Incubation

The Denitive Guide

to Science and Practice

The heart of the MacGyver Secret is incubation, plain and simple. It's the time that brings together
all of the other steps and taps into your inner MacGyver to solve any problem you ask. But, it's also
the most mysterious part of the system and the part most people understand the least.

So, to help you navigate, based on all the comments I received, I decided to put together an in depth
look at the science of incubation. This information is referenced in the book but has never before
been compiled in one place.

I want to warn you: a lot of this is dense science so be prepared for challenging reading. BUT, this is
top shelf material from the best minds in the eld.

Creativity - The Unconscious Foundations of the Incubation Period Page 1-10

Simone M. Ritter and Ap Dijksterhuis, Radboud University Nijmegen

Does Incubation Enhance Problem Solving? A Meta-Analytic Review Page 11-37

Ut Na Sio and Thomas C. Ormerod, Lancaster University

Dont Wait to Incubate: Immediate Versus Delayed Incubation in Divergent Page 38-47
Thinking Kenneth J. Gilhooly & George J. Georgiou & Jane Garrison & Jon D. Reston & Miroslav Sirota

Driven to Distraction: The Impact of Disaster Type on Unconscious Decision Making Page 48-63
Kibby McMahon, Betsy Sparrow, Ljubica Chatman, and Travis Riddle, Columbia University

Inspired by Distraction: Mind Wandering Facilitates Creative Incubation Page 64-70

Benjamin Baird, Jonathan Smallwood, Michael D. Mrazek, Julia W. Y. Kam, Michael S. Franklin and Jonathan W. Schooler

Neural Correlates of Subliminal Language Processing Page 71-80

Vadim Axelrod, Moshe Bar, Geraint Rees and Galit Yovel

On Making the Right Choice: The Deliberation-Without-Attention Eect Page 81-84

Ap Dijksterhuis

Probabilistic Word Pre-Activation During Language Comprehension Inferred Page 85-89

From Electrical Brain Activity
Katherine A DeLong, Thomas P Urbach and Marta Kutas

Some Experiments on the Recognition of Speech, with One and with Two Ears Page 90-94
E. Colin Cherry

The Automaticity of Everyday Life John A. Bargh, New York University Page 95-126

The Cognitive Unconscious John F. Kohlstrom

Think Dierent: The Merits of Unconscious Thought in Preference Development Page 127-135
and Decision Making Ap Dijksterhuis, University of Amsterdam

Thinking Fast and Slow Page 136-152

Daniel Kahneman, (New York: Farrar, Straus and Giroux, 2011), 418 pp. Reviewed by Frank J. Babetski
 REVIEW ARTICLE
published: 11 April 2014
HUMAN NEUROSCIENCE doi: 10.3389/fnhum.2014.00215

Creativitythe unconscious foundations of the incubation

period
Simone M. Ritter* and Ap Dijksterhuis
Behavioural Science Institute, Radboud University Nijmegen, Nijmegen, Netherlands

Edited by:
Matthijs Baas, University of
Amsterdam, Netherlands
Reviewed by:
Kirsten G. Volz, Werner Reichardt
Centre for Integrative Neuroscience,
Germany
Ut Na Sio, Carnegie Mellon University,
USA
Stephanie Lichtenfeld, University of
Munich, Germany
*Correspondence:
Simone M. Ritter, Behavioural Science
Institute, Radboud University
Nijmegen, Montessorilaan 3, P.O. Box
9104, 6500 HE Nijmegen, Netherlands
e-mail: S.Ritter@psych.ru.nl
Creativity is one of the most important assets we have to navigate through the fast
changing world of the 21st century. Anecdotal accounts of creative individuals suggest
that oftentimes, creative discoveries result from a process whereby initial conscious
thought is followed by a period during which one refrains from task-related conscious
thought. For example, one may spend an embarrassing amount of time thinking about
a problem when the solution suddenly pops into consciousness while taking a shower.
Not only creative individuals but also traditional theories of creativity have put a lot of
emphasis on this incubation stage in creative thinking. The aim of the present article
is twofold. First, an overview of the domain of incubation and creativity is provided by
reviewing and discussing studies on incubation, mind-wandering, and sleep. Second, the
causes of incubation effects are discussed. Previously, little attention has been paid to
the causes of incubation effects and most findings do not really speak to whether the
effects should be explained by unconscious processes or merely by consequences of a
period of distraction. In the latter case, there is no need to assume active unconscious
processes. The findings discussed in the current article support the idea that it is not
merely the absence of conscious thought that drives incubation effects, but that during an
incubation period unconscious processes contribute to creative thinking. Finally, practical
implications and directions for future research will be discussed.

Keywords: creativity, problem solving, incubation, mind wandering, sleep, unconscious processes

Important achievements in the arts and sciences depend on cre-
ativity (Feist and Gorman, 1998; Kaufman, 2002), and creativity is
associated with the development of new social institutions (Bass,
1990; Mumford, 2002) and economic growth (Amabile, 1997;
Simonton, 1999). It is generally accepted that a creative idea or
a creative solution to a problem has to be novel (i.e., original)
and useful (Amabile, 1983; Runco and Pritzker, 1999). Creativity
is not limited to the realms of greatness, but can also be found
in daily life, for example, when one has to accomplish a task in a
new way (Cropley, 1990) or when one has to adapt to changes
(Runco, 2004). Todays world of continuous change thrives on
creative individuals. Creativity has been related to cognitive abili-
ties, expertise, and practice (Patrick, 1986; Amabile, 1996; Runco,
2004; Ericsson, 2006; Sawyer, 2012), and one may expect that cre-
ativity mainly thrives on extensive conscious thought. However,
creative individuals, in describing their work habits or the process
of creative problem solving, have suggested that oftentimes, cre-
ative ideas result from a period of incubationa process whereby
initial conscious thought is followed by a period during which
one refrains from task-related conscious thought (for anecdotal
accounts, see Ghiselin, 1952). The most frequently cited anecdote
is probably the one from the mathematician Poincar:
[. . .] I left Caen, where I was living, to go on a geologic excursion
under the auspices of the School of Mines. The incidents of the
travel made me forget my mathematical work. Having reached
Coutances, we entered an omnibus to go some place or other.
At the moment when I put my foot on the step, the idea came
to me, without anything in my former thoughts seeming to have
paved the way for it, that the transformations I had used to define
the Fuchsian functions were identical with those of non-Euclidian
geometry. I did not verify the idea; I should not have had the
time, as, upon taking my seat in the omnibus, I went on with a
conversation already commenced, but I felt a perfect certainty. On
my return to Caen, for conscience sake, I verified the result at my
leisure. (Poincar quoted in Hadamard, 1945, p. 13).
In addition, several famous anecdotes suggest that sleep facil-
itates creativity, ranging from musical compositions to scientific
insights (Mazzarello, 2000). In speaking of the attainment of
solutions, Beatle Paul McCartney announced that he came up
with the melody for Yesterday in a dream, and the Nobel Prize
winner Loewi woke up with the idea for how to experimentally
prove his theory of chemical neurotransmission. The idea that
a period of incubation might facilitate creativity has not only
been suggested by creative minds, but has also been stressed
in creativity models. Wallas (1926) proposed that the creative
process entails four stages: Preparation (acquisition of knowledge
to some task), Incubation (process that occurs when conscious
attention is diverted away from the task), Illumination (creative
idea flashes into sight), and Verification (creative idea is subjected
to evaluation). Certainly a creative idea may be found before a
decrease in conscious effort, that is, before the Incubation stage.

Frontiers in Human Neuroscience www.frontiersin.org April 2014 | Volume 8 | Article 215 | 1

