Content uploaded by Petter Johansson
Author content
All content in this area was uploaded by Petter Johansson
Content may be subject to copyright.
142
Psychologia, 2008, 51, 142–155
FROM CHANGE BLINDNESS TO CHOICE BLINDNESS
Petter JOHANSSON1), Lars HALL2)
1)The University of Tokyo, Japan, 2)Harvard University, U.S.A.
and
Sverker SIKSTRÖM3)
3)Lund University, Sweden
The phenomenon of change blindness has received a great deal of attention during
the last decade, but very few experiments have examined the effects of the
subjective importance of the visual stimuli under study. We have addressed this
question in a series of studies by introducing choice as a critical variable in change
detection (see Johansson, Hall, Sikström, & Olsson, 2005, Johansson, Hall, Sikström,
& Tärning, 2006). In the present study, participants were asked to choose which of
two pictures they found more attractive. For stimuli we used both pairs of abstract
patterns and female faces. Sometimes the pictures were switched during to choice
procedure, leading to a reversal of the initial choice of the participants. Surprisingly,
the subjects seldom noticed the switch, and in a post-test memory task, they also
often remembered the manipulated choice as being their own. In combination with
our previous findings, this result indicates that we often fail to notice changes in the
world even if they have later consequences for our own actions.
Key words: Choice Blindness, Change Blindness, Intention, Decision, Action
Even if naïve participants often express bewilderment and disbelief during change
blindness experiments, the results of these experiments no longer surprise scientists
working in the field. In the last decade, a mass of empirical studies of change blindness
have been published in the journals of cognitive science and vision research (for
overviews, see Rensink 2000a; Rensink 2002; Simons & Rensink 2005), and change
blindness now serves as a standard example in cognitive science, on a par with the Stroop
effect and the Kaniza triangle.
The common denominator in experiments on change blindness is that participants
fail to detect changes in a scene when the change is accompanied by some other visual
disturbance. If the same changes had occurred in plain sight, with no interruptions in the
visual stream, they would have been detected instantaneously. While the mechanisms
behind this effect has not yet been universally agreed upon (Simons, 2000), experiments
The authors thank Ronald Rensink and an anonymous reviewer for helpful comments and suggestions.
Petter Johansson thanks the Japan Society for the Promotion of Science, Lars Hall thanks the Swedish
Research Council and the Erik-Philip Sörensen Foundation, and Sverker Sikström thanks The Swedish
Research Council for financial support.
Correspondence concerning this article should be addressed to Petter Johansson, the University of Tokyo,
4-6-1 Komaba, Meguro-ku, Tokyo, 153-8904, Japan (e-mail: petter@fennel.rcast.u-tokyo.ac.jp)
FROM CHANGE BLINDNESS TO CHOICE BLINDNESS 143
on change blindness has been put to great use in mapping out the fine-grained properties
of attention, and has lead to a deepened understanding of the various stages of visual
processing (Rensink, 2000b; Tse, Sheinberg & Logothetis, 2003). More controversially,
change blindness has also served as a focal point in the debate about the nature of visual
consciousness (the so called Grand Illusion Debate, see Noë, 2002), where a proposal
have been made that change blindness shows that we all have a drastically false
conception of our own visual experiences (e.g. Blackmore, 2002). A less radical
conclusion to draw from these experiments is that we represent the world in much less
detail than what was previously thought. Instead, when we need to be informed, we just
direct our attention toward those features of the visual environment that is of current
importance (as Brooks 1991, and later O’Regan and Noë 2002, put it: “allowing the world
to be its own best model”). Thus, in this process, we rely on the stability of the world, and
we implicitly assume that it does not change in undetectable ways.
While the study of change blindness has proliferated into various sub-fields, both
with respect to the theoretical outlook (Mitroff, Simons, & Franconeri, 2002; Rensink,
2002), and to the techniques used (Grimes, 1996; O’Regan, Rensink, & Clark, 1999;
Smilek, Eastwood, & Merikle, 2000) there has been surprisingly little research aimed at
investigating our ability to detect changes when the stability of the world is of particular
importance to us—i.e. when changes in the visual environment have effects in relation to
our intentions and actions. As Rensink (2002) writes:
The study of change detection has evolved over many years, proceeding through
phases that have emphasized different types of stimuli and different types of tasks.
All studies, however, rely on the same basic design. An observer is initially shown a
stimulus ... a change of some kind is made to this stimulus ... and the response of the
observer is then measured.
