down feedback mechanisms. If
ATTENTIONAL CAPTURE BY ABRUPT ONSETS INDEPENDENT OF ACS II
122
underadditivity is a strict prerequisite for attentional capture by the onset distractor in
previous experiments, then contingent capture by the red distractor in the present
experiment should not have been additive to the effect of the color cue. In other words, if
we accept that the distractor captured attention in the current experiment, then
additivity per se is not a diagnostic for or against attentional capture. Thus, the
argument that additivity between cue validity and onset presence is better explained by
filtering costs and thus excludes attentional capture is doubtful.
Experiment 2
The important question remains why later distractors (whether color-based or onset-
based) did not obliterate the earlier cueing effect (i.e. cause an underadditive interaction
of cueing and distractor effects). We have already suggested that the attentional priority
induced by the cue may linger, thus even exerting effects after observers have already
visited the next distractor. This would still mean that if there exists a distractor powerful
enough to capture attention away from the location of the contingent pre-cue, plus then
hold attention long enough for the cueing effect to finally dissipate, underadditivity may
be observed. With 10 to 20 ms interference effect the onset distractor in Schreij et al.
(2008) or the attentional set-matching red distractor of Experiment 1 may not have
been strong enough. We expected that if anything, a distractor that was both salient in a
stimulus-driven fashion (i.e. it featured an abrupt onset) and that carried a feature also
possessed by the target at the same time (i.e. its bounding box was red), might do the
job, as it would evoke both contingent top-down and stimulus-driven attentional
resources. It has been argued before that attentional disengagement from items
possessing task-relevant features is more time-consuming than from irrelevant items
(Theeuwes, et al., 2000). We argued that if the onset distractor manages to occupy
attentional resources long enough, any residual effects of the pre-cue will have
dissipated once attention has been able to disengage from the distractor location. The
consequence would be that attention will not return to the previous position of this cue,
eliminating the cueing effects as found in previous experiments, this time resulting in an
underadditive relation between onset presence and cue validity. At the same time, this
would again demonstrate a crucial contribution of the onset to attentional capture.
CHAPTER 6
123
METHOD
Participants
13 students, between 18 and 30 (average 24) years old, of the Vrije Universiteit of
Amsterdam participated in this experiment in return for money or course credits. All
reported no color blindness and normal or corrected-to-normal vision.
Apparatus, stimuli, design and procedure
The apparatus and stimuli were largely similar to Experiment 1, except that there was
no longer a distractor cue and that in a new contingent onset distractor condition the
distractor was a white onset character appearing in a red bounding box positioned
between two old bounding boxes (in addition to the irrelevant onset condition). The
target itself never appeared through an abrupt onset. In addition, the ISI between the
cue and target display was reduced from 150 ms (as in Experiment 1) to 100 ms, as it
was in Folk et al. (1992). Thus, the important factors of this experiment were Cue
Validity (invalid, valid) and Onset Presence (No Onset, Onset), which were varied within
blocks, and Onset Distractor Type (Irrelevant Onset, Contingent Onset), which was
varied between blocks. The experiment consisted of 2 practice blocks and 10
experimental blocks, each consisting of 80 trials, which took participants around 45
minutes in total to complete.
Results and discussion
Incorrect responses were removed from the dataset, discarding 3.7% of the trials, as
were responses with RTs below and above 2.5 SD from the mean (another 2.1%). The
remaining data is depicted in Figure 21 and was submitted to an ANOVA with Cue
Validity (valid, invalid) and Onset Presence (no onset, onset) for both conditions of
Onset Distractor Type (contingent vs. irrelevant).
When the onset had an irrelevant white colored bounding box, cue validity and
onset presence were significant as before, F(1,12) = 17.66, p < .001 and F(1,12) = 13.69.
p < .001 respectively, demonstrating that a valid cue yielded considerable response
benefits and the presence of an onset slowed down participants. The interaction
between these two factors was far from significant (p > .6). In the blocks where the onset
had a red colored bounding box, there was no main effect of Cue Validity, F(1,12) = 1.07,
p = .321, indicating that there was no difference between RTs for a valid and invalid cue.
ATTENTIONAL CAPTURE BY ABRUPT ONSETS INDEPENDENT OF ACS II
124
B)
450
470
490
510
530
550
570
590
610
Mea
n C
orre
ct R
T (m
s)
No OnsetIrrelevant Onset
0%
4%
8%
invalid validCue Validity
Erro
r
Figure 21: Results of Experiment 2. The RTs and error percentages as a function of cue validity, in situations where there was no onset and where the onset appeared in A) a white box (normal onset) or B) a red box (contingent onset).
A)
0%
4%
8%
invalid validCue Validity
Erro
r450
470
490
510
530
550
570
590
610
Mea
n C
orre
ct R
T (m
s)
No OnsetContingent Onset
The main effect of Onset Presence was significant, F(1,12) = 13.08, p < .001. Participants’
responses were delayed when an onset was present. Importantly, the interaction
between Cue Validity and Onset Presence was significant F(1,12) = 6.59, p < .05. When
present, a contingent red onset annihilated the validity effect of the earlier color cue.
Pair-wise comparisons of the valid and invalid cue conditions revealed that there was an
effect of the cue in the no-onset condition [t(12) = 2.379, p < 0.05], but not when a
contingent onset was present [t(12) = -0.471, p = .646].
Comparing the no-onset trials in the contingent-onset with the irrelevant-onset
blocks, it was found that participants were significantly slower during the contingent
onset blocks, F(1,12) = 5.62, p < .05. Conducting an ANOVA with Onset Distractor Type
(Contingent, Irrelevant) and Cue validity(Invalid, Valid) as factors on all onset-present
trials revealed that participants were significantly slower when a contingent red onset
CHAPTER 6
125
was present, than when a irrelevant onset was present, F(1,12) = 10.21, p < .001. The
interaction between Cue Validity and Onset Presence was significant as well, F(1,12) =
8.52, p < .001, showing that the cueing effect got attenuated by the contingent red onset
distractor.
For completeness, we also investigated if the onset character caused a response
compatibility effect with the target, for both levels of Onset Distractor Type. An ANOVA
with Cue Validity (Invalid, Valid) and Compatibility (Incompatible, Compatible) as
factors revealed a significant main effect for Compatibility in both Irrelevant and
Contingent Onset conditions. F(1,12) = 32.597, p < 0.01 and F(1,12) = 6.708, p < 0.05,
respectively. Participants thus responded slower with an incompatible onset, than with
a compatible one, regardless if it contained task relevant features. In neither Onset
Distractor Type condition was there an interaction of Compatibility and Cue Validity (p >
0.1 for both conditions).
Analysis on the error rates in the irrelevant onset blocks revealed no significant
effects. In the contingent onset blocks, only Onset Presence reached significance, F(1,12)
= 82.94, p < .001. Participants made significantly more errors when a contingent red
onset was present, than when it was absent.
Thus, distractors matching the participants’ attentional set inflicted larger RT
costs and were accompanied by larger error rates than distractors that were not
contingent on the attentional set. This supports the notion that attentional capture
towards, and/or attentional disengagement from task-relevant items is more time
consuming than from items that are deemed irrelevant (Theeuwes, et al., 2000).
Additionally, a contingent red onset distractor managed to completely eliminate the
effect of the pre-cue, as opposed to the onset-only distractor. Apparently, the
combination of bottom-up salience and top-down task-relevance gave the distractor the
necessary boost to completely overrule a lingering cue validity effect.
Another interesting finding is that response times were generally higher in
contingent distractor blocks than in irrelevant distractor blocks, even on no-onset trials.
Participants apparently became more conservative overall in responding when they
knew that the distractor could share critical features with the target. This overall delay
may also have allowed for further decay of lingering cueing effects.
The interaction between cue validity and the contingent red onset as found in the
current experiment exhibits the stronger notion of capture proposed by Folk et al.
ATTENTIONAL CAPTURE BY ABRUPT ONSETS INDEPENDENT OF ACS II
126
(2009) in that it meets the criterion of underadditivity. However, it also adds a
constraint in that the capturing item has to be sufficiently strong to overcome the earlier
cueing effects. In the present case, adding a strong bottom-up signal – an abrupt onset –
was sufficient. If the distractor lacks a strong bottom-up signal (as in Experiment 1) or
does not contain a task-relevant feature (as in Schreij et al., 2008), residual activation of
a pre-cue will be able overcome capture by the irrelevant item and draw attention back
to its location, with additivity as a consequence. In any case, the conclusion is that
bottom-up signals strongly contribute to attentional capture.
Experiment 3
If the filtering account is correct, onsets do not capture attention when the observer is
not looking for onsets. If anything, the observer is set against onsets, as they need to be
filtered out from competition in order to prevent interference with target processing.
The prediction then is that occasional targets featuring an abrupt onset should not be
processed any faster than normal targets that feature no abrupt onset. In contrast, if
onsets automatically draw attention, as stated by the capture account, then a target
appearing with an abrupt onset should allow for faster responses than for no-onset
targets.
Earlier work by Yantis and Jonides (1984) has indeed shown that onset targets
are given priority even when on the majority of trials the abrupt onsets are distractors.
In their experiments, observers viewed a display of pre-masks, of which one was likely
to change into a target. Together with the display change, a new object was added to the
display, which only occasionally was the target. Nevertheless, search was faster and
more efficient when the new object indeed turned out to be the target, consistent with
the idea that it captured attention (see also Becker, 2007) . Here we wished to apply
exactly the same logic to the paradigm used by Folk et al (1992). In the current
experiment the onset coincided with a distractor on the vast majority of trials (as in the
previous experiments), but now it could also accidentally coincide with the target
(which was again a red item) Hence, observers had no incentive whatsoever to attend to
the abrupt onset, and would be expected to employ an attentional set only for red (as
indicated once more by a cueing effect). If the filtering account is correct, and the onset
does not capture attention, there should be no benefit for onset over no-onset targets. If
the capture account is correct, and following Yantis and Jonides (1984), response
CHAPTER 6
127
benefits should be observed for onset targets because onsets involuntarily capture
attention.
METHOD
Participants
Eight students aged between 20 and 30 (average 24) from the Vrije Universiteit
Amsterdam participated in a half hour session in exchange for course credit or money.
Apparatus and stimuli
Apparatus and stimuli were largely the same as in Experiment 2, with the exception of
the following adaptations: In the onset distractor condition an extra light grey bounding
box containing a bright white character (either “X” or “=”) suddenly appeared in the
search display at the same moment that the other characters were revealed. In the onset
target condition this character was presented in red.
Design and procedure
The design and procedure were mostly identical to Experiment 2, except for the
following: There were no more contingent distractor blocks. A new factor Onset Type
consisted of the following levels: In the no onset condition, there was no additional onset
present. In the onset distractor condition, the new onset contained a distractor item (‘X’
or ‘=’). In the onset target condition, the new onset contained the target. This implied
that in both onset distractor and onset target condition, the effective set size was 5,
while in the no-onset condition it was 4. Cue Validity was varied as in Experiment 2, but
note that in the onset target condition, the cue was always invalid (since the empty
location in which the onset target appeared could not be cued). There were eight blocks
of 80 trials each, preceded by a practice block of 40 trials. Only 6 of the 80 trials per
block would contain an onset target. The onset was a distractor on 34 trials, and there
was no onset on the remaining 40 trials.
Results and discussion
Incorrect responses were removed from the dataset, resulting in a loss of 3% of the
trials, as were responses with RTs below and above 2.5 SD from the mean (another 2%).
The remaining data is depicted in Figure 22 and was first submitted to an ANOVA with
Onset Type (no-onset, onset distractor) and Cue Validity (valid, invalid) as factors. The
ATTENTIONAL CAPTURE BY ABRUPT ONSETS INDEPENDENT OF ACS II
128
onset target level was omitted in this analysis, because it could not be fully crossed with
Cue Validity (as onset targets could only be invalidly cued, see Method section). We will
return to onset targets below. There was a significant main effect of Cue Validity, F(1,7)
= 49.48, p < .001. Participants were slower after an invalid cue than after a valid cue. The
presence of an onset distractor also slowed the responses, Onset Type, F(1,7) = 40.80 , p
< .001. There was no interaction between Onset Type and Cue Validity, F(1,7) < 1.9, p >
.2, pointing towards additive effects.