1
Ritter and Dijksterhuis
Sometimes, however, a period of incubation seems to precede
creative breakthroughs as illustrated above for several scientific
discoveries and artistic compositions. Sparked by the anecdotal
accounts on incubation and creativity, various attempts have
been made to investigate incubation effects. As demonstrated
by a Google Scholar search (Sio and Ormerod, 2009), with the
search restricted to the years 19972007 and the subject areas to
social sciences, arts, and humanities, the term incubation along
with either creativity, insight, or problem, yielded more than 5000
articles. Empirical research has shown that a period of incubation
indeed helps creativity (Dodds et al., 2003; Sio and Ormerod,
2009).
However, it is not yet clear why incubation is helpful. The
moderators discovered thus far do not really speak to whether the
effects should be explained by unconscious processes or merely by
other consequences of a period of distraction (e.g., relaxation, for-
getting of fixating elements, mental set-shifting) without the need
to assume active unconscious processes (see also Orlet, 2008). The
aim of the current article is to provide an overview of the domain
of incubation and creativity, and to review and discuss findings
that speak to whether during an incubation period unconscious
processes contribute to creative thinking, or whether it is merely
the absence of conscious thought that drives creativity. Finally,
practical implications and directions for future research will be
discussed.
INCUBATION
Many anecdotal accounts and traditional theories of creativity
have put emphasis on incubation. The basic phenomenon is a
familiar one: we are working on a problem, we cant solve the
task, we leave it aside for some period of timethe incubation
periodand when we return attention to the task we have some
new insight that helps us to solve the problem. In general, there are
two frequently used methods to conduct incubation experiments.
In the interpolated activity method, participants work on a task
or problem for a period of time, are then given an incubation
period, and finally return to the task. Participants performance is
compared with that of a control group of people who worked on
the same problem continuously. In the multiple trial or multiple
item method, multiple tasks or problems are presented and,
afterwards, items that have not been solved are re-administered.
For example, participants work for one minute apiece on several
problems. Then, the unsolved problems are re-administered for
one minute apiece. It is presumed that the time between the first
and second encounter with the tasks or problems allows incuba-
tion to occur. Some researchers using this approach also insert
an incubation period between the first and second encounter
with the problems. For example, after the first encounter with
the problems participants perform a distractor task and after this
incubation interval return to work for a certain time on problems
that they did not solve. Participants performance is compared
with that of a control group, and if a within-participant design
is used, the increase in number of problems is used.
Whereas some studies reported strong incubation effects (e.g.,
Kaplan, 1989; Smith and Blankenship, 1989; Smith and Dodds,
1999; Dodds et al., 2003), others have failed to find any effects at
all (e.g., Olton and Johnson, 1976). To resolve the uncertainties
Frontiers in Human Neuroscience www.frontiersin.org
2
Creativityunconscious foundations of incubation
surrounding incubation effects, Dodds et al. (2003) conducted
a review of experimental literature on incubation in problem
solving and creativity, and revealed that 29 out of 39 experi-
ments have found a significant effect of incubation. The authors
suggested that incubation length and preparatory activities can
increase incubation effects. Moreover, the authors demonstrated
that presenting a clue during the incubation period can either
have strong positive (if the clue is useful) or negative effects (if
the clue is misleading). For example, ocean or floor could
be a misleading clue when trying to find a fourth word that
functions as an associative link between the three items ship,
outer, crawl, whereas space could be a useful clue. Sio and
Ormerod (2009) conducted a statistical meta-analytic review
of empirical studies of incubation. In their meta-analysis 117
independent studies were included, and the contributions of
moderators such as problem type, presence of cues, and lengths
of preparation and incubation periods were investigated. Over-
all, a positive incubation effect was found. In a recent study,
Gilhooly et al. (2013) investigated interactions between the type
of creativity task (verbal or spatial) and the type of incubation
activity (verbal or spatial) on creative performance. Experimental
groups, after 5 min of conscious work on a verbal creativity
task (Alternative Uses Task) or a spatial creativity task (Mental
Synthesis), had a 5-min incubation period that involved either
spatial (Mental Rotation) or verbal (Anagrams) tasks. Following
incubation, participants resumed their main task for a further
5 min. Control groups undertook Alternative Uses or Mental
Synthesis for 10 min without any incubation periods. Significant
incubation effects were found overall and there were interac-
tions in that spatial incubation benefited verbal fluency and
verbal-rated creativity, and verbal incubation benefited spatial-
task fluency and spatial-rated creativity but not vice versa. These
findings suggest that an interpolated incubation activity of a
dissimilar nature to the target task leads to stronger effects of
incubation as compared to an interpolated activity similar to the
target task.
Not only the task that is performed during an incubation
period, but also the time interval of an incubation period can
vary. It can vary from a few moments or a night of sleep
through days or weeks away from the problem. An example of
a relatively short incubation period is mind-wanderinga state
of mind that occurs spontaneously, and largely autonomously,
whenever an awake individual is not engaged in a cognitively
demanding task. Research on mind-wandering has a long his-
tory, and was recently popularized by Smallwood et al. (2003)
who used thought sampling and questionnaires to investigate
mind-wandering. In past and recent literature, alternative names
to the term mind-wandering (Smallwood and Schooler, 2006;
Mason et al., 2007) have been used, such as day dreaming
(Giambra, 1979), spontaneous thought (Christoff et al., 2011),
task-unrelated thought (Giambra and Grodsky, 1989; Small-
wood et al., 2003), and stimulus independent thought (Teas-
dale et al., 1995). In a recent study, Baird et al. (2012) exam-
ined whether creative performance was facilitated differentially
by engaging in mind-wandering (i.e., a 0-back task, an unde-
manding task without memory load that has been shown to
elicit mind-wandering, Smallwood et al., 2009), a demanding
April 2014 | Volume 8 | Article 215 | 2
Ritter and Dijksterhuis
task (i.e., a 1-back working memory task), a rest period, or no
break between creativity problems. To measure creative perfor-
mance, the Unusual Uses Task (a task that requires participants
to generate as many unusual uses as possible for a common
object) was used. All participants performed two Unusual Uses
Task problems (2 min per problem) to measure baseline creative
performance. Subsequently, participants were assigned to one
of the four between-subjects conditions. After the incubation
interval (or following the baseline measure, in the case of the
no-break condition), participants worked on the Unusual Uses
Task again. Four problems (2 min per problem) were presented
in a random order: two problems that were identical to the
problems presented at baseline and two new problems. Engag-
ing in an undemanding task during an incubation period led
to significant increases in creative solutions to the target prob-
lems as compared to the demanding task, rest, and no break
conditions. This improvement was observed only for repeated-
exposure problems, which demonstrates that it resulted from
an incubation process rather than a general increase in creative
problem solving. The unrelated thoughts that occur during mind
wandering uniquely seem to facilitate incubation. According to
Baird et al., one possible explanation may be that mind wan-
dering enhances creativity by increasing unconscious associative
processing, as predicted by the spreading-activation account of
incubation (e.g., Yaniv and Meyer, 1987; Dijksterhuis and Meurs,
2006).
In recent years, functional Magnetic Resonance Imaging
(fMRI) research has been used to focus on understanding how
the brain generates the spontaneous and relatively unconstrained
thoughts that are experienced when the mind wanders. One
candidate neural mechanism for mind-wandering is a network
of regions in the frontal and parietal cortex known as the default
mode network (Mason et al., 2007; Christoff et al., 2009). The
default mode network, also called the default network, default
state network, or task-negative network, is defined as a set of inter-
connected brain regions including the medial prefrontal cortex
(MPFC), posterior cingulate cortex (PCC), and lateral and medial
temporal lobes (Spreng et al., 2010). It is a brain system that is
especially active when an individual is not focused on the outside
world (Buckner et al., 2008) and when cognitive control is low
(Andreasen, 1995). Moreover, it has been related to complex, eval-
uative and unconscious forms of information processing (Vincent
et al., 2007; Yang et al., 2010), and it contrasts with the cognitive
control network (Fox et al., 2005)a set of brain regions includ-
ing the anterior cingulate cortex (ACC), dorsolateral prefrontal
cortex (DLPFC), inferior frontal junction (IFJ), anterior insular
cortex (AIC), dorsal pre-motor cortex (dPMC), and posterior
parietal cortex (PPC; Cole and Schneider, 2007). Indeed, when
one network is activated, the other is deactivated (Fox et al.,
2005). In addition to these findings, structural MRI research has
provided a first indication that the default mode network may
be involved in creativity. Jung et al. (2010) have linked cortical
thickness measures to psychometric measures of creativity and
found a negative correlation between creative performance and
activity in the lingual gyrus and a positive correlation between
creative performance and grey matter volume in the right PCC,
a brain area that is part of the default mode network.
Frontiers in Human Neuroscience www.frontiersin.org
3
Creativityunconscious foundations of incubation
In a recent structural MRI study, Khn et al. (2013) provided
further support for the involvement of the default mode network
in creativity. Participants performed a well-established creativity
task by which a participants cognitive flexibility and the average
uniqueness and average creativity of a participants ideas were
assessed. For all psychometric measures of creativity a positive
correlation was observed between inter-individual differences in
creative performance and inter-individual differences in volume
of the default mode network. Based on these findings, it can be
assumed that greater volume in the default mode network (i.e., in
the counterpart of the cognitive control network) provides more
neural resources for generating creative ideas. These findings
suggest that less controlled processes such as mind-wandering are
important in creativity. One relatively controversial finding is that
periods of mind-wandering are associated with increased activa-
tion in both the default and executive system, a result that implies
that mind-wandering may often be goal oriented (Smallwood
and Schooler, 2006; Smallwood et al., 2009). Apart from studies
about the default mode network, there are several important other
studies on neuroimaging and creativity. For example, the research
from Reverberi et al. (2005) demonstrates that the lateral frontal
cortex impairs problem solving, and the research by Kounios
and Jung-Beeman (2009) on the cognitive neuroscience of insight
suggests that insight is the culmination of a series of brain states
and processes operating at different time scales. Recently, Dietrich
and Kanso (2010) reviewed 72 neuroimaging studies on creativity
and insight and concluded that the neuroscientific literature on
creativity, thus far, is self-contradicting and that creative thinking
does not appear to critically depend on any single mental pro-
cess or brain region. The default mode network can, therefore,
be considered one, but not the single neural underpinning of
creativity.
Whereas mind-wandering can be considered a relatively short
incubation period, sleep can be considered an incubation period
that covers a longer period of time. Sleep is divided into two broad
types, rapid eye movement (REM) sleep and non-rapid eye move-
ment (NREM) sleep. Each type has a distinct set of associated
physiological and neurological features. REM sleep is a stage of
sleep characterized by the rapid and random movement of the
eyes, and typically occupies 2025% of total sleep. During REM,
the activity of the brains neurons is quite similar to that during
waking hours and subjects vividly recalled dreams mostly occur
during REM sleep. Unlike REM sleep, during NREM sleep there
is usually little or no eye movement and dreaming is rare. The
differences in the REM and NREM activity reported is believed
to arise from differences in the memory stages that happen
during the two methods of sleep (Manni, 2005). For example,
Stickgold et al. (1999) have shown that cognition during REM
sleep is qualitatively different from that of waking and NREM
sleep, and may reflect a shift in associative memory systems. They
suggest that this shift in cognitive processing is responsible, in
large part, for the bizarre nature of dreams and may serve to
enhance the strength of associations between weakly associated
memories, an important skill underlying creative thinking. The
mental activity that takes place during NREM sleep is believed
to be thought-like, whereas REM sleep includes hallucinatory
and bizarre content (Manni, 2005). Thus far, sleep research has
April 2014 | Volume 8 | Article 215 | 3
Ritter and Dijksterhuis
mainly been focused on memory performance. A prominent
finding is that sleep, and certain stages of sleep in particular, are
important in memory processing, resulting in delayed learning
without the need for further practice or task engagement (Stick-
gold et al., 2001). These findings of sleep-dependent learning
are now strongly supported by cellular and molecular evidence
of sleep-dependent plasticity across a broad range of phylogeny
(Bennington and Frank, 2003). Yet memory consolidation is only
one of many cognitive virtues possessed by the human brain,
another is creativity.
The link between creativity and sleep, especially dreaming,
has long been a topic of intense speculation (Stickgold and
Walker, 2004). In recent years, the facilitatory effect of sleep
on creativity has also received empirical support. Research from
Barrett (1993) has shown that college students incubated answers
to real-life homework and other objective problems on which
they were working, finding that in one weeks time, half of the
students had dreamed about their topic and 25% had a dream
that provided an answer. Barrett (2001) also interviewed modern
artists and scientists (including Nobel Prizes winners) about
their use of their dreams and concluded that while anything
math, musical composition, business dilemmasmay get solved
during dreaming, the two areas dreams are especially likely to
help are anything where vivid visualization contributes to the
solution and any problem where the solution lies in thinking
outside the boxi.e., where the person is stuck because the
conventional wisdom on how to approach the problem is wrong.
Moreover, in an experimental study Wagner et al. (2004) have
shown that sleep inspires creative insight. Subjects completed a
number reduction task, and each numerical sequence could be
completed in a slow, stepwise way, but the trials could also be
completed according to a hidden, more abstract rule that would
speed up participants responses. The initial training was followed
by 8 h of nighttime sleep, nighttime wakefulness, or daytime
wakefulness. Of the people who slept before they resumed, almost
60% discovered the rule, as opposed to 23% of the people
in the two groups that did not sleep. Thus, participants who
got several hours of sleep were more than two times as likely
during retesting to gain insight into a hidden rule built into
the task.
In addition, sleep has been shown to enhance important
aspects of creativity, including cognitive flexibility and the ability
to find remote associations. In a study on cognitive flexibility
across the sleep-wake cycle, Walker et al. (2002) found that when
woken from REM sleep, participants had a 32% advantage in the
number of anagrams solved compared with NREM awakenings,
which were equal to that of wake time performance. These find-
ings suggest that REM sleep may offer a different mode of problem
solving compared with wake and NREM. The authors hypothe-
sized that REM sleep is highly conducive to fluid reasoning and
flexible thought due to the lack of aminergic dominance in REM
sleep. In a study on the ability to find remote associations, Sio
et al. (2012) participants were presented with a set of Remote
Associates Test (RAT) items. Each RAT item contains a triplet of
words presented horizontally along with a blank space. For each
item, the participant has to find a fourth word that functions
as an associative link between these three words (e.g., cookies,
Frontiers in Human Neuroscience www.frontiersin.org
4
Creativityunconscious foundations of incubation
sixteen, heart: . . . . . . . . .; the answer to this item is sweet: cookies
are sweet, sweet sixteen, sweetheart). Reaching a solution requires
creative thought as the first, most probable associate to each of
the items is often not correct, so the participant must think of
more remote associations (i.e., distantly related information) to
connect the three words. In the current study the RAT items
varied in difficulty as a function of the strength of the stimuli
answer associations. After a period of sleep, wake, or no delay,
participants reattempted earlier unsolved problems. The sleep
group solved a greater number of difficult RAT items than did
the other groups, but no difference was found for easier RAT
items. These findings suggest that sleep facilitates creative think-
ing for harder problems. While evidence for the role of sleep in
creative problem-solving has been looked at by prior research,
underlying mechanisms such as different stages of sleep had
not been explored. Cai et al. (2009) used the RAT, and tested
participants in the morning, and again in the afternoon, after
either a nap with REM sleep, one without REM or a quiet rest
period. Participants grouped by REM sleep, non-REM sleep and
quiet rest were indistinguishable on measures of memory. Most
importantly, although the quiet rest and NREM sleep groups
received the same prior exposure to the task, they displayed no
improvement on the RAT test, whereas the REM sleep group
improved by almost 40% over their earlier performances. The
authors hypothesized that the formation of associative networks
from previously unassociated information in the brain, leading to
creative problem-solving, is facilitated by changes to neurotrans-
mitter systems during REM sleep. Thus, REM sleep is assumed to
enhance the integration of unassociated information for creative
problem solving.
To recap, various attempts have been made to investigate
incubation effects in creativity and creative problem solving. The
conclusion of the literature is that overall, a positive incubation
effect can be observed. Especially the work by Dodds et al. (2003)
and Sio and Ormerod (2009), who conducted reviews of empirical
studies of incubation, justify the conclusion that incubation can
enhance creative performance. This is also supported by research
on mind-wandering and sleep, which can be seen as short and
relatively long periods of incubation. However, the process(es)
underlying incubation effects remain unclear. In the next section,
we aim to shed light on the question whether during an incuba-
tion period unconscious processes contribute to creative thinking,
or whether it is merely the absence of conscious thought that
drives incubation effects.
MECHANISMS UNDERLYING INCUBATION EFFECTS
Whereas the effects of incubation are generally accepted (Sio
and Ormerod, 2009), its causes are controversial. The main
debate between different theories is about whether during an
incubation period unconscious processes contribute to creative
thinking (unconscious work theory), or whether it is merely the
absence of conscious thought that drives creativity (conscious
work theory). Historically, incubation effects refer to the idea
that setting a problem aside for a while helps creative thought
and problem solving as unconscious processes are working on the
problem while the individual is not consciously thinking about
the problem (see Wallas, 1926, as well as, e.g., Hadamard, 1945;
April 2014 | Volume 8 | Article 215 | 4
Ritter and Dijksterhuis
Kris, 1952; Rugg, 1963; Kubie, 1985). That is, the unconscious
actively thinks and contributes to solving a problem (see also
Koestler, 1964; Claxton, 1997). In contrast, conscious work the-
ories have ascribed incubation effects on creative performance
to relaxation (being well-rested, one can do better the next
time one engages in the problem; Helmholtz, 1896; Woodworth
and Schlosberg, 1954) and to the effects of facilitating cues
from the environment (environmental cues trigger retrieval of
previously un-retrieved relevant information; e.g., Yaniv and
Meyer, 1987; Langley and Jones, 1988). Moreover, sometimes
old and inappropriate ideas can cause mental fixation, imped-
ing the generation of new and appropriate ideas (Smith, 2003).
Therefore, in addition to relaxation and facilitating cues, it has
been suggested that incubation effects can lead to forgetting of
fixating elements (Smith and Blankenship, 1989; Segal, 2004)
and to mental set-shifting (wrong cues become less accessible,
leading to a fresh, new and unbiased start; Schooler and Melcher,
1995).
Recently, Gupta et al. (2012) investigated whether high-
frequency candidate answers should be avoided in order to find
creative solutions in for instance a RAT. They tested individual
differences in creativity as measured with a complex problem-
solving task, and developed a computational model of the RAT.
Findings showed that individuals performed poorly on the RAT
when they were biased to consider high-frequency candidate
answers. Storm and Angelo (2010) investigated whether inhibi-
tion may facilitate creative problem solving by providing a mech-
anism by which to bypass fixation. They measured participants
retrieval-induced forgetting and, thereafter, participants had to
solve RAT problems. Half of the participants were exposed to
misleading associates prior to problem solving (fixation condi-
tion) and half were not (baseline condition). Correlating the
retrieval-induced forgetting measure with performance on the
RAT revealed that the propensity to inhibit irrelevant information
comes at a price, as potentially relevant information may be
inhibited. However, inhibition can also provide a means by which
to overcome fixation and, thereby, facilitate creativity. There is
no denying that a period of distraction allows for forgetting of
fixation and/or mental set-shifting, relaxation, and exposure to
environmental cues, and that these effects can contribute to cre-
ative thoughts or problem solving. However, it can be questioned
whether these effects are the only benefit of an incubation period,
or whether during an incubation period unconscious processes
contribute to creative thinking.1
Research from Bowers et al. (1990) suggests that the uncon-
scious is able to close in on the correct answer some time before
the answer is accessible to consciousness. They asked participants
to find a target word while they were given successive hints, such
as an associated word. Individuals felt clueless for some time
and then suddenly came up with the correct answer. However,
analysing the prior guesses revealed that individuals were slowly
getting closer to the right solution before the solution reached
consciousness. Participants successive guesses, thus, converged
towards the correct answer. Moreover, a study conducted by
1 Note that the section below greatly overlaps with Dijksterhuis (submitted for
publication).
Frontiers in Human Neuroscience
Creativityunconscious foundations of incubation
Betsch et al. (2001) demonstrated that people can unconsciously
integrate large amounts of information. Participants watched
TV ads shown on a computer screen and simultaneously the
numerical increases and decreases of hypothetical shares were
shown at the bottom of the screen. Participants could not cor-
rectly answer specific questions about the shares, but they had
developed a liking or disliking towards each of the shares. These
findings suggest that participants processed and integrated the
information while they were attending to the TV ads. Recently,
Ric and Muller (2012) have shown that people can unconsciously
initiate and follow arithmetic rules, such as addition. In several
studies participants were instructed to detect whether a symbol
was a digit, and this symbol was preceded by two digits and a
subliminal instruction (i.e., the add instruction or a control
instruction). Participants were faster at identifying a symbol as
a number when the symbol was equal to the sum of the two digits
and they received the instruction to add the digits. In line with
these findings, Sklar et al. (2012) demonstrated that presenting
participants with additions or subtractions subliminally leads to
higher accessibility of correct answers (i.e., answers could be ver-
balized faster) than incorrect answers. A recent review on uncon-
scious higher-order cognition conducted by Van Gaal et al. (2012)
revealed strong evidence for unconscious response-inhibition,
conflict resolution, as well as for error detection. Importantly, they
also concluded that people can unconsciously integrate multiple
pieces of information across space and time.To resume, evidence
from various research areas demonstrates that processes that we
consider thought processes can ensue unconsciously. This makes
it reasonable to assume that thought processes in the service of
creativity and problem solving can, in principle at least, also take
place unconsciously.
The idea that during an incubation period unconscious pro-
cesses are active was one of the building blocks of Unconscious
Thought Theory (Dijksterhuis and Nordgren, 2006). Uncon-
scious thought, that is, deliberation in the absence of conscious
attention directed at the problem, (Dijksterhuis et al., 2006,
p.1005) has mainly been studied in the context of decision-
making (Strick et al., 2011; see also Dijksterhuis and Nordgren,
2006; Bargh, 2011; Nieuwenstein and van Rijn, 2012). In the
literature on unconscious thought in decision-making, partici-
pants are typically first presented with information pertaining
to a decision. Thereafter, they are distracted for a while, before
they make a decision. For example, Bos et al. (2008) compared
participants decision performance after three conditions, a con-
scious thought condition and two incubation conditions, that
is, an unconscious thought condition and a mere distraction
condition. Whereas participants in the unconscious thought con-
dition were told that they would engage in an unrelated task
before returning to the actual task, participants in the mere
distraction condition were told that they had finished the task
and would move on to unrelated tasks. In the mere distraction
condition participants were, thus, distracted just as in the uncon-
scious thought condition, but did not have a problem-solving
goal. A period of distraction only improved decision-making in
the unconscious thought condition, that is, when participants
expected to make a decision following the distraction period.
Comparing an unconscious thought condition with a mere
www.frontiersin.org April 2014 | Volume 8 | Article 215 | 5
5
Ritter and Dijksterhuis
distraction condition provides evidence for true, active thought
taking place unconsciously. Given the evidence for unconscious
thought processes that we have from multiple research areas (e.g.,
decision making, lie detection), the question raises whether there
is also evidence for unconscious thought effects in the domain of
creativity.
Dijksterhuis and Meurs (2006) investigated the relation
between different thought processes and the generation of cre-
ative ideas. In several experiments participants were asked to
generate a list of items (new names for products, names of
places beginning with a certain letter, things one can do with a
brick), and three conditions were compared. In the immediate
condition (i.e., the baseline condition) participants started right
after receiving the instruction. In the conscious thought condi-
tion, participants were given three minutes to consciously think
about the items before they were given time to list them. In
the unconscious thought condition, people were first given the
instruction, and were then distracted for three minutes before
they were given the opportunity to list the items. Conscious
thought led to more accessible items and to items in line with
a cue, whereas unconscious thought led to more inaccessible
items and to items diverging from the cue. Moreover, unconscious
thought led to more creative and unusual items than conscious
thought. In all experiments, unconscious thinkers also differed
significantly from participants who were not given time to think
at all. These findings suggest that whereas conscious thought may
be focused and convergent, unconscious thought may be more
associative and divergent. Ritter et al. (2012b) investigated the
role of unconscious thought for both idea generation and idea
selection. Participants generated creative ideas immediately, after
conscious thought, or after a period of unconscious thought.
After having listed their ideas, participants selected their most
creative idea. Performance in idea generation was similar between
conscious and unconscious thought; however, individuals who
had unconsciously thought about ideas were better in selecting
their most creative idea. These findings are in support of the
idea that unconscious processes actively contribute to creativity,
as it is unlikely that these findings are the consequence of set-
shifting or relaxation. During task instruction no examples were
provided and no hints or cues were given, meaning that no
fixating elements or specific mental sets were induced that could
have become less accessible, changed, or forgotten altogether
during a period of distraction. Recovering from fatigue is also
unlikely to account for the current findings, as incubation effects
also occurred in the study when a cognitively demanding task (n-
back task) was used as distracter task (Dijksterhuis and Meurs,
2006).
Zhong et al. (2008) investigated the effect of unconscious
thought on the ability to find remote associations, as measured by
the RAT. Participants were presented difficult RAT triads (selected
from Bowden and Jung-Beeman, 2003). Afterwards, participants
in the conscious thought condition were told that they had 5 min
to think about these triads, and during this time, they were shown
the screen containing all triads, but were not allowed to write
down notes or answers. Participants in the unconscious-thought
condition were told that they would engage in an unrelated task
before returning to the word task. Participants in the distraction
Frontiers in Human Neuroscience www.frontiersin.org
6
Creativityunconscious foundations of incubation
condition were told that they had finished the task and would
move on to unrelated tasks. To prevent conscious thought about
the RAT items in the unconscious thought and mere distraction
conditions participants completed a 2-back task for 5 min (see
Dijksterhuis (2004)). After 5 min of conscious thought or dis-
traction, all participants engaged in a lexical decision task (Bargh
et al., 1995). Strings of letters appeared on the center of the
screen, and participants indicated whether or not each string
constituted an English word by pressing one of two buttons. The
letter strings included the RAT answers plus control words. After
completing the lexical decision task, participants in all three con-
ditions were again shown the RAT items and were asked to report
their answers. In the current research two separate outcomes of
the RAT test were assessed: implicit accessibility of correct RAT
answers (i.e., mental accessibility of RAT answers, as measured by
a lexical decision task) versus expression of those correct answers
(Wegner and Smart, 1997). A period of incubation, compared
with the same duration of conscious thought, did not increase the
reporting of correct answers. The results on accessibility, however,
revealed a striking difference: Unconscious thought, compared
with conscious thought and mere distraction, increased the men-
tal accessibility of RAT answers. These results are consistent with
unconscious thought theory, which systematically differentiates
conscious and unconscious thought processes, and suggest that
unconscious processing is more adept at associating and integrat-
ing information than conscious processing is (Dijksterhuis and
Nordgren, 2006). Importantly, in the unconscious-thought con-
dition the level of activation of RAT answers was higher than in
the mere-distraction condition, which suggests that the increased
accessibility after unconscious thought was not due to relaxation,
forgetting or the release of incorrect associations (i.e., mental
set-shifting). These findings indicate that unconscious processes
may actively facilitate the discovery of remote associations, an
important mental skill underlying creative thinking, and may
contribute to divergent thinking.
Yang et al. (2012) investigated under what conditions uncon-
scious thought can outperform conscious thought on creativity
tasks. Their results demonstrated that unconscious thought did
not provide creative advantage over conscious thought when
deliberation duration was either short or long (1 or 5 min,
respectively). However, when deliberation duration was of a mod-
erate length (3 min), the creative output of unconscious thought
exceeded that of conscious thought. These findings suggest that
the duration of unconscious thought has an inverted-U shaped
relationship with creativity. However, as different tasks require
different amounts of mental effort, the appropriate duration of
a moderate length can be assumed to be task dependent. In
line with these findings, a meta-analysis on unconscious thought
effects on decision-making (Strick et al., 2011) has shown that
unconscious thought effects are larger with moderate uncon-
scious thought intervals. Moreover, unconscious thought effects
in decision-making have been shown to be larger when a task is
used that does not require much processing capacity, that is, a
relatively undemanding task. Similarly, Sio and Ormerods (2009)
statistical meta-analytic review of incubation effects revealed that
the benefits of an incubation period are greater when participants
are occupied by an undemanding task than when they engage in
April 2014 | Volume 8 | Article 215 | 6
Ritter and Dijksterhuis
either a demanding task or no task at all (Sio and Ormerod, 2009).
Moreover, Ellwood et al. (2009) demonstrated that the type of
break during the incubation period effects later solutions. As the
functional fixedness theory as well as the general fatigue theory
predict that a break, independent of its content, should be equally
effective in producing an incubation effect, these findings suggest
that systematic effects beyond relief from functional fixedness or
general fatigue are at play.
Gallate et al. (2012) and Ritter et al. (2012b) investigated
whether one can manipulate unconscious thought processes. In
the study from Gallate et al., participants were either aware
or unaware that they would soon be returning to a divergent
thinking task. During the break period, all participants were
distracted from the task (they did an arithmetic task), ensur-
ing that any ongoing problem solving was not conscious, but
unconscious. Immediately after finishing the arithmetic task, par-
ticipants returned to the divergent thinking task. Participants in
the aware condition had significantly higher post-break creativity
scores than those in the unaware condition. Ritter et al. (2012a)
investigate whether one can actively enhance the beneficial effect
of sleep on creativity by covertly reactivating the creativity task
during sleep. Individuals creative performance was compared
after three different conditions: sleep-with-conditioned-odor;
sleep-with-control- odor; or sleep-with-no-odor. In the evening
prior to sleep, all participants were presented with a problem that
required a creative solution. In the two odor conditions, a hidden
scent-diffuser spread an odor while the problem was presented.
In the sleep-with-conditioned-odor condition, task reactivation
during sleep was induced by means of the odor that was also
presented while participants were informed about the problem. In
the sleep-with-control-odor condition, participants were exposed
to a different odor during sleep than the one diffused during
problem presentation. In the no odor condition, no odor was
presented. After a night of sleep with the conditioned odor, par-
ticipants were found to be more creative and better able to select
their most creative idea than participants who had been exposed
to a control odor or no odor while sleeping. Task reactivation
during sleep seems to actively trigger creativity-related processes
during sleep. These findings give a first indication that one can
manipulate unconscious thought processes and, thereby, facilitate
creative performance.
The idea that unconscious processes work on a problem in the
absence of conscious guidance has been described by many great
artists and thinkers, and the above mentioned findings provide
first scientific evidence for the idea that a period of incuba-
tion benefits from unconscious processes. This may be related
to the fact that unconscious thought organizes information.
Representations become better organized and more polarized,
and memory becomes more gist-based. Moreover, unconscious
thought theory postulates that unconscious thought leads to a
process of weighting whereby the importance of information
is assessed. However, this idea awaits further study, as it was
supported in some experiments (Bos et al., 2011; Usher et al.,
2011), but not in others (Ashby et al., 2011; Pachur and Forrer,
2013). These findings may suggest that unconscious thought is
a process whereby disorganized information becomes more and
more organized until some kind of equilibrium is reached, and
Frontiers in Human Neuroscience www.frontiersin.org
7
Creativityunconscious foundations of incubation
the conclusions can be transferred to consciousness. Recently,
the first neuroscientific evidence into unconscious thought was
provided. As in earlier studies on unconscious thought and
decision-making, Creswell et al. (2013) showed that uncon-
scious thinkers made better decisions than conscious thinkers
and than immediate decision makers. Moreover, their fMRI
data demonstrated that participants who thought unconsciously
while doing a distraction task showed more activity in the right
DLPFC and left intermediate visual cortex than participants who
merely performed the same distraction task. These areas were
already involved in the initial encoding of the information in
the first place, and the authors proposed a neural reactivation
account for unconscious thought, indeed demonstrating uncon-
scious processing to continue after encoding. Importantly, neural
reactivation in the right DLPFC and left intermediate visual
cortex was predictive of decision quality of unconscious thinkers.
Further neuroscientific research on creativity and incubation
should investigate brain activity during incubation to shed fur-
ther light on the underlying cognitive mechanisms of incubation
effects.
To conclude, several studies suggest that it is not merely the
absence of conscious thought that drives creativity, but that dur-
ing an incubation period unconscious processes can contribute to
creative thinking. Often, it takes time to come up with creative
ideas and solutions. It is reasonable to assume that most thought
processes underlying creative thought are neither fully conscious
nor fully unconscious. Instead, prolonged creative thought pro-
cesses may have both conscious and unconscious elements, and
conscious and unconscious thought may alternate. You think
about a problem consciously, you get stuck and perform another
task, you think some more, you sleep on it for a while, you
then think a bit more after youve encountered relevant new
information, et cetera.
PRACTICAL APPLICATIONS AND FUTURE RESEARCH
Previous studies have shown that creativity training can enhance
everyday creative performance (e.g., Scott et al., 2004), and many
tactics have been identified to facilitate creative thinking skills,
such as set-shifting, questioning assumptions, and using analogies
(i.e., finding correspondence of inner relationship or function
between different concepts). The application of unconscious
processes, however, has not been systematically introduced to
educational, innovation and business contexts. By demonstrat-
ing that unconscious processes can be important for creativity,
the current findings may encourage practitioners to use uncon-
scious processes in order to enhance creative thinking. Applying
unconscious processes could, for example, entail that people
set a goal to find creative solutions for a problem before they
are distracted from the problem by doing something different.
What people do in the meantime should be chosen carefully.
Sio and Ormerods (2009) meta-analytic review revealed that the
benefits of an incubation period are greater when participants are
occupied by an undemanding task than when they engage in a
demanding task or no task at all. Moreover, Gilhooly et al. (2013)
found that spatial incubation benefited verbal-rated creativity,
and verbal incubation benefited spatial-rated creativity but not
vice versa. Therefore, when stuck on a creative task, during an
April 2014 | Volume 8 | Article 215 | 7
Ritter and Dijksterhuis
incubation period one should do something undemanding that
is very different from the main task, before returning to it.
Although unconscious processes can be a powerful source to
facilitate creativity, only engage in daydreaming or sleeping to
produce groundbreaking discoveries or great artistic creations will
not do the trick. A plethora of raw materials has to be available to
be connected and one has to be able to focus on some options
out of an array of options. In this sense, conscious processing is
needed to establish a knowledge base, to know what problems to
tackle, and to verify and implement new ideas. Future research
may investigate what combination of conscious and unconscious
processes is most fruitful for creativity. One could think about
the order of the two processes (e.g., a period of task-related
conscious thought that is followed by a period during which one
refrains from task-related conscious thought, or repeatedly switch
between the two modes of thought), and the optimal duration
of each of the two processes. People are likely to benefit more
from an incubation period when they get stuck and, therefore,
one can assume that a relatively long period of conscious thought
should be preferred above a short period of conscious thought.
Also for unconscious processes the duration of the incubation
period seems to be of importance. In a recent study from Yang
et al. (2012), 3 min (as compared to 1 min and 5 min) seemed
to be the optimal duration of unconscious thought. However, it
is likely that 3 min of incubation is not the most appropriate
duration for all creativity tasks. It can be assumed that the optimal
duration is contingent on the task (Weisberg, 1999). Besides
exploring the optimal duration of unconscious processes as a
function of task characteristics and the optimal combination of
conscious and unconscious processes, future research could also
focus on the similarities and disparities between the different
unconscious processes (i.e., incubation, unconscious thought,
mind-wandering and sleep) and could investigate which process is
most beneficial for creativity and for the distinct mental processes
underlying creative thought (Baer, 1998). Finally, future research
may study potential moderators, for example, whether experts
and people with ample prior knowledge exhibit a different pattern
of creative performance as a result of unconscious and conscious
processes.
The present article aimed to provide an overview of the
domain of incubation and creativity, and to shed more light on
the causes of incubation effects. Research on incubation, mind
wandering, and sleep was presented and discussed, and it was
investigated whether people can think unconsciously and whether
unconscious processes can contribute to creativity. The current
findings provide first empirical support for the idea that during
an incubation period unconscious processes contribute to creative
thinking, and that it is not merely the absence of conscious
thought that drives incubation effects. We hope that the current
article inspires researchers to further tackle the unconscious foun-
dations of creativity. This will not only increase our theoretical
knowledge on the role of unconscious processes in creativity, but
will also offer valuable insights for practical implication. Under-
standing and facilitating creativity is important, as the ability to
think creatively plays an important role in many areas of our life,
such as education, arts, sciences, and the economic sector.
Frontiers in Human Neuroscience www.frontiersin.org
8
Creativityunconscious foundations of incubation
REFERENCES
Amabile, T. M. (1983). The Social Psychology of Creativity. New York, NY: Springer-
Verlag.
Amabile, T. M. (1996). Creativity in Context: Update to the Social Psychology of
Creativity. Boulder, CO: Westview Press.
Amabile, T. M. (1997). Entrepreneurial creativity through motivational synergy. J.
Creat. Behav. 31, 1826. doi: 10.1002/j.2162-6057.1997.tb00778.x
Andreasen, N. C. (1995). The Creating Brain: The Neuroscience of Genius. New York
and Washington DC: Dana Press.
Ashby, N. J. S., Glckner, A., and Dickert, S. (2011). Conscious and unconscious
thought in risky choice. Testing the capacity principle and the appropriate
weighting principle of unconscious thought theory. Front. Psychol. 2:261.
doi: 10.3389/fpsyg.2011.00261
Baer, J. (1998). The case for domain specificity of creativity. Creat. Res. J. 11, 173
177. doi: 10.1207/s15326934crj1102_7
Baird, B., Smallwood, J., Mrazek, M. D., Kam, J. W., Franklin, M. S., and Schooler,
J. W. (2012). Inspired by distraction mind wandering facilitates creative incuba-
tion. Psychol. Sci. 23, 11171122. doi: 10.1177/0956797612446024
Bargh, J. A. (2011). Unconscious thought theory and its discontents: a critique of
the critiques. Soc. Cogn. 29, 629647. doi: 10.1521/soco.2011.29.6.629
Bargh, J. A., Raymond, P., Pryor, J., and Strack, F. (1995). Attractiveness of
the underling: an automatic power sex association and its consequences for
sexual harassment and aggression. J. Pers. Soc. Psychol. 68, 768781. doi: 10.
1037//0022-3514.68.5.768
Barrett, D. (1993). The committee of sleep: a study of dream incubation for
problem solving. Dreaming 3, 115122. doi: 10.1037/h0094375
Barrett, D. (2001). The Committee of Sleep: How Artists, Scientists and Athletes use
their Dreams for Creative Problem Solvingand How you can too. NY: Crown
Books/Random House.
Bass, B. (1990). From transactional to transformational leadership: learning to
share the vision. Organ. Dyn. 18, 1931. doi: 10.1016/0090-2616(90)90061-s
Bennington, J. H., and Frank, M. G. (2003). Cellular and molecular connections
between sleep and synaptic plasticity. Prog. Neurobiol. 69, 71101. doi: 10.
1016/s0301-0082(03)00018-2
Betsch, T., Plessner, H., Schwieren, C., and Gtig, R. (2001). I like it but i dont
know why: a value-account approach to implicit attitude formation. Pers. Soc.
Psychol. Bull. 27, 242253. doi: 10.1177/0146167201272009
Bos, M. W., Dijksterhuis, A., and van Baaren, R. B. (2008). On the goal-dependency
of unconscious thought. J. Exp. Psychol. 44, 11141120. doi: 10.1016/j.jesp.2008.
01.001
Bos, M. W., Dijksterhuis, A., and van Baaren, R. B. (2011). The benefits of sleeping
on things: unconscious thought leads to automatic weighting. J. Consum.
Psychol. 21, 48. doi: 10.1016/j.jcps.2010.09.002
Bowden, E. M., and Jung-Beeman, M. (2003). Aha! Insight experience correlates
with solution activation in the right hemisphere. Psychon. Bull. Rev. 10, 730
737. doi: 10.3758/bf03196539
Bowers, K. S., Regehr, G., Balthazard, C., and Parker, K. (1990). Intuition
in the context of discovery. Cogn. Psychol. 22, 72110. doi: 10.1016/0010-
0285(90)90004-n
Buckner, R. L., Andrews-Hanna, J. R., and Schacter, D. L. (2008). The brains default
network: anatomy, function and relevance to disease. Ann. N Y Acad. Sci. 1124,
138. doi: 10.1196/annals.1440.011
Cai, D. J., Mednick, S. A., Harrison, E. M., Kanady, J. C., and Mednick, S. C.
(2009). REM, not incubation, improves creativity by priming associative net-
works. Proc. Natl. Acad. Sci. U S A 106, 1013010134. doi: 10.1073/pnas.09002
71106
Christoff, K., Gordon, A. M., Smallwood, J., Smith, R., and Schooler, J. W. (2009).
Experience sampling during fMRI reveals default network and executive system
contributions to mind wandering. Proc. Natl. Acad. Sci. U S A 106, 87198724.
doi: 10.1073/pnas.0900234106
Christoff, K., Gordon, A. M., and Smith, R. (2011). The role of spontaneous
thought in human cognition, in Neuroscience of Decision Making, eds O.
Vartanian and D. R. Mandel (New York, NY: Psychology Press), 259284.
Claxton, G. (1997). Hare Brain, Tortoise Mind: How Intelligence Increases When you
Think Less. New York, NY: Harper Collins.
Cole, M. W., and Schneider, W. (2007). The cognitive control network: integrated
cortical regions with dissociable functions. Neuroimage 37, 343360. doi: 10.
1016/j.neuroimage.2007.03.071
April 2014 | Volume 8 | Article 215 | 8
Ritter and Dijksterhuis
Creswell, J. D., Bursley, J. K., and Satpute, A. B. (2013). Neural reactivation
links unconscious thought to decision-making performance. Soc. Cogn. Affect.
Neurosci. 8, 863869. doi: 10.1093/scan/nst004
Cropley, A. J. (1990). Creativity and mental health in everyday life. Creat. Res. J. 3,
167187. doi: 10.1080/10400419009534351
Dietrich, A., and Kanso, R. (2010). A review of EEG, ERP and neuroimaging studies
of creativity and insight. Psychol. Bull. 136, 822848. doi: 10.1037/a0019749
Dijksterhuis, A. (2004). Think different: the merits of unconscious thought in
preference development and decision making. J. Pers. Soc. Psychol. 87, 586598.
doi: 10.1037/0022-3514.87.5.586
Dijksterhuis, A., Bos, M. W., Nordgren, L. F., and Van Baaren, R. B. (2006). On
making the right choice: the deliberation-without-attention effect. Science 311,
10051007. doi: 10.1126/science.1121629
Dijksterhuis, A., and Meurs, T. (2006). Where creativity resides: the generative
power of unconscious thought. Conscious Cogn. 15, 135146. doi: 10.1016/j.
concog.2005.04.007
Dijksterhuis, A., and Nordgren, L. F. (2006). A theory of unconscious thought.
Perspect. Psychol. Sci. 1, 95109. doi: 10.1111/j.1745-6916.2006.00007.x
Dodds, R. D., Ward, T. B., and Smith, S. M. (2003). Incubation in problem solving
and creativity, in The Creativity Research Handbook (Vol. 3), ed M. A. Runco
(Cresskill, NJ: Hampton).
Ellwood, S., Pallier, G., Snyder, A., and Gallate, J. (2009). The incubation effect:
hatching a solution? Creat. Res. J. 21, 614. doi: 10.1080/10400410802633368
Ericsson, K. A. (2006). The influence of experience and deliberate practice on the
development of superior expert performance. Camb. Handb. Expertise Expert
Perform. 683703. doi: 10.1017/cbo9780511816796.038
Feist, G. J., and Gorman, M. E. (1998). The psychology of science: review and
integration of a nascent discipline. Rev. Gen. Psychol. 2, 347. doi: 10.1037/1089-
2680.2.1.3
Fox, M. D., Snyder, A. Z., Vincent, J. L., Corbetta, M., Van Essen, D. C., and
Raichle, M. E. (2005). The human brain is intrinsically organized into dynamic,
anticorrelated functional networks. Proc. Natl. Acad. Sci. U S A 102, 96739678.
doi: 10.1073/pnas.0504136102
Gallate, J., Wong, C., Ellwood, S., Roring, R. W., and Snyder, A. (2012). Creative
people use nonconscious processes to their advantage. Creat. Res. J. 24, 146151.
doi: 10.1080/10400419.2012.677282
Ghiselin, B. (1952). The Creative Process. New York, NY: New American Library.
Giambra, L. M. (1979). Sex differences in daydreaming and related mental activity
from the late teens to the early nineties. Int. J. Aging Hum. Dev. 10, 134. doi: 10.
2190/01bd-rfne-w34g-9eca
Giambra, L. M., and Grodsky, A. (1989). Task-unrelated images and thoughts while
reading. Imagery 2731. doi: 10.1007/978-1-4899-0876-6_3
Gilhooly, K. J., Georgiou, G., and Devery, U. (2013). Incubation and creativity: do
something different. Think. Reason. 19, 137149. doi: 10.1080/13546783.2012.
749812
Gupta, N., Yang, Y., Mednick, S. C., and Huber, D. E. (2012). The road not taken:
creative solutions require avoidance of high-frequency responses. Psychol. Sci.
23, 288294. doi: 10.1177/0956797611429710
Hadamard, J. (1945). Essay on The Psychology of Invention in the Mathematical Field.
Princeton, NJ: Princeton University Press.
Helmholtz, H. von. (1896). Vortraege Und Reden. Brunswick: Friedrich Viewig und
Sohn.
Jung, R. E., Segall, J. M., Bockholt, H. J., Chavez, R. S., Flores, R., and Haier, R. J.
(2010). Neuroanatomy of creativity. Hum. Brain Mapp. 31, 398409. doi: 10.
1002/hbm.20874
Kaplan, C. A. (1989). Hatching a Theory of Incubation: Does Putting a Problem Aside
Really Help? If so, Why? (Pittsburgh, PA: Doctoral dissertation Carnegie-Mellon
University).
Kaufman, J. C. (2002). Dissecting the golden goose: components of studying
creative writers. Creat. Res. J. 14, 2740. doi: 10.1207/s15326934crj1401_3
Koestler, A. (1964). The Act of Creation. New York, NY: Penguin Books.
Kounios, J., and Jung-Beeman, M. (2009). Aha! The cognitive neuroscience
of insight. Curr. Dir. Psychol. Sci. 18, 210216. doi: 10.1111/j.1467-
8721.2009.01638.x
Kris, E. (1952). Psychoanalytic Explorations in Art. New York, NY: International
Universities Press.
Kubie, L. S. (1985). Neurotic Distortion of the Creative Process. Lawrence, Kansas:
University of Kansas Press.
Frontiers in Human Neuroscience www.frontiersin.org
9
Creativityunconscious foundations of incubation
Khn, S., Ritter, S. M., Mller, B. C. N., van Baaren, R. B., Brass, M., and
Dijksterhuis, A. (2013). The importance of unconscious processes in creativity
a structural MRI study. J. Creat. Behav. doi: 10.1002/jocb.45
Langley, P., and Jones, R. (1988). A computational model of scientific insight,
in The Nature of Creativity: Contemporary Psychological Perspectives, ed R. J.
Sternberg (New York, NY: Cambridge University Press), 177201.
Manni, R. (2005). Rapid eye movement sleep, non-rapid eye movement
sleep, dreams and hallucinations. Curr. Psychiatry Rep. 7, 196200. doi: 10.
1007/s11920-005-0053-0
Mason, M. F., Norton, M. I., Van Horn, J. D., Wegner, D. M., Grafton, S. T., and
Macrae, C. N. (2007). Wandering minds: the default network and stimulus-
independent thought. Science 315, 393395. doi: 10.1126/science.1131295
Mazzarello, P. (2000). What dreams may come? Nature 408, 523. doi: 10.
1038/35046170
Mumford, M. D. (2002). Social innovation: ten cases from Benjamin Franklin.
Creat. Res. J. 14, 253266. doi: 10.1207/s15326934crj1402_11
Nieuwenstein, M., and van Rijn, H. (2012). The unconscious thought advantage:
further replication failures from a search for confirmatory evidence. Judgm.
Decis. Mak. 7, 779798.
Olton, R. M., and Johnson, D. M. (1976). Mechanisms of incubation in
creative problem solving. Am. J. Psychol. 89, 617630. doi: 10.2307/14
21461
Orlet, S. (2008). An expanding view on incubation. Creat. Res. J. 20, 297308.
doi: 10.1080/10400410802278743
Pachur, T., and Forrer, E. A. (2013). Selection of decision strategies after conscious
and unconscious thought. J. Behav. Decis. Mak. 5, 477488. doi: 10.1002/bdm.
1780
Patrick, A. S. (1986). The role of ability in creative incubation. Pers. Indiv. Diff. 7,
169174. doi: 10.1016/0191-8869(86)90052-8
Reverberi, C., Toraldo, A., DAgostini, S., and Skrap, M. (2005). Better without
(lateral) frontal cortex? Insight problems solved by frontal patients. Brain 128,
28822890. doi: 10.1093/brain/awh577
Ric, F., and Muller, D. (2012). Unconscious addition: when we unconsciously
initiate and follow arithmetic rules. J. Exp. Psychol. Gen. 141, 222226. doi: 10.
1037/a0024608
Ritter, S. M., Strick, M., Bos, M. W., Van Baaren, R. B., and Dijksterhuis, A. (2012a).
Good morning creativity: task reactivation during sleep enhances beneficial
effect of sleep on creative performance. J. Sleep Res. 21, 643647. doi: 10.1111/j.
1365-2869.2012.01006.x
Ritter, S. M., van Baaren, R. B., and Dijksterhuis, A. (2012b). Creativity: the role of
unconscious processes in idea generation and idea selection. Think. Skills Creat.
7, 2127. doi: 10.1016/j.tsc.2011.12.002
Rugg, H. (1963). Imagination. New York, NY: Harper and Row.
Runco, M. A. (2004). Creativity. Annu. Rev. Psychol. 55, 657687. doi: 10.
1146/annurev.psych.55.090902.141502
Runco, M. A., and Pritzker, S. (1999). The Encyclopedia of Creativity. San Diego:
Academic Press.
Sawyer, R. K. (2012). Explaining Creativity: The Science of Human Innovation. 2nd
Edn. New York, NY: Oxford University Press.
Schooler, J. W., and Melcher, J. (1995). The ineffability of insight, in The Creative
Cognition Approach, eds S. M. Smith, T. B. Ward, and R. A. Fink (Cambridge,
MA: MIT Press), 97134.
Scott, G., Leritz, L. E., and Mumford, M. D. (2004). The effectiveness of cre-
ativity training: a quantitative review. Creat. Res. J. 16, 361388. doi: 10.
1080/10400410409534549
Segal, E. (2004). Incubation in insight problem solving. Creat. Res. J. 16, 141148.
doi: 10.1207/s15326934crj1601_13
Simonton, D. K. (1999). Origins of Genius: Darwinian Perspectives on Creativity.
New York, NY: Oxford University Press.
Sio, U. N., and Ormerod, T. C. (2009). Does incubation enhance problem
solving? A meta-analytic review. Psychol. Bull. 135, 94120. doi: 10.1037/a00
14212
Sio, U. N., Monaghan, P., and Ormerod, T. (2012). Sleep on it, but only if it is
difficult: effects of sleep on problem solving. Mem. Cognit. 41, 159166. doi: 10.
3758/s13421-012-0256-7
Sklar, A. Y., Levy, N., Goldstein, A., Mandel, R., Maril, A., and Hassin, R. R. (2012).
Reading and doing arithmetic unconsciously. Proc. Natl. Acad. Sci. U S A 109,
1961419619. doi: 10.1073/pnas.1211645109
April 2014 | Volume 8 | Article 215 | 9
Ritter and Dijksterhuis
Smallwood, J., Nind, L., and OConnor, R. C. (2009). When is your head at? An
exploration of the factors associated with the temporal focus of the wandering
mind. Conscious. Cogn. 18, 118125. doi: 10.1016/j.concog.2008.11.004
Smallwood, J., Obonsawin, M., and Heim, D. (2003). Task unrelated thought: the
role of distributed processing. Conscious. Cogn. 12, 169189. doi: 10.1016/s1053-
8100(02)00003-x
Smallwood, J., and Schooler, J. W. (2006). The restless mind. Psychol. Bull. 132,
946958. doi: 10.4324/9781315037509
Smith, S. M., and Blankenship, S. E. (1989). Incubation effects. Bull. Psychon. Soc.
27, 311314. doi: 10.3758/BF03334612
Smith, S. M., and Dodds, R. A. (1999). Incubation. Encyclopedia Creat. 2, 3943.
doi: 10.1787/9789264173781-en
Smith, S. M. (2003). The constraining effects of initial ideas, in Group Creativity:
Innovation through Collaboration, eds P. Paulus and B. Nijstad (New York:
Oxford University Press). 3146.
Spreng, R. N., Stevens, W. D., Chamberlain, J., Gilmore, A. W., and Schacter, D.
L. (2010). Default network activity, coupled with the frontoparietal control
network, supports goal-directed cognition. Neuroimage 53, 303317. doi: 10.
1016/j.neuroimage.2010.06.016
Stickgold, R., Hobson, J. A., Fosse, R., and Fosse, M. (2001). Sleep, learning
and dreams: off-linememory reprocessing. Science 294, 10521057. doi: 10.
1126/science.1063530
Stickgold, R., Scott, L., Rittenhouse, C., and Hobson, J. A. (1999). Sleep-induced
changes in associative memory. J. Cogn. Neurosci. 11, 182193. doi: 10.
1162/089892999563319
Stickgold, R., and Walker, M. P. (2004). To sleep, perchance to gain creative insight?
Trends Cogn. Sci. 8, 191192. doi: 10.1016/j.tics.2004.03.003
Storm, B. C., and Angelo, G. (2010). Overcoming fixation: creative problem
solving and retrieval-induced forgetting. Psychol. Sci. 21, 12631265. doi: 10.
1177/0956797610379864
Strick, M., Dijksterhuis, A., Bos, M. W., Sjoerdsma, A., van Baaren, R. B., and
Nordgren, L. F. (2011). A meta-analysis on unconscious thought effects. Soc.
Cogn. 29, 738762. doi: 10.1521/soco.2011.29.6.738
Teasdale, J. D., Segal, Z., and Williams, J. M. G. (1995). How does cognitive therapy
prevent depressive relapse and why should attentional control (mindfulness)
training help? Behav. Res. Ther. 33, 2539. doi: 10.1016/0005-7967(94)e0011-7
Usher, M., Russo, Z., Weyers, M., Brauner, R., and Zakay, D. (2011). The impact
of mode of thought in complex decisions: intuitive decisions are better. Front.
Psychol. 2:37. doi: 10.3389/fpsyg.2011.00037
Van Gaal, S., de Lange, F. P., and Cohen, M. X. (2012). The role of consciousness
in cognitive control and decision making. Front. Hum. Neurosci. 6:121. doi: 10.
3389/fnhum.2012.00121
Vincent, J. L., Patel, G. H., Fox, M. D., Snyder, A. Z., Baker, J. T., Van Essen,
D. C., et al. (2007). Intrinsic functional architecture in the anaesthetized monkey
brain. Nature 447, 8386. doi: 10.1038/nature05758
Frontiers in Human Neuroscience www.frontiersin.org
10
Creativityunconscious foundations of incubation
Wagner, U., Gais, S., Haider, H., Verleger, R., and Born, J. (2004). Sleep inspires
insight. Nature 427, 352355. doi: 10.1038/nature02223
Walker, M. P., Liston, C., Hobson, J. A., and Stickgold, R. (2002). Cognitive
flexibility across the sleepwake cycle: REM-sleep enhancement of anagram
problem solving. Brain Res. Cogn. Brain Res. 14, 317324. doi: 10.1016/s0926-
6410(02)00134-9
Wallas, G. (1926). The Art Of Thought. New York, NY: Harcourt Brace.
Wegner, D. M., and Smart, L. (1997). Deep cognitive activation: a new approach to
the unconscious. J. Consult. Clin. Psychol. 65, 984995. doi: 10.1037/0022-006x.
65.6.984
Weisberg, R. W. (1999). Creativity and knowledge: a challenge to theories, in
Handbook of Creativity, ed R. J. Sternberg (Cambridge, England: Cambridge
University Press), 226250.
Woodworth, R. S., and Schlosberg, H. (1954). Experimental Psychology (Rev. ed.).
Oxford, UK: Holt.
Yang, H., Chattopadhyay, A., Zhang, K., and Dahl, D. W. (2012). Unconscious
creativity: when can unconscious thought outperform conscious thought? J.
Consum. Psychol. 22, 573581. doi: 10.1016/j.jcps.2012.04.002
Yang, J., Weng, X., Zang, Y., Xu, M., and Xu, X. (2010). Sustained activity within
the default mode network during an implicit memory task. Cortex 46, 354366.
doi: 10.1016/j.cortex.2009.05.002
Yaniv, I., and Meyer, D. E. (1987). Activation and metacognition of inaccessible
stored information: potential bases for incubation effects in problem solving.
J. Exp. Psychol. Learn. Mem. Cogn. 13, 187205. doi: 10.1037/0278-7393.13.
2.187
Zhong, C. B., Dijksterhuis, A. J., and Galinsky, A. D. (2008). The merits of
unconscious thought in creativity. Psychol. Sci. 19, 912918. doi: 10.1111/j.1467-
9280.2008.02176.x
Conflict of Interest Statement: The authors declare that the research was conducted
in the absence of any commercial or financial relationships that could be construed
as a potential conflict of interest.
Received: 31 October 2013; accepted: 26 March 2014; published online: 11 April
2014.
Citation: Ritter SM and Dijksterhuis A (2014) Creativitythe unconscious founda-
tions of the incubation period. Front. Hum. Neurosci. 8:215. doi: 10.3389/fnhum.2014.
00215
This article was submitted to the journal Frontiers in Human Neuroscience.
Copyright 2014 Ritter and Dijksterhuis. This is an open-access article distributed
under the terms of the Creative Commons Attribution License (CC BY). The use, dis-
tribution or reproduction in other forums is permitted, provided the original author(s)
or licensor are credited and that the original publication in this journal is cited, in
accordance with accepted academic practice. No use, distribution or reproduction is
permitted which does not comply with these terms.
April 2014 | Volume 8 | Article 215 | 10
Psychological Bulletin 2009 American Psychological Association
2009, Vol. 135, No. 1, 94 120 0033-2909/09/$12.00 DOI: 10.1037/a0014212