(p. 251, our emphasis)
Research on change blindness has occasionally contained elements of interaction
(most notably, the real-person interactions in Simons & Levin, 1998, and Levin, Simons,
Angelone, & Chabris, 2002), and at least one task in which the actions of the participants
have functional relevance has been investigated (Triesch, Ballard, Hayhoe, & Sullivan,
2003), but the full potential of change blindness as a tool for studying the human mind is
far from realized. Why should change blindness be used only to study perceptual aspects
of cognition?
In a series of studies, we have modified the basic design of change blindness
experiments to incorporate other non-perceptual elements of cognition. The result is a
novel research tool we call choice blindness, in which we surreptitiously manipulate the
relationship between the choice and outcome that our participants experience (Hall,
Johansson, Tärning, Sikström, & Deutgen, submitted; Johansson et al., 2005; Johansson et
al., 2006). In particular, we have been interested in the relationship between intention,
choice, and introspection. For example, in Johansson et al. (2005), the participants were
shown pairs of pictures of female faces, and were instructed to point at the face they found
most attractive. After pointing, the chosen picture was given to the participants, and they
were asked to explain why they preferred the picture they now held in their hand.
JOHANSSON, HALL, & SIKSTRÖM144
Unknown to the participants, on certain trials, a double-card ploy was used to covertly
exchange one face for the other. Thus, on these trials, the outcome of the choice became
the opposite of what they intended.
The number of manipulated trials detected by the participants was surprisingly low.
Even when they were given unlimited time to deliberate upon their choice no more than
30% of all manipulated trials were detected. But not only were the participants often blind
to the manipulation of their choices, they also offered introspectively derived reasons for
preferring the alternative they were given instead. In addition to this, manipulated and non-
manipulated reports were compared on a number of different dimensions, such as the level
of emotionality, specificity and certainty expressed, but no substantial differences were
found (see Johansson et al., 2006). Choice blindness also extends to other modalities than
vision. In a recent consumer choice study we have demonstrated the effect for the taste of
jam and the smell of tea. Even for such remarkably different jams as spicy cinnamon apple
vs bitter grapefruit, or for the smell of teas like sweet mango vs liquorice pernod, were no
more than half of the manipulated trials detected (see Hall et al., submitted).
In change blindness experiments participants are usually more likely to notice
changes when they concern features of particular relevance to the scene, or if they are of
central interest to the participants, or if the participants are particularly knowledgeable
about them (Rensink, 2002; Triesch et al., 2003). For choices it would almost seem to be
a defining feature that they concern properties of high relevance and interest, or things we
are very knowledgeable about. But in our experiments, in the great majority of trials, our
participants were blind to the mismatch between choice and outcome. While intending to
choose X (a central-interest, non-peripheral, valenced stimuli), they failed to take notice
when ending up with Y. This is a result that ought to be surprising even to the most
seasoned change blindness researcher.
Before implementing the card-based paradigm of Johansson et al. (2005) we ran a
series of studies exploring the phenomenon of choice blindness. These studies are hitherto
unpublished, but importantly add to the evidential base for choice blindness by
demonstrating the effect in a different medium, with a different design, with different
types of stimuli, and with additional post-test controls to guarantee that the manipulated
images have been adequately processed by the participants1. We will here present this
work in greater detail, and relate it to our previously published results.
In these studies, the participants had to choose which one of two abstract patterns
they found most appealing (Experiment 1) and which one of two female faces they found
most attractive (Experiment 2 and 3). The alternatives were presented on a computer
screen, and the participants had to indicate their choice by moving the cursor to the chosen
picture. Each experiment consisted of 15 trials, three of which were manipulated (i.e.
after the choice was indicated the participants were presented with the non-chosen
alternative instead). When all the choice trials were completed, an unannounced memory
test was introduced. The participants had to look at all the pairs again, presented in a
1 These experiments have previously only been reported in summary form in the Supporting Online
Material for Johansson et al. (2005).
FROM CHANGE BLINDNESS TO CHOICE BLINDNESS 145
randomized order and without time-constraint, and try to remember which face or pattern
they previously preferred.
EXPERIMENT 1
METHOD
Participants. Twenty undergraduate students (12 female) at Lund University participated in the study.
They received a cinema ticket for their participation. The experiment was described as a test of rapid,
intuitive judgment of aesthetic beauty. All participants were naïve about the actual purpose of the
experiment.