To assess the effect of an onset target, we examined at invalidly cued trials only
for onset target, onset distractor, and no onset trials. A one-way ANOVA revealed a main
effect of Onset Type, F(2,14) = 26.05, p < 0.001. Separate comparisons revealed a
significant benefit for onset target trials relative to no onset trials and onset distractor
430
440
450
460
470
480
490
Mea
n C
orre
ct R
T (m
s)
Onset distractorNo onsetOnset target
0%1%2%3%4%5%
Invalid Valid
Cue Validity
Erro
r
Figure 22: Results of Experiment 3. The RTs and error percentages as a function of cue validity, when the onset was a target, a distractor, or there was no onset at all.
CHAPTER 6
129
trials, t(7) = 4.13, p < .01, t(7) = 5.70, p = 0.001, respectively. RTs on onset distractor
trials were significantly slower than on no onset trials, t(7) = 5.65, p = 0.001.
Analysis for compatibility effects with Cue Validity( Valid, Invalid ) and
Compatibility (Incompatible, Compatible) as factors once more revealed a significant
effect of target-distractor compatibility, F(1,7) = 32.739, p < .001. Participants
responded slower to a target when it was accompanied by an incompatible distractor
than by a compatible distractor. There was no interaction between Cue Validity and
Compatibility (p > .8).
The error pattern largely followed that of the RTs and the same analyses revealed
no significant effects (although Cue Validity approached significance, F(1,7) = 6.156, p =
0.056).
The results clearly demonstrate a performance benefit over no-onset trials when
the target featured an abrupt onset (in the invalid cue condition), as opposed to costs
when the onset coincided with a distractor. This provides direct evidence for an
attentional capture account: Despite its irrelevance to the task, the abrupt onset results
in a local enhancement of processing, leading to benefits when it is a target, and costs
when it is a distractor. No such benefits would be predicted by the filtering account:
According to this account, the onset is irrelevant. As the cue validity effects once more
demonstrate, the observers were set for red, not for onsets. In defense of a filter account,
one might argue that the transient of the onset made the target more salient and hence
easier to orient to. However, if this is the case, it would argue against the primary idea of
contingent capture, namely that the allocation of attention does not depend on salience,
but only on task relevance. The idea that high salience facilitates (or invokes
involuntary) orientation of attention towards objects is usually a claim made by the
attentional capture account. Another possibility is that the onset benefit stems from
differential masking as caused by the pre-masks for old objects. However, we consider
this unlikely. Using the current displays, Schreij et al. (2008; Experiments 1 & 2) directly
compared performance under the presence of premasks vs. no premasks and found no
influence of this on the interference caused by abrupt onsets.
General Discussion
There has been a longstanding discussion on whether irrelevant visual objects demand
spatial attention or demand additional filtering operations (Becker, 2007; Folk &
ATTENTIONAL CAPTURE BY ABRUPT ONSETS INDEPENDENT OF ACS II
130
Remington, 1998; Kahneman, et al., 1983; Theeuwes, 1994b; Theeuwes & Burger, 1998).
These two mechanisms are by no means mutually exclusive, and, as pointed out by
Becker (2007) different paradigms may result in different emphasis on one or the other.
This is why here and in Schreij et al (2008) we chose to integrate several attentional
capture paradigms (Folk et al., 1992; Theeuwes, 1992; Yantis & Jonides, 1984) and see if
particular findings generalize. Experiments 1 and 2 investigated the claim made by Folk
et al (2009) that if a distractor captures attention away from the cued location, its effect
should be underadditive with the effect of cue validity. This implies that if the onset
distractor captures attention, there should be little to no response benefits for a valid
preceding cue. Experiment 1 demonstrated that a contingent non-onset distractor was
able to capture attention away from the pre-cued location, but that a target appearing at
this location still enjoyed considerable response benefits. The contingent distractor
effect was additive with the contingent cueing effect, suggesting that a) the effects of
cueing were prolonged well into the target display, and b) additivity per se is not a good
diagnostic for or against attentional capture. Experiment 2 then showed that when an
abrupt onset was added to the contingent distractor, the cueing effects disappeared. We
infer that only a sufficiently strong distractor is capable of eliminating lingering cueing
effects, in this case when it both has a strong bottom-up signal and is contingent on the
participants’ attentional set. In Experiment 3, onset costs turned into benefits when the
onset coincided with the target location, indicating preferential processing of onsets.
According to the useful taxonomy of capture and filter effects outlined by Becker (2007),
especially the latter effect passes the criterion for attentional capture: Prioritized
processing of an item is not expected if the response time costs it generates when it is a
distractor are the result of a filtering operation. Since a filtering operation is assumed to
only suppress elements, a facilitated response to objects that should normally be filtered
out is exactly the opposite of what one would expect and can be better explained by
attentional capture.
Another possible explanation for the priority given to the abrupt onset here and
in Schreij et al. (2008) might be that here and in the previous experiment an onset
distractor was disruptive because participants operated in “singleton detection mode”
(Bacon & Egeth, 1994) in search for the target. Since the target was the only red
character among white distractors, it constituted a color singleton, a property which
participants might have actively used to find it. An onset however may also temporarily
CHAPTER 6
131
obtain a singleton status, due to its strong transient. Therefore the onset and color
singleton might be direct competitors as the only salient elements in the display, when
participants search for any salient object. However, Experiment 4 of Schreij et al. (2008)
precludes this explanation. Even when the color target was not a singleton, the onset
still interfered. In this case, all distractors were heterogeneously colored and
participants were required to actively search for redness, forcing them to use “feature
detection mode” (Bacon & Egeth, 1994). Furthermore, the current set-up closely
matched that of the original Folk et al. (1992) experiments, in which the targets were
also singletons, yet there was no evidence of singleton detection mode there, since
singletons that did not match the attentional set did not result in cueing effects.
While we believe that our findings support the claim that onsets have a special
status of being able to capture attention regardless of one’s attentional set, the results
appear to contradict the findings of Folk et al (1992), showing that an abrupt onset cue
did not cause any validity effect (and hence did not appear to capture attention), when
participants searched for a target defined by color. Theeuwes et al. (2000) explained this
lack of capture in the Folk et al. (1992) paradigm by reasoning that during the short SOA
between cue and target appearance, any capture effect of an irrelevant cue might have
already dissipated, especially since the irrelevant cue contained no task-relevant
features. This made swift disengagement from the distractor location possible. In the
Folk et al. (1992) study the onset cue appeared 150 ms prior to the target, while in most
studies where capture by irrelevant onsets was found, the onset appeared simultaneous
with the target, leaving no time for the capturing power of the onset to diminish. After
all, Schreij et al. (2008) found the costs inflicted by an onset distractor appearing
simultaneously with the target to be only 15 ms on average, which is short enough to
have already fleeted in the SOA between cue and target as used in the experiments of
Folk et al.
However, Lamy and colleagues (Lamy, 2005; Lamy & Egeth, 2003 Exp. 3-4) have
shown that an irrelevant onset cue was actually able to capture attention even when it
appeared before the target, like in Folk et al.’s original experiments. These studies
identified onset salience and the predictability of onset-to-target SOA as important
factors for evoking involuntary capture by irrelevant onsets. An onset cue managed to
capture attention when it was relatively salient and preceded the target at an
unpredictable SOA. These parameters were fixed in the original Folk et al. study, and
ATTENTIONAL CAPTURE BY ABRUPT ONSETS INDEPENDENT OF ACS II
132
seemed important for modulating capture by onsets. Because the onset distractor was
presented before the target display, the measured effects were unambiguously spatial.
The fact that Theeuwes et al. (2000) only found capture for short SOAs, in spite of using
an unpredictable distractor to target SOA, could possibly be attributed to the fact that
they used a less salient, static (color-based) distractor, instead of an onset distractor.
Related to the idea that the onset has to be sufficiently salient, previous research
has shown that the capturing power of an onset is the largest when it also makes up a
new perceptual object (Enns, et al., 2001; Yantis & Hillstrom, 1994). The onset of a
complete box including a letter as used by us here can be regarded as a new object, but
this may not be true for the four small white dots which Folk et al. used as their onset
pre-cue. These might have been perceived as a property change of an old object, namely
the bounding box which they surrounded. For this reason, the capturing power of the
cue as used in Folk et al. paradigms might not have been sufficient for longstanding
effects, being the reason they found a lack of capture by an onset cue when participants
searched for a color target (though see Folk & Remington (1999) for a study of
contingent capture with new-object onsets).
Other support that new object onsets capture attention comes from research by
Brockmole and Henderson (2005). They conducted a study in which they tracked eye
movements over natural scenes. Incidentally a new object was abruptly presented either
during fixation or during a saccade. In both situations the new object was fixated more
often than chance. However, onsets that appeared during fixation were fixated sooner
and more often than those coinciding with saccades. This made the authors conclude
that a new object does not need to have a perceivable transient to capture the eyes; a
non-transient new object is capable of doing so as well, though to a lesser extent. The
fact that these effects were not modulated by observers’ expectations concerning the
appearance of new objects, strengthens the notion that onset prioritization is
involuntary and the finding that our eyes consistently move to new objects, underpins
that this prioritization is spatial in nature (see also Theeuwes, et al., 1998).
Identity intrusion
Schreij et al. (2008) found additional evidence for attentional capture by the onset in an
experiment using the so-called identity intrusion technique first introduced by
Theeuwes (1996; see also Theeuwes & Burger, 1998) Instead of presenting a neutral
CHAPTER6
133
abruptonset,theabruptonsetwaseithercompatibleorincompatiblewiththeresponse
totheidentityofthetarget.Thistechniquerestsontheassumptionthat ifattentionis
shifted to the locationof theabruptonset, its identitywillbepreferentiallyprocessed
(e.g. Kramer& Jacobson, 1991). An incompatible distractor identitywould thenmore
greatlydegradetaskperformancethanacompatibleone.Consistentwiththis,Schreijet
al.(2008)foundaresponsecompatibilityeffect.
Folk et al. (2009) however, have argued that this response compatibility effect
canalsobeexplainedbyparallelprocessingofthetargetandthedistractor.Insupport
of thisargument, theyconductedanexperiment inwhichparticipantswerepresented
withadisplaywithtwobox‐shapedplaceholdersontheleftandrightsideoffixation,in
whichtwocharactersappeared.Oneofthesecharacterswasaredtarget;theotherwas
a white distractor which could either be compatible or incompatible with the target.
Priortothetargetdisplay,oneofthepositionswas indicatedbyanuninformativered
cue.Theyfoundthatevenwhenthetargetwascorrectlycued,andthusattentionshould
befocusedonthetargetlocation,thecompatibilityofthedistractorstillaffectedRTs–
despitethefact,asarguedbyFolketal.,thatthisdistractordidnotappearasanabrupt
new onset. Since participants had no reason whatsoever to attend to the irrelevant
character when the cue was valid, but nonetheless its identity was registered, the
conclusion was that both characters were processed in parallel, contrary to what a
spatialattentionalcaptureaccountwouldpredict.
Note that this explanation assumes that filtering is not always perfect. If the
distractor is indeed processed in parallelwith the target to the extent that it directly
interfereswiththeresponsetothat targetthenthismeansthat thedistractorwasnot
filtered out successfully. If filtering were successful, there would be no interference
(whatwouldotherwisebethepurposeoffiltering).Butjustlikefilteringmaynotalways
beperfect,participantsmayalsonotalwaysperformperfectlyinattendingtothecue.In
otherwords,attentionmaynotalwaysbeperfectlyfocusedonthetargetasassumedby
Folketal.(2009)intheirexperiment.Observersmayoccasionallyaccidentallyattendto
thedistractor,alsobecausewhenitappeareditshowedasimilardynamicchangeasthe
target. The two stimulimight even group on the basis of such commondynamics. All
suchfactorsmightcontributetoaresponsecompatibilityeffect.