Does Incubation Enhance Problem Solving? A Meta-Analytic Review

Ut Na Sio and Thomas C. Ormerod

Lancaster University

A meta-analytic review of empirical studies that have investigated incubation effects on problem solving
is reported. Although some researchers have reported increased solution rates after an incubation period
(i.e., a period of time in which a problem is set aside prior to further attempts to solve), others have failed
to find effects. The analysis examined the contributions of moderators such as problem type, presence of
solution-relevant or misleading cues, and lengths of preparation and incubation periods to incubation
effect sizes. The authors identified a positive incubation effect, with divergent thinking tasks benefiting
more than linguistic and visual insight tasks from incubation. Longer preparation periods gave a greater
incubation effect, whereas filling an incubation period with high cognitive demand tasks gave a smaller
incubation effect. Surprisingly, low cognitive demand tasks yielded a stronger incubation effect than did
rest during an incubation period when solving linguistic insight problems. The existence of multiple types
of incubation effect provides evidence for differential invocation of knowledge-based vs. strategic
solution processes across different classes of problem, and it suggests that the conditions under which
incubation can be used as a practical technique for enhancing problem solving must be designed with
care.

Keywords: incubation, problem solving, insight, meta-analysis

Anecdotal reports of the intellectual discovery processes of
individuals hailed as geniuses (see, e.g., Ghiselin, 1985; Wallas,
1926; Woodworth & Schlosberg, 1954) share a common theme: A
flash of insight pops unexpectedly into the mind of the individual
after he or she has put an unsolved problem aside for a period of
time, having failed in initial attempts to solve it. This temporary
shift away from an unsolved problem that allows a solution to
emerge seemingly as if from no additional effort is termed an
incubation period (Wallas, 1926). Its importance in current think-
ing and practice is illustrated by a recent search in Google Scholar
for the term incubation along with either creativity, insight, or
problem that yielded 5,510 articles, with the search restricted to the
years 1997 to 2007 and the subject areas to social sciences, arts,
and humanities. An additional 1,970 articles were yielded when
the subject areas business, administration, and economics were
included. Yet there are many conflicting accounts of incubation,
with some studies reporting strong effects (e.g., Smith & Blanken-
ship, 1989) and others failing to find any effect at all (Olton &
Johnson, 1976). This article aims to resolve the uncertainties
surrounding the phenomenon by providing a statistical meta-
analytic review of empirical studies of incubation. We suggest that
it is only armed with the results of an integrative and quantitatively
based review that progress in understanding the mechanisms that
might underlie the phenomenon can be made.
One theoretical reason for studying incubation is because it is
closely associated with insightful thinking. Indeed, Wallas (1926)
Ut Na Sio and Thomas C. Ormerod, Department of Psychology, Lan-
caster University, Lancaster, United Kingdom.
Correspondence concerning this article should be addressed to Thomas
C. Ormerod, Department of Psychology, Lancaster University LA1 4YF,
United Kingdom. E-mail: t.ormerod@lancaster.ac.uk
proposed incubation as the second of four phases in problem
solving (the others being preparation, illumination, and verifica-
tion). Insight may be characterized as a sudden, unpredictable, and
nonverbalizable solution discovery (e.g., Metcalfe & Weibe,
1987). Some researchers see the apparently unconscious nature of
solution discovery as evidence that the processes required to
achieve insight in problem solving are qualitatively different from
those used to tackle problems that do not require insight (e.g.,
Jung-Beeman & Bowden, 2000; Wertheimer, 1985). Incubation
might serve a valuable role in arbitrating between theories of
insight, in particular between special-process theories based on
unconscious mechanisms of spreading activation (e.g., Knoblich,
Ohlsson, Haider, & Rhenius, 1999) and theories of insight as
normal problem-solving processes based on conscious mecha-
nisms of search (e.g., MacGregor, Ormerod, & Chronicle, 2001).
Understanding the role of incubation periods may also allow us
to make use of them effectively to promote creativity in areas such
as individual problem solving, classroom learning, and work en-
vironments. Educational researchers have tried to introduce incu-
bation periods in classroom activity, and positive incubation ef-
fects in fostering students creativity have been reported (Lynch &
Swink, 1967; Medd & Houtz, 2002; Rae, 1997; Webster, Camp-
bell, & Jane, 2006). However, in the absence of a comprehensive
theory or model that can explain how and why positive incubation
effects might emerge and under what conditions they are best
fostered, no general pedagogic recommendations can be made.
Several hypotheses have been proposed to account for the
alleged positive effects of incubation periods on problem solving,
and they can be divided into two main kinds: conscious work and
unconscious work. The conscious-work hypothesis holds that in-
cubation effects are due to issues such as reduction of mental
fatigue (Posner, 1973) or additional covert problem solving during
the incubation period (Browne & Cruse, 1988; Posner, 1973). Both

94

11
 INCUBATION AND PROBLEM SOLVING
sources implicate changes in consciously controlled problem-
solving activities during an incubation period. In contrast, the
unconscious-work hypothesis suggests that positive incubation
effects are the result of gradual and unconscious problem-solving
processes that occur during an incubation period (Bowers, Regehr,
Balthazard, & Parker, 1990; Seifert, Meyer, Davidson, Patalano, &
Yaniv, 1995; Simon, 1966; Smith, 1995; Smith & Blankenship,
1991; Yaniv & Meyer, 1987).
Three different unconscious processes have been proposed to
account for incubation effects. The first involves eliciting new
knowledge: Over time, activation will spread toward previously
ignored but relevant memory items. Even if relevant items do not
receive above-threshold activation, this process can still sensitize
individuals to related concepts, and thus they will be more likely
to make use of external cues to solve a problem. In addition,
partially activated concepts may combine with others to yield
fortuitous insightful ideas (Bowers et al., 1990; Smith, 1995; Smith
& Blankenship, 1991; Yaniv & Meyer, 1987). The second hypoth-
esis is selective forgetting: An incubation period will weaken the
activation of inappropriate solution concepts that distract individ-
uals during initial attempts, allowing a fresh view of the problem
(Smith, 1995; Smith & Blankenship, 1991). The third hypothesis is
problem restructuring: An individuals mental representation of a
problem will be reorganized into a more appropriate and stable
form after initial unsuccessful attempts. The individual is then
more able to capitalize upon relevant external information or to
rearrange problem information in a manner that allows a solution
to be found more readily (Seifert et al., 1995). Problem restruc-
turing might emerge either from switching the strategy used to
search for moves to attempt (e.g., MacGregor et al., 2001) or from
relaxing self-imposed inappropriate constraints on the problem
representation (Knoblich et al., 1999). Studies of metacognition
indicate that strategy switching can be unconscious (Newton &
Roberts, 2005; Reder & Schunn, 1996; Siegler & Stern, 1998) and
that different strategies compete for activation during the strategy
selection process (Siegler & Stern, 1998).
The conscious- and unconscious-work accounts generate differ-
ent predictions concerning the effects of activities that individuals
engage in during an incubation period. According to the conscious-
work account, individuals benefit most from an unfilled incubation
period, as this gives them an opportunity either to relax, reduce
fatigue, or continue working on the problem. In contrast,
unconscious-work accounts suggest that unconscious problem-
solving processes occur when individuals shift their attention away
from the problem to other mental activities. Thus, a certain level of
involvement in other tasks during an incubation period should
facilitate postincubation problem solving.
A number of experimental studies have examined the role of
task type during an incubation period. The experimental paradigms
of these incubation studies are fairly uniform: The work of one
group of participants is interrupted with an incubation period
(having a break or performing other tasks) while solving a prob-
lem, whereas the other group works on the problem continuously.
Performance differences between these two groups are then com-
pared. The findings of the published studies do not give uncondi-
tional support to either the unconscious-work or the conscious-
work account. Patrick (1986) found that participants who had a
filled incubation period outperformed those who had an unfilled
incubation period. However, Browne and Cruse (1988) reported
12
95
the opposite pattern: Participants who took a rest during an incu-
bation period performed better than those who had to perform
tasks during an incubation period. There are also studies that report
the same level of performance by participants with filled and
unfilled incubation periods (Olton & Johnson, 1976; Smith &
Blankenship, 1989). However, these studies vary in terms of the
length of incubation period, the target problems tackled, and the
nature of the interpolated tasks during the incubation period.
Because of inconsistent findings concerning incubation, some
researchers have doubted the existence of the effect, in particular
rejecting the unconscious-work hypotheses (Browne & Cruse,
1988; Olton & Johnson, 1976; Perkins, 1995). However, one
explanation for conflicting findings is that there are procedural
moderators other than task type that influence the occurrence of
problem-solving processes during an incubation period, such as the
length of the incubation period or the nature of the problem. The
field lacks a comprehensive review that summarizes and evaluates
these studies. There have been two reviews to date of relations
between different procedural variables and the incubation effect,
and both are qualitative in nature. Oltons (1979) review of past
incubation studies led him to question the existence of an incuba-
tion effect, given that no experimental paradigm appeared to
demonstrate an incubation effect reliably. Yet, a limited number of
studies were available at that time: Only 10 incubation studies
were included in his review. A recent review by Dodds, Ward, and
Smith (in press) with more studies included suggested that several
variables may interact to influence the effectiveness of an incuba-
tion period. However, the qualitative nature of their review led
them to conclude that findings of past studies are too divergent and
that more studies are needed to assess the impact of each variable
and to identify the optimum settings for an incubation effect.
The wide variation in experimental parameters among studies
makes it difficult to draw cross-experiment conclusions from a
qualitative review. To overcome these problems, a systematic
meta-analytic review is needed. Meta-analytic review allows a
quantitative evaluation of research domains that describes the
typical strength of the effect or phenomenon and also the relation
of each moderator to the size of the effect by using statistical
analysis methods (Rosenthal, 1995). The objectives of the current
study were to carry out the first statistical meta-analysis of incu-
bation studies to assess the effect size of the experimental incu-
bation effect, and more important, the impact of potential moder-
ators on the incubation effect size. However, in order to undertake
the meta-analysis, we first need to identify the likely key moder-
ators, which we achieve in the next section by reviewing the
methods used in previous studies. A particular focus of this review
is to identify moderators that might discriminate conscious-work
and unconscious-work hypotheses and also the mechanisms (re-
duction of fatigue, additional work, activation of new information,
forgetting, restructuring) that might underlie each hypothesis.
Potential Moderators
The Interpolated Task Used During the Incubation Period
Various types of interpolated task have been used in past stud-
ies, and they can be divided into tasks of high or low cognitive
demand. Examples of high cognitive demand tasks include mental
rotation, counting backwards, and visual memory tests, whereas
96 SIO AND ORMEROD
reading is commonly adopted as a low cognitive demand task.
High demand tasks should fully occupy the individuals mind and
prevent further conscious work on the unsolved problem. Some
studies have reported that undertaking a high cognitive demand
task during an incubation period is beneficial to the problem-
solving process (Kaplan, 1990; Patrick, 1986; Segal, 2004). None-
theless, studies using low cognitive demand tasks that do not
require individuals to focus their conscious attention on task per-
formance have reported similar benefits (Beck, 1979; Silveira,
1972; Smith & Blankenship, 1989).
Length of the Incubation Period
Longer incubation periods may allow additional problem-
solving activity or a greater degree of forgetting of misleading
items or spreading of activation memory. Thus, problem solvers
may show a larger performance improvement when they return to
the problem after a long incubation period than after a short one.
Some studies have reported evidence supporting this contention
(Beck, 1979; Fulgosi & Guilford, 1968; Silveira, 1972; Smith &
Blankenship, 1989). However, it is difficult to draw cross-experiment
conclusions, because there is no standard operationalization of what
constitutes long and short incubation periods. In Smith and
Blankenships (1989) study, for example, a 15-min incubation
period was defined as a long incubation period, and they reported
that participants receiving this length of incubation period per-
formed better than those receiving a 5-min incubation period.
However, in Becks (1979) study, a 20-min incubation period was
considered short, and participants performance in this group did
not differ from that of the control group. Kaplan (1990) suggested
that to judge whether the incubation period is short or long, the
length of time that problem solvers spend on initial attempts to
solve (named the preparation period by Wallas, 1926) should also
be taken into account. Kaplan found that a larger incubation effect
was observed after increasing the ratio of the length of the prep-
aration period to the incubation period. Thus, in addition to in-
cluding incubation and preparation periods as separate moderators
in the meta-analysis reported below, we also undertook a second-
ary analysis using the ratio of preparation to incubation time as an
alternative moderator.
Length of the Preparation Period
During the preparation period, problem solvers gather informa-
tion to formulate a problem representation and make initial at-
tempts to solve, which may lead to an impasse. Although a
problem may not be solved during the preparation period, this does
not mean that the effort the problem solver spends on the problem
is fruitless. Schank (1982, 1999) and VanLehn (1988) both sug-
gested that failure in problem solving is important in the human
learning process. Studies by Patalano and Seifert (1994) and
Seifert et al. (1995) have found evidence of a Zeigarnik effect in
insight problem solving (Zeigarnik, 1927, 1938), wherein individ-
uals remembered the problems on which they got stuck better
than those they solved immediately. Seifert et al. hypothesized that
having a better memory for failed problems might help individuals
return efficiently to the problem once relevant new information is
encountered during an incubation period, thereby maximizing the
chance of solving. Evidence concerning this prediction has been
13
obtained in an empirical study carried out by Silveira (1972),
showing that problem solvers performed better with longer prep-
aration and incubation periods.
Nature of the Problem
Various different types of problem have been used in incubation
studies. Some problems, which we term creative problems here,
require individuals to produce multiple new ideas to meet a spe-
cific brief. For example, a verbal divergent-production task is the
consequences task (e.g., What would be the result if everyone
suddenly lost the ability to read and write; Fulgosi & Guilford,
1968). Typically, there is no right or wrong answer to this kind of
problem, and performance is assessed in terms of the number of
solution ideas that are generated.
Other problems require individuals to discover a specific target
solution that is known in advance by the experimenter. Problems
of this kind that have been studied in the literature on incubation
are generally of a type described as insight problems, in that they
require the solver to reject initial solution ideas by achieving
insight into an alternative strategy or knowledge domain. The
insight problems used in the incubation studies can be divided into
visual problems, which typically require the solver to consider a
visuospatial array (e.g., the nine-dot problem; Scheerer, 1963), and
linguistic problems, which typically require the solver to consider
linguistic information related to the problem. The remote associ-
ates task (RAT; S. A. Mednick, 1962) is one of the most com-
monly used linguistic problems in incubation studies. In each
RAT, three stimulus words are presented to individuals, who then
have to think of a fourth word that can form an association with
each of the three words. For example, if the three stimulus words
of a RAT are electric, wheel, and high, the fourth word can be
chair. Bowden and Jung-Beeman (2003b) have developed a pool
of remote associates problems and collected normative data re-
garding the resolution rates and response times for the problems.
The classification of insight problems into visual- and linguistic-
based categories is supported by research findings from Gilhooly
and Murphy (2005) showing that solving visual and linguistic
insight problems requires different types of cognitive skills.
In the remainder of this article, we refer to problem types as
creative, visual, and linguistic. Descriptions of the types of prob-
lem used in incubation studies are illustrated in Appendix A.
Problem type is likely to be an important determinant of incuba-
tion, because it seems likely that each type creates different task
demands. For instance, the nine-dot problem appears to require the
solver to restructure an initially faulty or incomplete problem
representation in searching for a representation that allows solu-
tion, whereas the consequences task appears to require the activa-
tion of as wide a range of different concepts as possible. One
question the meta-analysis allows us to address is whether an
incubation period favors one type of problem more than another.
The Presence of Solution-Relevant Cues
Some unconscious processes proposed to explain incubation
effects are purely internal and thus independent of the external
environment, such as the inhibition of irrelevant memory (Smith,
1995; Smith & Blankenship, 1991) and the recombination of
partially activated concepts (Bowers et al., 1990). Others stress
 INCUBATION AND PROBLEM SOLVING
interactions with the external environment, such as the proposal
that spreading activation can partially activate previously ignored
relevant memory and therefore sensitize the problem solver to
chance encounters with related stimuli (Seifert et al., 1995). A few
studies have examined the effects of the presence of cues during an
incubation period (Browne & Cruse, 1988; Dodds, Smith, & Ward,
2002; Dorfman, 1990; Dreistadt, 1969; Olton & Johnson, 1976).
Most failed to find any positive effect of cues on the incubation
effect. However, the failure reported in these studies may be due to
other factors, such as the difficulty of the unsolved problems. In
order to have a fair evaluation of the impact of this moderator, we
first have to isolate the effect of other moderators on the incubation
effect.
Misleading Cues
Another factor that may influence the occurrence of incubation
effects is the presence of misleading cues. Smith and Blankenship
(1989) carried out a series of experiments to examine the effect of
an incubation period on solving RATs, in which participants had to
find a word that might accompany each of three presented words.
Smith and Blankenship presented cues (shown here in italics)
comprising misleading associates and the target word next to each
of the three stimulus words. An example of a misleading RAT is:
SHIP ocean, OUTER space, CRAWL floor. The target solution is
space. Performance improvements after an incubation period were
observed only when participants solved tasks containing mislead-
ing cues. They concluded that a problem solver who is fixated on
misleading information benefits more from an incubation period.
The misleading cues data provide critical support for forgetting-
based explanations of incubation. The presence of misleading cues
is therefore one of the potential moderators examined in this
meta-analysis.
The Meta-Analysis
The variables mentioned above were the potential moderators of
incubation chosen for this meta-analysis. Note that other potential
moderators might have been included (e.g., number of trials,
participant characteristics), but we focused on those we believe are
fundamental to discriminating differing theoretical accounts of
incubation.
The statistical meta-analysis that follows addressed two ques-
tions: First, is there reliable evidence for incubation; and second,
what are the most influential moderators? To address the first
question, we computed the effect size of the incubation effect
reported in each available study. Given that the variability among
effect sizes is likely to be greater than that resulting from subject-
level sampling, a random-effects model was adopted in this meta-
analysis. A heterogeneity test was carried out to verify this as-
sumption, and then the weighted mean under the assumption of
random-effects model was computed and assessed to determine if
it was significantly larger than zero.
To address the second question, we carried out weighted least-
squares linear regressions using the aforementioned moderators as
predictor variables and the incubation effect size in each study
weighted by the inverse of its variance as the criterion variable.
The results of the regression show the independent contributions of
each potential moderator to the incubation effect, while controlling
14
97
for all other moderators. This approach allows us to summarize the
past studies systematically even though they vary widely in num-
ber and type of experimental parameters. In addition, interactions
between different moderators, such as the nature of the interpola-
tion task during the incubation period and the nature of the prob-
lem, were examined.
Literature Search
We collected publications that contained studies relevant to a
meta-analysis of incubation through a search of the ERIC, Psy-
cINFO, PsycARTICLES, and MEDLINE databases using the key-
word incubat!, intersected with one of fixation, creativ!, diver-
gent!, insight!, or problem. Then, references given in all the
obtained articles were systematically searched for additional rele-
vant publications. There is a concern that studies with statistically
significant results are more likely to get published than those
without significant results, and this may lead to a biased retrieval
of studies. To ameliorate this to some extent, we carried out similar
literature searches in the ProQuest Digital Dissertations database
and used Google Scholar for retrieving doctoral dissertations,
unpublished articles, and conference articles concerning the incu-
bation effect. In total, 37 relevant publications were identified and
obtained. Studies meeting the following criteria were assimilated
in the analysis:
1. The settings and difficulty of the problems were the same
among all the experimental conditions.
2. The total length of time that participants could spend on
solving the problem consciously was the same among all
the conditions.
3. The study included a control (no-incubation) group, and
participants in that group worked on the problem contin-
uously.
4. Participants problem-solving performance in pre- and
postincubation periods was measured.
5. The study reported information that allowed the compu-
tation of an effect size.
The first and second selection criteria ensured that tasks were
presented in an identical way among different conditions and that
any between-conditions performance differences could be attrib-
uted to differences in settings of the incubation period. The inclu-
sion of Criterion 3, a control condition (no break between the first
and the second attempts at the problem), is essential to provide a
baseline against which performance in incubation conditions can
be compared. Only publications that assessed the problem-solving
performance in both first and second attempts were included in the
analysis (Criterion 4). Therefore, some studies (e.g., Sio &
Rudowicz, 2007; Yaniv & Meyer, 1987) that did not assess postin-
cubation problem-solving performance were excluded. The infor-
mation required for computing effect sizes is discussed in the
section Estimation of Effect Sizes. Eight publications were ex-
cluded because the experimental studies contained within them
failed to meet one or more of the mentioned criteria. The specific
reasons for excluding the publications are described in Table 1,
 Table 1 98
Summary of Incubation Studies