Material. As stimulus material we used abstract patterns collected from various websites containing
“artistic” computer wallpaper for non-commercial use. The pictures were organized in pairs, roughly
matched for similarity and attractiveness, covering a range from “similar” to “not so similar”. The matching
was performed by the authors. The presentation size on the screen was around 5.0° × 5.0° visual angle (non-
fixated viewing distance around 60 cm with a picture size of 5 × 5 cm)
Procedure. Experiment 1 consisted of a simple binary choice task, where participants had to choose
which one of two abstract patterns presented on a computer screen they found most aesthetically appealing
(see Figure 1). Each trial began when the participants clicked on a left-aligned start-icon that made two
patterns appear on the right side of the screen. Participants were given 1500 ms to consider their choice, then
a beep was played, and they had to move the cursor to the preferred pattern. In addition, the cursor trajectory
had to pass through one of two small, color-coded, intermediate squares corresponding to either the upper or
the lower pattern on the right. These two squares only became visible after the sound was played, and to
prevent learning-effects the vertical position of the squares was randomized within their half of the screen.
The upper square was always red and the lower square was always blue, and when the participants passed
through one of these squares, the entire screen flashed in matching color for 50 ms. The intermediate square
and the screen flash were explained to the participants as a way to help them keep the “pace” of the
experiment.
After the participants completed their choice, the indicated pattern was framed in the same color as the
prior intermediate box, and the non-chosen picture was removed from the screen. The chosen picture
remained on the screen for an additional 1500 ms after the choice was completed. If the participants had not
yet managed to complete a choice 1500 ms after the sound alert, the trial ended, and was categorized as a
mistrial. The full experiment consisted of 15 trials2.
For each participant, on 3 of these trials a change manipulation was introduced (see Figure 1c). On a
manipulation trial, the attention-grabbing properties of the midway square and the 50 ms screen flash were
used to conceal the fact that the two choice alternatives switched places while the participants were moving
the cursor across the screen. The manipulation always occurred on trial 7, 10 and 14, but the presentation
order of the pairs was randomized.
After all 15 trials had been completed, the participants were given an unannounced memory test. The
same pairs of patterns were once again presented, and the participants were asked to indicate which one of the
two patterns they had previously found most appealing. In this phase, no time constraints were imposed.
Before the experiment started, the participants were given 10 practice trials. After the experiment all
participants were debriefed, and asked whether they consented to have the data from their trials included in
the analysis.
A trial was classified as detected if participants showed signs of detection concurrent with the switch
(such as explicitly reporting that the patterns had been switched, or that something went wrong with their
2 149 of 900 trials (16.5%) in the three experiments were classified as mistrials and were removed from
further analyses. There were no differences between manipulated and non-manipulated trials in the number of
mistrials.
JOHANSSON, HALL, & SIKSTRÖM146
choice, or by showing signs of confusion and surprise), or if they later in post-experiment interviews claimed
to have detected a switch or sensed that something went wrong. For participants that did not show any
concurrent signs of detection, a series of increasingly specific questions were asked to make sure their
responses were not misclassified as non-detected: “What did you think about the experiment?”, “Did you find
anything odd with the experiment?” and “Did you notice anything strange with the stimuli presented in the
experiment?”. At this point, if the participants still revealed no sign of having noticed anything odd with the
experiment, they were told that we planned a follow-up study in which the patterns presented sometimes
would switch place in mid-trial (i.e. an actual description of the current experiment), and asked if they
believed that they would have noticed such a switch. Finally, participants were asked if they had noticed
anything in the current experiment resembling the hypothetical switches that we had just described. If they
answered no to this question, we concluded that they did not consciously notice any of the manipulations
made during the experiment.
Previously, we have made a distinction between concurrent and retrospective detection of the
manipulation, i.e. if the participants indicated that they detected a manipulation during the experiment or in
the post-test interview. As we did not verbally interact with the participants during the present experiments,
the category of concurrent detection is not as reliable as in our previous work. Therefore, in the current
article, we report detection as a single measure based on an overall interpretation of the participant’s
immediate reactions and their answers in the post-test interview.
The post-test memory task was included to measure if a (non-detected) manipulation would influence
what the participants remembered as their “own” choice, i.e. if the original choice or the manipulated
outcome would be remembered as the picture preferred. The memory task also served as an independent
measure that the pictures were processed after the manipulation was performed. If we were to find no
Fig. 1. Step-by-step progression of a manipulated trial. A. The participants press the start icon and the two
pictures appear on the right hand side. B. After 1500 ms a beep is played, and the participants
moves the cursor to the midway square corresponding to the chosen picture. C. When the cursor
hits the square the screen is occluded for 50 ms. D. The participants continue the movement to the
chosen (but now altered) picture, and when it is reached the non-chosen alternative is removed from
the screen. The chosen picture is then framed and remains visible for 1500 ms. Note, for purposes
of illustration the pictures are here somewhat magnified compared to their size in the experiment.