ATTENTIONALCAPTUREBYABRUPTONSETSINDEPENDENTOFACSII
134
Additivity
Note that just as in Schreij et al. (2008), we again found that the contingent cueing
effectsandthestandardonsetdistractoreffectwereadditive.Folketal.(2009)argued
thattheadditivityoftheonsetdistractoreffectsandthespatialcueingeffectsmeantthat
they could not both operate on the same level of spatial attention. If we accept that
contingent cueing affects spatial orienting, then onset capture cannot do so too.
However,we have arguments against this claim. It is conceivable that activation at a
cuedlocationishigherandmoresustainedovertime,becausethecuecontainedatask
relevantfeature.Forthisreason,attentionmight ‘snapback’tothecuedlocation,after
theabruptonset in the subsequent targetdisplaywasbrieflyable to attract attention
away fromthere.Theshortexcursionofattentioncausesanadditiveeffect.Folketal.
argue against this “rubber band” explanation on the ground that the cue has already
disappearedafterthetargetdisplayandonsetarepresentedandthusshouldnolonger
beabletoattractattention.Ifanything,attentionshouldberepelledbytheoldlocation
duetothemechanismknownasinhibitionofreturn(Posner&Cohen,1984).
We however believe that it would not be the first time that an already
disappearedstimulus still exerts considerableeffects in thecognitive system.Forone,
the(red)cueisnotmaskedandsoanycue‐relatedactivityisexpectedtoautomatically
lingerforawhile.Moreover,thisactivityislikelytobemaintainedorevenstrengthened
bymeritofitbeingrelevanttotheobserver(whoafterallislookingforsomethingred).
Inhibitionofreturn(IOR)isthenquiteunlikelybecausea)thecueisconsideredrelevant
by the system (IOR is found to be weak under these conditions, Pratt, Sekuler, &
McAuliffe,2001)andb)thetimebetweencueandtargetdisplayisonly150ms,whichis
typicallytooshorttogenerateIOR(Posner&Cohen,1984).
Furthermore, the cue may already have triggered higher order processes
involvingworkingmemoryandnonspecificresponsepreparation,processesthatarenot
easilydisengagedorwipedoutbythepresentationoftheonsetdistractorandthatmay
demand that resources return to the original location. In fact, Godijn and Theeuwes
(2002)conductedaneye‐movementstudy thatdirectlysupports the ideaof sustained
activation for locations of interest in the visual field, even when the stimulus has
disappeared. Participants searched for a color singleton in a circular array of disks,
while eye movements were recorded. On some of the trials, an irrelevant onset
distractor appeared which often captured the participants’ gaze. Important for the
CHAPTER6
135
present discussion, during the eyemovement to the onset distractor, the target color
singleton had moved to another location. Nevertheless, on 82% of the trials, the
participants’nexteyemovementwas towards theold target location,not the(bynow
moresalient)newtargetlocation.
Evenifweacceptthattheremustbetwodifferentstagesinvolved,canonereally
claimthatcontingentcueingeffectsonlyreflectspatialcapture,andnootherprocesses?
Forexample,thedisengagementtimesfrominvalidcuesasfoundbyFolketal.(1992)
areofamuchgreatermagnitudethantheonesusuallyfoundforitemsthatareassumed
toexogenouslyhavecapturedattention(Theeuwes,etal.,2000).Ashasbeenarguedby
Theeuwes et al. (2000), top‐down effects on attention such as those found in the
contingentcaptureparadigmmayatleastpartlyinvolvedisengagementand/ordecision
mechanisms, rather than just initial orienting. Such effects may well turn out to be
considered additive with the spatial capture effects caused by an abrupt onset, thus
reversingtheargument.
Conclusion
It is clear that many questions remain about the exact mechanisms of attentional
capture, and that more sophisticated paradigms as well as better definitions are
necessary.Whatwehavearguedhereisthatabruptonsetssummonattention,despite
observersemployinganattentionalsetforacertaincolor.Wedonotdenyarolefortop‐
downinfluencesonattentionalcapture,buthereweshowagainthatsomestimuliare,to
acertainextent,immunetothistop‐downset.
Acknowledgements
ThisworkwassupportedbyVIDIgrant452‐06‐007fromtheNetherlandsOrganization
forScientificResearch(NWO)grantedtoCNLO.Wewould liketo thankChipFolk, Jim
BrockmoleandDominiqueLamyfortheirconstructivereviewsandsuggestions.
Chapter 7 Abrupt Irrelevant onsets cause
inhibition of return regardless of attentional set
Schreij, D., Theeuwes, J. & Olivers C.N.L (2010) Abrupt Irrelevant onsets cause inhibition of return regardless of attentional set
Attention, Perception & Psychophysics 72(7), 1725-1729
IOR FOR IRRELEVANT ONSETS REGARDLESS OF ACS
138
Abstract
It is disputed if onsets capture spatial attention either in a purely stimulus-driven
fashion, or only when they are contingent on one’s attentional set. According to the
latter assumption, interference from irrelevant onsets may result from non-spatial
filtering costs. In the current study we used Inhibition of Return (IOR) as a marker for
spatial attention. IOR mainly occurs for locations attention has visited before.
Participants searched for a red object among white objects. An attentional set for
redness was demonstrated by a spatial validity effect of red cues on response times.
However, a stronger validity effect was found for irrelevant white onsets, as they slowed
responses when being a distractor, but speeded them when being a target. Most
importantly, this onset benefit for targets turned into a deficit at longer SOAs, indicating
IOR. We conclude that onset distractors capture spatial attention regardless of the
observer’s attentional set.
CHAPTER 7
139
Introduction
There is ongoing debate about whether abrupt visual onsets automatically capture
attention or not (Folk, et al., 1992; Folk, et al., 2009; Schreij, et al., 2008; Yantis &
Jonides, 1984). In a central study, Folk et al. (1992) provided evidence that capture by
abrupt onsets is contingent on the observer’s top-down attentional set. In their
paradigm, participants responded to a target character appearing in one of four possible
boxes positioned at equal distances from fixation. A spatial cue (four small dots
surrounding a box) was presented 150 ms before the target display. This cue was
uninformative, only coinciding with the target location at chance level. The target could
either consist of a single white element (onset type) or of a red element among white
elements (color type). Despite its irrelevance, Folk et al. found an effect of the cue, but
only when it was of the same type as the target. When participants searched for onset
targets, only onset cues drew attention, whereas color cues did not. Vice versa, when
participants searched for color targets, color cues captured attention whereas onsets did
not. Folk et al. (1992) concluded that in order for an object to capture attention, it has to
possess a feature that participants are actively looking for. In other words, attentional
capture is under top-down control.
However, in an adaptation of the classic Folk et al. (1992) paradigm, Schreij et al.
(2008) recently found evidence consistent with automatic, stimulus-driven capture by
abrupt onsets. In their version, participants were always instructed to look for a red
target, which was always preceded by a red cue. As in Folk et al., Schreij et al. found a
spatial validity effect of the cue, demonstrating that participants had adopted an
attentional set for color. The crucial manipulation was the presence of an abrupt onset
distractor simultaneous with the target display. If capture by abrupt onsets entirely
depends on the observer’s attentional set, then no interference from this distractor
should be expected. This turned out not to be the case: The presence of an abrupt onset
caused search times to increase, indicating stimulus-driven attentional capture by
abrupt onsets. To save the idea that capture by abrupt onsets is under top-down control,
Folk, Remington, and Wu (2009) proposed an alternative explanation for the Schreij et
al. (2008) results. The onset-related costs could be due to non-spatial filtering
operations (Folk & Remington, 1998; though see Schreij, Theeuwes, & Olivers, 2010).
Non-spatial filtering is a mechanism that was first described by Kahneman, Treisman
IOR FOR IRRELEVANT ONSETS REGARDLESS OF ACS
140
and Burkell (1983), who found that a response to a target was delayed when other
objects appeared concurrently with the target, even though the distinction between
targets and distractors was clear. They reasoned that these simultaneously appearing
distractors compete for attention and need to be filtered out. On this view, these objects
do not attract spatial attention themselves (i.e. there is no capture), but nevertheless
delay the moment at which attentional focus can be deployed to the target location.
A mechanism that could be used to determine more precisely whether or not
irrelevant onsets capture spatial attention is Inhibition of Return (IOR). IOR is
operationalized as the slowing of responses to previously attended locations (i.e. more
than about 300 ms ago), compared to previously unattended locations (Posner & Cohen,
1984). It is assumed that when attention is drawn to a location in space, and is
subsequently disengaged from that location, an inhibitory mechanism is implemented,
slowing the return of attention (see Klein, 2000 for a review). Thus, an important aspect
of IOR is that it is assumed to be spatial in nature (although it has also been shown that
IOR can be object-based, see Tipper, et al., 1994). The purpose of the present study was
therefore to determine whether the irrelevant onsets as used by Schreij et al (2008)
indeed cause IOR, while at the same time participants have an attentional set for the
color red. If so, this would provide strong evidence that onsets capture attention in a
spatially specific manner, despite an attentional set for a different property. On the other
hand, if any of the previously found onset-related costs are due to non-spatial filtering,
then no IOR would be expected. For this purpose, the colored search target could
occasionally (at less than chance level) appear inside the irrelevant abrupt onset, after
either a short or a long SOA. On the basis of the previous IOR literature (Posner & Cohen,
1984), we predicted that the abrupt onset would cause facilitation of target
identification at the short SOA, relative to a no onset condition, whereas at the long SOA
it would cause inhibition. No such IOR is expected under a filtering account, since it
assumes attention has never visited the onset location in the first place.
METHOD
Participants
Twenty students aged 18 to 27 years (average 20.3) participated. None reported color
blindness, or any other visual deficit.
CHAPTER 7
141
Apparatus
The experiment was run on a PC with a 19” , 120 Hz CRT at 1024 x 768 pixel resolution,
viewed from approximately 75 cm in a dimly lit, soundproof room. Stimulus
presentation and response recording were done in E-prime 1.2 (Psychological Software
Tools, 2003).
Stimuli
All backgrounds were black. First a bright white (CIE(0.286, 0.311), 59,50 cd/m2)
fixation cross at the center of the screen was surrounded by four light gray placeholder
boxes (CIE(0.285, 0.306), luminance 28.28 cd/m2 ; dimensions: 3.4o visual angle) . These
were positioned above, below, to the left and to the right of the fixation cross, at a
distance of 6.6o visual angle. In the cue display four sets of four dots (0.5o visual angle )
surrounded the boxes. Most sets were bright white (CIE(0.286, 0.311), 59,50 cd/m2),
except for the cued location, which was indicated by a red set (CIE(0.621, 0.345), 10.43
cd/m2). Each box contained a bright white masking figure consisting of overlapping
Figure 23: An illustration of a trial where the target appeared inside the onset. It is important to note that this occurred at less than chance level, and the target appeared inside one of the four already present boxes in the majority of cases. First a fixation display was presented of which the fixation cross was flashed for 50 ms to alert participants about the start of the trial. Then a fixation display was presented again for 1000 ms, followed by a color cue which was flashed for 50 ms. Then the fixation display was presented again for 100 ms. Depending on the block type, the target either directly appeared inside the onset element, or the onset appeared during the fixation display after which the target was revealed inside the onset 900 ms later. Black lines in reality were white against a black background, and the grey items were red.
IOR FOR IRRELEVANT ONSETS REGARDLESS OF ACS
142
“X”,”|” and “=” symbols (Myriad Roman). In the target display, the irrelevant line
segments were removed, revealing an “X” or an “=”. Simultaneously, the color of the
target character turned from white to red. There were always two ‘X’s and two ‘=’ s
present. In the onset condition an extra light grey box was added to the display.
containing either a placeholder, distractor or target character depending on the trials
conditions.
Design and procedure
Figure 23 shows the experimental procedure. The experiment took around 45 minutes.