Misleading cues Preparation Incubation Cues during Incubation effect Included in

Year Author Problem in problems period period Incubation task incubation period Other factors reported meta-analysis

1964 Mednick et al. RAT No 1 min Not specified Analogy vs. analogy Yes Yes, in analogy No, no control group
(Experiment 1) ! cues ! cues condition
1964 Mednick et al. RAT No 1 min Not specified Analogy vs. analogy Yes Cue relevance and Yes, in high-ability No, no effect size
(Experiment 2) ! cues correctness; problem- group in analogy information
solving ability ! cues condition
1967 Gall & RAT No 2 min 0, 25 min Nonverbal vs. free Yes No Yes, estimated from
Mendelsohn associates ! cues p " value
1968 Fulgosi & Consequences task No 2 min 0, 10, 20 min Number series (10 and 20 No Yes, in 20-min Yes, estimated from
Guilford min) condition p " value and
statement of
significance
1969 Dreistadt Farm & tree planting No 5 min 0, 8 min Guess card vs. guess card Yes In control condition, half Yes, with cues, farm Yes
problems ! cues the participants problem only
received cues
1969 Murray & Denny Saugstads ball problem No 5 min 0, 5 min Syllogisms, and tracing No Problem-solving ability Yes, in low-ability Yes
complex sequences of group
digits/numbers
1972 Fulgosi & Consequences Task No 2 min 0, 30, 60 min Number series (30 and 60 No Yes, in 30-min No, no effect size
Guilford min) condition information
1972 Dominowski & Hat rack problem No 5 min 0, 10 min Free association vs. Yes One group received cues No Yes, estimated from
Jenrick anagram throughout statement of
(Experiment 1) significance
1972 Dominowski & Hat rack problem No 3 min 0, 3 min Free association vs. No One group received No Yes, estimated from
Jenrick anagram cues; problem-solving statement of

15
(Experiment 2) ability significance
1972 Silveira Necklace problem No 3 min vs. 13 0, 30, Read ! free activity No Yes Yes
(Experiment 1) min 210 min
1972 Silveira Necklace problem No 13 min 0, 210 min Read ! free activity No Yes Yes
SIO AND ORMEROD

(Experiment 2)
1972 Silveira Consequences task No 2 min 210 min Read ! free activity No No No, no control group
(Experiment 3)
1974 Peterson Anagram No 0.33 min 1.8 min Other anagram No Task difficulty Yes Yes
1975 Bennett RAT No 1 min 10 min Hear music vs. mathematics No No No, no control group
problems
1976 Olton & Johnson Farm problem No 10 min 0, 15 min Rest vs. Stroop ! count Yes No Yes
backward vs. review
problem vs. lecture vs.
hear music
1979 Beck Verbal divergent- No 12 min 0, 20, 30 min Relax vs. write essay No Yes, with longer Yes
thinking task incubation period
1985 Brockett Brick task and RAT No 10 min (brick 0, 20 min Questionnaires No Yes Yes
task), 0.33
min (RAT)
1986 Patrick RAT No 2 min 0, 5 min Conversation vs. mental No Yes, in mental Yes, estimated from
rotation rotation condition p " value and
statement of
significance
1987 Yaniv & Meyer Rare-word definition No Not specified Not specified Questionnaire No Yes No, no postincubation
task performance
measure
1988 Browne & Cruse Farm problem No 20 min or 25 0, 5 min Hear music, draw graph, Yes No Yes
(Experiment 2) min memorize text
(table continues)
 Table 1 (continued )

Misleading cues Preparation Incubation Cues during Incubation effect Included in

Year Author Problem in problems period period Incubation task incubation period Other factors reported meta-analysis

1989 Smith & Rebus Misleading or 0.5 min 0, 5, 15 min Rest vs. music perception No Useful cues presented Yes, 15-min ! 5- Yes
Blankenship useful cues for second attempt min incubation
(Experiment 1) and control group
1989 Smith & Rebus Misleading or 0.5 min 0, 5, 15 min Rest vs. music perception No Yes, 15- and 5-min Yes
Blankenship useful cues incubation !
(Experiment 2) control group
1989 Smith & Rebus Misleading or 0.5 min 0, 10, 15 min Rebus with music No Yes, no difference Yes, estimated from
Blankenship useful cues perception, mathematics, among incubation p " value
(Experiment 3) or rest conditions
1989 Smith & Rebus Yes, misleading 0.5 min 0, 5 min Read vs. mathematics No Yes, in read story Yes, estimated from
Blankenship or useful condition p " value and
(Experiment 4) cues statement of
significance
1990 Dorfman Word puzzle task No 0.49 min 0, 5, 15 min Word problems vs. word No Yes, in both Yes
(Experiment 3) problems incubation
# cues conditions
1990 Dorfman Word puzzle task No 0.49 min 0, 3, 8, 13 Number series # cues No Yes, estimated from
(Experiment 4) min statement of
significance
1990 Kaplan Consequences task No 2 min 0, 30 min Psychometric test battery No Yes Yes
(Experiment 1)
1990 Kaplan Consequences task No 2 min 0, 32 min Mathematics # lecture No Yes Yes
(Experiment 2)
1990 Kaplan Consequences task No 2 min 0, 30 min Lecture No No Yes, estimated from
(Experiment 3) statement of
significance

16
1990 Kaplan Consequences task No 4.57 min 28.08 min Mathematics and insight No Yes Yes
(Experiment 4) problems
1991 Smith & RAT Yes 0.5 min 0, 5 min Read science fiction No Problem-solving ability Yes, with misleading Yes
Blankenship cues, greater for
(Experiment 1) low-ability
participants
1991 Smith & RAT Yes 1 min 0, 5 min Read science fiction No Problem-solving ability Yes, with misleading Yes
Blankenship cues
(Experiment 2)
INCUBATION AND PROBLEM SOLVING

1991 Smith & RAT Yes 0.5 min 0, 0.5, 2 min Free association No Problem-solving ability Yes, with misleading Yes, estimated from
Blankenship cues, less for low- p " value and
(Experiment 5) ability participants statement of
significance
1992 Goldman et al. Anagram No 0.25 min 0, 20, 1,440 General knowledge No Yes, in 1,440-min Yes
min # free activity condition
1992 Houtz & Frankel Life-relevant problem No 10 min 10 min Anagram task No Yes Yes
1997 Torrance-Perks Word completion Yes 0.5 min 0, 10 min Rest # lexical decision vs. Yes No Yes
(Experiment 1) problems rest # lexical decision #
cues
1997 Torrance-Perks Candle & radiation Yes, misleading 1 min (RAT 0, 8 min Memory (candle); read Yes No Yes
(Experiment 2) problems, RAT cues and candle (radiation); analogy
problem), (RAT)
5 min
(radiation
problem)
1998 Hansberry Riddles Yes 1 min 0, 15 min RAT No No Yes
(Experiment 2)
(table continues)
99
 Table 1 (continued )
100

Misleading cues Preparation Incubation Cues during Incubation effect Included in

Year Author Problem in problems period period Incubation task incubation period Other factors reported meta-analysis

1998 Hansberry RAT Yes 0.5 min 0, 10 min RAT No Problem-solving ability No Yes
(Experiment 3)
1999 Henley Anagram No 0.25 min 0, 1,440 min Free activity No No Yes, estimated from
(Experiment statement of
3.2) significance
1999 Henley Anagram No 0.93 min 0, 1,440 min Free activity No No Yes, estimated from
(Experiment 4) statement of
significance
1999 Jamieson RAT Yes 0.33 min 0, 5 min Mathematics No No Yes
(Experiment 1)
1999 Jamieson RAT Yes 0.33 min 0, 5 min Mathematics No No Yes
(Experiment 2)
2002 Dodds et al. RAT No 0.5 min 15 min Insight problem with: Yes Cues (answer, relevant Yes, when answer or No, no control group
(Experiment 1) Drawing vs. make a information, or related unrelated word
word test vs. make a word) presented during
word test ! cues incubation
2002 Dodds et al. RAT No 0.5 min 0, 15 min Insight problem with: Yes Cues (answer, relevant Yes, when answer Yes
(Experiment 2) Drawing vs. make a information, or related presented during
word test vs. make a word) incubation
word test ! cues
2002 Medd & Houtz Creative writing Yes 10 min 0, 10 min Unrelated writing task vs. No Prompt to think about Yes, if working on Yes, estimated from
related writing task problem during related task during p " value and
incubation incubation statement of
significance
2002 Moss RAT No 0.5 min 15 min Verbal reasoning task Yes Tasks described as Yes, interaction No, no control group
! cues solvable or between ability

17
unsolvable; problem- and problem type
solving ability
2003 Seabrook & Anagram No 0.25 min 7 min Word generation tasks Yes Cues (irrelevant or Yes, in relevant-cue Yes, estimated from
Dienes relevant) group p " value.
SIO AND ORMEROD

2004 Both et al. Anagram No 1.67 min 0, 6 min Letter search; questionnaire No Yes Yes
2004 Both et al. Anagram Yes 1.67 min 0, 6 min Letter and digit search; Yes Yes Yes
questionnaire
2004 Penney et al. Anagram No 5.75 min 15, 180 min Word completion No Yes, in 15-min No, no control group
(Experiment 1) ! cues vs. inactivity condition
2004 Penney et al. Anagram No 5.75 min 15, 1,440 Word completion No Yes, in 15-min No, no control group
(Experiment 2) min ! cues vs. inactivity condition
2004 Penney et al. Anagram No 5.75 min 0, 30, 120 Word completion No Yes Yes
(Experiment 3) min ! cues vs. inactivity
2004 Segal Insightful mathematic No 20 min 0, 4, 12 min Read vs. word puzzle No Yes, with word Yes
puzzle puzzle
2004 Snyder et al. Divergent-thinking task No 5 min 5 min Conversation No Unknown No
2005 Christensen & Insight puzzle No Not specified Not specified Analogy or distractor rating Yes Cues (analogous cue or Yes, in analogous No, no control group
Schunn task distractor cue) cue condition
2007 Sio & Rudowicz RAT Yes 1 min 0, 2 min Mental rotation task No Chess expertise; cues to Yes, experts more No, no postincubation
! mathematics task vs. chess expertise were sensitive to performance
listening to music related, neutral, or relevant concepts measure
misleading after incubation
2007 Vul & Pashler Anagram No 1 min 0, 5 min Video game No Task difficulty No Yes
(Experiment 1)
2007 Vul & Pashler RAT Yes 1 min 0, 5 min Video game No Task difficulty Yes, in the fixation Yes
(Experiment 2) group

Note. RAT # remote associates task.

 INCUBATION AND PROBLEM SOLVING
which also describes the settings of studies included in the meta-
analysis. Of the remaining 29 publications, 20 were refereed
journal articles, 8 were doctoral dissertations, and 1 was a confer-
ence article. The ratio of the refereed to other studies is 2.2:1,
which is within the suggested range of 128:1 to 1:1 for including
unpublished studies in an effort to avoid publication bias (Thorn-
ton & Lee, 2000). Most publications included multiple experi-
ments, thereby allowing a reasonable sample size of independent
studies (N ! 117) to be achieved.
Coding Procedure
Many of the experiments reported in the selected publications
had two or more experimental conditions, such as incubation
periods of different lengths or different types of task in the incu-
bation period. For the sake of the meta-analysis, experiments with
more than one incubation condition were broken down into inde-
pendent studies with one incubation condition and one control
condition. The same control group may be included in more than
one independent study and compared with more than one incuba-
tion condition. For example, in Goldman, Wolters, and Wino-
grads (1992) experiment, there were control, short incubation
period, and long incubation period conditions. The experiment was
decomposed into two studies, one consisting of the control and
short incubation period conditions and the other consisting of the
control and long incubation period conditions. To avoid inflating
the degrees of freedom available, we split the number of partici-
pants in the control condition across studies entered into the
analysis, a method advocated by Bar-Haim, Lamy, Pergamin,
Bakermans-Kranenburg, and van IJzendoorn (2007).
There were also studies that had more than one control condi-
tion. In such cases, the control condition that had the most similar
setting to the incubation condition was chosen. For example, in
Hansberrys (1998) third experiment, participants had to solve a
list of RATs under one of three conditions: two control and one
incubation. In one control condition, the RATs were presented
individually for 60 s. In the other control condition, as well as in
the incubation condition, each RAT was presented in two separate
30-s blocks. Data from the latter control condition were therefore
used in computing the effect size, because this control condition
and the incubation condition had the closest settings in terms of
RAT presentation.
After separating the experiments into independent studies, a
standard system was used to code each study. Background infor-
mation on each independent study (author, publication year, and
the number of participants in each condition), as well as potential
moderator variables, were extracted. Table 2 presents the coding
system used in this meta-analysis. Appendix B presents the infor-
mation we extracted from each independent study by using the
coding system.
Estimation of Effect Sizes
The effect size, Cohens d, was computed for each study entered
into the meta-analysis. Cohens d in this meta-analysis comprised
the difference in mean problem-solving performance scores be-
tween the control and incubation conditions divided by their
pooled standard deviation (Hedges & Olkin, 1985). In some cases,
effect sizes had to be calculated from t and F values, frequencies,
18
101
Table 2
Coding System
Variable Coding description
Author Author(s) of the study
Year Year the study was published
Total Total number of participants
Problem type 0 ! creative problem (e.g., consequences task)
1 ! visual problem (e.g., farm problem, radiation
problem)
2 ! linguistic problem (e.g., remote associates
task, anagram, rebus)
Misleading cues 0 ! no misleading cues
1 ! misleading cues embedded in the problem
Preparation period Amount of time spent on each problem before the
incubation period (in minutes)
Incubation period Length of the incubation period (in minutes)
Incubation task 0 ! rest
1 ! low cognitive demand task (e.g., drawing a
picture, reading)
2 ! high cognitive demand task (e.g., mental
rotation task, memory test)
Cues Presence of relevant cues during the incubation
period
0 ! no cue
1 ! yes
or p values. If a p-less-than value was given instead of an exact p
value, the p-less-than value was treated as an exact value, and an
estimate of Cohens d was generated. For studies that did not
include any of the abovementioned information but only provided
statements of nonsignificant differences between the control and
incubation groups, the Cohens d was assumed to be zero. Among
the studies that included multiple incubation conditions, some
provided a statement of nonsignificant performance differences
among the incubation conditions and reported only the overall
performance difference between the control and the incubation
conditions. In such cases, all incubation conditions were assumed
to have generated the same magnitude of incubation effect sizes.
Of the 117 effect sizes, 88 were extracted directly from the means
and standard deviations, t value, F value, frequencies, or p value;
8 were computed from a p-less-than value; and 21 were estimated
from statements of significance.
In some incubation studies, problem-solving performance was
assessed along more than one dimension. For example, in the study
carried out by Vul and Pashler (2007), participants performance
on RATs was measured in terms of the time spent on solving
RATs and the number of correct solutions. In such cases, we
computed a single effect size by averaging the effect size from
each measure (cf. Durlak & Lipsey, 1991).
Following Hedge and Olkins (1985) suggestion for removing
bias caused by small-sample studies, we computed an unbiased
effect size estimate by multiplying the effect size of each single
study by a factor of 13/[4(total N " 2) " 1], where total N is the
total number of participants of that study. Any unbiased effect size
larger than 2 SDs from the group mean was considered an outlier
and was recoded to the value of the effect size found at 2 SDs,
following a procedure for reducing the bias caused by extreme
effect sizes reported by Lipsey and Wilson (2001).
102 SIO AND ORMEROD
Heterogeneity Analysis
In this analysis, we predicted that the variance in magnitude of
the unbiased effect sizes among studies was not due simply to
sampling error but instead to the difference in settings of each
study (e.g., length of incubation period, nature of incubation task,
presence of cues). Therefore, analyses of the effect sizes should be
carried out under the assumption of a random-effects model. To
confirm the assumption of a heterogeneous distribution of effect
sizes, we carried out a heterogeneity test before running any
analyses on the effect sizes. The standard measure of heterogeneity
is Cochrans Q test. The Q statistic is the weighted1 sum of the
squared difference between the unbiased effect size estimate of
each independent study and the weighted average unbiased effect
size estimate across studies. Q is distributed as a chi-square sta-
tistic with k ! 1 degrees of freedom, where k is the number of
independent studies. If Cochrans Q test for heterogeneity is sta-
tistically significant (Q is larger than the chi-square value with k !
1 degrees of freedom), the assumption of the random-effects model
is supported.
Publication Bias
Prior to investigating the impact of potential moderators, we
undertook a preliminary analysis to assess if a publication bias
existed in the selection of studies despite the inclusion of unpub-
lished studies. A funnel plot of sample size against unbiased effect
size estimates was created. In the absence of any publication bias,
it is expected that the plot would be a funnel shape, such that the
amount of scatter about the mean effect size deceases with increas-
ing sample size. In addition to checking the presence of publication
bias qualitatively, we carried out a weighted least-squares linear
regression using the unbiased effect size estimates as the depen-
dent variable and the sample sizes weighted by the inverse of the
variance in a random-effects model, which is the sum of the
between-studies variance2 (random variance component) and
within-study variance of the unbiased effect size. The regression
slope (unstandardized coefficient of the predictive variable) would
be expected to approach zero if there is no publication bias
(Macaskill, Walter, & Irwing, 2001). The outcome of this analysis
is reported below.
Regression Model Testing
Due to the wide variation in experiment settings among incu-
bation studies, observed incubation effect size differences may
reflect the combined impact of different moderators. Hence,
weighted least-squares regression analyses were carried out to
reveal the true impact of each moderator on incubation effects. The
regression analyses were organized into two main sections. In the
first section, the incubation studies were first classified into dif-
ferent groups, in terms of the type of problem used, the cognitive
load of the incubation tasks, the presence of misleading cues, and
the presence of relevant cues during an incubation period. Within
each subgroup, the random variance component of the studies was
computed. A larger than zero random variance component implies
that the variability of effect sizes within these studies is not simply
due to subject-level sampling error (Lipsey & Wilson, 2001). A
weighted least-squares regression analysis was carried out as a
19
follow-up analysis to model the effect sizes. The unbiased effect
size estimate weighted by the inverse of the variance was the
outcome variable of the regression analysis. The predictor vari-
ables included problem type, misleading cues, (solution-relevant)
cues, incubation task, preparation period, and incubation period.
The categorical variables were represented with an appropriate
number of dummy-coded vectors. The categories (with the asso-
ciated predictor variable following in parentheses) of creative
problem (problem type), rest (incubation task), no misleading cues
(misleading cues), and no cue (cues) were used as reference groups
in the analysis, and their coefficients were restricted to zero.
The second section of the analyses investigated the general
impact of the moderators on the incubation effect sizes. In this
section, all the incubation studies were grouped together, and a
weighted least-squares regression analysis was carried out to in-
vestigate the general impact of each moderator. Again, the
weighted unbiased effect size was the outcome variable of the
regression analysis. The predictor variables were problem type,
misleading cues, cues, incubation task, preparation period, and
incubation period. Another weighted least-squares regression was
carried out to examine the interaction between the categorical
variables problem type and incubation task. The predictor vari-
ables of the regression model were the appropriate number of
dummy-coded vectors and the multiplicative terms of these two
categorical variables, as well as the variables misleading cues,
preparation period, incubation period, and cues. A more detailed
description of the selection of the dummy-coded vectors and the
multiplicative terms of problem type and incubation task is pre-
sented in Appendix C.
Results
There were 117 studies included in this meta-analysis. The total
number of participants was 3,606, and the median number of
participants per study was 25. An unbiased effect size estimate was
computed for each independent study. Among these studies, 85
reported positive effect sizes. The unbiased effect size estimates
ranged from !0.71 to 4.07, and the median effect size was 0.26.
The unweighted mean of the unbiased effect size estimate was
0.41, with a standard deviation of 0.71. The upper and lower
bounds of the 95% confidence interval were 0.54 and 0.28. Unbi-
ased effect sizes larger than 2 SDs from the mean were recoded to
1
The weighting was the inverse of the within-study variance of the
effect estimate, and the formula for the within-study variance was [(2 "
square root of total N) # (N of experimental " N of control " square root
of the unbiased effect size)]/(2 " total N " N of experimental " N of
control), where N is the number of participants in that condition (Cooper &
Hedges, 1994). Three studies were excluded when computing the weighted
average unbiased effect size estimate and the Cochrans Q value because
they had a within-subjects design, and thus all participants were involved
in both control and incubation conditions. Thus, the weighting formula
could not apply to them.
2
The between-studies variance was equal to [Q ! (k ! 1)]/c, where Q
is the Cochrans Q value and k is the number of studies. The formula for
c was [(the sum of the inverse of the within-study variance) (the sum of
the square of the inverse of the within-study variance)]/(the sum of the
inverse of the within-study variance) and was suggested by Cooper and
Hedges (1994).
 INCUBATION AND PROBLEM SOLVING 103

the value of the effect size found at 2 SDs. Table 3 gives the 300
stem-and-leaf display showing the distribution of the unbiased
effect sizes. The unweighted mean of the adjusted unbiased effect 250
size estimate was 0.36, with a standard deviation of 0.51. The
200

Sample Size
upper and the lower bounds of the 95% confidence interval were
0.26 and 0.45. The confidence interval does not include zero, 150
implying that the estimate of the mean unbiased effect size is
significantly larger than zero. 100
The heterogeneity statistic, Cochrans Q, was 173.99, and was
significantly larger than the chi-square critical value, df ! 113, 50
p " .001. This supports the use of the random-effects model. The
0
variance of each unbiased effect size in the random-effects model
was the sum of the between-studies variance and within-study -1 -0.5 0 0.5 1 1.5 2
variance of the unbiased effect size. The between-studies variance, Unbiased Effect Size
also called the random variance component, among these incuba-
tion studies was .0834. The mean of the weighted unbiased effect Figure 1. Funnel plot of the studies included in this meta-analysis. The
dotted line indicates the mean unbiased effect size.
size was 0.29, with a standard deviation3 of 0.04, and the 95%
confidence interval was 0.21, 0.39. The nonzero confidence inter-
val implies that the weighted mean is significantly larger than zero.
suggesting the absence of publication bias. Thus, no correction has
This answers our first question, showing the existence of a positive
been made for publication bias.
incubation effect.
Table 4 presents the weighted mean, standard deviation, 95%
Figure 1 presents the funnel plot of sample size against the
confidence interval, and random variance component in each sub-
estimated unbiased effect size of each study in the meta-analysis.
group of each categorical moderator. Six of the subgroups (lin-
We carried out a weighted least-squares regression using unbiased
guistic problems, creative problems, absence of misleading cues,
effect sizes weighted by the inverse of the variance as the depen-
absence of relevant cues, high cognitive load task, unoccupied
dent variable and sample size as the predictive variable. The
incubation period) had larger than zero random variance compo-
regression coefficient of the predictive variable was not signifi-
nents. New weightings, under the assumption of a random-effects
cantly different from zero (standardized # ! $.08, p ! .41), model, were computed for each of the subgroups. Weighted least-
squares regression analyses were carried out to find the moderators
that accounted for the effect size variability among these sub-
Table 3 groups. Small numbers of studies using creative tasks and studies
Stem-and-Leaf Display of 117 Unbiased Effect Sizes having unoccupied incubation periods preclude the possibility of
regression analyses with these moderators. An effect of applying a
Stem Leaf weighting to this regression analysis is to underestimate the orig-
inal standard error of each unstandardized coefficient. Thus, we
$0.7 1
$0.5 9, 8, 8
computed an adjusted standard error by dividing the original
$0.4 1 standard error by the square root of the mean square residual, a
$0.3 8, 5, 3 procedure suggested by Lipsey and Wilson (2001). The corrected
$0.2 standard error was used in the significance test (z test) of each
$0.1 8, 7, 4, 4, 0 unstandardized coefficient.
$0.0 3
0.0 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, Tables 5,6,7, and 8 present a summary of the regression analysis
0, 0, 0, 3, 4, 5, 5, 6, 6, 7, 9, 9 results of each subgroup. With the subgroup of studies using
0.1 0, 0, 0, 0, 0, 1, 1, 3, 4, 4, 4, 5, 7, 7 linguistic problems, low cognitive load tasks generated larger
0.2 1, 4, 5, 8, 8 incubation effects than did rest alone (# ! .54, p " .05). Also,
0.3 1, 4, 6, 6, 7
0.4 0, 0, 1, 2, 2, 4, 5, 7, 7, 8, 8, 9
there was an interaction between problem type and incubation task
0.5 0, 2, 2, 6, 6, 6, 6, 6, 6, 7 with this subgroup, such that that low cognitive load tasks facili-
0.6 2, 2, 4, 5, 6, 6, 6 tated the incubation effect only when solving linguistic problems.
0.7 0, 1, 2, 4, 4 With the absence of misleading cue and absence of relevant cue
0.8 6 subgroups, regression analyses revealed that, in the absence of
0.9 9, 9
1.0 5, 5, 7, 7, 8, 9 these cues, individuals solving visual problems had smaller incu-
1.1 7 bation effects than did those solving creative and linguistic prob-
1.2 lems. There was also a positive impact of longer preparation periods
1.3 on the incubation effect sizes. The cognitive load of the incubation
1.4
1.5
1.6 8 3
1.7 0 The standard deviation of the weighted mean, also known as the
1.8 2, 2, 2, 2 standard error, was calculated as the square root of 1/%wi, and the 95%
confidence interval was calculated as the weighted mean &1.96 times the
Note. Outliers were recoded. standard deviation of the weighted mean (Hedges & Olkin, 1985).