FROM CHANGE BLINDNESS TO CHOICE BLINDNESS 147
differences between the manipulated and the non-manipulated trials on the memory task, it might mean that
the pictures were not fully processed after the switch.
Results. In Experiment 1, only 19% of the manipulated trials were categorized as
detected (see Figure 3). One participant detected all of the switches, three participants
detected two of the switches, one participant detected one switch, and 15 of the participant
did not detect any of the switches. Of the 15 participants that did not detect any of the
switches, 12 believed themselves to be able to do so, had any manipulations been made.
Of the non-manipulated choices, 86% were remembered correctly in the post-test memory
task. For the manipulated trials, the original choice was remembered in 61% of the trials
(see Figure 4). The distribution differs significantly between non-manipulated and
manipulated trials (Chi square (1, N=20)=6.95, p< 0.0084), showing that the
manipulation influenced what the participants remembered as being their own choice.
EXPERIMENT 2
In the post-test interviews in Experiment 1, most participants described the choice
task as being both “real” and meaningful. Nevertheless, it could still be argued that there
is something slightly artificial about evaluating abstract patterns, as it is something most
people have very little experience of 3. To provide a more critical test of our approach, we
therefore chose to use human faces as the stimuli in Experiment 2. In contrast to abstract
patterns, most people have had a lot of practice in evaluating faces, and they often have
strong opinions about attractiveness. Given this, it seems likely that we would be better at
detecting manipulations of faces than most other stimuli.
METHOD
Participants. Twenty undergraduate students (11 female), at Lund University participated in the study.
They were given a cinema ticket for their participation. The experiment was described as a test of rapid,
intuitive judgment of attractiveness. All participants were naïve about the actual purpose of the experiment.
Material. Experiment 2 used gray-scale pictures of female faces (taken from the University of Stirling
database (PICS), see Figure 2). The pictures were organized in pairs, roughly matched for similarity and
attractiveness. The matching was performed by the authors. The presentation size on the screen was around
5.0° × 5.0° visual angle (non-fixated viewing distance around 60 cm with a picture size of 5 × 5 cm)
Procedure. As in Experiment 1, participants were given the task to choose the picture they preferred
the most. However, the exact wording of the instructions was changed from “choose the pattern you find
most aesthetically appealing” to “choose the face you find most attractive”. The procedure employed was the
same as that in Experiment 1, using 15 trials, three of which were manipulated.
Results. In Experiment 2, the detection rate for the manipulated trials was 12% (see
3 But this is not true for all participants. For instance, an architect student had very strong views on the use
of symmetry and what colours could be mixed without unbalancing the picture etc. When the actual
procedure was revealed she simply refused to believe that something like that could have taken place.
JOHANSSON, HALL, & SIKSTRÖM148
Figure 3). This detection rate does not differ statistically from Experiment 1. Two
participants detected two of the switches, two participants detected one switch, and 16 of
the participants did not detect any of the switches. Of the 16 participants that did not
detect any of the switches, 14 believed themselves to be able to do so, had any
manipulations been made. The participants remembered their choices in 87% of the trials
in the post-test memory task. For the manipulated trials, the participants indicated their
“original” choice as being what they chose for 76% of the trials (see Figure 4). This
number does not differ significantly from the results of the non-manipulated trials.
EXPERIMENT 3
At the outset, it seemed likely to us that the change of stimulus material would lead
to a difference in detection rate between Experiment 1 and Experiment 2. However, this
was not the case. One possible explanation is that there are other factors than the nature of
the stimuli that are more important in determining the detection rate. For instance, it may
be the case that the relative “distance” between the items paired are not equivalent in the
two experiments (e.g. that the face pairs differed less in similarity or attractiveness
compared to the pairs of patterns used in Experiment 14). Another possible explanation is
that the participants did not fully process the faces after the switch in Experiment 2. In
Experiment 1 we found a difference between the manipulated and non-manipulated trials
Fig. 2. Examples of pairs of pictures used.
4 The use of two sorts of stimuli should be seen more as a further test of the general phenomenon of choice
blindness with this presentation technique rather than a thorough comparison between the likelihood of
detection for patterns and faces.
FROM CHANGE BLINDNESS TO CHOICE BLINDNESS 149
in the post-test memory task. The only candidate explanation for this result is the
manipulation itself—i.e. it can be assumed that the participants did in fact look at the
pictures after the switch, but did not realize they had been switched, and then later
remembered the altered alternative as the one they preferred. But in Experiment 2 there
were no differences in the memory test, and therefore we have no independent measure
that the participants actually attended to the faces after the switch. And if this was not the
case, then it is unsurprising that very few manipulation trials were reported as detected.