Oral instructions were given to familiarize participants with the task. Participants were
instructed to look for a red X or = inside one of the placeholders and press “X” or “M”
correspondingly. They were told to keep an index finger on each response button and to
not move their eyes from fixation during a trial. Initially the fixation display was
presented for 500 ms, after which the fixation cross blinked for 100 ms, to indicate the
start of a trial. 1000 ms after this event, the red cue appeared for 50 ms. Following a
variable SOA, the search display was presented until response (with a maximum of
2,000 ms). When the given response was incorrect, the experiment paused for 5 seconds
and a corresponding sound was played. The onset, when present, always appeared 150
ms after the cue display, but its contents were only revealed simultaneously with the
target and remained masked until that moment. Between blocks, participants had to
take a mandatory break of 30 seconds.
There were three main factors, both varied within subjects: Onset Type (no onset,
onset distractor, onset target), Cue Validity (valid or invalid) and SOA (short or long).
When short, the Cue-to-Target SOA was 150 ms and the Onset-to-Target SOA was 0 ms;
In the long SOA condition these were 1050 ms and 900 ms respectively (see Figure 23
for full timeline). The red cue prior to the target was uninformative about the location of
the upcoming target. When the cue was valid, its location corresponded with the target’s
(25% of the cases). The onset only appeared in 50% of the trials randomly at one of the
four possible locations between two placeholders. In the onset distractor condition, it
contained a distractor character and in the onset target condition, it contained the target.
Note that in the onset target condition, the cue could only be invalid, since the empty
location in which the onset target appeared could not be cued. Cue Validity and Onset
Type were randomly mixed within 8 blocks of 80 trials each, preceded by two practice
CHAPTER 7
143
blocks. Only 6 of the 80 trials (7.5%) per block contained an onset target. The onset was
a distractor on 34 trials, and there was no onset on the remaining 40 trials. SOA was
counterbalanced across subjects: half of the subjects performed the short SOA condition
first and the long SOA in the second half of the experiment; for the other subjects this
order was switched.
Results
Incorrect responses (4.7% of the trials), as well as responses with RTs below and above
2.5 SD from the mean (another 2.5%) were removed from the dataset. The mean of the
remaining RTs and the error percentages are depicted in Figure 24.
First, an ANOVA was performed with SOA (short, long), Onset Type (no-onset,
onset distractor) and Cue Validity (valid, invalid) as factors. Note that the onset target
level was omitted in this particular analysis, because it could not be fully crossed with
Cue Validity (as onset targets could only be combined with invalid cues, see Method
section). This analysis revealed that Onset Type interacted with SOA, F(1,19) = 10.57, p
A. Short SOA B. Long SOA
Figure 24: Response times and error rates for the different onset distractor conditions with an invalidly or validly cued target. A) Data for the short SOA (150 ms) between cue and target onset. B) Data for the long SOA (1050 ms)
IOR FOR IRRELEVANT ONSETS REGARDLESS OF ACS
144
< .05. Onset interference was larger in the short SOA than in the long SOA condition.
There was also an interaction between Cue Validity and SOA, F(1,19) = 20.767, p < .001,
indicating that the effect of the red cue diminished with a longer SOA. There was no
three-way interaction between these factors (p > .2). A separate ANOVA for each SOA
condition, with Cue Validity (valid, invalid) and Onset Presence (no onset, onset
distractor) as factors revealed a significant main effect of Cue Validity in the short SOA
condition, F(1,19) = 35.44, p < .001, as participants were slower after an invalid cue than
after a valid cue. The presence of an onset distractor slowed responses too, Onset Type,
F(1,19) = 29.65 , p < .001. There was no interaction between Onset Type and Cue
Validity, F(1,19) < 1.0, p > .5, suggesting additive effects. In the long SOA condition, the
validity effect of the cue completely disappeared, F(1,19) = 0.62, p = .44. The cost
associated with the onset distractor was also reduced, to the extent that it was no longer
significant, F(1,19) = 2.635, p = .146. There was no interaction between these factors
either, F(1,19) = 2.293, p = .17. An analysis on the error rates revealed no significant
effects.
Second, we assessed the effects of onset targets for invalidly cued trials. An
ANOVA with Onset Type (no-onset, onset distractor, onset target) and SOA (short, long)
as factors revealed a significant interaction, F(2,38) = 25.40, p < .001, reflecting the fact
that, compared to the no onset condition, onset targets led to faster responses at the
short SOA, but to slower responses that the long SOA. This was confirmed by separate
analyses. The short SOA, there was a significant benefit for onset target trials relative to
no onset trials and onset distractor trials, t(19) = 4.75, p < .001 and t(19) = 5.69, p <
.001, respectively. RTs on onset distractor trials were significantly slower than on no-
onset trials, t(19) = 4.55, p < .001. In the long SOA condition, RTs for onset targets were
significantly slower than for no-onset trials, t(19) = -2.61, p < .05. There was no
significant difference between response times in onset target and distractor conditions,
t(19) = 1.339, p = 0.20. There was a significant difference between the onset distractor
and no-onset conditions, t(19) = -1.34, p < .05. Analyses on the error pattern revealed no
significant effects or speed/accuracy trade-offs.
Discussion
The results of this experiment replicate Schreij et al. (2008) in that onset distractors
presented together with color targets led to interference, even though participants were
CHAPTER 7
145
looking for color. The experiment also shows that when the onset happens to coincide
with the target location, detection is speeded. This in itself suggests that an abrupt onset
is preferentially processed, rather than filtered out. The filtering operation’s main goal is
to prevent an irrelevant onset from acquiring attention and therefore one would not
expect an irrelevant onset to receive priority in processing (also see Schreij, et al., 2010).
More importantly, we show that this RT benefit turns into a cost when the interval
between the onset and the target at the same location is prolonged, thus revealing the
typical characteristics of IOR. Our data also show that the color cue loses its capacity to
drive attention with a long SOA. This was to be expected, because participants have
ample time in this condition to disengage their attention from the cue, which is known to
be uninformative. In addition, all responses in the long SOA condition are as fast as for a
validly cued target in the short SOA condition. This pattern of overall decreasing RTs
with increasing SOA is one that is found in most IOR studies (Klein, 2000).
Finding IOR at the onset location is important for two reasons. First, the
occurrence of IOR is a strong indicator that spatial attention was directed to the onset
location – something which cannot be explained by a mere filtering operation. Second,
IOR is thought to mainly follow reflexive shifts of attention (Posner & Cohen, 1984;
Rafal, Calabresi, Brennan, & Sciolto, 1989), and thus the current results support the idea
that onsets drive attention automatically, in a stimulus-driven fashion. Even though it is
generally assumed that IOR only occurs when attention is captured reflexively (for an
overview see Klein, 2000) recently a few exceptions were reported For example, Weger,
Abrams, Law, and Pratt (2008), as well as Berlucchi, Chelazzi, and Tassinari (2000),
showed that t it is possible to obtain IOR-like effects under certain conditions of
endogenous orienting (also see Berlucchi, 2006).
Note that at the long SOA, the onset distractor still inflicted RT costs when the cue
was invalid. Thus, while we found evidence for onset targets to be inhibited, the onset
distractor still caused an RT cost. One might explain these remaining costs as the result
of a filtering operation still working at the long SOA. However, in their original
conception of filtering, Kahneman et al. (1983) have argued that filtering costs are tied
to perceptual events like the sudden appearance of an object, rather than to the mere
presence of an object. They showed that filtering costs disappeared when objects had
been present for a while, as was the case here in the long SOA condition. Another
possibility is that onset distractors are suppressed, and that this suppression is
IOR FOR IRRELEVANT ONSETS REGARDLESS OF ACS
146
accompanied by a surrounding gradient of inhibition. Bennett and Pratt (2001)
demonstrated that IOR is indeed not constrained to the inhibited location, but spreads
out to the surrounding area. This inhibitory annulus then suppresses no-onset targets
appearing nearby. This would lead to slowed target-related responses, especially when
attention is not cued towards the target.
The occurrence of IOR in the Folk et al. (1992) contingent cueing paradigm has
been demonstrated before in studies of Gibson and Amelio (2000) and Pratt, Sekuler
and McAuliffe (2001). In contrast to the present study, these studies looked at IOR
towards the cue, rather than towards an additional distractor, by prolonging the usual
SOA of around 150 ms between the presentation of the cue and the target displays to an
SOA that would permit IOR. Both studies showed that an onset cue could evoke IOR, but
only when participants were also looking for an onset target, not when looking for a
color target – thus providing support for contingent capture rather than stimulus-driven
capture. However, a later study by Pratt and McAuliffe (2002) demonstrated that the
onset cue can elicit IOR even when the target is defined by color. Pratt and McAuliffe
(2002) argued that the absence of IOR in their earlier study was due to outliers in the
data set, and that the absence of IOR in the Gibson and Amelio (2000) study may have
been due to the peculiarities of their displays. For example, Gibson and Amelio used a
short brightening of the placeholder box as an onset cue instead of the usual four
surrounding dots. In the presence of an attentional control setting for a color-defining
target, perhaps such a brightening of an existing object may not be a sufficiently strong
stimulus to capture attention. There is substantial evidence that the effectiveness of
luminance transients largely depends on whether a new perceptual object is being
created (Enns, et al., 2001). In the present study, the abrupt onset stimulus was always a
new object, and, together with the Pratt and McAuliffe (2002) study, it now
unequivocally shows that IOR – and thus involuntary attentional orienting – towards an
onset is possible even when observers look for color. Our findings further extend those
of Pratt and McAuliffe (2002) by demonstrating capture by onsets simultaneously with
(i.e. occurring on the very same trials as) the presumed capture by the color cues, as is
demonstrated through the cue validity effect. This indicates that contingent and
stimulus-driven capture do not need to be mutually exclusive. Thus, unlike what Folk et
al. (2009) argue, it is not necessary to assume an attentional orienting mechanism for
CHAPTER 7
147
the one effect, and a completely different filtering mechanism for the other. The current
results cannot be explained by filtering, since filtering would not predict IOR.
This is not to say that contingent capture and stimulus-driven capture reflect one
and the same mechanism. Note that in the Gibson and Amelio, the Pratt and McAuliffe, as
well as in the present study, a color cue never elicited IOR, even when participants had
adopted an attentional set for color. This raises strong doubts if contingent capture
really is a form of capture, often receiving the additional labels of being automatic and
exogenous in nature (Folk & Remington, 1998). Instead, one could argue that when
observers are instructed to look for the red object that will appear shortly, they will do
exactly that, and take the occasional selection of the near simultaneous cue for granted
(see also Belopolsky, et al., 2010; Pratt, et al., 2001). In other words, attention would be
endogenously driven. The fact that there was no IOR for the cued locations fits with a
more endogenous source of attention (Klein, 2000; Posner & Cohen, 1984). This is not to
say that colors cannot capture attention in an exogenous, stimulus-driven fashion.
Theeuwes and Godijn (2002) and Folk and Remington (2006) have shown IOR to a color
singleton when this color singleton is not relevant to the observer.
In conclusion, we have shown that even when observers have an attentional set
for color, irrelevant onsets cause IOR. Since IOR can be taken as a measure of stimulus-
driven allocation of spatial attention, we conclude that onsets capture spatial attention
in an automatic fashion regardless of attentional set.
Summary in Dutch Nederlandse samenvatting
SUMMARY IN DUTCH
150
Aandacht voor verschijnende objecten
De wereld om ons heen bevat een immens aantal objecten. Om een coherent beeld te
houden van de omgeving, onderhoudt ons brein representaties van de objecten die het
waarneemt onder de aanname dat deze objecten continue zijn in ruimte en tijd. Een
object zal doorgaans niet ergens plots verdwijnen en op een geheel andere plek weer
tevoorschijn komen, maar moet zich gradueel van het ene naar het andere punt
verplaatsen. Deze eigenschap van spatietemporele continuïteit stelt ons in staat om ons
bewust te blijven van objecten die we voor korte momenten niet kunnen waarnemen,
omdat we bijvoorbeeld simpelweg in een andere richting kijken en een bepaald object
niet meer in ons visueel veld valt, of wanneer een bewegend object tijdelijk achter een
ander object verdwijnt en zo aan het zicht ontnomen wordt. Dit proefschrift onderzoekt
hoe deze objectrepresentaties de manier waarop we onze aandacht op zo’n object
richten beïnvloeden.