20
104 SIO AND ORMEROD

Table 4
The Random Variance Component, Weighted Mean, Standard Deviation, Standard Error, and 95% Confidence Interval of the Effect
Size Estimate by Each Categorical Moderator

Problem type Misleading cues Incubation task Relevant cues

Moderator Linguistic Verbal Creativea Yes Noa High load Low load Resta Yes Noa

Number of studies 65 35 14 29 85 76 22 16 32 82
Random variance
component .00281 0 .37418 0 .12004 .06478 0 .30000 0 .10409
Mean 0.22 0.26 0.29 0.35 0.32 0.24 0.52 0.46 0.24 0.34
Standard deviation 0.05 0.08 0.14 0.09 0.06 0.05 0.10 0.18 0.09 0.06
95% confidence interval 0.32, 0.13 0.41, 0.10 0.55, 0.02 0.53, 0.17 0.43, 0.20 0.35, 0.13 0.72, 0.32 0.82, 0.11 0.41, 0.07 0.45, 0.22
Mean comparison with
the reference group t(77) ! "0.43 t(47) ! "0.19 t(112) ! 0.30 t(90) ! 1.17 t(36) ! 0.28 t(112) ! 0.90

Note. The lower confidence intervals of all the weighted unbiased effects are larger than zero, suggesting that the mean is significantly larger than zero, p % .05.
a
The reference groups in mean comparisons.

tasks did not have any impact on the magnitude of the incubation
effects in these two subgroups. Note, however, that the presence or
absence of these effects with these specific subgroups does not
necessarily imply converse effects in other subgroups, hence a
shift to analysis of general impacts of each moderator.
In the second stage of the analysis, a weighted least-squares
regression analysis was carried out to look at the general impact of
each moderator on the incubation effect sizes. A summary of the
regression analysis results is presented in Table 9. The negative
coefficients associated with visual problems and linguistic prob-
lems indicate that individuals solving these two types of insight
problem showed a smaller incubation effect than did individuals
solving creative problems. A z test was carried out to compare the
coefficients of the visual problem and linguistic problem groups.
The result was not statistically significant (z ! "1.25, p # .05),
suggesting the magnitude of the incubation effect for visual and
linguistic insight problems was comparable.
The length of the preparation period was found to have a
significant impact on the magnitude of the incubation effect ($ !
.03, p % .05). Three bivariate correlations were carried out to
check for positive relationships between the length of the prepa-
ration period and the magnitude of the weighted incubation effect
when solving the three types of problem. There was a statistically
significant positive correlation between the weighted incuba-
tion effect size and the length of the preparation period with
visual problems, r(35) ! .40, p ! .02, and creative problems,
r(14) ! .60, p ! .03, but not with linguistic problems, r(65) !
".04, p ! .75.
We carried out another weighted least-squares regression anal-
ysis to examine the interaction between problem type and incuba-
tion task, using misleading cues, preparation period, incubation
period, cues, and an appropriate number of dummy vectors and the
multiplicative terms of the variables problem type and incubation
task as predictor variables. Table 10 presents the results of this

Table 5
Regression Model for Linguistic Problem Studies (N ! 65)

Variable Sum of squares df Mean square F p value

ANOVA significance test

Model 13.42 5 2.68 3.15! .01
Residual 5.24 59 0.85
Total 63.66 64

Summary of the regression model

Unstandardized $ Standardized $ Corrected SE$ z score

Incubation task
High cognitive load task .06 .06 .21 0.27
Low cognitive load task .54! .43 .25 2.15
Misleading cues .15 .17 .12 1.24
Ratio of the preparation period to the
incubation period .15 .12 .19 0.80
Relevant cue ".04 ".04 .13 "0.30

Note. Random variance component !.00281, R2 ! .21. Another regression with the same predicting variables,
except replacing the variable ratio of length of the preparation to the incubation period by the variables
incubation period and preparation period, was carried out. The pattern of the results was comparable, but it had
lower explanatory power, and a lower significance level, R2 ! .20, F(6, 58) ! 2.47, p ! .034. Neither the
variable preparation period nor incubation period was significant. ANOVA ! analysis of variance.
!
p % .05.

21
 INCUBATION AND PROBLEM SOLVING 105

Table 6
Regression Model for Misleading Cue Studies (N " 85)

Variable Sum of squares df Mean square F p value

ANOVA significance test

Model 2.25 7 2.89 3.56! .002
Residual 62.62 77 0.81
Total 82.87 84

Summary of the regression model

Unstandardized $ Standardized $ Corrected SE$ z score
Problem type
Visual problem !.59! !.51 .19 !3.02
Linguistic problem !.31 !.29 .18 !1.74
Incubation task
High cognitive load task !.18 !.14 .22 !0.80
Low cognitive load task .06 .04 .26 0.23
Length of the incubation period #.001 .055 #.001 0.45
Length of the preparation period .03! .36 .01 2.16
Relevant cue !.03 !.02 .16 !0.20

Note. Random variance component ".12004, R2 " .24. ANOVA " analysis of variance.
!
p # .05.

regression analysis. To examine the interaction effects, we exam-
ined the coefficient differences between their multiplicative terms
to see if they were significantly larger than zero by using z tests.
The details of equations for computing the coefficient difference
can be found in Appendix C. Table 11 presents the coefficient
differences between multiplicative terms.
With creative problems, undertaking high cognitive load tasks
was associated with smaller incubation effects than with low
cognitive load tasks or rest during the incubation period (CjHj !
CjRj " !.91, p # .05; CjHj ! CjLj " !.79, p " .08, where C
refers to creative problems, H to high cognitive demand tasks, R to
rest, and L to low cognitive demand tasks, all in study j). When
solving linguistic and visual problems, no differences were found
for incubation periods filled with low or high cognitive load tasks
or rest. However, this regression model has a problem in exploring
the interaction between problem type and incubation task. Among
studies examining the role of an incubation period with creative
problems, 12 out of 14 employed high cognitive load tasks, and the
remaining 2 studies employed rest during the incubation period.
This unbalanced distribution may cause bias when examining the
interaction between incubation task and problem type. Hence,
another regression analysis was carried out that excluded studies
using creative problems. This third regression model included the
variables misleading cues, preparation period, cues, the dummy
variables of problem type (excluding creative problem) and incu-
bation task, and their multiplicative variables. Appendix D pre-

Table 7
Regression Model for No Relevant Cue Studies (N " 82)

Variable Sum of squares df Mean square F p value

ANOVA significance test

Model 29.34 6 4.89 7.35! #.001
Residual 49.94 75 0.67
Total 79.28 81

Summary of the regression model

Unstandardized $ Standardized $ Corrected SE$ z score
Problem type
Visual problem !.79!! !.66 .19 !4.10
Linguistic problem !.22 !.21 .18 1.24
Incubation task
High cognitive load task !.25 !.22 .21 !1.20
Low cognitive load task .16 .12 .25 0.64
Length of the incubation period #.001 .02 0 0.13
Length of the preparation period .05!! .51 .02 3.01
Misleading cue .09 .07 .18 0.53

Note. Random variance component ".10409, R2 " .37. ANOVA " analysis of variance.
!
p # .05. !! p # .001.

22
106 SIO AND ORMEROD

Table 8
Regression Model for High Cognitive Load Incubation Task Studies (N " 75)

Variable Sum of squares df Mean square F p value

ANOVA significance test

Model 7.36 6 1.47 1.53 .19
Residual 66.19 69 0.96
Total 73.55 74

Summary of the regression model

Unstandardized # Standardized # Corrected SE# z score
Problem type
Visual problem !.41! !.34 .20 !2.01
Linguistic problem !.28 !.30 .17 !1.58
Length of the incubation period !.01 !.12 .01 !0.87
Length of the preparation period .02 .21 .02 1.31
Misleading cue .08 .06 .16 0.51
Relevant cue .09 .08 .14 0.62

Note. Random variance component ".06478, R2 " .10. ANOVA " analysis of variance.
!
p $ .05.

sents a detailed description of the selection of the dummy-coded
vectors and the multiplicative terms. The results of this analysis are
presented in Table 12, and the coefficient differences between the
multiplicative terms are presented in Table 13.
The regression results indicated an interaction between prob-
lem type and incubation task. When solving linguistic prob-
lems, a low cognitive load task generated significantly larger
incubation effects than did rest (LijLj ! LijRj " .45, p " .05,
where Li refers to linguistic problems, L to low cognitive
demand tasks, and R to rest, all in study j). The difference
between low and high cognitive loads was in the same direction
but did not reach significance, and there was no significant
difference between the rest condition and the high cognitive
load condition. When solving visual problems, the effect sizes
among the three incubation conditions were comparable. This
pattern of findings is consistent with the previous regression
analysis results. In addition, the exclusion of creative problem
studies makes the positive impact of a low cognitive load
incubation period on linguistic problems more significant.
The model using the variables preparation period and incubation
period was found to be not significant, and the variable preparation
period was not significant in the analysis. This may be due to the
decrease in the number of studies included in the current regres-
sion model. Moreover, as mentioned above, a positive association
between the length of preparation period and incubation effect
size was found when solving visual and creative problems. Thus,
the exclusion of creative problem studies appears to decrease the
significance of the variable preparation period and the significance

Table 9
Regression Model for All Studies (N " 114)

Variable Sum of squares df Mean square F p value

ANOVA significance test

Model 25.56 8 3.20 4.08 $.001
Residual 82.22 105 0.78
Total 107.79 113

Summary of the regression model

Unstandardized # Standardized # Corrected SE# z score
Problem type
Visual problem !.60!! !.52 .15 !3.61
Linguistic problem !.31! !.31 .14 !1.96
Misleading cues .16 .13 .12 1.16
Incubation task
Low cognitive load .14 .10 .18 0.72
High cognitive load !.16 !.15 .14 !1.02
Preparation period .03! .35 .01 2.29
Incubation period $.001 .04 $.001 $0.001
Cue presented !.01 !.01 .11 !0.09

Note. Random variance component " .0834, R2 " .24. ANOVA " analysis of variance.
!
p $ .05. !! p $ .001.

23
 INCUBATION AND PROBLEM SOLVING 107

Table 10
Regression Model for All StudiesInteractions (N ! 114)

Variable Sum of squares df Mean square F p value

ANOVA significance test

Model 35.60 12 2.97 4.15 ".001
Residual 72.18 101 0.72
Total 107.78 113

Summary of the regression model

Unstandardized # Standardized # Corrected SE# z score
Problem Type $ Incubation Task
Visual Problem $ Low
Cognitive Load Task 0.17 0.08 0.35 0.50
Visual Problem $ High
Cognitive Load Task 0.90 !
0.61 0.35 2.60
Linguistic Problem $ Low
Cognitive Load Task 0.72 0.43 0.38 1.93
Linguistic Problem $ High
Cognitive Load Task 1.00 !
0.97 0.38 2.59
Problem type
Visual problem %1.29!! %1.10 0.32 %4.00
Linguistic problem %1.24!! %1.21 0.40 %3.14
Incubation task
High cognitive load task %0.91!! %0.81 0.30 %3.00
Low cognitive load task %0.12 %0.09 0.33 %0.38
Misleading cues 0.24 0.19 0.15 1.54
Length of the incubation period "0.001 0.018 "0.001 "0.001
Length of the preparation period 0.03! 0.01 0.01 1.97
Answer was presented 0.03 0.03 0.13 0.26

Note. Random variance component ! .08340, R2 ! .33. ANOVA ! analysis of variance.

!
p " .05. !! p " .001.

level of the regression model that used preparation period as one of
the predictor variables.
Discussion
The meta-analysis supports the existence of incubation effects
and also identifies some potential moderators, including the prob-
lem type, length of the preparation period, and the incubation task.
Individuals solving creative problems were more likely to benefit
from an incubation period than individuals solving linguistic and
visual problems. Longer preparation periods gave rise to larger
incubation effects. When solving linguistic problems, a low cog-
nitive load task gave the strongest incubation effects.
We suggest that the positive incubation effects found with
creative problems are a direct reflection of their multiple-solution
nature. When solving a creative problem, individuals benefit from
performing a wide search of their knowledge to identify as many
relevant connections as possible with the presented stimuli. Each
time individuals reapproach the problem, they improve their per-
formance by extending the search to previously unexplored areas
of their knowledge network. Incubation appears to facilitate the
widening of search of a knowledge network in this fashion.
Linguistic and visual problems typically have only one possible
solution. In order to solve them, individuals have to explore their
memory or environment to look for specific relevant knowledge or
to adapt a specific strategy. Widening search to new items of
knowledge may not be facilitative if the solution to a problem lies
within already activated knowledge that is currently represented
inappropriately. Under this account, incubation supports knowl-
edge activation, but it does not support restructuring.
Another finding of the meta-analysis was the beneficial effect of
an incubation period filled with low demand tasks on solving
linguistic problems. A positive effect of a filled incubation period
on problem solving compared with rest during the incubation
period undermines the conscious-work hypothesis that incubation
effects are due to the mental fatigue reduction (Posner, 1973).
There remains a possibility, of course, that a sufficiently light load
might allow additional covert problem solving compared with a
heavier task load (Browne & Cruse, 1988; Posner, 1973), but this
does not explain why a light load should be more facilitative than
rest alone.
The positive effect of a light cognitive load may indicate com-
petition between controlled and automatic processes in solving
linguistic problems. It has been suggested with remote associates
task (RAT) performance that only strong (and in this context,
incorrect) associates are accessed when individuals focus their
attention on seeking solutions, whereas remote associates are more
likely to be accessed when an individuals cognitive resources are
allocated in a diffuse manner (Ansburg & Hill, 2003; Finke, Ward,
& Smith, 1992; Martindale, 1995). During an incubation period,
low demand tasks may occupy part of the problem solvers atten-
tion, preventing the focused concentration that yields strong asso-
ciates. Resting during an incubation period may allow individuals
to continue consciously working on the problem, whereas perform-
ing high demand tasks may shift attention entirely to that interpo-

24
108 SIO AND ORMEROD

Table 11 and Brooks (1993), in which the act of verbalizing can impair
Summary of Coefficient Differences of the Original Regression performance by focusing individuals attention inappropriately on
Model verbalizable components of a task. The suggested role of light-load
incubation tasks receives indirect support from recent findings that
Multiplicative terms Coefficient difference Corrected SE z score show that visual search can be more efficient when performed
LijHj LijRj 0.08 .49 0.17 concurrently with an unrelated task than when performed alone
LijLj LijRj 0.60 .50 1.21 (Smilek, Enns, Eastwood, & Merikle, 2006). They suggest that the
LijLj LijHj 0.52 .70 0.74 dual-task condition prevents a narrow attentional focus in search-
VjHj VjRj !0.01 .46 !0.03 ing for stimuli.
VjLj VjRj 0.05 .48 0.10
VjLj VjHj 0.06 .66 0.09 With visual problems, the magnitude of the incubation effect
CjHj CjRj !0.91! .30 !3.00 was independent of the setting of an incubation period (filled or
CjLj CjRj !0.12 .33 !0.38 unfilled). Differences between visual and linguistic insight prob-
CjLj CjjHj 0.79 .45 1.77 lems may arise through a greater reliance on strategic search rather
VjHj CjHj !0.39 .47 !0.82
LijHj CjHj !0.25 .55 !0.45
than knowledge activation in the former than the latter. MacGregor
LijHj VjHj 0.14 .72 0.19 et al. (2001) proposed that in solving the nine-dot problem, indi-
VjLj CjLj !1.11! .47 !2.36 viduals select and execute moves that maximally reduce the dis-
LijLj CjLj !0.52 .55 !0.95 tance between current and goal states, essentially drawing lines
LijLj VjLj 0.60 .72 0.82
that connect as many dots as possible. While there remain moves
VjRj CjRj !1.29!! .32 !4.00
LijRj CjRj !1.24!! .40 !3.14 available that satisfy a criterion of satisfactory progress (in this
LijRj VjjRj 0.04 .51 0.08 case, the ratio of dots cancelled to lines available), individuals will
persevere with an initial representation of the problem that, in the
Note. Li " linguistic problem; H " high cognitive demand task; R " case of the nine-dot problem, does not include consideration of
rest; L " low cognitive demand task; V " visual problem; C " creative
problem; j " study j. space outside the dot array. According to MacGregor et al.s
!
p # .05. !! p # .001. p " .08. account, individuals must experience a failure to find moves that
meet a criterion of satisfactory progress before they change their
initial representation of the problem, thereby including space out-
lated task, leading to a narrow rather than diffused attentional side the dot array in their attempts. An incubation period would be
focus. The impact of performing high demand tasks is analogous helpful only if the problem solvers became aware of the necessity
to the verbal overshadowing effect reported by Schooler, Ohlsson, of a strategy shift, but according to MacGregor et al. they are

Table 12
Regression Model for Studies Excluding Creative Problem (N " 100)

Variable Sum of squares df Mean square F p value

ANOVA significance test

Model 15.04 8 1.88 2.04! .05
Residual 83.92 91 0.92
Total 98.96 99

Summary of the regression model

Unstandardized $ Standardized $ Corrected SE$ z score
Problem Type % Incubation Task
Visual Problem % Low
Cognitive Load Task !.21 !.12 .28 !0.77
Visual Problem % High
Cognitive Load Task .09 .08 .24 0.38
Problem type
Visual problem !.15 !.15 .21 !0.68
Incubation task
High cognitive load task !.01 !.01 .18 !0.05
Low cognitive load task .45! .41 .20 2.27
Misleading cues .13 .13 .11 1.10
Length of the preparation period/
incubation .08 .20 .05 1.56
Answer was presented .01 .01 .11 0.13

Note. Random variance component " .00031, R2 " .15. Another regression was carried out with the same
variables, replacing the variable ratio of length of preparation to incubation period with incubation period and
preparation period. The regression model was not significant, F(9, 90) " 1.27, p " .10. Neither the variable
preparation period nor incubation period was significant. ANOVA " analysis of variance.
!
p # .05.

25
 INCUBATION AND PROBLEM SOLVING 109

Table 13 larger incubation effect, but an opposite pattern of results was

Summary of Coefficient Differences of the Original Regression found in studies employing visual problems, suggesting that the
Model impact of misleading cues may be modality specific. In order to
test whether the presence of misleading cues affects incubation
Coefficient Corrected with linguistic problems alone, we ran a weighted one-way anal-
Multiplicative terms difference SE z score
ysis of variance to compare the incubation effect sizes of studies
LijHj LijRj ".01 .17 "0.05 using linguistic problems that included misleading cues (25 stud-
LijLj LijRj .45! .20 2.27 ies) against studies using linguistic problems that did not include
LijLj LijHj .46 .26 1.76
misleading cues (40 studies), F(1, 65) ! 3.04, MSe ! 3.00, p !
VjHj VjRj .08 .30 0.28
VjLj VjRj .24 .34 0.69 .08. Thus, where problem materials involve linguistic stimuli,
VjLjj VjHj .15 .45 0.34 getting rid of misleading concepts may be the key to task solution,
VjHj LijHj ".06 .32 "0.17 in contrast with visual problems in which the key may be to
VjLj LijLj ".36 .35 "1.03 restructure the knowledge that is currently active. The effect of
VjRj LijRj ".15 .21 "0.68
misleading cues also offers some support for the selective-
Note. Li ! linguistic problem; H ! high cognitive demand task; R ! forgetting hypothesis, but this effect may be task specific.
rest; L ! low cognitive demand task; V ! visual problem; j ! study j. In contrast with previous reports (e.g., Dominowski & Jenrick,
!
p # .05. 1972; Dreistadt, 1969; Mednick, Mednick, & Mednick, 1964), the
moderator presence of solution-relevant cues was not found to be
a significant predictor of incubation effects. It has been hypothe-
unlikely to do so unless they encounter criterion failure as a result sized that, during an incubation period, unconscious processes
of reaching impasse. Seifert et al. (1995) offered an alternative such as spreading activation sensitize individuals to solution con-
account that also points to the criticality of experiencing failure cepts and make them more likely to utilize externally presented
and impasse for eventual success in insight problem solving. cues. To examine this hypothesis, researchers have presented the
If the hypothesis that visual problems require impasse for a answers of unsolved problems during an incubation period and
strategy switch to occur is correct, a long preparation period (i.e.,
compared their postincubation performance with that of partici-
preincubation problem solving) should be more likely to yield
pants not receiving any cues during the incubation period (e.g.,
benefits from subsequent incubation with visual problems because
Dominowski & Jenrick, 1972; Dodds, Smith, & Ward, 2002).
it allows individuals to reach impasse prior to incubation. The
Findings of these studies are equivocal, but it does appear that
results of the regression analysis and the follow-up bivariate cor-
problem solvers do not always make use of solution-relevant cues,
relations are consistent with this prediction, showing a statistically
even when the cue includes the solution itself. For example,
significant positive correlation between the incubation effect size
Chronicle, Ormerod, and MacGregor (2001) found that presenting
and the length of the preparation period with visual problems.
the nine-dot problem with a shaded background in the shape of the
A positive correlation between the length of the preparation
solution did not lead to significant levels of facilitation, even when
period and incubation effect size was also found with creative
the relevance of the shading was drawn to participants attention.
problems. A long preparation period may allow individuals to
exhaust search in one domain, making it more likely for them to However, because of the small number of studies that presented
explore a new domain in the second phase of solving. A positive solution-relevant cues (three with linguistic, seven with visual
correlation was not found when solving linguistic problems, problems, none with creative problems) and the wide variation in
though this may simply reflect the small variability in length of experimental parameters among these studies, it is impossible to
preparation period among studies using linguistic problems (the carry out further statistical analysis.
preparation period of 82% of these studies ranged from 0.5 In summary, the meta-analysis results support the existence of
to 1 min). incubation effects, though there appears to be a range of effects
The meta-analysis reveals that embedding misleading cues in specific to particular tasks and performance conditions. When
the problems was not a significant predictor overall. This result is attempting to solve creative problems that require a wide search of
in contrast with previous reports (e.g., Smith & Blankenship, 1989, knowledge, individuals benefit from an incubation period. Prob-
1991), which suggested that incubation effects arise through for- lems that involve reaching some kind of insight into a unique
getting of inappropriate information. The true effect of misleading solution do not always benefit from incubation under all condi-
cues is underestimated in our regression analyses, as we examined tions. In the case of linguistic problems, such as the RAT, there is
only the overall effect of misleading cues on problem solving in a modest incubation effect but only where the incubation period is
general. There were 29 independent studies included in this meta- filled with a low cognitive demand task. One possible explanation
analysis that examined the impact of misleading cues; 25 of them is that performing low cognitive demand incubation tasks allows
examined the impact on linguistic problems, the rest on visual the occurrence of some unconscious problem-solving processes,
problems. The weighted means of the effect size estimates of these such as spreading activation and selective forgetting. In the case of
studies for each problem type by the presence of misleading cues visual problems, incubation effects arise only where there has been
were as follows: linguistic with misleading cue: M ! 0.36, SD ! a sufficiently long preparation period prior to incubation for the
0.09; linguistic without misleading cue: M ! 0.17, SD ! 0.06; problem solver to have entered a state of impasse. Only under
visual with misleading cue: M ! 0.18, SD ! 0.39; visual without these conditions can an incubation period contribute to the strate-
misleading cue: M ! 0.26, SD ! 0.08. For studies employing gic shift needed to restructure a problem representation. Thus, the
linguistic problems, the presence of misleading cues induced a theoretical positions of spreading activation, selective forgetting,