Thus, to make sure that the manipulated item was fully processed after the manipulation,
in Experiment 3, we included a rating task of the chosen and non-chosen faces directly
after the choice was completed. Now the pictures stayed visible on the screen until they
were rated for attractiveness.
METHOD
Participants. Twenty undergraduate students (10 female) at Lund University participated in the study.
They received a cinema ticket for their participation. The experiment was described as a test of rapid,
intuitive judgment of attractiveness. All participants were naïve of the actual purpose of the experiment.
Material. Experiment 3 used the same set of female faces as in Experiment 2.
Procedure. The procedure was the same as Experiment 2, with the following exceptions. After the
choice had been indicated on the screen, the chosen picture stayed visible and the participants were asked to
rate the face on scale for attractiveness from 1 to 9. The picture remained on the screen until the participants
had typed their numerical rating in a box next to the picture. After the chosen picture was rated it was
removed, and the non-chosen picture appeared, and the participants were asked to rate this alternative as well.
After the participants had done so, the next trial began. As in the previous experiments, the full set consisted
of 15 trials, three of which where manipulated.
Results. The detection rate in Experiment 3 was 39%. This is a significantly higher
level of detection compared to Experiment 2 (Chi Square (1, N= 20) = 8.75, p< 0.0031).
Fig. 3. Detection frequency in the three experiments.
JOHANSSON, HALL, & SIKSTRÖM150
Four participants detected all of the switches, two participants detected two of the
switches, three participants detected one switch, and 11 of the participant did not detect
any of the switches. Of the 11 participants that did not detect any of the switches, 9
believed themselves to be able to do so, had any manipulations been made.
The result on the memory test differed markedly when comparing non-detected
manipulated and non-manipulated trials. In the non-manipulated trials, the participants
remembered their original choice in 92% of the trials. But for the manipulated trials, only
33% of the originally chosen pictures were later remembered as being what the
participants preferred initially. The difference between manipulated and non-manipulated
trials is significant, (Chi Square (1, N= 20) = 69.62, p< 0.00001).
We also analyzed the attractiveness rating task. In the rating phase, when the faces
were presented again, the participants “knew” that the first face to be rated was the face
they originally chose. This means that the participants ought to rate the first face higher,
as that was the alternative they thought was the more attractive just a few seconds ago.
This is also what we found. In 89% of the non-manipulated trials, the ratings of the
participants were consistent with their initial choice. The same was true for the
manipulated trials. In 67% of the manipulated trials the participants rated the first picture
higher, even though this picture was not the one originally chosen.
DISCUSSION
We have described three experiments involving a simple choice task in combination
with a covert manipulation of the outcome of the choices made. The participants in our
Fig. 4. Memory of original choice.
FROM CHANGE BLINDNESS TO CHOICE BLINDNESS 151
experiments often failed to notice that the outcome of their choice became the opposite of
what they intended, thereby demonstrating the effect we have termed choice blindness. In
the experiments presented we varied both the stimuli used and the choice procedure. The
first experiment used abstract patterns, and in the second and third experiment we used
pictures of female faces. In all three experiments, the majority of the manipulations
remained undetected, indicating that choice blindness is a robust phenomenon. In relation
to our previously published studies, we have shown that it is possible to generate similar
results on a computerized test as with a real-world magic trick. This result demonstrates
that the prevalence of choice blindness is not strongly tied to the prior “believability” of
the switch (i.e. a card-based magic trick is generally perceived to be “impossible”, while
almost anything can happen in the virtual world of computer presentation). By using
abstract patterns we have also demonstrated the effect for a new type of stimuli. Finally,
by adding a memory test we have obtained a clear indication both that the participants
attend to the pictures after the switch, and that they tend to remember the manipulated
choice as being the alternative they choose themselves.
But given the counter-intuitive nature of the result, we need to carefully consider
some objections and alternative interpretations.
To be certain that the pictures were attended to after the switch, we amended the
procedure somewhat in the final study. In Experiment 3, we left the faces on the screen
until an attractiveness rating was made by the participants. It is very difficult to imagine
that the participants did not look at the pictures when performing this task.
A similar question is how can we know that 1500 ms is enough to form an opinion
about aesthetic preference. According to recent research we are remarkably fast at
forming opinions about the appearance of faces (Todorov, Mandisodza, Goren, & Hall,
2005). For example, it has been shown that an attractiveness evaluation of a face made
after as short exposure as 100 ms correlates highly with judgments made after free
viewing time (Willis & Todorov, 2006). This indicates that 1500 ms is sufficient time to
decide at least which of two faces are the more attractive.