In het eerste deel bestuderen we of de kennis die iemand heeft opgedaan tijdens
een eerste waarneming van een object beïnvloedt hoe hij z’n aandacht toewijst aan
ditzelfde object wanneer hij het naderhand opnieuw tegenkomt. Met ditzelfde bedoelen
we overigens dat het object als één en dezelfde wordt gezien omdat het continue is in
ruimte en tijd; het heeft dus geen betrekking op identiekheid qua uiterlijke kenmerken.
Denk bijvoorbeeld aan een situatie waarin we een schotel waarnemen met allerlei
hapjes, en we vinden kaasstengels aan de linkerbovenkant van deze schotel, zijn we dan
ook weer geneigd om de kaasstengels op deze zelfde plek te zoeken wanneer we
dezelfde schotel een tijdje later weer zien? En hoe gaan we dan om met een schotel die
er precies hetzelfde uitziet, maar waarvan we weten dat het een andere, nieuwe schotel
is? Met andere woorden, bewaren wij zogenaamde “aandachtsinstellingen” bij de
representatie van een object en beïnvloeden deze instellingen vervolgens ons
zoekgedrag met betrekking tot dit object? Verder vragen we of deze instellingen alleen
bewaard worden aan de hand van de spatietemporele eigenschappen van een een
object, of ook aan de hand van het uiterlijk.
In het onderzoek beschreven in Hoofdstuk 2 lieten we mensen zoeken naar een
specifieke doelvorm (ook wel target genoemd) dat gepresenteerd werd ergens binnen
een groter object. We bestudeerden of de aandachtsinstelling die mensen aannemen
voor de locatie van het target binnen dit object ervoor zorgen dat ze onbewust het target
NEDERLANDSE SAMENVATTING
151
op dezelfde plaats verwachten te vinden wanneer ze naderhand weer hetzelfde object te
zien krijgen. Indien dit het geval is, kan men snellere reactietijden verwachten wanneer
de target ook inderdaad weer op z’n oude locatie te vinden is. In het experiment waren
er twee identiek uitziende objecten die fungeerden als displays en grotendeels
verborgen waren achter twee van de muren die zich aan elke rand van het scherm
bevonden. Gedurende iedere meting of ‘trial’ schoof een van de twee displays naar het
midden van het scherm en bevatte een verzameling vormen waartussen de
proefpersoon het target moest zoeken. Wanneer hij deze gevonden had schoof het
display weer terug achter een van de muren en begon de volgende trial. We vonden
inderdaad dat proefpersonen sneller reageerden op het target, wanneer deze zich op
dezelfde plek in het zoekveld bevond als in de trial ervoor, maar belangrijker was dat ze
nog sneller waren wanneer het zoekveld ook weer werd gepresenteerd op hetzelfde
object als de vorige trial, in vergelijking met presentatie in het andere object (dat verder
dus niet te onderscheiden was qua uiterlijk). Dit duidt er dus op dat mensen inderdaad
aandachtsinstellingen verbinden aan de spatietemporele representatie van een object en
dat deze instellingen consequent zoekgedrag m.b.t. dit object beïnvloed.
Hoofdstuk 3 onderzocht welke spatietemporele factor belangrijker is om een
object als hetzelfde te beschouwen: continuïteit in ruimte of continuïteit in tijd. Wanneer
een zoekdisplay achter een voorwerp bewoog en aan de andere kant weer tevoorschijn
kwam op de plaats waar men het zou verwachten, dan suggereerde het
reactietijdpatroon als gevolg van een herhaling of verandering van de targetlocatie dat
proefpersonen het verschenen display als dezelfde zagen als degene die zojuist achter
het verhullende voorwerp verdween. Wanneer het display vanachter een onverwacht
deel van de muur verscheen, dan was een dergelijk patroon nagenoeg afwezig. Het
maakte niet uit voor de reactietijden of het display een moment achter het voorwerp stil
bleef staan of in een vloeiende beweging naar z’n eindpunt bewoog. Dit duidt erop dat
objectcontinuïteit meer afhangt van continuïteit in de ruimte dan in de tijd.
Hoofdstuk 4 toont aan dat naast aandachtsinstellingen voor de locatie van een
target er ook instellingen bewaard blijven voor diens uiterlijke kenmerken zoals vorm of
kleur (ook wel features genoemd). Wanneer target features veranderen t.o.v. de trial
ervoor en het target in hetzelfde object verschijnt, dan zijn responstijden hoger dan
wanneer het veranderde target in een ander object verschijnt. Dit duidt erop dat target
feature informatie bewaard blijft en dat zoekprestaties vervolgens worden aangetast
SUMMARY IN DUTCH
152
wanneer de features van de target niet meer overeenkomen met degene die in de
representatie waren opgeslagen. Verder onderzochten we of het uiterlijk van het
displayobject invloed heeft op de handhaving van aandachtsinstellingen die aan een
representatie gebonden zijn. We toonden de twee aanwezige zoekdisplays ieder in het
schermgebied van een mobiel apparaat, zoals een iPod of mobiele telefoon. Dit frame
kon tussen trials veranderen wanneer het zoekdisplay verborgen was achter een van de
muren aan weerszijden van het scherm. Het zoekdisplay kon op die manier met een
ander frame verschijnen als toen het verdween. We vonden dat een verandering in of
herhaling van het frame van het displayobject geen invloed had op de handhaving en
hergebruik van aandachtsinstellingen. Deze bleken dus sterker verbonden aan de
spatietemporele representatie van het object dan aan diens uiterlijk.
Het tweede deel van dit proefschrift gaat in op een controverse die onderzoekers
van aandacht al een geruime tijd bezig houdt. Het is geregeld aangetoond dat sommige
objecten of gebeurtenissen de aandacht van een toeschouwer kunnen trekken buiten
zijn intenties om, wat ook wel “attentional capture” wordt genoemd. Enerzijds bestaat er
het standpunt dat een object alleen de aandacht kan vangen wanneer het een eigenschap
deelt met hetgeen waar de toeschouwer eigenlijk naar op zoek is. Als iemand
bijvoorbeeld op zoek is naar aardbeien, dan kan hij hun rode kleur gebruiken om ze
makkelijker te kunnen vinden. De toeschouwer heeft dan zogezegd een
aandachtsinstelling (of attentional set) voor de kleur rood. In dit geval is echter de kans
ook groot dat de toeschouwer onïntentioneel zijn aandacht op tomaten zal richten
gezien deze net als aardbeien een rode kleur hebben. Deze vorm van attentional capture
wordt contingent capture genoemd. Anderzijds zijn er studies die aangetoond hebben
dat een uit het niets verschijnend object (of onset) altijd de aandacht van een
toeschouwer vangt, ongeacht wat zijn attention set is. Deze studies laten zien dat een
onset altijd prioriteit krijgt tijdens een zoekproces.
Onze veronderstelling was dat deze tegenstrijdige bevindingen ten grondslag
zouden kunnen liggen aan verschillen in de ontwerpen van experimenten die gebruikt
zijn om beide standpunten aan te tonen. Bij paradigma’s die aantoonden dat onsets altijd
prioriteit krijgen in aandacht, werden deze vaak tegelijkertijd met het target
gepresenteerd. In de paradigma´s die voornamelijk contingent capture aantoonden
verscheen een onset echter vaak een moment voor de presentatie van de target,
waardoor een potentieel capture effect van de onset al uitgewerkt kon zijn op het
NEDERLANDSE SAMENVATTING
153
moment dat de target verscheen. Om deze discrepantie te overbruggen, namen wij een
typisch paradigma dat vaak gebruikt is om contingent capture aan te tonen over en
presenteerden de onset daarin gelijktijdig met het target.
In het onderzoek beschreven in Hoofdstuk 5 onderwierpen wij proefpersonen
aan een zoektaak waarin ze doelgericht moesten zoeken naar een rode letter (de color
target) die in één van vier aanwezige vakjes op het scherm kon verschijnen.
Ondertussen kon op een lege locatie uit het niets een nieuwe letter verschijnen die door
zijn witte kleur compleet irrelevant was voor de taak (de onset distractor). Kort voordat
de color target en onset distractor beiden verschenen, werd een van de potentiële target
locaties kort omringd door een cue bestaande uit vier rode balletjes. Omdat de cue dus
dezelfde kleur bezat als de target, konden proefpersonen deze cue vrijwel niet negeren,
ondanks dat ze wisten dat de cue- en targetpositie slechts op kansniveau samenvielen,
en reageerden ze sneller als de target dan ook toevalligerwijs op de gecuede locatie
verscheen. Deze cues waren dus van belang omdat deze aantoonden dat proefpersonen
daadwerkelijk op zoek waren naar de kleur rood en deze hun attention set vormde. Wij
vonden dat de onset distractor ondanks dat hij vanwege zijn witte kleur geen deel
uitmaakte van de attention set toch de aandacht ving en zodoende de reactietijden op de
target beïnvloedde, onafhankelijk van het effect van de cue.
Verder controleerden wij of aandacht ook echt naar de locatie van de onset
distractor ging en de verhoging in reactietijd niet te wijten was aan factoren die geen
spatiële aandacht betroffen. Men neemt over het algemeen aan dat hetgeen waar wij ons
aandacht op richten tot een dieper niveau (of niveau van betekenis) verwerkt wordt,
wat dan ook zou moeten gelden voor de letter op de onsetlocatie. In dit geval kan men
een grotere interferentie verwachten door de onset distractor, wanneer diens letter
verschilt met die van de color target dan wanneer beide letters identiek zijn. Wij vonden
inderdaad dat een verschil in letters hogere reactietijden veroorzaakte dan bij identieke
letters, hetgeen erop duidt dat aandacht ook werkelijk naar de onset locatie ging.
In Hoofdstuk 6 vonden we dat een distractor die een relevante targeteigenschap
droeg (opnieuw de kleur rood) net als een irrelevante onset distractor een additief effect
vertoont met de cue. Dit duidt erop dat een object of gebeurtenis dat de aandacht vangt,
niet per sé de effecten van een eerder ‘capture event’ hoeft te elimineren. Stel dat dit wel
het geval zou zijn, dan zou aandacht niet meer naar de origineel gecuede locatie terug
moeten gaan nadat de tweede distractor de aandacht ervan ‘weggevangen’ heeft. Echter,
SUMMARY IN DUTCH
154
ook wanneer deze distractor aanwezig was dan waren reactietijden flink sneller
wanneer de target op de locatie verscheen die eerder gecued was. In een tweede
experiment toonde we aan dat er wel degelijk situaties mogelijk zijn waarin een object
of gebeurtenis dat de aandacht vangt een eerder attentional capture effect teniet kan
doen, maar dat dit aandachtsvangende object een hoge saillantie moet hebben en
tegelijkertijd taakrelevant moet zijn om in staat te zijn dit te doen. Tenslotte
demonstreerden we dat de reactietijden voor een color target nog sneller zijn
(vergelijkbaar met die voor een valide gecuede target positie) wanneer deze in de onset
zelf verschijnt, hetgeen verder bewijs vormt dat de onset zelf daadwerkelijk de aandacht
vangt.
Hoofdstuk 7 toont daarnaast aan dat er Inhibition of Return (IOR) optreedt voor
een target die met een vertraging in de onset verschijnt. IOR is een fenomeen dat
aandacht er langer over doet om terug te keren naar een eerder bezochte locatie; in het
bijzonder wanneer de tijd tussen de twee bezoeken meer dan 300 ms bedraagt en het
eerste bezoek plaatsvond omdat de stimulus onvrijwillig de aandacht ving. Door IOR te
vinden voor een target die op de onset positie verschijnt tonen we aan dat aandacht de
onset locatie bezocht heeft op het moment dat hij verscheen en dat dit gebeurde buiten
de intentie van de proefpersoon.