26
110 SIO AND ORMEROD
and restructuring each receive support. However, evidence for
each appears to be specific to particular problem types.
Spreading activation and strategic search are basic mechanisms
underlying different types of general cognitive process. For in-
stance, the adaptive control of thoughtrational (ACT-R) com-
putational framework (Anderson, 1994) utilizes both these mech-
anisms. Different researchers have successfully adapted this model
to simulate a wide range of noninsight problem-solving processes.
Examining the occurrence of incubation effects in terms of fun-
damental cognitive processes would offer important findings for
developing a computational model of insight problem solving.
Although the conscious-work hypothesis receives little support
here, the meta-analysis leaves open the possibility that uncon-
scious processes may reflect forgetting, activation of new knowl-
edge, or restructuring. Further experimental studies might focus on
comparing the occurrence of these different unconscious processes
during an incubation period in different experimental settings. One
methodology that might allow such comparisons was employed by
Sio and Rudowicz (2007), who examined the occurrence of
spreading activation by measuring individuals sensitivity to an-
swers of the unsolved RATs before and after filled and unfilled
incubation periods. They found that the spreading-activation pro-
cess occurred only in the filled incubation period condition and in
a fixated mind, though this study did not measure postincubation
period performance.
Given the outcome that both spreading activation and restruc-
turing might arise from an incubation period, depending on the
task, then proponents of both the current views of insight problem
solving may take some comfort from the results. The evidence
from the linguistic problems is consistent with release from inap-
propriate constraints (Knoblich et al., 1999), whereas evidence
from the visual problems is consistent with restructuring that
results from a strategic shift following impasse (MacGregor et al.,
2001). Given that these theoretical alternatives have to date been
explored only with different task sets, it seems quite possible that
a complete account of insight might need both theoretical compo-
nents. Such a view may well be consistent with the account of
insight offered by Kershaw and Ohlsson (2004), who proposed a
multiple-source account of the difficulties individuals encounter in
insight problem solving.
The finding of a positive impact of an incubation period on
solving creative thinking problems supports the contention that
incubation periods help the elicitation of new ideas. Incubation is
a concept central to many methodologies for encouraging creative
decision making, especially among management science and busi-
ness communities (e.g., Rickards, 1991), and this result may be
taken as supporting the inclusion of an incubation phase in such
methodologies.
It should be noted that, despite efforts to include studies from
sources other than peer-reviewed journals, the meta-analysis may
be influenced by a bias in favor of reporting significant effects at
the expense of null effects. Thus, incubation effects may be to
some extent overstated in this meta-analysis, a problem common to
all meta-analyses. Nonetheless, the reasonably large effect sizes
found with creative problems indicates that, with this class of
problem at any rate, incubation is a potentially valuable mecha-
nism for fostering creative thought.
Clearly, the empirical data on incubation are not straightfor-
ward. As a consequence, the traditional narrative review approach
27
is not amenable to drawing strong cross-study conclusions. It is
perhaps disappointing that relatively few published studies met the
necessary criteria for inclusion in the meta-analysis, since a failure
either to measure postincubation performance or to provide effect
size information limits the conclusions that can be drawn from
them. Nonetheless, sufficient studies remain for the meta-analysis
to be undertaken and to reveal some intriguing results.
One remaining problem is the relatively narrow range of prob-
lem types that have been explored. For instance, the majority of
studies that explored incubation effects with linguistic problems,
which is the majority of studies overall, used the RAT. It is unclear
whether the RAT can be considered an insight problem or a
linguistic completion task, suggesting it may not be representative
of all linguistic problem-solving tasks. Bowden and Jung-Beeman
(2003a) have found that participants claimed that they solved
RATs sometimes with insight and sometimes without insight.
Further studies should aim to explore task-specific experimental
settings for maximizing the incubation effect with a wider range of
tasks. A further research issue of value might also be to explore
individual differences in incubation effects. For example, if the
role of incubation is to encourage diffused attention, then individ-
uals who show a propensity toward allocating attention broadly
(e.g., as measured via field dependence) may benefit differentially
from an incubation period. Also, studies have revealed that strat-
egy switching is related to working memory capacity (Geary,
Hoard, Byrd-Craven, & DeSoto, 2004). Thus, memory capacity
may also interact with incubation effects in solving visual prob-
lems.
One the whole, the results of this meta-analysis support the
existence of multiple types of problem-specific incubation effect.
We suggest that the concept of incubation can be understood only
through a close examination of the problems to which it is applied
and the conditions under which it is elicited.
References
References marked with an asterisk indicate studies included in the
meta-analysis.
Anderson, R. J. (1994). Representations and requirements: The value of
ethnography in system design. HumanComputer Interaction, 9, 151
182.
Ansburg, P. I., & Hill, K. (2003). Creative and analytic thinkers differ in
their use of attentional resources. Personality and Individual Differ-
ences, 34, 11411152.
Bar-Haim, Y., Lamy, D., Pergamin, L., Bakermans-Kranenburg, M. J., &
van IJzendoorn, M. J. (2007). Threat-related attentional bias in anxious
and nonanxious individuals: A meta-analytic study. Psychological Bul-
letin, 133, 124.
!Beck, J. (1979). The effect of variations in incubation activity and
incubation time on creative response production. Unpublished doctoral
dissertation, New York University.
!Bennett, S. (1975). Effect of incubation on the associative process of
creativity. Unpublished doctoral dissertation, University of Notre Dame.
!Both, L., Needham, D., & Wood, E. (2004). Examining tasks that facil-
itate the experience of incubation while problem solving. Alberta Jour-
nal of Educational Research, 50(1), 57 67.
Bowden, E. M., & Jung-Beeman, M. (2003a). Aha! Insight experience
correlates with solution activation in the right hemisphere. Psychonomic
Bulletin & Review, 10, 730 737.
Bowden, E. M., & Jung-Beeman, M. (2003b). Normative data for 144
 INCUBATION AND PROBLEM SOLVING
compound remote associate problems. Behavior Research Methods,
Instruments, & Computers, 35(5), 634 639.
Bowers, K. S., Regehr, G., Balthazard, C. G., & Parker, K. (1990).
Intuition in the context of discovery. Cognitive Psychology, 22, 72110.
!Brockett, C. (1985). Neuropsychological and cognitive components of
creativity and incubation. Unpublished doctoral dissertation, Virginia
Commonwealth University.
!Browne, B. A., & Cruse, D. F. (1988). The incubation effect: Illusion or
illumination. Human Performance, 1(3), 177185.
!Christensen, B. T., & Schunn, C. D. (2005). Spontaneous access and
analogical incubation effects. Creativity Research Journal, 17(23),
207220.
Chronicle, E. P., Ormerod, T. C., & MacGregor, J. N. (2001). When insight
just wont come: The failure of visual cues in the nine-dot problem.
Quarterly Journal of Experimental Psychology, 54(A), 903919.
Cooper, H., & Hedges, L. V. (1994). The handbook of research synthesis.
New York: Russell Sage Foundation.
!Dodds, R. A., Smith, S. M., & Ward, T. B. (2002). The use of environ-
mental clues during incubation. Creativity Research Journal, 14(3 4),
287305.
Dodds, R. A., Ward, T. B., & Smith, S. M. (in press). Incubation in
problem solving and creativity. In M. A. Runco (Ed.), Creativity re-
search handbook (Vol. 3). Cresskill, NJ: Hampton Press.
!Dominowski, R. L., & Jenrick, R. (1972). Effect of hints and interpolated
activity on solution of an insight problem. Psychonomic Science, 26(6),
335338.
!Dorfman, J. (1990). Metacognitions and incubation effects in insight
problem solving. Unpublished doctoral dissertation, University of Cali-
fornia, San Diego.
!Dreistadt, R. (1969). The use of analogies and incubation in obtaining
insights in creative problem solving. The Journal of Psychology, 71,
159 175.
Durlak, J. A., & Lipsey, M. W. (1991). A practitioners guide to meta-
analysis. American Journal of Community Psychology, 19, 291332.
Finke, R. A., Ward, T. B., & Smith, S. M. (1992). Creative cognition:
Theory, research, and applications. Cambridge, MA: MIT Press.
!Fulgosi, A., & Guilford, J. P. (1968). Short-term incubation in divergent
production. American Journal of Psychology, 81, 241246.
!Fulgosi, A., & Guilford, J. (1972). A further investigation of short-term
incubation. Acta Instituti Psychologici, 64 73, 6770.
!Gall, M., & Mendelsohn, G. A. (1967). Effects of facilitating techniques
and subject experimenter interaction on creative problem solving. Jour-
nal of Personality and Social Psychology, 5, 211216.
Geary, D. C., Hoard, M. K., Byrd-Craven, J., & DeSoto, M. C. (2004).
Strategy choices in simple and complex addition: Contributions of
working memory and counting knowledge for children with mathemat-
ical disability. Journal of Experimental Child Psychology, 88, 121151.
Ghiselin, B. (1985). The creative process: A symposium. Berkeley: Uni-
versity of California Press.
Gilhooly, K. J., & Murphy, P. (2005). Differentiating insight from non-
insight problems. Thinking and Reasoning, 11, 279 302.
!Goldman, W., Wolters, N., & Winograd, E. (1992). A demonstration of
incubation in anagram problem solving. Bulletin of the Psychonomic
Society, 30, 36 38.
!Hansberry, M. T. (1998). Fixation and incubation effects in problem
solving. Unpublished doctoral dissertation, University of Hampshire,
United Kingdom.
Hedges, L. V., & Olkin, I. (1985). Statistical methods for meta-analysis.
Orlando, FL: Academic Press.
!Henley, R. J. (1999). Priming incubated problems. Unpublished doctoral
dissertation, University of Sussex, United Kingdom.
!Houtz, J. C., & Frankel, A. D. (1992). Effects of incubation and imagery
training on creativity. Creativity Research Journal, 5(2), 183189.
!Jamieson, B. A. (1999). Incubation and aging: The nature of processing
28
111
underlying insight. Unpublished doctoral dissertation, University of
Georgia.
Jung-Beeman, M., & Bowden, E. (2000). The right hemisphere maintains
solution-related activation for yet-to-be solved insight problems. Mem-
ory and Cognition, 28, 12311241.
!Kaplan, C. (1990). Hatching a theory of incubation: Does putting a
problem aside really help? If so, why? Unpublished doctoral disserta-
tion, Carnegie Mellon University.
Kershaw, T. C., & Ohlsson, S. (2004). Multiple causes of difficulty in
insight: The case of the nine-dot problem. Journal of Experimental
Psychology: Learning, Memory, and Cognition, 30, 313.
Knoblich, G., Ohlsson, S., Haider, H., & Rhenius, D. (1999). Constraint
relaxation and chunk decomposition in insight problem solving. Journal
of Experimental Psychology: Learning, Memory, and Cognition, 25,
1534 1555.
Lipsey, M. W., & Wilson, D. B. (2001). Practical meta-analysis. Thousand
Oaks, CA: Sage.
Lynch, M. D., & Swink, E. (1967). Some effects of priming, incubation
and creative aptitude on journalism performance. Journal of Communi-
cation, 17(4), 372382.
Macaskill, P., Walter, S. D., & Irwing, L. (2001). A comparison of methods
to detect publication bias in meta-analysis. Statistics in Medicine, 20,
641 654.
MacGregor, J. N., Ormerod, T. C., & Chronicle, E. P. (2001). Information-
processing and insight: A process model of performance on the nine-dot
and related problems. Journal of Experimental Psychology: Learning,
Memory, and Cognition, 27, 176 201.
Martindale, C. (1995). Creativity and connectionism. In S. M. Smith, T. B.
Ward, & R. A. Finke (Eds.), The creative cognition approach (pp.
249 268). Cambridge, MA: MIT Press.
!Medd, E., & Houtz, J. C. (2002). The effects of facilitated incubation on
fourth graders creative writing. Educational Research Quarterly, 26(2),
1316.
!Mednick, M., Mednick, S., & Mednick, E. (1964). Incubation of creative
performance and specific associative priming. Journal of Abnormal and
Social Psychology, 69, 84 88.
Mednick, S. A. (1962). The associative basis of the creative process.
Psychological Review, 69, 220 232.
Metcalfe, J., & Weibe, D. (1987). Intuition in insight and non-insight
problem-solving. Memory and Cognition, 15, 238 246.
!Moss, S. A. (2002). The impact of environmental clues in problem
solving and incubation: The moderating effect of ability. Creativity
Research Journal, 14(2), 207211.
!Murray, H. G., & Denny, J. P. (1969). Interaction of ability level and
interpolated activity (opportunity for incubation) in human problem
solving. Psychological Reports, 24(1), 271276.
Newton, E. J., & Roberts, M. J. (2005). The window of opportunity: A
model for strategy discovery. In M. J. Roberts & E. J. Newton (Eds.),
Methods of thought: Individual differences in reasoning strategies (pp.
129 158). Hove, United Kingdom: Psychology Press.
Olton, R. M. (1979). Experimental studies of incubation: Searching for the
elusive. Journal of Creative Behavior, 13, 9 22.
!Olton, R. M., & Johnson, D. M. (1976). Mechanism of incubation in
creative problem solving. American Journal of Psychology, 89(4), 617
630.
Patalano, A. L., & Seifert, C. M. (1994). Memory for impasses during
problem solving. Memory and Cognition, 22, 234 242.
!Patrick, A. S. (1986). The role of ability in creative incubation. Per-
sonality and Individual Differences, 7(2), 169 174.
!Penney, C. G., Godsell, A., Scott, A., & Balsom, R. (2004). Problem
variables that promote incubation effects. Journal of Creative Behavior,
38, 3555.
Perkins, D. N. (1995). Insight in minds and genes. In R. J. Sternberg & J. E.
112 SIO AND ORMEROD
Davidson (Eds.), The nature of insight (pp. 495533). Cambridge, MA:
MIT Press.
!Peterson, C. (1974). Incubation effects in anagram solution. Bulletin of
the Psychonomic Society, 3, 29 30.
Posner, M. I. (1973). Cognition: An introduction. Glenview, IL: Scott,
Foresman.
Rae, C. M. (1997). The creative power of doing nothing. Writer, 110(7),
1315.
Reder, L. M., & Schunn, C. D. (1996). Metacognition does not imply
awareness: Strategy choice is governed by implicit learning and mem-
ory. In L. M. Reder (Ed.), Implicit memory and metacognition (pp.
4578). Hillsdale, NJ: Erlbaum.
Rickards, T. (1991). Innovation and creativity: Woods, trees and pathways.
R&D Management, 21(2), 97108.
Rosenthal, R. (1995). Writing meta-analytic reviews. Psychological Bul-
letin, 118, 183192.
Schank, R. C. (1982). Dynamic memory: A theory of reminding and
learning in computers and people. Cambridge, United Kingdom: Cam-
bridge University Press.
Schank, R. C. (1999). Dynamic memory revisited. Cambridge, United
Kingdom: Cambridge University Press.
Scheerer, M. (1963). Problem solving. Scientific American, 208(4), 118
128.
Schooler, J. W., Ohlsson, S., & Brooks, K. (1993). Thoughts beyond
words: When language overshadows insight. Journal of Experimental
Psychology: General, 122, 166 183.
!Seabrook, R., & Dienes, Z. (2003, July). Incubation in problem solving as
a context effect. In R. Alterman & D. Kirsh (Eds.), Proceedings of the
25th annual meeting of the Cognitive Science Society (pp. 10651069).
Austin, TX: Cognitive Science Society. Available from http://
www.cogsci.rpi.edu/CSJarchive/Proceedings/2003/pdfs/203.pdf
!Segal, E. (2004). Incubation in insight problem solving. Creativity Re-
search Journal, 16(1), 141149.
Seifert, M. C., Meyer, D. E., Davidson, N., Patalano, A. L., & Yaniv, I.
(1995). Demystification of cognitive insight: Opportunistic assimilation
and the prepared-mind perspective. In R. J. Sternberg & J. E. Davidson
(Eds.), The nature of insight (pp. 65124). Cambridge, MA: MIT Press.
Siegler, R. S., & Stern, E. (1998). Conscious and unconscious strategy
discoveries: A microgenetic analysis. Journal of Experimental Psychol-
ogy: General, 127, 377397.
!Silveira, J. (1972). Incubation: The effect of interruption timing and
length on problem solution and quality of problem processing. Unpub-
lished doctoral dissertation, University of Oregon.
Simon, H. A. (1966). Scientific discovery and the psychology of problem
solving. In R. G. Colodny (Ed.), Mind and cosmos: Essays in contem-
porary science and philosophy (pp. 22 40). Pittsburgh, PA: University
of Pittsburgh Press.
29
!Sio, U. N., & Rudowicz, E. (2007). The role of an incubation period in
creative problem solving. Creativity Research Journal, 19(23), 307
318.
Smilek, D., Enns, J. T., Eastwood, J. D., & Merikle, P. A. (2006). Relax!
Cognitive style influences visual search. Visual Cognition, 14, 543564.
Smith, S. M. (1995). Fixation, incubation, and insight in memory and
creative thinking. In S. M. Smith, T. B. Ward, & R. A. Finke (Eds.), The
creative cognition approach (pp. 135146). Cambridge, MA: MIT
Press.
!Smith, S. M., & Blankenship, S. E. (1989). Incubation effects. Bulletin of
the Psychonomic Society, 27(4), 311314.
!Smith, S. M., & Blankenship, S. E. (1991). Incubation and the persistence
of fixation in problem solving. American Journal of Psychology, 104,
61 87.
!Snyder, A., Mitchell, D. J., Ellwood, S., Yates, A., & Pallier, G. (2004).
Nonconscious idea generation. Psychological Reports, 94, 13251330.
Thornton, A., & Lee, P. (2000). Publication bias in meta-analysisIts
causes and consequences. Journal of Clinical Epidemiology, 53(2),
207216.
!Torrance-Perks, J. (1997). The incubation effect: Implication for under-
lying mechanisms. Unpublished doctoral dissertation, University of Wa-
terloo, Ontario, Canada.
VanLehn, K. (1988). Towards a theory of impasse-driven learning. In H.
Mandl & A. Lesgold (Eds.), Learning: Issues for intelligent tutoring
systems (pp. 19 41). New York: Springer-Verlag.
!Vul, E., & Pashler, H. (2007). Incubation is helpful only when people are
misled. Memory and Cognition, 35(4), 701710.
Wallas, G. (1926). The art of thought. London: Cape.
Webster, A., Campbell, C., & Jane, B. (2006). Enhancing the creative
process for learning in primary technology education. International
Journal of Technology & Design Education, 16(3), 221235.
Wertheimer, M. (1985). A Gestalt perspective on computer simulations of
cognitive processes. Computers in Human Behavior, 1, 19 33.
Woodworth, R. S., & Schlosberg, H. (1954). Experimental psychology.
New Delhi, India: Oxford & IBH.
!Yaniv, I., & Meyer, D. E. (1987). Activation and meta-cognition of
inaccessible stored information: Potential bases for incubation effects in
problem solving. Journal of Experimental Psychology: Learning, Mem-
ory, and Cognition, 13(2), 187205.
Zeigarnik, B. (1927). Untersuchungen zur Handlungs- und Affektpsy-
chologie: III. Das Behalten erledigter und unerledigter andlungen [In-
vestigations on the psychology of action and affection: III. The memory
of completed and uncompleted actions]. Psychologische Forschung, 9,
1 85.
Zeigarnik, B. (1938). On finished and unfinished tasks. In W. D. Ellis
(Ed.), A sourcebook of Gestalt psychology (pp. 300 314). New York:
Humanities Press.

Task type
Creative problems
Brick task
Consequences task
Creative writing
Visual insight problems
Candle problem
Farm problem
Hat rack problem
Insightful mathematic problem
Necklace problem
INCUBATION AND PROBLEM SOLVING
Appendix A
Task Descriptions
Description
Participant has to list all the uses he
or she can think of for a brick
Participant has to list as many
consequences of an event as he
or she can foresee
Participant has to support a candle
on a wall by using some matches
and a box of tacks
Participant has to divide an L-
shaped farm into four parts that
have the same size and shape
Participant has to construct a stable
hat rack by using two boards and
a clamp
Participant has to compute
separately the area of the square
and that of the parallelogram
shown below:
Participant is given four pieces of
chain, each made up of three
links; he or she has to join all the
pieces by opening and closing
only three links
30
113
Solution or sample task
Sample task: What would be the results
if everyone suddenly lost the ability
to read and write?
Sample task: Write about three concrete
objects: a Koosch ball, a wooden
type of propeller, and a triangular
frisbee
Solution: Use a tack to attach box to
the wall, then drip some wax onto
the box and mount the candle on the
box
Solution:
Solution: Wedge the two boards
between the ceiling and the floor,
holding them in place with the
clamp, with the clamp also serving
as a hook
Solution: Restructure the given shape
into partially overlapping triangles
ABG and ECD. The sum of their
area is 2 ! ab/2 " ab
Solution: Open all three links of one
chain, and join the other three chains
together
(table continues)
114 SIO AND ORMEROD
Table (continued)
Task type Description
Radiation problem A patient has an inoperable tumor
in the middle of the body. There
is a ray at a strong intensity that
can destroy the tumor, but the ray
also harms the healthy tissue that
it travels through. At low
intensities, the ray will spare the
healthy tissue but will not destroy
the tumor. Participant has to
think out a way to use the ray to
destroy the tumor without
damaging healthy tissue.
Saugstads ball problem Participant has to transfer steel balls
from a drinking glass to a
cylinder from a distance of 8 ft
by using the following objects: a
nail, a pair of pliers, a length of
string, a pulley, elastic bands, and
a newspaper. The glass sits on a
moveable frame.
Tree problem Participant has to plant 10 trees in
five rows with four trees in each
row
Linguistic insight problems
Anagram Participant has to rearrange the
scrambled letters to form a new
word
Remote associates task (RAT) Three words are presented to the
participant, and he or she has to
think of a word that can form
associations with each one
Rebus Participant has to figure out the
phrase portrayed by the
pictogram
Riddle
Word associates task Six words are presented to the
participant; he or she has to think
out a word that can form an
association with each of them
Word fragment completion Participant has to complete a word
task that has various letters missing
31
Solution or sample task
Solution: Direct multiple low-intensity
rays simultaneously toward the tumor
from different directions
Solution: Bend the nail into a hook
then attach it to the string. Use it to
drag back the frame and remove the
balls. Transfer balls into the cylinder
using a tube constructed of
newspaper and elastic band.
Solution:
(The dots represent trees)
Sample task: The scambled letters are
t s l t i n e, and one possible
solution is silent
Sample task: The three stimulus words
are blue, cake, and cottage, and one
possible solution is the word cheese
Sample task: The pictogram
The answer is first aid
Sample task: A wine bottle is half-
filled and corked. How can you
drink all of the wine without
removing the cork from the bottle?
Answer: The cork can be opened by
pushing it in.
Sample task: The six stimulus words
are school, chair, jump, noon, heels,
and wire, and one possible solution
is the word high
Sample task: The stimulus is OC_
_N and one possible answer is
ocean
 INCUBATION AND PROBLEM SOLVING 115

Appendix B

Information Extracted From Each Independent Study

Presence of Preparation Incubation Presence of
ID Total misleading period in period in relevant Unbiased
No. Year Author N Problem typea cuesb minutes minutes Incubation task typec cuesd effect size

1 1967 Gall & Mendelsohn 60 2 (RAT) 0 2 25 2 (nonverbal task) 0 !0.58

2 1967 Gall & Mendelsohn 60 2 (RAT) 0 2 25 2 (associations 1 !0.58
training)
3 1968 Fulgosi & Guilford 50 1 (consequences task) 0 2 10 2 (number series task) 0 0
4 1968 Fulgosi & Guilford 49 1 (consequences task) 0 2 20 2 (number series task) 0 0.52
5 1969 Dreistadt 20 1 (farm problem) 0 5 8 1 (guess playing card) 0 0.34
6 1969 Dreistadt 20 1 (farm problem) 0 5 8 1 (guess playing card) 1 1.68
7 1969 Dreistadt 20 1 (tree problem) 0 5 8 1 (guess playing card) 0 0.21
8 1969 Dreistadt 20 1 (tree problem) 0 5 8 1 (guess playing card) 1 2.80
9 1969 Murray & Denny 36 1 (Saugstads ball 0 5 5 2 (multiple choice 0 0.62
(Experiment 1) problem, low syllogisms and
ability group) traced sequences of
numbers and
letters)
10 1969 Murray & Denny 36 1 (Saugstads ball 0 5 5 2 (multiple choice 0 !0.59
(Experiment 2) problem, high syllogisms and
ability group) traced sequences of
numbers and
letters)
11 1972 Dominowski & Jenrick 27 1 (hat rack problem) 5 10 2 (free association) 0 0
12 1972 Dominowski & Jenrick 30 1 (hat rack problem) 3 3 2 (free association) 0 0
13 1972 Silveira (Experiment 1) 18 1 (necklace problem) 0 3 30 1 (read book) 0 0.11
14 1972 Silveira (Experiment 1) 18 1 (necklace problem) 0 3 210 1 (read book for 30 0 0.06
min and free
activity for 3 hr)
15 1972 Silveira (Experiment 1) 18 1 (necklace problem) 0 13 30 1 (read book) 0 0.17
16 1972 Silveira (Experiment 1) 18 1 (necklace problem) 0 13 210 1 (read book for 30 0 0.42
min and free
activity for 3 hr)
17 1972 Silveira (Experiment 2) 32 1 (necklace problem) 0 13 210 1 (read book for 30 0 0.44
min and free
activity for 3 hr)
18 1974 Peterson 24 1 (anagram) 0.33 1.8 Anagram 1 0.65
19 1976 Olton & Johnson 21 1 (farm problem) 0 10 15 0 (rest) 0 0.10
20 1976 Olton & Johnson 21 1 (farm problem) 0 10 15 2 (Stroop test " 0 0.11
counting backward)
21 1976 Olton & Johnson 21 1 (farm problem) 0 10 15 2 (review the 1 0.10
problem)
22 1976 Olton & Johnson 21 1 (farm problem) 0 10 15 2 (have lecture) 1 0
23 1976 Olton & Johnson 20 1 (farm problem) 0 10 15 0 (listen to music) 0 !0.10
24 1976 Olton & Johnson 20 1 (farm problem) 0 10 15 0 (rest) 1 0.10
25 1976 Olton & Johnson 20 1 (farm problem) 0 10 15 2 (have lecture) 1 !0.03
26 1979 Beck 60 0 (verbal divergent- 0 12 20 0 (relax) 0 2.19
thinking task)
27 1979 Beck 60 0 (verbal divergent- 0 12 30 2 (write essay) 0 1.07
thinking task)
28 1979 Beck 60 0 (verbal divergent- 0 12 20 0 (relax) 0 4.07
thinking task)
29 1979 Beck 60 0 (verbal divergent- 0 12 30 2 (write essay) 0 4.04
thinking task)
30 1985 Brockett 30 0 (brick task) 0 10 20 2 (questionnaire) 0 0.42
31 1985 Brockett 30 2 (RAT) 0 0.33 20 2 (questionnaire) 0 0.37
32 1986 Patrick 30 2 (RAT) 0 2 5 2 (conversation) 0 0
33 1986 Patrick 30 2 (RAT) 0 2 5 2 (mental rotation 0 0.66
task)
34 1988 Browne & Cruse 60 1 (farm problem) 0 20 5 0 (listen to music) 0 0.47
(Experiment 2)
35 1988 Browne & Cruse 53 1 (farm problem) 0 20 5 1 (graph drawing) 1 0.24
(Experiment 2)
(table continues)

32
116 SIO AND ORMEROD

Table (continued)
Presence of Preparation Incubation Presence of
ID Total misleading period in period in relevant Unbiased
No. Year Author N Problem typea cuesb minutes minutes Incubation task typec cuesd effect size

36 1988 Browne & Cruse 55 1 (farm problem) 0 20 5 2 (memory test) 0 !0.18

(Experiment 2)
37 1989 Smith & Blankenship 26 2 (rebus) 1 0.5 5 0 (rest) 0 0.45
(Experiment 1)
38 1989 Smith & Blankenship 26 2 (rebus) 1 0.5 5 2 (music perception 0 0.48
(Experiment 1) task)
39 1989 Smith & Blankenship 26 2 (rebus) 1 0.5 15 0 (rest) 0 0.49
(Experiment 1)
40 1989 Smith & Blankenship 26 2 (rebus) 1 0.5 15 2 (music perception 0 0.50
(Experiment 1) task)
41 1989 Smith & Blankenship 25 2 (rebus) 1 0.5 5 0 (rest) 1 0.05
(Experiment 2)
42 1989 Smith & Blankenship 25 2 (rebus) 1 0.5 5 2 (music perception 1 0.05
(Experiment 2) task)
43 1989 Smith & Blankenship 25 2 (rebus) 1 0.5 15 0 (rest) 1 0.40
(Experiment 2)
44 1989 Smith & Blankenship 25 2 (rebus) 1 0.5 15 2 (music perception 1 0.40
(Experiment 2) task)
45 1989 Smith & Blankenship 29 2 (rebus) 1 0.5 10 2 (rebus) 1 0.56
(Experiment 3)
46 1989 Smith & Blankenship 29 2 (rebus) 1 0.5 15 0 (rebus " rest) 1 0.56
(Experiment 3)
47 1989 Smith & Blankenship 29 2 (rebus) 1 0.5 15 2 (rebus " math) 1 0.56
(Experiment 3)
48 1989 Smith & Blankenship 29 2 (rebus) 1 0.5 15 0 (rebus " music) 1 0.56
(Experiment 3)
49 1989 Smith & Blankenship 29 2 (rebus) 1 0.5 15 2 (rebus) 1 0.56
(Experiment 3)
50 1989 Smith & Blankenship 49 2 (rebus) 1 0.5 5 1 (read story) 0 0
(Experiment 4)
51 1989 Smith & Blankenship 49 2 (rebus) 1 0.5 5 2 (math task) 0 0.49
(Experiment 4)
52 1990 Dorfman (Experiment 15 2 (word associate 0 0.49 15 2 (number series task) 0 0
3) task)
53 1990 Dorfman (Experiment 15 2 (word associate 0 0.49 5 2 (number series task) 0 1.07
3) task)
54 1990 Dorfman (Experiment 27 2 (word associate 0 0.49 3 2 (number series task) 0 0
4) task)
55 1990 Dorfman (Experiment 27 2 (word associate 0 0.49 8 2 (number series task) 0 0
4) task)
56 1990 Dorfman (Experiment 26 2 (word associate 0 0.49 13 2 (number series task) 0 0
4) task)
57 1990 Kaplan (Experiment 1) 278 0 (consequences task) 0 2 30 2 (psychometric test 0 0.06
battery)
58 1990 Kaplan (Experiment 2) 64 0 (consequences task) 0 2 32 2 (division problem) 0 0.71
59 1990 Kaplan (Experiment 3) 36 0 (consequences task) 0 2 30 2 (lecture) 0 0
60 1990 Kaplan (Experiment 4) 20 0 (consequences task) 0 4.57 28.08 2 (division and insight 0 1.08
problem)
61 1991 Smith & Blankenship 18 2 (RAT) 1 0.5 5 1 (read science 0 1.70
(Experiment 1) fiction)
62 1991 Smith & Blankenship 21 2 (RAT) 0 0.5 5 1 (read science 0 0.64
(Experiment 1) fiction)
63 1991 Smith & Blankenship 30 2 (RAT) 0 1 5 1 (read science 0 0.36
(Experiment 2) fiction)
64 1991 Smith & Blankenship 30 2 (RAT) 1 1 5 1 (read science 0 0.72
(Experiment 2) fiction)
65 1991 Smith & Blankenship 16 2 (RAT) 0 0.5 2 2 (free associations 0 0
(Experiment 5) task)
66 1991 Smith & Blankenship 18 2 (RAT) 0 0.5 0.5 2 (free associations 0 0
(Experiment 5) task)
67 1991 Smith & Blankenship 17 2 (RAT) 1 0.5 0.5 2 (free associations 0 0.99
(Experiment 5) task)
(table continues)