But the most obvious objection to the results is that perhaps the participants actually
did notice all the manipulations, but for some reason they just did not tell us. We find this
to be quite unlikely. As was described in the procedure section of the first experiment, the
debriefing after the experiment involved asking the participants a series of questions, the
last one being if they thought they would have noticed if a switch had been made during a
“similar” experiment. Of the participants that did not notice any manipulations during the
experiment, 85% believed that they would have detected such a switch if it had been
performed. When the actual purpose of the experiment was finally revealed, the
participants showed considerable surprise, and sometimes even questioned our claim that
we had switched the pictures. This type of strong reaction is very difficult to reconcile
with participants actually knowing about the manipulations but not telling us about it. By
answering yes to the meta-question about whether they think they would have noticed a
manipulation, the participants in a sense also set the norm for what should be expected of
them. To answer yes to the first question and then deliberately “lie” when asked whether
they detected any manipulations seems a very strange thing to do. This also has an
JOHANSSON, HALL, & SIKSTRÖM152
interesting parallel in the change blindness literature, in that people tend to overestimate
their ability to detect visual changes (the so called change blindness blindness effect, see
Levin, Momen, Drivdahl, & Simons, 2000).
In comparison with change blindness experiments, it is also important to point out
that even though the current experiment involves an element of cognitive load and a
limited exposure time of the stimuli, this is not a prerequisite to obtain the choice
blindness effect. In Johansson et al. (2005) the choice alternatives were presented for 2 s,
4 s, or for as long as the participants wanted, and even in the free viewing time condition a
large majority of the manipulations remained undetected. The memory test in the current
experiment also shows that for the non-manipulated trials, the initial exposure (pre- and
post-choice presentation) is enough time to form a relatively stable memory of which
alternative was preferred when the choice was made.
But to what extent is choice blindness something new and different in comparison
with change blindness? Given that neither the current nor our previous studies were
designed to address the neuro-cognitive underpinnings of choice and change blindness, it
would be premature to offer any speculations whether the actual mechanisms are identical
or not. However, as we see it, our methodology is perfectly positioned to bridge the
disconnected research areas of choice/intentionality and change blindness, and to create
some productive friction between the two. This can be seen clearly by a brief exposition
of what intentional choice is supposed to entail. Sirigu Daprati, Ciancia, Giraux,
Nighoghossian, Posada, & Haggard (2004) state: “voluntary action implies a subjective
experience of the decision and the intention to act/.../For willed action to be a functional
behavior, the brain must have a mechanism for matching the consequences of the motor
act against the prior intention” (p. 80, our emphasis, see also Ullsperger & Cramon, 2004;
Ridderinkhof, van den Wildenberg, Segalowitz, & Carter, 2004; Haggard, 2005). But if
this is the case, how can it be that the participants in our studies so often fail to detect the
glaring discrepancy between the prior intention and the outcome of their choice? Pairing
this question with the most common explanations for change blindness offered in the
literature does not seem to produce any satisfactory answers. In fact, in our view, given
the almost complete lack of reference to mechanisms of decision making and
intentionality in the change blindness literature, choice blindness would be an even more
remarkable phenomenon if it turned out to be qualitatively identical to change blindness.
For example, the prevalence of choice blindness in our experiment might be due to a
failure to sufficiently encode the choice alternatives during the deliberation phase
(O’Regan & Noe, 2002). But from the perspective of a decision researcher it would
amount to a strangely maladaptive decision process not to encode the features that are
supposed to be the very basis of the choice (such as the gross identity of the two
alternatives). Another option is that the intentions simply are forgotten during the brief
occlusion time. But intentions are not supposed to be instantly forgotten. As Sirigu et al.
(2004) contend, they are supposed to be the guiding structures behind our actions (and
phenomenologically speaking, this is what many people claim them to be), which makes
this option equally unattractive to decision theorists. Similar things can be said for the
other common explanations for change blindness: that initial representations might be
FROM CHANGE BLINDNESS TO CHOICE BLINDNESS 153
disrupted or overwritten by the feedback (Beck & Levin, 2003), that change blindness
results from a failure to compare pre- and post-change information (Mitroff, Simons, &
Levin, 2004; Hollingworth, 2003), or that explicit change detection is impossible because
the representations are in a format inaccessible to consciousness (Simons & Silverman,
2004). They are all viable candidates to explain choice blindness, but also more or less
incompatible with popular theories of choice and intentionality. If our task can be seen as
a good example of willed action, involving perfectly standard intentions and choices (and
currently we can see no reason why this should not be the case), but the outcome of the
experiment could be fully explained by the conceptual apparatus of change blindness
research, then something would seem to be seriously amiss in current theories of decision
making and cognitive control.