In het licht van onze bevindingen komen we tot de conclusie dat men
aandachtsinstellingen bewaart bij representaties die ze van waargenomen objecten
onderhouden. Deze aandachtsinstellingen bepalen hoe we onze aandacht toewijzen aan
interne eigenschappen van die objecten als we deze later opnieuw aantreffen. We
gebruiken spatietemporele continuïteit als voornaamste factor om te bepalen of een
object dat we waarnemen hetzelfde is als voorheen en laten dit verder weinig afhangen
van continuïteit in de uiterlijke kenmerken van het object. Verder tonen we aan dat
objecten die uit het niets verschijnen prioriteit krijgen bij aandachtstoewijzing, zelfs als
men doelgericht op zoek is naar eigenschappen die het verschijnende object niet bezit.
REFERENCES
155
References
Abrams, R. A., Davoli, C. C., & Suszko, J. W. (2007). New objects can capture attention without a unique luminance transient. Psychonomic Bulletin & Review, 14(2), 338-343.
Bacon, W. F., & Egeth, H. E. (1994). Overriding Stimulus-Driven Attentional Capture. Perception & Psychophysics, 55(5), 485-496.
Becker, S. I. (2007). Irrelevant singletons in pop-out search: Attentional capture or filtering costs? Journal of Experimental Psychology-Human Perception and Performance, 33(4), 764-787.
Belopolsky, A. V., Schreij, D., & Theeuwes, J. (2010). What is top-down about contingent capture? Attention, Perception & Psychophysics, 72(2), 326-341.
Belopolsky, A. V., & Theeuwes, J. (2010). No capture outside the attentional window. Vision Research, 50(23), 2543-2550.
Belopolsky, A. V., Theeuwes, J., & Kramer, A. F. (2005). Prioritization by transients in visual search. Psychonomic Bulletin & Review, 12(1), 93-99.
Belopolsky, A. V., Zwaan, L., Theeuwes, J., & Kramer, A. F. (2007). The size of an attentional window modulates attentional capture by color singletons. Psychonomic Bulletin & Review, 14(5), 934-938.
Bennett, P. J., & Pratt, J. (2001). The spatial distribution of inhibition of return. Psychological Science, 12(1), 76-80.
Berlucchi, G. (2006). Inhibition of return: A phenomenon in search of a mechanism and a better name. Cognitive Neuropsychology, 23(7), 1065-1074.
Berlucchi, G., Chelazzi, L., & Tassinari, G. (2000). Volitional covert orienting to a peripheral cue does not suppress cue-induced inhibition of return. Journal of Cognitive Neuroscience, 12(4), 648-663.
Boot, W. R., Brockmole, J. R., & Simons, D. J. (2005). Attention capture is modulated in dual-task situations. Psychonomic Bulletin & Review, 12(4), 662-668.
Bower, T. G. R. (1967). Development of Object-Permanence - Some Studies of Existence Constancy. Perception & Psychophysics, 2(9), 411-418.
Bower, T. G. R., Broughto.J, & Moore, M. K. (1971). Development of Object Concept as Manifested in Changes in Tracking Behavior of Infants between 7 and 20 Weeks of Age. Journal of Experimental Child Psychology, 11(2), 182-&.
Breitmeyer, B. G., & Ganz, L. (1976). Implications of Sustained and Transient Channels for Theories of Visual-Pattern Masking, Saccadic Suppression, and Information-Processing. Psychological Review, 83(1), 1-36.
Broadbent, D. E. (1958). Perception and Communication. London: Pergamon Press. Brockmole, J. R., & Henderson, J. M. (2005). Prioritization of new objects in real-world
scenes: Evidence from eye movements. Journal of Experimental Psychology-Human Perception and Performance, 31(5), 857-868.
Burke, L. (1952). On the Tunnel Effect. Quarterly Journal of Experimental Psychology, 4, 121-138.
REFERENCES
156
Burnham, B. P. (2007). Displaywide visual features associated with a search display's appearance can mediate attentional capture. Psychonomic Bulletin & Review, 14(3), 392-422.
Cave, K. R., & Wolfe, J. M. (1990). Modeling the Role of Parallel Processing in Visual-Search. Cognitive Psychology, 22(2), 225-271.
Christ, S. E., & Abrams, R. A. (2006). Abrupt onsets cannot be ignored. Psychonomic Bulletin & Review, 13(5), 875-880.
Christ, S. E., & Abrams, R. A. (2008). The Attentional Influence of New Objects and New Motion. Journal of Vision, 8(3), 1-8.
Chun, M. M. (2000). Contextual cueing of visual attention. Trends in Cognitive Sciences, 4(5), 170-178.
Chun, M. M., & Jiang, Y. H. (1998). Contextual cueing: Implicit learning and memory of visual context guides spatial attention. Cognitive Psychology, 36(1), 28-71.
Desimone, R., & Duncan, J. (1995). Neural Mechanisms of Selective Visual-Attention. Annual Review of Neuroscience, 18, 193-222.
Di Lollo, V., Enns, J. T., & Rensink, R. A. (2000). Competition for consciousness among visual events: The psychophysics of reentrant visual processes. Journal of Experimental Psychology-General, 129(4), 481-507.
Donk, M., & van Zoest, W. (2008). Effects of salience are short-lived. Psychological Science, 19(7), 733-739.
Duncan, J. (1985). Visual search and visual attention. In M. I. Posner & O. S. Marin (Eds.), Attention and performance XI (pp. 85-106). Hillsdale: NJ:Erlbaum.
Egeth, H. E., & Yantis, S. (1997). Visual attention: Control, representation, and time course. Annual Review of Psychology, 48, 269-297.
Egly, R., Driver, J., & Rafal, R. D. (1994). Shifting Visual-Attention between Objects and Locations - Evidence from Normal and Parietal Lesion Subjects. Journal of Experimental Psychology-General, 123(2), 161-177.
Enns, J. T., Austen, E. L., Di Lollo, V., Rauschenberger, R., & Yantis, S. (2001). New objects dominate luminance transients in setting attentional priority. Journal of Experimental Psychology-Human Perception and Performance, 27(6), 1287-1302.
Eriksen, C. W., & Hoffman, J. E. (1972). Temporal and Spatial Characteristics of Selective Encoding from Visual Displays. Perception & Psychophysics, 12(2B), 201-&.
Flombaum, J. I., Kundey, S. M., Santos, L. R., & Scholl, B. J. (2004). Dynamic object Individuation in rhesus macaques - A study of the tunnel effect. Psychological Science, 15(12), 795-800.
Flombaum, J. I., & Scholl, B. J. (2006). A temporal same-object advantage in the tunnel effect: Facilitated change detection for persisting objects. Journal of Experimental Psychology-Human Perception and Performance, 32(4), 840-853.
Flombaum, J. I., Scholl, B. J., & Pylyshyn, Z. W. (2008). Attentional resources in visual tracking through occlusion: The high-beams effect. Cognition, 107(3), 904-931.
Folk, C. L., & Annett, S. (1994). Do Locally Defined Feature Discontinuities Capture Attention. Perception & Psychophysics, 56(3), 277-287.
REFERENCES
157
Folk, C. L., Leber, A. B., & Egeth, H. E. (2002). Made you blink! Contingent attentional capture produces a spatial blink. Perception & Psychophysics, 64(5), 741-753.
Folk, C. L., & Remington, R. (1998). Selectivity in distraction by irrelevant featural singletons: Evidence for two forms of attentional capture. Journal of Experimental Psychology-Human Perception and Performance, 24(3), 847-858.
Folk, C. L., & Remington, R. (1999). Can new objects override attentional control settings? Perception & Psychophysics, 61(4), 727-739.
Folk, C. L., & Remington, R. (2006). Top-down modulation of preattentive processing: Testing the recovery account of contingent capture. Visual Cognition, 14(4-8), 445-465.
Folk, C. L., Remington, R. W., & Johnston, J. C. (1992). Involuntary Covert Orienting Is Contingent on Attentional Control Settings. Journal of Experimental Psychology-Human Perception and Performance, 18(4), 1030-1044.
Folk, C. L., Remington, R. W., & Wright, J. H. (1994). The Structure of Attention Control - Contingent Attentional Capture by Apparent Motion, Abrupt Onset, and Color. Journal of Experimental Psychology-Human Perception and Performance, 20(2), 317-329.
Folk, C. L., Remington, R. W., & Wu, S. C. (2009). Additivity of abrupt onset effects supports nonspatial distraction, not the capture of spatial attention. Attention Perception & Psychophysics, 71(2), 308-313.
Franconeri, S. L., Hollingworth, A., & Simons, D. J. (2005). Do new objects capture attention? Psychological Science, 16(4), 275-281.
Franconeri, S. L., & Simons, D. J. (2003). Moving and looming stimuli capture attention. Perception & Psychophysics, 65(7), 999-1010.
Franconeri, S. L., Simons, D. J., & Junge, J. A. (2004). Searching for stimulus-driven shifts of attention. Psychonomic Bulletin & Review, 11(5), 876-881.
Gellatly, A. (1999). Perception and information processing. Behavioral and Brain Sciences, 22(3), 377-+.
Gellatly, A., Cole, G., & Blurton, A. (1999). Do equiluminant object onsets capture visual attention? Journal of Experimental Psychology-Human Perception and Performance, 25(6), 1609-1624.
Gibson, B. S., & Amelio, J. (2000). Inhibition of return and attentional control settings. Perception & Psychophysics, 62(3), 496-504.
Gibson, B. S., & Kelsey, E. M. (1998). Stimulus-driven attentional capture is contingent on attentional set for displaywide visual features. Journal of Experimental Psychology-Human Perception and Performance, 24(3), 699-706.
Gordon, R. D., & Irwin, D. E. (1996). What's in an object file? Evidence from priming studies. Perception & Psychophysics, 58(8), 1260-1277.
Gordon, R. D., & Irwin, D. E. (2000). The role of physical and conceptual properties in preserving object continuity. Journal of Experimental Psychology-Learning Memory and Cognition, 26(1), 136-150.
Hickey, C., McDonald, J. J., & Theeuwes, J. (2006). Electrophysiological evidence of the capture of visual attention. Journal of Cognitive Neuroscience, 18(4), 604-613.
REFERENCES
158
Hickey, C., Olivers, C., Meeter, M., & Theeuwes, J. (2011). Feature priming and the capture of visual attention: Linking two ambiguity resolution hypotheses. Brain Research, 1370, 175-184.
Hoffman, J. E. (1979). 2-Stage Model of Visual-Search. Perception & Psychophysics, 25(4), 319-327.
Hollingworth, A., Richard, A. M., & Luck, S. J. (2008). Understanding the function of visual short-term memory: Transsaccadic memory, object correspondence, and gaze correction. Journal of Experimental Psychology-General, 137(1), 163-181.
Hommel, B. (1998). Event files: Evidence for automatic integration of stimulus-response episodes. Visual Cognition, 5(1-2), 183-216.
Hommel, B. (2004). Event files: feature binding in and across perception and action. Trends in Cognitive Sciences, 8(11), 494-500.
Hommel, B. (2007). Feature integration across perception and action: event files affect response choice. Psychological Research-Psychologische Forschung, 71(1), 42-63.
Hommel, B., Musseler, J., Aschersleben, G., & Prinz, W. (2001). The Theory of Event Coding (TEC): A framework for perception and action planning. Behavioral and Brain Sciences, 24(5), 849-+.
Huang, L. Q., Holcombe, A. O., & Pashler, H. (2004). Repetition priming in visual search: Episodic retrieval, not feature priming. Memory & Cognition, 32(1), 12-20.
Jonides, J. (1981). Voluntary versus automatic control over the mind's eye's movement. In J. Long & A. Baddeley (Eds.), Attention and performance (Vol. IX, pp. 187-203). Hillsdale, NJ: Erlbaum.
Jonides, J., & Yantis, S. (1988). Uniqueness of Abrupt Visual Onset in Capturing Attention. Perception & Psychophysics, 43(4), 346-354.
Juola, J. F., Koshino, H., & Warner, C. B. (1995). Tradeoffs between Attentional Effects of Spatial Cues and Abrupt Onsets. Perception & Psychophysics, 57(3), 333-342.