33
 INCUBATION AND PROBLEM SOLVING 117

Table (continued)
Presence of Preparation Incubation Presence of
ID Total misleading period in period in relevant Unbiased
No. Year Author N Problem typea cuesb minutes minutes Incubation task typec cuesd effect size

68 1991 Smith & Blankenship 18 2 (RAT) 1 0.5 2 2 (free associations 0 0.99

(Experiment 5) task)
69 1992 Goldman et al. 36 2 (anagram) 0 0.25 20 2 (general knowledge 0 0.13
questionnaire)
70 1992 Goldman et al. 36 2 (anagram) 0 0.25 1,440 2 (general knowledge 0 0.66
questionnaire !
free activity)
71 1992 Houtz & Frankel 105 1 (creative writing) 0 10 10 1 (anagram) 0 0.03
72 1997 Torrance-Perks 15 2 (word fragment 0 0.5 10 0 (rest ! lexical 0 "0.38
(Experiment 1) completion) decision test)
73 1997 Torrance-Perks 15 2 (word fragment 0 0.5 10 0 (rest ! lexical 1 "0.41
(Experiment 1) completion task) decision test)
74 1997 Torrance-Perks 15 2 (word fragment 1 0.5 10 0 (rest ! lexical 0 0.14
(Experiment 1) completion) decision test)
75 1997 Torrance-Perks 15 2 (word fragment 1 0.5 10 0 (rest ! lexical 1 0.07
(Experiment 1) completion) decision test)
76 1997 Torrance-Perks 8 1 (candle problem) 0 1 8 2 (memory test: cues 1 0.17
(Experiment 2) presented as one of
the stimuli)
77 1997 Torrance-Perks 8 1 (candle problem) 0 1 8 2 (memory test) 0 "0.71
(Experiment 2)
78 1997 Torrance-Perks 8 1 (radiation problem) 0 5 8 1 (read story: analogy 1 0.41
(Experiment 2) to the problem)
79 1997 Torrance-Perks 8 1 (radiation problem) 0 5 8 1 (read unrelated 0 0
(Experiment 2) story)
80 1997 Torrance-Perks 8 2 (RAT) 0 1 8 2 (analogy: have the 1 0.15
(Experiment 2) same solution as
the unsolved RAT)
81 1997 Torrance-Perks 7 2 (RAT) 0 1 8 2 (neutral analogy) 0 0.14
(Experiment 2)
82 1997 Torrance-Perks 7 1 (candle problem) 1 1 8 2 (memory test: cue 1 0.56
(Experiment 2) presented a one of
the stimuli)
83 1997 Torrance-Perks 7 1 (candle problem) 1 1 8 2 (memory test) 0 "0.33
(Experiment 2)
84 1997 Torrance-Perks 7 1 (radiation problem) 1 5 8 1 (read story: analogy 1 0.52
(Experiment 2) to the problem)
85 1997 Torrance-Perks 7 1 (radiation problem) 1 5 8 1 (read unrelated 0 0
(Experiment 2) story)
86 1997 Torrance-Perks 7 2 (RAT) 1 1 8 2 (analogy: have the 1 0.14
(Experiment 2) same solution as
the unsolved RAT)
87 1997 Torrance-Perks 7 2 (RAT) 1 1 8 2 (neutral analogy) 0 0.28
(Experiment 2)
88 1998 Hansberry (Experiment 32 2 (riddle) 0 1 15 2 (RAT) 0 0.66
2)
89 1998 Hansberry (Experiment 20 2 (RAT) 0 0.5 10 2 (RAT) 0 0.04
3)
90 1998 Hansberry (Experiment 20 2 (RAT) 1 0.5 10 2 (RAT) 0 0.28
3)
91 1999 Henley (Experiment 48 2 (anagram) 0 0.25 1,440 1 (free activity) 0 0
3.2)e
92 1999 Henley (Experiment 4)e 26 2 (anagram) 0 0.93 1,440 1 (free activity) 0 0
93 1999 Jamieson (Experiment 52 2 (RAT) 0 0.33 5 2 (math problem) 0 0
1)
94 1999 Jamieson (Experiment 52 2 (RAT) 0 0.33 5 2 (math problem) 0 0.10
2)
95 2002 Dodds et al. 45 2 (RAT) 0 0.5 0.5 2 (insight problem ! 2 0.31
(Experiment 2) make a word task)
96 2002 Dodds et al. 45 2 (RAT) 0 0.5 0.5 2 (insight problem ! 1 "0.14
(Experiment 2) make a word task)
97 2002 Dodds et al. 42 2 (RAT) 0 0.5 0.5 2 (insight problem ! 0 0.10
(Experiment 2) make a word task)
(table continues)

34
118 SIO AND ORMEROD

Table (continued)
Presence of Preparation Incubation Presence of
ID Total misleading period in period in relevant Unbiased
No. Year Author N Problem typea cuesb minutes minutes Incubation task typec cuesd effect size

98 2002 Dodds et al. 70 2 (RAT) 0 0.5 0.5 2 (insight problem ! 0 0

(Experiment 2) drawing test)
99 2002 Medd & Houtz 15 1 (creative writing) 0 10 10 2 (writing) 1 1.05
100 2002 Medd & Houtz 15 1 (creative writing) 0 10 10 2 (writing) 0 0
101 2003 Seabrook & Dienese 60 2 (anagram) 0 0.25 7 2 (word generation) 1 0.74
102 2004 Both et al. 98 2 (anagram) 0 1.67 6 2 (search letter and 0 0.48
answer
questionnaire)
103 2004 Both et al. 82 2 (anagram) 0 1.67 6 2 (search letter and 0 0.09
answer
questionnaire)
104 2004 Penney et al. 9 2 (anagram) 0 5.75 30 1 (free activity) 1 0.74
(Experiment 3)
105 2004 Penney et al. 9 2 (anagram) 0 5.75 120 1 (free activity) 1 1.05
(Experiment 3)
106 2004 Penney et al. 9 2 (anagram) 0 5.75 30 1 (free activity) 0 0.62
(Experiment 3)
107 2004 Penney et al. 9 2 (anagram) 0 5.75 120 1 (free activity) 0 0.70
(Experiment 3)
108 2004 Segal 20 1 (insightful 0 20 4 1 (read paper) 0 1.17
mathematic
problem)
109 2004 Segal 21 1 (insightful 0 20 12 1 (read paper) 0 1.09
mathematic
problem)
110 2004 Segal 20 1 (insightful 0 20 4 2 (word puzzle) 0 0.90
mathematic
problem)
111 2004 Segal 23 1 (insightful 0 20 12 2 (word puzzle) 0 0.57
mathematic
problem)
112 2007 Vul & Pashler 14 2 (anagram) 0 1 5 2 (video game) 0 "0.17
(Experiment 1)
113 2007 Vul & Pashler 14 2 (anagram) 0 1 5 2 (video game) 0 "0.35
(Experiment 1)
114 2007 Vul & Pashler 14 2 (anagram) 0 1 5 2 (video game) 0 0.36
(Experiment 1)
115 2007 Vul & Pashler 14 2 (anagram) 0 1 5 2 (video game) 0 0.25
(Experiment 1)
116 2007 Vul & Pashler 25 2 (RAT) 0 1 5 2 (video game) 0 0.86
(Experiment 2)
117 2007 Vul & Pashler 25 2 (RAT) 1 1 5 2 (video game) 0 "0.14
(Experiment 2)

Note. RAT # remote associates task.

a
0 # creative problem; 1 # visual problem; 2 # linguistic problem. b0 # no; 1 # yes, embedded in problem. c0 # rest; 1 # low cognitive demand task;
2 # high cognitive demand task. d0 # no; 1 # yes. eStudy had a within-subject design: An individual was in both incubation and control conditions; the
weighting formula would not apply to this study, hence, it would be excluded from the regression analyses.

Appendix C

Description of the Regression Model for All StudiesInteractions

The regression model describing the interaction effect of nature
of problem and incubation task on the weighted unbiased effect
size estimate included the multiplicative terms between nature of
problem (visual [V], linguistic [Li], creative [C]) and incubation
task (high cognitive demand [H], low cognitive demand [L], rest
[R]) as predictors. The regression model is specified as follows:
Yj # ajXj ! b1LijHj ! b2LijLj ! b3LijRj ! b4VjHj ! b5VjLj
! b 6V jR j ! b 7C jH j ! b 8C jL j ! b 9C jR j ! k j,
where Yj is the weighted unbiased effect size estimation of the
study j and Xj is a vector of other categorical and explanatory
variables (misleading cues, cue, preparation period, and incu-

35
 INCUBATION AND PROBLEM SOLVING 119

bation period) of that study. aj is the corresponding vector of
coefficients. b1 to b9 are the coefficients of the multiplicative
terms, and kj is the error term. The dummy variables Rj and Cj
are eliminated through substituting equations Lij ! Vj ! Cj "
1 and Rj ! Lj ! Hj " 1 into the model. The transformed
regression model is as follows:
Yj " ajXj ! (b3 # b9)Lij ! (b6 # b9)Vj ! (b7 # b9)Hj
! (b8 # b9)Lj ! (b1 # b7 # b3 ! b9)LijHj
! (b2 # b8 ! b3 ! b9)LijLj ! (b4 # b7 # b6 ! b9)VjHj
! (b5 # b8 # b6 ! b9)VjLj.
A regression analysis was carried out to find out the coefficient
of each variable in the transformed model. To examine the inter-
action effect of nature of the problem and incubation task, we
compared the difference between coefficients multiplicative terms
in the original model. For example, the coefficients of the com-
pound dummies VjRj and CjRj were compared to check if an
incubation period filled with low cognitive demand tasks would
improve performance on visual problems more than on divergent
thinking tasks. A comparison between the original and the trans-
formed model indicated that the coefficient of each variable in the
transformed model was actually the combination of the coeffi-
cients in the original model. The difference in the coefficients
between the compound dummies in the original model could be
found by reinterpreting the coefficient in the transformed model.
Table C1 displays a list of coefficient differences between com-
pound dummies in the original model and the equivalent combi-
nation of the coefficients in the transformed model.

Table C1

The Regression Coefficients in the Original and Transformed Regression Models

Coefficient difference between
compound dummies in the Combination of coefficients
original regression model in the regression analysis

LijHj#LijRjj LijHj ! Hj
LijLj#LijRjj LijLj ! Lj
VjHj#VjRjj VjHj ! Hj
VjLj#VjRjj VjLj ! Lj
CjHj#CjRjj Hj
CjLj#CjRjj Lj
VjHj#CjHjj VjHj ! Vj
LijHj#CjHjj LijHj ! Lij
VjLj#CjLjj VjLj ! Vj
LijLj#CjLjj LijLj ! Lij
VjRj#CjRjj Vj
LijRj#CjRjj Lij

Note. Li " linguistic problem; H " high cognitive demand task; R " rest; L " low cognitive demand task; V " visual
problem; C " creative problem; j " study j.

Appendix D
Description of the Regression Model for Studies Excluding Creative Problem

The third regression model describing the interaction effect of
nature of problem and incubation task on the weighted unbiased
effect size estimate included the multiplicative terms between
nature of problem (visual [V], linguistic [Li]) and incubation task
(high cognitive demand [H], low cognitive demand [L], rest [R]) as
predictors. The regression model is specified as follows:
variables (misleading cues, cue, preparation period, and incubation
period) of that study. aj is the corresponding vector of coefficients.
b1 to b9 are the coefficients of the multiplicative terms, and kj is the
error term. The dummy variables Li and Rj were eliminated
through substituting equations Lij ! Vj " 1 and Rj ! Lj ! Hj "
1 into the model. The transformed regression model is as follows:

Yj " ajXj ! b1LijHj ! b2LijLj ! b3LijRj ! b4VjHj ! b5VjLj Yj " ajXj ! (b6 # b3)Vj ! (b1 # b3)Hj ! (b2 # b3)Lj

! b 6V jR j ! k j, ! (b3 # b1 # b6 ! b4)VjHj ! (b5 # b2 # b6 ! b3)VjLj.

where Yj is the weighted unbiased effect size estimation of the A regression analysis was carried out to find out the coeffi-
study j, and Xj is a vector of other categorical and explanatory cient of each variable in the transformed model. To examine the
(Appendixes continue)

36
120 SIO AND ORMEROD

interaction effect of nature of the problem and incubation task,
we compared the difference between coefficients multiplica-
tive terms in the original model. For example, the coefficients
of the compound dummy variables VjRj and LijRj were com-
pared to check if an incubation period filled with low cognitive
demand tasks would improve performance on visual problems
more than on linguistic insight tasks. A comparison between the
original and the transformed models indicated that the coeffi-
cient of each variable in the transformed model was actually the
combination of the coefficients in the original model. The
difference in the coefficients between the compound dummies
in the original model could be found by reinterpreting the
coefficient in the transformed model. Table D1 displays a list of
coefficient differences between compound dummies in the orig-
inal model and the equivalent combination of the coefficients in
the transformed model.

Table D1

The Regression Coefficients in the Original and Transformed Regression Models

Coefficient difference between
compound dummies in the Combination of coefficients
original regression model in the regression analysis

LijHjj!LijRjj Hj
LijLjj!LijRjj Lj
LijLjj!LijHjj Lj ! Hj
VjHjj!VjRjj VjHjj " Hj
VjLjj!VjRjj VjLjj " Lj
VjLjj!VjHjj VjLjj " Lj ! VjHjj ! Hj
LijHjj!VjHjj VjHjj " Vj
LijLjj!VjLjj VjLjj " Vj
LijRjj!VjRjj Vj

Note. Li # linguistic problem; H # high cognitive demand task; R # rest; L # low cognitive demand task; V # visual
problem; C # creative problem; j # study j.

Received December 11, 2007

Revision received August 13, 2008
Accepted August 14, 2008 !

37
Mem Cogn (2012) 40:966975
DOI 10.3758/s13421-012-0199-z
Dont wait to incubate: Immediate versus delayed incubation
in divergent thinking
Kenneth J. Gilhooly & George J. Georgiou &
Jane Garrison & Jon D. Reston & Miroslav Sirota
Published online: 2 March 2012
# Psychonomic Society, Inc. 2012
Abstract Previous evidence for the effectiveness of imme-
diate incubation in divergent creative tasks has been weak,
because earlier studies exhibited a range of methodological
problems. This issue is theoretically important, as a demon-
stration of the effects of immediate incubation would
strengthen the case for the involvement of unconscious
work in incubation effects. For the present experiment, we
used a creative divergent-thinking task (alternative uses) in
which separate experimental groups had incubation periods
that were either delayed or immediate and that consisted of
either spatial or verbal tasks. Control groups were tested
without incubation periods, and we carried out checks for
intermittent conscious work on the target task during the
incubation periods. The results showed significant incuba-
tion effects that were stronger for immediate than for
delayed incubation. Performance was not different between
the verbal and spatial incubation conditions, and we found
no evidence for intermittent conscious working during the
incubation periods. These results support a role for uncon-
scious work in creative divergent thinking, particularly in
the case of immediate incubation.
Keywords Creativity . Problem solving
Creative problems are generally defined as problems that
require the production of new approaches and solutions,
where by new we mean novel to the solver (Boden,
2004). Explaining how such personally novel solutions are
reached is still a major challenge for the psychology of
K. J. Gilhooly (*) : G. J. Georgiou : J. Garrison : J. D. Reston :
M. Sirota
School of Psychology, University of Hertfordshire,
Hatfield, Hertfordshire AL10 9AB, UK
e-mail: k.j.gilhooly@herts.ac.uk
38
thinking. In analyses of creative problem solving, it has often
been claimed that setting creative problems aside for a while
can lead to novel ideas about the solution, either spontaneous-
ly while attending to other matters or very rapidly when the
previously intractable problem is revisited. Personal accounts
by eminent creative thinkers in a range of domains have
attested to this phenomenon (e.g., Csikszentmihalyi, 1996;
Ghiselin, 1952; Poincar, 1913). In his well-known four-
stage analysis of creative problem solving, Wallas (1926,
p. 80) labeled a stage at which the problem is set aside and
not consciously addressed as incubation, and this stage is
the focus of the present study.
Following Wallas (1926), a substantial body of experi-
mental research on incubation effects has accumulated using
both insight problemsto which there is a single solution,
but the solver has to develop a new way of representing or
structuring the task in order to reach that solutionand
divergent problemsto which there is no single correct
solution, but the solution process encourages seeking as
many novel and useful ideas as possible. The prototypical
divergent task, which was the one used in the present study,
is the alternative-uses task, in which participants are asked
to generate as many uses as possible that are different from
the normal uses of one or more familiar objects, such as a
brick (Guilford, 1971; Guilford, Christensen, Merrifield, &
Wilson, 1978; Gilhooly, Fioratou, Anthony, & Wynn,
2007). In the classic laboratory paradigm for studying incu-
bation effects, which we will label the delayed-incubation
paradigm, participants in the incubation condition work on
the target problem for an experimenter-determined amount
of time (preparation time), are then given an interpolated
activity away from the target task for a fixed time (incuba-
tion period), and finally return to the target problem for a
postincubation work period. The performance of the incu-
bation group is contrasted with that of a control group, who
Mem Cogn (2012) 40:966975
have worked continuously on the target task for a time equal
to the sum of the preparation time and the postincubation
conscious working time among the incubation group. A
recently developed variant (the immediate-incubation para-
digm) employs an interpolated task for a fixed period
immediately after instructions on the target problem and
before any conscious work has been undertaken, followed
by uninterrupted work on the target problem (Dijksterhuis &
Meurs, 2006).
Previous studies of delayed- and immediate-incubation
effects
Considerable evidence has now emerged from laboratory
studies for the efficacy of delayed incubationthat is, that
setting a problem aside after a period of work is beneficial
(see Dodds, Ward, & Smith, in press, for a qualitative review).
A recent meta-analysis of 117 studies by Sio and Ormerod
(2009) identified a positive effect of delayed incubation, in
which the overall average effect size was in the low
medium band (mean d 0 0.32) over a range of insight
and divergent tasks. For divergent tasks considered sep-
arately, the mean d was larger, 0.65, which may be
considered to be in the highmedium band of effect sizes.
Overall, the existence of delayed-incubation effects can
now be regarded as well established, particularly in the
case of divergent problem solving.
Regarding the efficacy of immediate incubation, Dijksterhuis
and Nordgren (2006) reported studies in which better decisions
and more creative solutions were found when immediate
incubation breaks were given after the decision problems
or divergent tasks had been presented. In the realm of
decision problems, Nordgren, Bos, and Dijksterhuis (2011)
found that delayed incubation produced better decisions than
did immediate incubation, and both were better than no
incubation.
However, the beneficial effects of immediate incubation
on decision making have proven difficult to reproduce, and
a number of unsuccessful replication attempts have now
been reported (e.g., Acker, 2008; Newell, Wong, Cheung,
& Rakow, 2009; Rey, Goldstein, & Perruchet, 2009; Payne,
Samper, Bettman, & Luce, 2008).
The present study concerns creative thinking using a
divergent task, and Dijksterhuis and Meurs (2006) did report
that, in their Experiment 3, participants produced responses
of higher rated average creativity when the instructions to
list things one can do with a brick were followed immedi-
ately by a 3-min distractor task (immediate incubation)
before generating uses, relative to participants who began
generating uses right away. It may be noted that the instruc-
tions did not ask for unusual uses, which is the norm in
divergent-thinking tasks, and so it is not clear whether
39
967
participants had a goal of being creative. They may
have been reporting infrequent uses that they happened
to know rather than generating uses novel to them.
Raters tend to score infrequent responses as creative,
although such uses may have been preknown, and there-
fore could reflect memory retrieval rather than the
generation of subjectively novel responses (Quellmalz,
1985). Gilhooly et al. (2007) developed a self-report
method for assessing subjective novelty that addresses
the issue of individually creative responses as against
rare responses, and this method was used in the present
experiment. In this method, participants indicate which
of their responses were first thought of while doing the
task, and so were subjectively novel. Gilhooly et al.
found converging evidence for the validity of this method of
assessing responses as personally old or new. Self-judged new
responses were rated as significantly more creative by inde-
pendent judges and were more frequently produced by partic-
ipants with higher executive-functioning test scores. Self-
judged novel responses occurred later in the sequences of
responses, which is consistent with a reliance on memory for
the retrieval of early responses, followed by executively
demanding processes for generation of novel ideas when the
pool of already known uses is exhausted.
Zhong, Dijksterhuis, and Galinsky (2008), using the
immediate-incubation paradigm with the remote associates
task (RAT)in which participants have to retrieve an asso-
ciate common to three given words (e.g., cottage, blue,
mouse? Answer: cheese)found that, although immediate
incubation did not facilitate actual solution, it appeared
to activate solution words on unsolved trials, as indicated
by lexical decision measures, as compared to unsolved trials
without immediate incubation. However, it may be not-
ed that some theorists (e.g., Weisberg, 2006, p. 468)
have disputed whether the RAT is a creative task, as the
solutions are already-known associations rather than
novel responses. A normal criterion for a creative task
is that it requires the participant to generate a response
that is novel for the participant rather than one already
known.
Overall, the evidence in favor of a beneficial effect of
immediate incubation in creative tasks is rather weak, as
it is based on one study of a divergent task that did not
require novel responses (Dijksterhuis & Meurs, 2006)
and another study (Zhong et al., 2008) using a conver-
gent task (the RAT) in which the responses are not
themselves creative. The question of whether immediate
incubation is effective in creative tasks is important for
its bearing on theories of incubation, and the present
study aimed to provide more solid evidence regarding
the efficacy, or otherwise, of immediate incubation than
has been available hitherto. We now outline the main
theories regarding incubation effects.
968
Theories of incubation effects
Intermittent conscious work This theory suggests that
although incubation is intended to be a period without
conscious work on the target task, nevertheless participants
may carry out intermittent conscious work (Seifert, Meyer,
Davidson, Patalano, & Yaniv, 1995, p. 82; Weisberg, 2006,
pp. 443445). Any conscious work during the supposed
incubation period would reduce the time required when the
target problem was readdressedbut would be expected to
impair performance on the interpolated task. As a check
against the possibility of intermittent conscious work, per-
formance on the interpolated task during the incubation
period should be compared with the performance of a con-
trol group working on the same interpolated task without
being in an incubation condition. A deficit in the interpolat-
ed task on the part of the incubation group would be con-
sistent with the hypothesis of some conscious work on the
target task occurring during incubation. Although this seems
a rather basic methodological check, surprisingly, it does not
appear to have been carried out in previous research (Dodds
et al., in press; Sio & Ormerod, 2009). The study reported
here, on the other hand, did incorporate suitable checks for
intermittent conscious work on the target task during the
incubation period.
Fresh look This view (e.g., Segal, 2004; Simon, 1966;
see also Dijksterhuis & Meurs, 2006) proposes an important
role for automatic passive reduction in idea strength or
activation during the incubation period. The proposal is that
misleading strategies, mistaken assumptions, and related
mental sets weaken through forgetting during the incuba-
tion period, and thus a fresh start or set shifting is facili-
tated when the problem is resumed. On this view, incubation
works by allowing the weakening of misleading approaches
to the task during a break after a period of work (delayed
incubation), thus allowing a fresh start. This approach
would not predict a beneficial effect of immediate incuba-
tion, because with immediate incubation there is no time for
sets or fixations to develop, so that forgetting of misleading
approaches cannot occur.
Unconscious work This approach proposes that incubation
effects occur through active but unconscious processing of
the problem materials (as against the passive forgetting
processes envisaged in the fresh-look approach.) The term
unconscious work seems to have first been used in the
context of problem solving by Poincar (1913, p. 393).
Other phrases referring to the same notion include noncon-
scious idea generation (Snyder, Mitchell, Ellwood, Yates,
& Pallier, 2004) and unconscious thought (Dijksterhuis &
Nordgren, 2006), but we will generally use the term
unconscious work in this article. The question naturally
40
Mem Cogn (2012) 40:966975
arises of what form unconscious work might take. Is it
possible that unconscious work could be just like conscious
work, but carried out without conscious awareness? Or is it
better thought of as automatic spreading activation along
associative links, as against a rule- or strategy-governed
activity? We will consider the question of what form uncon-
scious work might take more fully in the Discussion section.
The possible mechanisms outlined above are not mu-
tually exclusive. A delayed-incubation condition could
conceivably evoke all three mechanisms, with the per-
son engaging in some intermittent conscious work when
attention wanders from the interpolated incubation task,
and with some beneficial forgetting and unconscious
work taking place when the person is attending to the
interpolated incubation task. However, an immediate-
incubation effect would not be consistent with a fresh-
look explanation but could involve some intermittent
conscious work and/or some unconscious work. The
present study aimed to clarify the contributions of the
three types of processes in explaining immediate and
delayed incubation, without assuming that one and only
one process can explain all of the findings.
Theories of incubation: Previous studies
What has previous research suggested regarding the possible
mechanisms of incubation?
Dijksterhuis and Meurs (2006) argued, as outlined above,
that in the immediate-incubation paradigm the fresh-look
approach may be ruled out, as there is no period of initial
work in which misleading fixations and sets could be
developed. Thus, if immediate incubation is shown to be
effective, the unconscious-work hypothesis must remain
in contention for immediate-incubation effects, and
would also be a candidate explanation for delayed incu-
bation. Dijksterhuis and Meurs took the beneficial effects
of the immediate-incubation paradigm on a divergent
task in their Experiment 3 as support for the role of
unconscious work in incubation. However, as already
mentioned, the task in this study did not clearly meet
the usual criteria for a creative task, and the scoring did not
distinguish infrequent from genuinely novel responses.
Hence, this study does not unequivocally address creative
thinking as against free recall of possibly rare, but previously
experienced, events from episodic and semantic memory.
Snyder et al. (2004) also found evidence consistent with
unconscious work from a study using the delayed-incubation
paradigm, but with a surprise return to the target task.
Although the return to the main task was unexpected, benefi-
cial effects were found, suggesting that automatic continua-
tion of unconscious work could have occurred when the task
was set aside. It should be noted, however, that Snyder et al.
Mem Cogn (2012) 40:966975
used a task that simply required production of uses for a
piece of paper, as against the generation of novel uses, so
their task did not necessarily involve creative thinking
rather than recall.
It is of interest that both Segal (2004) and Dijksterhuis
and Meurs (2006) used interpolated tasks during their incu-
bation periods that were different in character from the
target tasks. Segals target task was spatial, while the inter-
polated tasks were verbal; Dijksterhuis and Meurss target
task was verbal, but the interpolated task was spatial. From
Dodds et al.s (in press) extensive review, the issue of simi-
larity between the target and interpolated tasks does not appear
to have been addressed hitherto. The similarity relationship
between target and interpolated tasks could be important, in
that the main competing hypotheses suggest different effects
of similarity. If unconscious work is the main process, then
interpolated tasks similar to the target task should interfere
with any unconscious work using the same mental resources,
and so lead to weaker (or even reversed) incubation effects
when compared with the effects of dissimilar interpolated
tasks. On the other hand, the selective-forgetting mechanism
suggests that interpolated tasks similar to the target task would
cause greater interference, which would lead to more forget-
ting and enhanced incubation benefits.
Hlie, Sun, and Xiong (2008) found that more execu-
tively demanding interpolated tasks reduced reminiscence
scores for the free recall of pictures when a surprise free
recall was required after the interpolated task. In their study,
participants studied booklets of pictures for a set period,
freely recalled the items, and then did various interpolated
activities before being retested with free recall of the pic-
tures. The reminiscence score was the number of new items
recalled on the second test. The results were consistent with
Hlie and Suns (2010) explicitimplicit interaction model,
which can be applied to creative problem solving and which
allows for unconscious, implicit processes to occur in par-
allel with conscious, explicit processes. However, Hlie et
al.s (2008) target task was free recall rather than creative
thinking, so it does not speak directly to divergent thinking,
which is the focus of the present study.
Ellwood, Pallier, Snyder, and Gallate (2009) found a ben-
eficial effect on the number of postincubation responses of a
dissimilar interpolated task in a delayed-incubation experi-
ment. However, their study used a fluency-of-uses task rather
than a novel-uses task. Also, as Ellwood et al. pointed out,
although their findings are consistent with an explanation in
terms of unconscious work, an explanation in terms of selec-
tive relief of fatigue could also be invoked to account for the
effects of similarity between the incubation and target tasks.
On this view, for example, a spatial delayed-incubation task
very different from a main verbal task could allow for more
recovery from specific fatigue of verbal processes than would
an intervening verbal task. The present study includes tests of
41
969
the effects of incubationtarget task similarity in an immediate-
incubation paradigm, where fatigue can be ruled out, as well as
in a delayed-incubation paradigm in which fatigue relief could
be a factor.
Present study: Outline
The present study of the effects of varying incubation activi-
ties (verbal vs. spatial), detailed below, used a clearly creative
verbal divergent task (alternative uses), scored for novelty as
well as fluency, unlike the tasks of Ellwood et al. (2009) or
Hlie et al. (2008). Thus, the present study is clearly focused
on incubation effects in creative thinking. The study used both
immediate and delayed incubation with spatial and verbal
intervening tasks, so that the resource overlap predictions of
the selective-forgetting and unconscious-work hypotheses, as
well as the issue of the possible effects of differential fatigue
relief, could be addressed. The main aims of the study were to
determine the extent to which immediate incubation is indeed
helpful in divergent creative tasks (which previous research
had not clearly addressed) and to assess the relative contribu-
tions of intermittent-work, unconscious-work, and fresh-look
mechanisms of incubation in such tasks.
Method
In this experiment, the target task was the divergent produc-
tion of alternative uses for a brick, which we classed as a
verbal task. The incubation period (which was 4 min long)
was positioned either after 5 min of conscious work or
immediately after the initial divergent-task instructions. The
activities during the incubation period were either verbal
(anagrams) or spatial (mental rotation tasks). All participants
were instructed after 5 min of divergent production to draw a
line after their last response up to that point.
Participants
A group of 184 (123 female, 61 male) students at the University
of Hertfordshire took part in the experiment.
Design
A 2 (incubation position: immediate vs. delayed.) 3 (interpo-
lated task: none vs. verbal vs. spatial) independent-groups
design was used. The ns per experimental group were as
follows: n 0 25 with 4 min delayed incubation and a spatial
interpolated task, n 0 22 with 4 min delayed incubation and a
verbal interpolated task, n 0 30 with 4 min immediate incuba-
tion and a spatial interpolated task, and n 0 30 with 4 min
immediate incubation and a verbal interpolated task. We also
970
ran separate control groups for the delayed- and immediate-
incubation conditions (ns 0 47 and 30, respectively) that pro-
vided baseline performance data for target and interpolated
tasks in the absence of incubation periods.
Procedure
In the delayed-incubation conditions, the participants were
told that they would be asked to write down possible uses
for a brick different from the usual use; after 5 min of
working, participants were told that they would be returning
to the brick-uses task later in the study. During the 4-min
incubation periods, participants either undertook verbal
tasks (anagrams) or spatial tasks (mental rotation items)
presented in booklets. Sets of 73 five-letter single-solution
anagrams (from Gilhooly & Hay, 1977) and 48 mental
rotation items (from Peters et al., 1995) were used, and
performance was scored in terms of correct solutions during
the period allowed. After the delayed incubation periods,
there were a further 2 min of work on the brick-uses task.
In the immediate-incubation conditions, participants were
given the standard instructions about the brick-uses task and
immediately told that the experimenter wanted them to do
another task first, after which they would return to the uses
task, and they were assigned randomly to anagrams or mental
rotation for 4 min. After the immediate incubation period, they
worked on the brick-uses task for 7 min without a break.
Control participants worked on the uses task for 7 min
without any incubation periods and carried out mental rota-
tions and anagrams for 4 min each. The order of the three
tasks in the control groups was randomized. The control
rotation and anagram measures were compared with perfor-
mance on the same tasks when they were used as interven-
ing activities during the incubation periods. The control
uses-task measures were compared with performance on
the uses task in the incubation conditions.
At the end of the brick-uses task, participants reviewed their
response sheets and were asked to indicate (by circling) which
of the uses that they had reported were subjectively novelthat
is, had first occurred to them during the task rather than being
previously known from past direct experience or through films,
books, television, and so on. Gilhooly et al. (2007) found that
this was a valid measure of personal originality.
Results
Incubation effects
Figure 1 shows the average numbers of brick uses, and
Fig. 2 the average numbers of self-judged novel uses pro-
duced over the total of 7 min on the uses task with 0 min of
incubation (control data) or 4 min of delayed or immediate
42
Mem Cogn (2012) 40:966975
Fig. 1 Mean numbers of alternative uses produced during delayed and
immediate incubation using verbal or spatial interpolated tasks. Error
bars represent 1 SEM
incubation with spatial or verbal interpolated tasks (mental
rotations or anagrams). From these figures, it seems that
both immediate and delayed incubation periods were bene-
ficial as compared to the control conditions and that imme-
diate incubation produced better performance than did
delayed incubation.
An analysis of variance (ANOVA) indicated a significant
effect of the type of interpolated activity (none/verbal/spatial)
on the number of uses reported [F(2, 178) 0 7.89, p < .001,
p2 0 .08] and on the number of self-judged novel uses
[F(2, 178) 0 11.49, p < .001, p2 0 .11]. Post hoc tests indicated
significant differences between no incubation and incubation
tasks of both mental rotations and anagrams for number of uses
(p < .05) and for self-judged novelty (p < .05). No significant
differences were found between the verbal and spatial incuba-
tion conditions.
An ANOVA indicated a significant effect of the position of
incubation (delayed/immediate) on the number of uses
reported [F(1, 178) 0 6.39, p < .05, p2 0 .04] and on the
number of self-judged novel uses [F(1, 178) 0 10.03, p < .01,
p2 0 .05], with immediate incubation being more beneficial
for both measures.
There were no significant interactions between the type
of incubation activity and the position of incubation activity
on the number of uses or on self-judged novelty.
Pre- and postincubation performance in delayed-incubation
conditions
In the delayed-incubation conditions, data were available for
the uses-task performance measures separately for the 5-min
preincubation period and the 2-min postincubation period and
for the first 5 min and last 2 min of use production in the
control (no-incubation) condition. These data were examined
Mem Cogn (2012) 40:966975
Fig. 2 Mean numbers of self-rated novel uses produced by the
delayed- and immediate-incubation groups using verbal or spatial
interpolated tasks. Error bars represent 1 SEM
to check that any benefits in performance relative to controls
were concentrated in the postincubation (last 2-min) period.
One-way ANOVAs were carried out on the effects of incuba-
tion activity (none/verbal/spatial) on uses totals and uses
novelty in the first and last 2 min of work (pre- and post-
incubation periods in the incubation conditions). The mean
scores are shown in Table 1.
The first-5-min scores for both uses totals and novelty
were not significantly different between the incubation ac-
tivity conditions (none/verbal/spatial). However, the meas-
ures in the last 2 min (postincubation in the incubation
conditions) were significantly different between the condi-
tions. For uses totals, F(2, 91) 0 3.45, p < .05, p2 0 .07, and
for uses novelty, F(2, 91) 0 6.54, p < .01, p2 0 .11. Thus,
the effects of the delayed-incubation manipulation were
concentrated in the postincubation period, during which
incubation produced more responses and more novel
responses than did no incubation, as would be expected.
Performance over the first 5 min and the last 2 min
in immediate-incubation conditions
All uses-task performance in the immediate-incubation con-
ditions was postincubation, but it was possible to compare
Table 1 Delayed incubation: Pre- and postincubation scores for total brick uses and brick use novelty over incubation task conditions (control/
verbal/spatial)
Control (N 0 47) Verbal (N 0 22) Spatial (N 0 25)
M SD M SD M SD
First 5 min/Preincubation Brick Uses
7.90 3.24 8.59 3.67 9.60 4.09
First 5 min/Preincubation Brick Novelty
2.31 1.77 2.86 1.91 3.52 2.51
43
971
the first 5 min (which corresponded to the preincubation
time in the delayed condition) and the last 2 min (which
corresponded to the postincubation time in the delayed
condition).
One way ANOVAs were carried out on the effects of
incubation activity (none/verbal/spatial) on both uses totals
and uses novelty for the first-5-min and the last-2-min work
periods. The mean scores are shown in Table 2.
Both the first-5-min and the last-2-min scores for uses
totals and novelty were significantly different between the
incubation activity conditions (none/verbal/spatial): for the
first-5-min uses totals, F(2, 87) 0 3.29, p < .05, p2 0 .07,
and for the last-2-min uses totals, F(2, 87) 0 7.01, p < .01,
p 2 0 .14; similarly, for the first-5-min novel scores,
F(2, 87) 0 4.54, p < .05, p2 0 .09, and for the last-2-min
novelty scores, F(2, 87) 0 5.78, p < .01, p2 0 .12. Thus, as
would be expected, the effects of the immediate incubation
manipulation were apparent immediately, in the first 5 min,
and persisted into the final 2 min.
Effects of interpolation on the interpolated incubation period
tasks
As a check on the intermittent-conscious-work hypothesis,
we compared performance on the rotation and anagram
tasks when they were carried out in control conditions for
4 min and as interpolated tasks for 4 min in the incubation
conditions. The intermittent-work hypothesis makes a one-
tailed prediction that performance should be impaired on a
task when it is used as an interpolated incubation activity,
relative to controls, as participants would be distracted from
the interpolated task by the main target task if they were
intermittently working on the main task during incubation.
However, from Fig. 3 it appears that carrying out mental
rotation as an interpolated task during incubation periods did
not impair correct mental-rotation performance, and t tests
found no significant differences between interpolated and
control performance. Also, there were no significant impair-
ments between anagram solution rates when anagrams were
done as an incubation activity or as a stand-alone activity.
The possibility of fatigue effects for the control groups,
who did the alternative-uses task, mental rotations, and
Control (N 0 47) Verbal (N 0 22) Spatial (N 0 25)
M SD M SD M SD
Last 2 min/Postincubation Brick Uses
1.85 1.25 2.21 1.06 2.76 1.85
Last 2 min/Postincubation Brick Novelty
0.77 0.78 1.36 0.95 1.56 1.22
972 Mem Cogn (2012) 40:966975