There are a large number of possible future experiments to make to further examine
the relationship between change and choice blindness. For example, the choice task as
can be altered in several different ways to examine what effect intention has on retention
and change detection. One dimension is the concreteness of the choice task, e.g. the
participants can be asked to select the larger of the two patterns, or decide which of two
faces is the roundest, or which person is older, or has highest education. Another is the
consequences of the choice, e.g. which of the two persons the participants will pick as
partner in a subsequent problem solving task, or which to go with on an afternoon date at
the campus café. From a common sense perspective, it seems like these factors would
influence both detection rate and memory of initial choice, but it remains to be empirically
decided.
Another aspect with parallels in change blindness research is to search for implicit
measures of detection, like GSR or ERP (Eimer & Mazza, 2005), gaze duration
(Hollingworth, Williams, & Henderson, 2001), or a “sense” of something being wrong
occurring independent of explicit change detection (Rensink, 2004).
In comparison with standard change blindness studies, another novel use of the
choice blindness methodology is to examine what happens after the choice (what Dennett,
1991 calls The Hard Question: And then what happens?). In the present study, the
memory test used after the completion of the choice experiment showed that for the
manipulated trials, the participants were much more likely to remember the originally
non-preferred face as being their actual choice. In addition, in Experiment 3, the
participants tended to rate the originally non-chosen picture as being the more attractive.
The question is what becomes of the participants preferences and attitudes; what would
for instance happen if they had to do the same choice again, would they pick the
alternative they initially thought was better or the mismatched option they ended up with?
We have recently begun to explore this question. In Hall, Johansson, Tärning & Sikström
(in preparation), the participants had to choose between two faces, pick the one they
preferred, and give either a short or a long verbal report explaining their choice. But in
addition to this, their later preferences were also probed in several different ways. All
participants were presented with the pairs a second time and had to choose the picture
preferred once again. In one condition, the participants also had to rate on a numerical
scale how attractive they thought both pictures were directly after having given their
JOHANSSON, HALL, & SIKSTRÖM154
verbal reports. The results showed that the participants were clearly influenced by the
manipulations made, as they were significantly more likely to pick the originally non-
preferred face the second time they had to evaluate a pair. But perhaps even more
interestingly, this tendency was correlated with the participants “involvement” in the
choice, i.e. if they had given short or long reports, and if they had numerically rated the
pictures after the first choice.
We think this is a very interesting avenue of exploration. What will happen with
these “induced” preferences over time? Will they transfer to more general attributes (like
preferring brunettes to blondes)? Will they be modulated by other choices? In a sense,
choice blindness can be used as an instrument to measure how much we influence
ourselves by the choices we make.
In summary, we have made an argument that the choice blindness paradigm can be
seen as a general research tool to study decision making, intentional action, introspection,
and the dynamics of preference formation and change. As such, it extends prior research
on change blindness, which has primarily been focused on the properties of vision and
attention. The question of the exact relation between the mechanisms of choice and
changes blindness remains to be determined. In our view this is a topic that could serve as
a critical catalyst for further connecting the research fields of decision making and vision
science.
REFERENCES
Beck, M. R., & Levin, D. T. 2003. The role of representational volatility in recognizing pre- and postchange
objects. Perception and Psychophysics, 65, 458–468.
Blackmore, S. 2002. There is no stream of consciousness. Journal of Consciousness Studies, 9, 17–28.
Brooks, R. A. 1991. Intelligence without representation. Artificial Intelligence, 47, 139–159.
Dennett, D. C. 1991. Consciousness explained. Boston: Little, Brown & Company.
Eimer, M., & Mazza, V. 2005. Electrophysiological correlates of change detection, Psychophysiology, 42,
328–342.
Grimes, J. 1996. On the failure to detect changes in scenes across saccades. In K. Akins (Ed.), Vancouver
studies in cognitive science: Perception (Vol. 2, pp. 89–110). New York: Oxford University Press.
Haggard, P. 2005. Conscious intention and motor control. Trends in Cognitive Sciences, 9, 290–295.
Hall, L., Johansson, P., Tärning, B., & Sikström, S. In preparation. Choice Blindness and Preference Change.
Hall, L., Johansson, P., Tärning, B., Sikström, S., & Deutgen, T. Submitted. Magic at the Marketplace:
Choice blindness for the taste of jam and the smell of tea.
Hollingworth, A. 2003. Failures of retrieval and comparison constrain change detection in natural scenes.