Kahneman, D., & Treisman, A. (1984). Changing views of attention and automaticity. In R. Parasuraman & D. A. Davies (Eds.), Varieties of attention (pp. 29-61). San Diego, CA: Academic Press.
Kahneman, D., Treisman, A., & Burkell, J. (1983). The Cost of Visual Filtering. Journal of Experimental Psychology-Human Perception and Performance, 9(4), 510-522.
Kahneman, D., Treisman, A., & Gibbs, B. J. (1992). The Reviewing of Object Files - Object-Specific Integration of Information. Cognitive Psychology, 24(2), 175-219.
Keizer, A. W., Colzato, L. S., & Hommel, B. (2008). Integrating faces, houses, motion, and action: Spontaneous binding across ventral and dorsal processing streams. Acta Psychologica, 127(1), 177-185.
Kim, M. S., & Cave, K. R. (1999). Top-down and bottom-up attentional control: On the nature of interference from a salient distractor. Perception & Psychophysics, 61(6), 1009-1023.
Klein, R. M. (2000). Inhibition of return. Trends in Cognitive Sciences, 4(4), 138-147. Kramer, A. F., & Jacobson, A. (1991). Perceptual Organization and Focused Attention -
the Role of Objects and Proximity in Visual Processing. Perception & Psychophysics, 50(3), 267-284.
REFERENCES
159
Kristjansson, A., Mackeben, M., & Nakayama, K. (2001). Rapid, object-based learning in the deployment of transient attention. Perception, 30(11), 1375-1387.
Kruschke, J. K., & Fragassi, M. M. (1996). The Perception of Causality: Feature binding in interacting objects. Paper presented at the Eighteenth Annual Conference of the Cognitive Science Society.
Kumada, T. (1999). Limitations in attending to a feature value for overriding stimulus-driven interference. Perception & Psychophysics, 61(1), 61-79.
Lamy, D. (2005). Temporal expectations modulate attentional capture. Psychonomic Bulletin & Review, 12(6), 1112-1119.
Lamy, D., & Egeth, H. E. (2003). Attentional capture in singleton-detection and feature-search modes. Journal of Experimental Psychology-Human Perception and Performance, 29(5), 1003-1020.
Lamy, D., Tsal, Y., & Egeth, H. E. (2003). Does a salient distractor capture attention early in processing? Psychonomic Bulletin & Review, 10(3), 621-629.
Lavie, N. (1995). Perceptual Load as a Necessary Condition for Selective Attention. Journal of Experimental Psychology-Human Perception and Performance, 21(3), 451-468.
Leber, A. B., & Egeth, H. E. (2006). It's under control: Top-down search strategies can override attentional capture. Psychonomic Bulletin & Review, 13(1), 132-138.
Logan, G. D. (1988). Toward an Instance Theory of Automatization. Psychological Review, 95(4), 492-527.
Lu, S., & Zhou, K. (2005). Stimulus-driven attentional capture by equiluminant color change. Psychonomic Bulletin & Review, 12(3), 567-572.
Maljkovic, V., & Nakayama, K. (1996). Priming of pop-out .2. The role of position. Perception & Psychophysics, 58(7), 977-991.
MartinEmerson, R., & Kramer, A. F. (1997). Offset transients modulate attentional capture by sudden onsets. Perception & Psychophysics, 59(5), 739-751.
Masson, M. E. J., & Loftus, G. R. (2003). Using confidence intervals for graphically based data interpretation. Canadian Journal of Experimental Psychology-Revue Canadienne De Psychologie Experimentale, 57(3), 203-220.
Michotte, A. (1963). The perception of causality (M. R. Miles & E. Miles, Trans.). Londen: Methuen.
Miller, J. (1989). The Control of Attention by Abrupt Visual Onsets and Offsets. Perception & Psychophysics, 45(6), 567-571.
Mitroff, S. R., & Alvarez, G. A. (2007). Space and time, not surface features, guide object persistence. Psychonomic Bulletin & Review, 14(6), 1199-1204.
Mitroff, S. R., Scholl, B. J., & Noles, N. S. (2007). Object files can be purely episodic. Perception, 36(12), 1730-1735.
Mitroff, S. R., Scholl, B. J., & Wynn, K. (2004). Divide and conquer - How object files adapt when a persisting object splits into two. Psychological Science, 15(6), 420-425.
Moore, C. M., & Enns, J. T. (2004). Object updating and the flash-lag effect. Psychological Science, 15(12), 866-871.
REFERENCES
160
Moore, C. M., Stephens, T., & Hein, E. (2010). Features, as well as space and time, guide object persistence. Psychonomic Bulletin & Review, 17(5), 731-736.
Muller, N. G., & Kleinschmidt, A. (2003). Dynamic interaction of object- and space-based attention in retinotopic visual areas. Journal of Neuroscience, 23(30), 9812-9816.
Noles, N. S., Scholl, B. J., & Mitroff, S. R. (2005). The persistence of object file representations. Perception & Psychophysics, 67(2), 324-334.
Nothdurft, H. C. (1993). Saliency Effects across Dimensions in Visual-Search. Vision Research, 33(5-6), 839-844.
Nothdurft, H. C. (2006). Salience and target selection in visual search. Visual Cognition, 14(4-8), 514-542.
Pashler, H. (1988). Cross-Dimensional Interaction and Texture Segregation. Perception & Psychophysics, 43(4), 307-318.
Pinto, Y., Olivers, C. N. L., & Theeuwes, J. (2005). Target uncertainty does not lead to more distraction by singletons: Intertrial priming does. Perception & Psychophysics, 67(8), 1354-1361.
Posner, M. I. (1980). Orienting of Attention. Quarterly Journal of Experimental Psychology, 32(Feb), 3-25.
Posner, M. I., & Cohen, Y. (1984). Components of Visual Orienting. Attention and Performance(10), 531-556.
Posner, M. I., Snyder, C. R. R., & Davidson, B. J. (1980). Attention and the Detection of Signals. Journal of Experimental Psychology-General, 109(2), 160-174.
Pratt, J., & McAuliffe, J. (2002). Determining whether attentional control settings are inclusive or exclusive. Perception & Psychophysics, 64(8), 1361-1370.
Pratt, J., & Sekuler, A. B. (2001). The effects of occlusion and past experience on the allocation of object-based attention. Psychonomic Bulletin & Review, 8(4), 721-727.
Pratt, J., Sekuler, A. B., & McAuliffe, J. (2001). The role of attentional set on attentional cueing and inhibition of return. Visual Cognition, 8(1), 33-46.
Pylyshyn, Z. W. (2000). Situating vision in the world. Trends in Cognitive Sciences, 4(5), 197-207.
Pylyshyn, Z. W. (2001). Visual indexes, preconceptual objects, and situated vision. Cognition, 80(1-2), 127-158.
Pylyshyn, Z. W., & Storm, R. W. (1988). Tracking multiple independent targets: Evidence for a parallel tracking mechanism. Spatial Vision, 3, 18.
Rafal, R. D., Calabresi, P. A., Brennan, C. W., & Sciolto, T. K. (1989). Saccade Preparation Inhibits Reorienting to Recently Attended Locations. Journal of Experimental Psychology-Human Perception and Performance, 15(4), 673-685.
Rauschenberger, R. (2003a). Attentional capture by auto- and allo-cues. Psychonomic Bulletin & Review, 10(4), 814-842.
Rauschenberger, R. (2003b). When something old becomes something new: Spatiotemporal object continuity and attentional capture. Journal of Experimental Psychology-Human Perception and Performance, 29(3), 600-615.
REFERENCES
161
Remington, R., Johnston, J. C., & Yantis, S. (1986). Do Abrupt Onsets Capture Attention Involuntarily. Bulletin of the Psychonomic Society, 24(5), 347-347.
Remington, R. W., Johnston, J. C., & Yantis, S. (1992). Involuntary Attentional Capture by Abrupt Onsets. Perception & Psychophysics, 51(3), 279-290.
Reynolds, J. H., & Desimone, R. (2003). Interacting roles of attention and visual salience in V4. Neuron, 37(5), 853-863.
Richard, A. M., Luck, S. J., & Hollingworth, A. (2008). Establishing object correspondence across eye movements: Flexible use of spatiotemporal and surface feature information. Cognition, 109(1), 66-88.
Ruz, M., & Lupianez, J. (2002). A review of attentional capture: On its automaticity and sensitivity to endogenous control. Psicologica(23), 289-309.
Scholl, B. J. (2001). Objects and attention: the state of the art. Cognition, 80(1-2), 1-46. Scholl, B. J. (2007). Object persistence in philosophy and psychology. Mind & Language,
22(5), 563-591. Schreij, D., & Olivers, C. N. L. (2009). Object representations maintain attentional control
settings across space and time. Cognition, 113(1), 111-116. Schreij, D., & Olivers, C. N. L. (submitted). Object representations maintain attentional
control settings for task-relevant information. Journal of Experimental Psychology-Human Perception and Performance.
Schreij, D., Owens, C., & Theeuwes, J. (2008). Abrupt onsets capture attention independent of top-down control settings. Perception & Psychophysics, 70(2), 208-218.
Schreij, D., Theeuwes, J., & Olivers, C. N. L. (2010). Abrupt onsets capture attention independent of top-down control settings II: additivity is no evidence for filtering. Attention, Perception & Psychophysics, 72(3), 672-682.
Simons, D. J., & Chabris, C. F. (1999). Gorillas in our midst: sustained inattentional blindness for dynamic events. Perception, 28(9), 1059-1074.
Sternberg, S. (1969). Discovery of Processing Stages - Extensions of Donders Method. Acta Psychologica, 30, 276-+.
Theeuwes, J. (1990). Perceptual Selectivity Is Task Dependent - Evidence from Selective Search. Acta Psychologica, 74(1), 81-99.
Theeuwes, J. (1991a). Cross-Dimensional Perceptual Selectivity. Perception & Psychophysics, 50(2), 184-193.
Theeuwes, J. (1991b). Exogenous and Endogenous Control of Attention - the Effect of Visual Onsets and Offsets. Perception & Psychophysics, 49(1), 83-90.
Theeuwes, J. (1992). Perceptual Selectivity for Color and Form. Perception & Psychophysics, 51(6), 599-606.
Theeuwes, J. (1994a). Endogenous and Exogenous Control of Visual Selection. Perception, 23(4), 429-440.
Theeuwes, J. (1994b). Stimulus-Driven Capture and Attentional Set - Selective Search for Color and Visual Abrupt Onsets. Journal of Experimental Psychology-Human Perception and Performance, 20(4), 799-806.
REFERENCES
162
Theeuwes, J. (1995a). Abrupt Luminance Change Pops out - Abrupt Color-Change Does Not. Perception & Psychophysics, 57(5), 637-644.
Theeuwes, J. (1995b). Perceptual selectivity for color and form: On the nature of the interference effect. In A. F. Kramer , M. G. H. Coles & G. D. Logan (Eds.), Converging operations in the study of visual selective attention (pp. 297-314). Washington, DC: American Psychological Association.
Theeuwes, J. (1996). Perceptual selectivity for color and form: On the nature of the interference effect. In A. F. Kramer , M. G. H. Coles & G. D. Logan (Eds.), Converging operations in the study of visual selective attention (pp. 297-314). Washington, DC: American Psychological Association.
Theeuwes, J. (2000). Attentional capture and oculomotor control. Perception, 29, 2-2. Theeuwes, J. (2004). Top-down search strategies cannot override attentional capture.
Psychonomic Bulletin & Review, 11(1), 65-70. Theeuwes, J. (2010). Top-down and bottom-up control of visual selection. Acta
Psychologica, 135(2), 77-99. Theeuwes, J., & Burger, R. (1998). Attentional control during visual search: The effect of
irrelevant singletons. Journal of Experimental Psychology-Human Perception and Performance, 24(5), 1342-1353.
Theeuwes, J., De Vries, G. J., & Godjin, R. (2003). Attentional and oculomotor capture with static singletons. Perception & Psychophysics, 65(5), 735-746.
Theeuwes, J., & Godijn, R. (2001). Attentional and oculomotor capture with static singletons. In C. L. Folk & B. S. Gibson (Eds.), Attraction, distraction and action: Multiple perspectives on attentional capture (pp. 121-149). New York: Elsevier.
Theeuwes, J., & Godijn, R. (2002). Irrelevant singletons capture attention: Evidence from inhibition of return. Perception & Psychophysics, 64(5), 764-770.
Theeuwes, J., Kramer, A. F., & Atchley, P. (2000). On the time course of top-down and bottom-up control of visual attention. In S. Monsell & J. Driver (Eds.), Attention & Performance (Vol. 18). Cambridge: MIT Press.
Theeuwes, J., Kramer, A. F., Hahn, S., & Irwin, D. E. (1998). Our eyes do not always go where we want them to go: Capture of the eyes by new objects. Psychological Science, 9(5), 379-385.
Theeuwes, J., Kramer, A. F., Hahn, S., Irwin, D. E., & Zelinsky, G. J. (1999). Influence of attentional capture on oculomotor control. Journal of Experimental Psychology-Human Perception and Performance, 25(6), 1595-1608.
Theeuwes, J., & van der Burg, E. (2008). The role of cueing in attentional capture. Visual Cognition, 16(2-3), 232-247.
Tipper, S. P., Brehaut, J. C., & Driver, J. (1990). Selection of Moving and Static Objects for the Control of Spatially Directed Action. Journal of Experimental Psychology-Human Perception and Performance, 16(3), 492-504.
Tipper, S. P., Grison, S., & Kessler, K. (2003). Long-term inhibition of return of attention. Psychological Science, 14(1), 19-25.
REFERENCES
163
Tipper, S. P., Weaver, B., Jerreat, L. M., & Burak, A. L. (1994). Object-Based and Environment-Based Inhibition of Return of Visual-Attention. Journal of Experimental Psychology-Human Perception and Performance, 20(3), 478-499.
Todd, J. T., & Van Gelder, P. (1979). Implications of a Transient-Sustained Dichotomy for the Measurement of Human-Performance. Journal of Experimental Psychology-Human Perception and Performance, 5(4), 625-638.
Treisman, A., & Gelade, G. (1980). Feature-Integration Theory of Attention. Cognitive Psychology, 12(1), 97-136.
Treisman, A., & Kahneman, D. (1983). The Accumulation of Information within Object Files. Bulletin of the Psychonomic Society, 21(5), 354-354.
Treisman, A., & Sato, S. (1990). Conjunction Search Revisited. Journal of Experimental Psychology-Human Perception and Performance, 16(3), 459-478.
Tsal, Y., & Benoni, H. (2010a). Diluting the Burden of Load: Perceptual Load Effects Are Simply Dilution Effects. Journal of Experimental Psychology-Human Perception and Performance, 36(6), 1645-1656.
Tsal, Y., & Benoni, H. (2010b). Much Dilution Little Load in Lavie and Torralbo's (2010) Response: A Reply. Journal of Experimental Psychology-Human Perception and Performance, 36(6), 1665-1668.
Tsal, Y., & Benoni, H. (2010c). Where have we gone wrong? Perceptual load does not affect selective attention. Vision Research, 50(13), 1292-1298.
Tsotsos, J. K., Culhane, S. M., Wai, W. Y. K., Lai, Y. H., Davis, N., & Nuflo, F. (1995). Modeling Visual-Attention Via Selective Tuning. Artificial Intelligence, 78(1-2), 507-545.
von Muhlenen, A., Rempel, M. I., & Enns, J. T. (2005). Unique temporal change is the key to attentional capture. Psychological Science, 16(12), 979-986.
Weger, U. W., Abrams, R. A., Law, M. B., & Pratt, J. (2008). Attending to objects: Endogenous cues can produce inhibition of return. Visual Cognition, 16(5), 659-674.
Wolfe, J. M. (1994). Guided Search 2.0 - a Revised Model of Visual-Search. Psychonomic Bulletin & Review, 1(2), 202-238.
Xu, F., & Carey, S. (1996). Infants' metaphysics: The case of numerical identity. Cognitive Psychology, 30(2), 111-153.
Xu, F., & Carey, S. (2000). The emergence of kind concepts: a rejoinder to Needham and Baillargeon (2000). Cognition, 74(3), 285-301.
Xu, Y. D., & Chun, M. M. (2006). Dissociable neural mechanisms supporting visual short-term memory for objects. Nature, 440(7080), 91-95.
Yantis, S. (1993). Stimulus-Driven Attentional Capture and Attentional Control Settings. Journal of Experimental Psychology-Human Perception and Performance, 19(3), 676-681.
Yantis, S. (1996). Attentional capture in vision. In A. F. Kramer, M. G. H. Coles & G. D. Logan (Eds.), Converging operations in the study of visual selective attention (pp. 45-76). Washington, DC: American Psychological Association.
REFERENCES
164
Yantis, S., & Egeth, H. E. (1999). On the distinction between visual salience and stimulus-driven attentional capture. Journal of Experimental Psychology-Human Perception and Performance, 25(3), 661-676.
Yantis, S., & Hillstrom, A. P. (1994). Stimulus-Driven Attentional Capture - Evidence from Equiluminant Visual Objects. Journal of Experimental Psychology-Human Perception and Performance, 20(1), 95-107.
Yantis, S., & Johnson, D. N. (1990). Mechanisms of Attentional Priority. Journal of Experimental Psychology-Human Perception and Performance, 16(4), 812-825.
Yantis, S., & Jonides, J. (1984). Abrupt Visual Onsets and Selective Attention - Evidence from Visual-Search. Journal of Experimental Psychology-Human Perception and Performance, 10(5), 601-621.
Yantis, S., & Jonides, J. (1990). Abrupt Visual Onsets and Selective Attention - Voluntary Versus Automatic Allocation. Journal of Experimental Psychology-Human Perception and Performance, 16(1), 121-134.
Yi, D. J., Turk-Browne, N. B., Flombaum, J. I., Kim, M. S., Scholl, B. J., & Chun, M. M. (2008). Spatiotemporal object continuity in human ventral visual cortex. Proceedings of the National Academy of Sciences of the United States of America, 105(26), 8840-8845.
ACKNOWLEDGEMENTS / DANKWOORD
165
Acknowledgements / Dankwoord
Over the past four years that I have been doing research for this thesis, many people have in their own way contributed to its realization. To these people I would like to dedicate a few words of appreciation.
First of all, I would of course like to thank Chris Olivers, my supervisor and co-promotor, for guiding me in taking the first steps in the world of science. Thanks to you, I have gained numerous invaluable skills which I would never have acquired on my own. Not only did you greatly contribute to all the things I have learned, you also made sure our scientific endeavors always were a lot of fun and remained interesting to pursue, even if experiments sometimes did not seem to work out as well as hoped in the beginning. Thanks for being such an awesome supervisor!
Second, I would like to thank Jan Theeuwes, my promotor and professor, for coming up with the idea to pursue a career in science while we were hiking through the Blue Mountains in Australia. Many times you have been able to steer shaky research ideas in the right direction, often leading to successful experiments. I always liked your witty and provocative jokes, together with the interesting discussions you often managed to incite during lunch breaks. And remember, if you ever need my help with getting your phone or laptop on the wireless network of the VU again, just call.
My time at the Cognitive Psychology department of course wouldn’t have been as enjoyable without all of its other members. I would like to thank Artem, Mieke, Dirk, Richard, Hannie, Adelbert, Martijn, Erik, Thomas, Yaïr, Manon and Durk for making the last four years as fun and cultivating as they were. I am going to miss our lunch-break volleyball matches! Wieske and Clayton, thanks for helping me with preparing my trip to Vancouver and assisting me in creating my first conference poster. Sander, I always appreciated the sound advice you were able to give me whenever I got lost in the wonderland of complex statistical analyses again. Jaap, I am glad you could often help me with important organizational matters and that we could always share our enthusiasm for technical stuff with each other. Sebastiaan, thanks for the great toolset you built from scratch for creating and conducting experiments. I could always count on your help and expertise when the occasional problem occurred. And of course I could not forget Janne, Anna, Lisette, Alisha, Judith, Mauricio, Isabel, Kim, Wouter, Marlou and Shanna. We have always had lots of fun at the office, but even more so during the activities and trips we undertook outside working hours. Thanks for the dinners, excursions, afternoons in the park and the great atmosphere you generally created at the department!
Furthermore, I would like to thank prof. Jim Enns, prof. Alan Kingstone and all the other people at the Vision and BAR labs at the University of British Columbia in Vancouver, Canada for making the five months that I visited the best I could imagine. Jim, a special thanks to you, for providing the support and fruitful ideas for the experiments I
ACKNOWLEDGEMENTS/DANKWOORD
166
conducted at your lab. I have very fond memories of my time in Vancouver and I hope I will be able visit all of you again sometime soon.
I also would like to thank René, Henk-Jan and Sam for having been such good friends as long as I can remember. Sam, I am very impressed by the cover that you have created for this thesis. I have always appreciated your artist’s perspective on things and am still grateful that you sparked my interest in photography years ago.
I would also like to thank my friends Paul and Ronald, who were willing to stand by me during the defense of my dissertation. Together with Jurgen and Dorine, you have been the most valuable friends for years. We have known many good times together when traveling, going out or just hanging out. Thanks for lightening me up with one of your crazy jokes whenever I had ‘one of those days’ again.
Kah-Kin, Esmeralda, Kah-Wai, Myrthe and Wasja, thanks for having been such wonderful flat mates and friends for so many years. Together we made sure that we had a place which really felt like home, to which we could return after work each day.
Furthermore, I’d like to say thanks to Hanneke and Suzan, for being the best surfing and travelling buddies. No wave or mountain has been too high for us (ok, the wave part is indeed a lie). To Fong and Emma, for lending me your hands to exhibit on the cover. To all the people of the SDVA, for the numerous swims, dives and just very gezellige days I have experienced with you for many years now.
And most of all I would like to thank my parents, Leonie and Gerard, and my sister Michèle for their unfaltering support of me no matter what decision I made, good or bad, and for just having been there all my life.
CURRICULUM VITAE
167
Curriculum Vitae
Daniel was born on July 14th, 1982 in Amsterdam, the Netherlands. He received a
bachelor’s degree in Artificial Intelligence from the VU University Amsterdam in August
of 2003 and continued his studies there to receive a Master’s degree in Cognitive Science
in August of 2006. He conducted part of the required research for his Master’s thesis at
the University of Sydney, Australia under the supervision of dr. Caleb Owens and prof.dr.
Jan Theeuwes. The thesis is titled “Involuntary Contingent Orienting is subject to
Bottom-Up Attentional Capture”. He began as a PhD student at the VU University
Amsterdam in September of 2007 and conducted part of his research at the University of
British Columbia in Vancouver, Canada in the first half of 2011 under supervision of
prof.dr. Jim Enns.
AUTHOR PUBLICATIONS
168
Author publications
Belopolsky, A. V., Schreij, D., & Theeuwes, J. (2010). What is top-down about contingent capture? Attention, Perception & Psychophysics, 72(2), 326-341.
Mathôt, S., Schreij, D., & Theeuwes, J. (in press). OpenSesame: An open-source, graphical
experiment builder for the social sciences. Behavior Research Methods.
Schreij, D., Owens, C., & Theeuwes, J. (2008). Abrupt onsets capture attention
independent of top-down control settings. Perception & Psychophysics, 70(2), 208-218.
Schreij, D., & Olivers, C. N. L. (2009). Object representations maintain attentional
control settings across space and time. Cognition, 113(1), 111-116. Schreij, D., Theeuwes, J., & Olivers, C. N. L. (2010). Abrupt onsets capture attention
independent of top-down control settings II: additivity is no evidence for filtering. Attention, Perception & Psychophysics, 72(3), 672-682.
Schreij, D., Theeuwes, J., & Olivers, C. N. L. (2010). Irrelevant onsets cause inhibition of
return regardless of attentional set. Attention Perception & Psychophysics, 72(7), 1725-1729.