Journal of Experimental Psychology: Human Perception and Performance, 29, 388–403.
Hollingworth, A., Williams, C. C., & Henderson, J. M. 2001. To see and remember: Visually specific
information is retained in memory from previously attended objects in natural scenes. Psychonomic
Bulletin & Review, 8, 761–768.
Johansson, P., Hall, L., Sikström, S., Tärning, B., & Lind, A. 2006. How something can be said about telling
more than we can know: On choice blindness and introspection. Consciousness and Cognition, 15,
673–692.
Johansson, P., Hall, L., Sikström, S., & Olsson, A. 2005. Failure to detect mismatches between intention and
outcome in a simple decision task. Science, 310, 116–119.
Levin, D. T., Momen, N., Drivdahl, S. B., & Simons, D. J. 2000. Change blindness blindness: The
metacognitive error of overestimating change-detection ability. Visual Cognition, 7, 397–412.
FROM CHANGE BLINDNESS TO CHOICE BLINDNESS 155
Levin, D. T., Simons, D. J., Angelone, B. L., & Chabris, C. F. 2002. Memory for centrally attended changing
objects in an incidental real-world change detection paradigm. British Journal of Psychology, 93, 289–
302.
Mitroff, S. R., Simons, D. J., & Franconeri, S. L. 2002. The siren song of implicit change detection. Journal
of Experimental Psychology: Human Perception and Performance, 28, 798–815.
Noë, A. (Ed.) 2002. Is the visual world a grand illusion? Thorverton: Imprint Academic.
O’Regan, J. K., & Noë, A. 2002. A sensorimotor account of vision and visual consciousness. Behavioral
and Brain Sciences, 24, 939–1011.
O’Regan, J. K., Rensink, R. A., & Clark, J. J. 1999. Change-blindness as a result of “mudsplashes”. Nature,
398, 34.
Rensink, R. A. 2000a. Seeing, sensing, and scrutinizing. Vision Research, 40, 1469–1487.
Rensink, R. A. 2000b. Visual search for change: A probe into the nature of attentional processing. Visual
Cognition, 7, 345–376.
Rensink, R. A. 2002. Change detection. Annual Review of Psychology, 53, 245–277.
Rensink, R. A. 2004. Visual sensing without seeing. Psychological Science, 15, 27–32.
Ridderinkhof, K. R., van den Wildenberg, W. P. M., Segalowitz, S. J., & Carter, C. S. 2004. Neurocognitive
mechanisms of cognitive control: The role of prefrontal cortex in action selection, response inhibition,
performance monitoring, and reward-based learning. Brain and Cognition, 56, 129–140.
Simons, D. J. 2000. Current approaches to change blindness. Visual Cognition, 7, 1–15.
Simons, D. J., & Levin, D. T. 1998. Failure to detect changes to people in a real-world interaction.
Psychonomic Bulletin & Review, 5, 644–649.
Simons, D. J., & Rensink, R. A. 2005. Change blindness: Past, present, and future. Trends in Cognitive
Sciences, 9, 16–20.
Simons, D. J., & Silverman, M. 2004. Neural and behavioral measures of change detection. In L. M.
Chalupa & J. S. Werner (Eds.), The Visual Neurosciences (pp. 1524–1537). Cambridge, MA: MIT
Press.
Sirigu, A., Daprati, E., Ciancia, S., Giraux, P., Nighoghossian, N., Posada, A., & Haggard, P. 2004. Altered
awareness of voluntary action after damage to the parietal cortex. Nature Neuroscience, 7, 80–84.
Smilek, D., Eastwood, J. D., & Merikle, P. M. 2000. Does unattended information facilitate change
detection? Journal of Experimental Psychology: Human Perception and Performance, 26, 480–487.
Todorov, A., Mandisodza, A. N., Goren, A., & Hall, C. C. 2005. Inferences of competence from faces predict
election outcomes. Science, 308, 1623–1626.
Triesch, J., Ballard, D., Hayhoe, M., & Sullivan, B. 2003. What you see is what you need. Journal of Vision,
3, 86–94.
Tse, P. U., Sheinberg, D. L., & Logothetis, N. K. 2003. Attentional enhancement opposite a peripheral flash
revealed by change blindness. Psychological Science, 14, 91–99.
Ullsperger, M., & von Cramon, D. Y. 2004. Neuroimaging of performance monitoring: Error detection and
beyond. Cortex, 40, 593–604.
Willis, J., & Todorov, A. 2006. First impressions: Making up your mind after a 100-ms exposure to a face.
Psychological Science, 17, 592–598.
(Manuscript received December 31, 2007; Revision accepted April 28, 2008)