Table 2 Immediate incubation: First-5-min and last-2-min scores for total brick uses and brick use novelty over incubation task conditions
(control/verbal/spatial)

Control (N 0 30) Verbal (N 0 30) Spatial (N 0 30) Control (N 0 30) Verbal (N 0 30) Spatial (N 0 30)

M SD M SD M SD M SD M SD M SD

First 5 min Brick Uses Last 2 min Brick Uses

8.67 3.97 11.23 3.92 11.54 6.02 1.50 1.41 3.03 1.84 2.83 2.14
First 5 min Brick Novelty Last 2 min Brick Novelty
3.00 2.90 4.37 2.91 5.53 3.88 0.73 0.86 1.83 1.53 1.63 1.50

anagrams, should be considered, as possibly depressing
their performance and thus masking any effects of intermit-
tent work for the experimental groups. The control partic-
ipants did the tasks in counterbalanced orders. An ANOVA
revealed no significant order effects for any of the tasks.
That is, the control scores were not depressed due to possi-
ble fatigue effects, and the lack of significant differences
between the control and incubation groups on the interpo-
lated tasks did not reflect fatigue. The control anagram and
rotation scores tended to be lower than the incubation
groups scores, but not significantly.
It may be suggested that the participants did not give their
full attention to the rotation task, given the correct rate of about
four rotations in 4 min. The numbers of rotation items attemp-
ted were, of course, higher than the correct rates, with means of
6.68 (SD 0 2.62) in the control and 7.12 (SD 0 3.14) in the
relevant incubation condition (delayed), and these figures were
not significantly different. In the case of anagrams, the delayed-
incubation group attempted more anagrams than did the con-
trols, with means of 18.91 (SD 0 9.02) and 14.06 (SD 0 5.86),
respectively, F(1, 67) 0 7.02, p < .01, although the groups did
not differ in numbers correct. In the immediate-incubation
conditions, again, slightly more items were attempted for rota-
tions in the incubation condition (M 0 9.17, SD 0 2.81) than in
the controls (M 0 8.87, SD 0 3.95), and the numbers of
anagrams attempted were very similar in the incubation
(M 0 12.58, SD 0 7.57) and control (M 0 12.70, SD 0 6.89)
conditions; these differences were not significant. We may
note that, as with the correct scores, these results for anagrams
and rotations attempted in the control and incubation condi-
tions are generally counter to the one-tailed prediction of the
intermittent-work hypothesis, that performance should be im-
paired on interpolated tasks relative to controls.
Although the type of interpolated activity in the incubation
periods did not seem to affect the level of the groups use
performance, it may be that, over participants, those who gave
more attention to the interpolated tasks might have done worse
on the target tasks, as they would have had less scope for
intermittent work on the target task during incubation than
those who attended less to the interpolated tasks. Thus,
according to the intermittent-work hypothesis, negative corre-
lations might be expected between performance on the inter-
polated and target tasks. In the immediate-incubation
conditions, the correlations for anagrams correct in incubation
were r(28) 0 .19, n.s., with total uses, and .11, n.s., with
novel uses; for rotations correct in incubation, the correlations
were r(28) 0 .31, n.s., with total uses, and .36, p < .05, with
novel uses. In the delayed-incubation conditions, the correla-
tions for uses totals after the incubation period were
r(23) 0 .11, n.s., with anagrams correct in incubation, and
r(20) 0 .03, n.s., with rotations correct in incubation; and
finally, for novel uses after the incubation period, the correla-
tions were r(23) 0 .07, n.s., with anagrams correct, and .18,
n.s., with rotations correct in incubation. The only significant
correlation (in two-tailed tests) out of the eight was in the
direction opposite from the one predicted by the intermittent-
work hypothesisthat is, positive rather than negative.

Discussion

First, it seems that the intermittent-work hypothesis can be

Fig. 3 Mental rotation and anagram performance when carried out
ruled out, since under that hypothesis we would have expected
as an interpolated (incubation) task or as a control task. Error bars an impairment of performance on the anagram and rotation
represent 1 SEM tasks when they were performed as interpolated activities

44
Mem Cogn (2012) 40:966975
during the incubation periods, relative to when they are per-
formed as control activities. No such negative effects were
found. If anything, the effects were in the direction opposite
the one predicted by the intermittent-work hypothesis. Fur-
thermore, the intermittent-work hypothesis predicts negative
correlations between performance on the interpolated tasks
and on the target-uses tasks, but no significant correlations
were found in the predicted direction in any of the incubation
conditions. Indeed, only two of the eight correlations were
even in the predicted, negative direction. Thus, we conclude
that the effects of incubation found here cannot be explained
by the intermittent-work hypothesis.
Immediate incubation produced better performance than
the control condition, which constructively replicates
Dijksterhuis and Meurss (2006) finding of immediate-
incubation effects with a clearly creative divergent-
thinking task requiring novel uses, and our result held over
two types of incubation activity (spatial and verbal). Fur-
thermore, immediate incubation was more efficacious than
delayed incubation for the creative task used here. The
different effects of immediate and delayed incubation sug-
gest that different process mixtures are involved in the two
forms of incubation. A possible interpretation is that, with
delayed incubation, in which conscious work is carried out
for a period before incubation, relatively strong sets could
build up, so the delayed incubation period could involve
both beneficial forgetting and unconscious work. Thus,
delayed incubation is handicapped relative to immediate
incubation, in which sets would be expected to be nonex-
istent, or at least weaker, because sets would have less time
to be established and strengthened. The immediate incuba-
tion period thus could involve only unconscious work,
without the need to overcome sets. We found immediate
incubation followed by conscious work to be better for
creative performance in the uses task than was delayed incu-
bation after conscious work, which is opposite Nordgren et
al.s (2011) finding with a decision task, and presumably
reflects differences between divergent creative thinking as
compared to convergent decision making. The decision task
required participants to absorb a number of facts about the
options, a stage that may benefit from initial conscious study;
in contrast, the uses task draws on already stored semantic
memory of object characteristics and of the requirements for
various functions.
Our conclusion in favor of the unconscious-work hypoth-
esis as a viable mechanism is based on the benefits of imme-
diate incubation, in which sets are unlikely to have been
developed and in which we found no evidence for intermittent
work. This leaves unconscious work as the likeliest explana-
tion for the benefits of immediate incubation.
As mentioned in the introduction, the question arises of
what form this unconscious work might take. Is it possible
that unconscious work could be just like conscious work,
45
973
but carried out without conscious awareness? Perhaps un-
conscious work would be better thought of as involving
spreading activation along associative links, rather than as
a rule- or strategy-governed activity?
To explore the idea that unconscious work might be a
subliminal version of conscious work, it would seem useful
to consider the nature of conscious processing in the
alternative-uses task. This issue was addressed in a think-
aloud study by Gilhooly et al. (2007), in which they found
that participants used strategies such as scanning an objects
properties (e.g., a brick is heavy) and using the retrieved
properties to cue uses (e.g., heavy objects can hold down
things like sheets, rugs, tarpaulin, etc., so a heavy brick
could do those things, too). However, it seems unlikely
that unconscious work could simply duplicate the form of
conscious work without awareness. The standard views in
cognitive science are (a) that mental contents vary in acti-
vation levels, (b) that, above some high activation level,
mental contents become available to consciousness, (c) that
we are conscious of only a limited number of highly acti-
vated mental elements at any one time (that is, the contents
of working memory), and (d) that strategy- or rule-based
processing, as found in Gilhooly et al.s think-aloud study,
requires such highly activated (conscious) material as inputs
and generates highly activated (conscious) outputs. That is,
the kind of processing involved in conscious work requires
the highly activated contents of working memory, of which
we are necessarily aware, given that material is in con-
sciousness if and only if it is above a high activation thresh-
old. Thus, it seems logically impossible that unconscious
processes could duplicate conscious processes in every re-
spect and yet remain unconscious. For example, using rules
and working memory to multiply two three-digit numbers
(e.g., 364 279 0 ?) seems impossible without having in
working memory highly activated representations of the
numbers, the goal, and the intermediate results, and such
representations are necessarily conscious. Unconscious
multiplication of even moderately large numbers, not previ-
ously practiced, seems impossible. (With practice, of course,
it would be possible to store many three-digit multiplication
results in long-term memory that could then be directly
retrieveda type of unconscious processbut this is not
the same as mental multiplication.) Overall, then, we dis-
count the idea that unconscious work or thought could be
just the same as conscious work minus awareness of any
mental content. What, then, might unconscious work consist
of? Many theorists, such as Poincar (1913), Campbell
(1960), and Simonton (1995), have argued that unconscious
work in incubation involves the quasi-random generation of
associations between mental elements to produce novel
combinations of ideas, some of which may be useful. Pro-
cesses such as parallel spreading activation through a
semantic network could serve to form remote and unusual
974
associations (Jung-Beeman et al., 2004) without requiring
activation levels to rise above the threshold of conscious-
ness. In Hlie and Suns (2010) recent explicitimplicit
interaction model, incubation is regarded as involving un-
conscious implicit associative processes that demand little
attentional capacity, in contrast with conscious, explicit,
rule-governed attentionally demanding processes. Accord-
ing to Dijksterhuis and Nordgrens (2006) unconscious-
thought theory (UTT), unconscious thought, or work, has
the following characteristics: It is parallel, bottom-up, inex-
act, andimportantly for the present studydivergent,
whereas conscious thought is serial, exact, and generally
convergent. There is broad agreement among a number of
theorists that unconscious thinking, or workin the form of
implicit associative processes based on spreading activation
is a possible explanation of incubation effects. According
to the unconscious-work view then, a beneficial effect of
immediate incubation would be expected, as a useful foun-
dation of novel associations could be formed by spreading
activation and could be highly accessible when the use-
reporting stage begins.
A possible difficulty of our results for the unconscious-
work hypothesis is that it predicts that an incubation period
on a presumed verbal task, such as the brick-uses task,
should be more beneficial if the interpolated task is nonver-
bal rather than verbal. The rationale for this prediction is that
verbal processing resources would be invoked in work on a
verbal interpolated task, thus depleting the resources avail-
able for simultaneous unconscious work on the target task.
A spatial interpolated incubation task would not compete
with simultaneous unconscious verbal activity, and so
should produce stronger incubation effects for a verbal main
task. Hlie and Suns (2010) explicitimplicit interaction
model of creative thinking explicitly makes this prediction
and draws on supporting results; however, these results
came from reminiscence memory rather than creative-
thinking tasks. The selective-forgetting mechanism of the
fresh-look approach, on the other hand, leads to opposite
predictions regarding the effects of interpolated tasks. How-
ever, neither hypothesis, both of which could apply to the
delayed-incubation condition, was supported, as the type of
interpolated activity did not affect target task performance.
Thus, the present results did not support the predictions of
the unconscious-work or of the fresh-look (selective-forget-
ting) hypotheses regarding the effects of the type of incuba-
tion activity. In this regard, our results on the effects of type
of interpolated activity are contrary to those of Ellwood et
al. (2009), which were in line with the unconscious-work
hypothesis. However, some differences between the present
study and that of Ellwood et al. may be relevant. For one
thing, Ellwood et al. did not inform their participants that the
target task would be returned to after incubation, whereas in
our study the goal of returning was stated; this might be an
46
Mem Cogn (2012) 40:966975
important factor affecting the incubation process. Future
studies will address this issue. A second major difference
is that Ellwood et al. used a fluency task that simply re-
quired the reporting of uses for a piece of paper, rather than
original or novel uses. That Ellwood et al.s task did not tap
creativity is indicated by the reported lack of correlation
between performance on their target task and the personality
characteristic of openness on a Big-5 personality test,
because openness typically correlates well with creative
divergent-test performance (Batey & Furnham, 2006). Also,
it may be noted that our results included novelty scores,
which Ellwood et al.s did not.
Another explanation for the lack of any effect of the type of
incubation activity in the present study is that we may have
misclassified the uses task as a purely verbal task. Indeed,
Gilhooly et al. (2007) did find protocol evidence of imagery
processes in the uses task, and it may be that the uses task is
better conceived as invoking both verbal and spatial processes.
If so, both types of incubation activity could have similar
effects, according to both the unconscious-work and
selective-forgetting hypotheses for delayed incubation. Future
research will aim to address this point by using creative tasks
that are more purely spatial (e.g., mental synthesis with shapes;
Pearson, Logie, & Gilhooly, 1999) or verbal (e.g., mental
synthesis with words; Haught & Johnson-Laird, 2003).
Finally, we note that our results have a clear practical
application. When faced with a task requiring that familiar
objects be used in new ways, it seems that it would be helpful
for respondents to put aside the task immediately and return to
it after a period, allowing unconscious incubation processes to
operate, before undertaking conscious work.
Author note This work was supported by a grant from the U.K.
Economic and Social Research Council (RES-000-22-2191) to K.J.G.
References
Acker, F. (2008). New findings on unconscious versus conscious
thought in decision making: Additional empirical data and meta-
analysis. Judgment and Decision Making, 3, 292303.
Batey, M., & Furnham, A. (2006). Creativity, intelligence, personality:
A critical review of the scattered literature. Genetic, Social, and
General Psychology Monographs, 132, 355429.
Boden, M. A. (2004). The creative mind: Myths and mechanisms (2nd
ed.). London: Routledge.
Campbell, D. T. (1960). Blind variation and selective retention in
creative thought as in other knowledge processes. Psychological
Review, 67, 380400.
Csikszentmihalyi, M. (1996). Creativity: Flow and the psychology of
discovery and invention. New York, NY: HarperCollins.
Dijksterhuis, A., & Meurs, T. (2006). Where creativity resides: The
generative power of unconscious thought. Consciousness and
Cognition, 15, 135146.
Dijksterhuis, A., & Nordgren, L. F. (2006). A theory of unconscious
thought. Perspectives on Psychological Science, 1, 95109.
Mem Cogn (2012) 40:966975
Dodds, R. A., Ward, T. B., & Smith, S. M. (in press). A review of the
experimental literature on incubation in problem solving and
creativity. In M. A. Runco (Ed.), Creativity research handbook
(Vol. 3). Cresskill, NJ: Hampton Press.
Ellwood, S., Pallier, P., Snyder, A., & Gallate, J. (2009). The incuba-
tion effect: Hatching a solution? Creativity Research Journal, 21,
614.
Ghiselin, B. (1952). The creative process: A symposium. New York,
NY: Mentor.
Gilhooly, K. J., Fioratou, E., Anthony, S. H., & Wynn, V. (2007). Divergent
thinking: Strategies and executive involvement in generating novel
uses for familiar objects. British Journal of Psychology, 98, 611625.
Gilhooly, K. J., & Hay, D. (1977). Rated characteristics of 205 five
letter words having single solution anagrams. Behavior Research
Methods & Instrumentation, 9, 1217.
Guilford, J. P. (1971). The nature of human intelligence. New York,
NY: McGraw-Hill.
Guilford, J. P., Christensen, P. R., Merrifield, P. R., & Wilson, R. C.
(1978). Alternate uses: Manual of instructions and interpreta-
tions. Orange, CA: Sheridan Psychological Services.
Haught, C., & Johnson-Laird, P. N. (2003). Creativity and constraints:
The production of novel sentences. In R. Alterman & D. Kirsh
(Eds.), Proceedings of the 25th Annual Meeting of the Cognitive
Science Society (pp. 528532). Mahwah, NJ: Erlbaum.
Hlie, S., & Sun, R. (2010). Incubation, insight, and creative problem
solving: A unified theory and a connectionist model. Psychological
Review, 117, 9941024. doi:10.1037/a0019532
Hlie, S., Sun, R., & Xiong, L. (2008). Mixed effects of distractor tasks
on incubation. In B. C. Love, K. McRae, & V. M. Sloutsky (Eds.),
Proceedings of the 30th Annual Meeting of the Cognitive Science
Society (pp. 12511256). Austin, TX: Cognitive Science Society.
Jung-Beeman, M., Bowden, E. M., Haberman, J., Frymiare, J. L.,
Arambel-Liu, S., Greenblatt, R., . . . Kounios, J. (2004). Neural
activity when people solve verbal problems with insight. PLoS
Biology, 2, e97. doi:10.1371/journal.pbio.0020097
Newell, B. R., Wong, K. Y., Cheung, J. C. H., & Rakow, T. (2009).
Think, blink or sleep on it? The impact of modes of thought on
complex decision making. Quarterly Journal of Experimental
Psychology, 62, 707732.
Nordgren, L. F., Bos, M. W., & Dijksterhuis, A. (2011). The best of both
worlds: Integrating conscious and unconscious thought best solves
complex decisions. Journal of Experimental Social Psychology, 47,
509511.
47
975
Payne, J., Samper, A., Bettman, J. R., & Luce, M. F. (2008). Boundary
conditions on unconscious thought in complex decision making.
Psychological Science, 11181123.
Pearson, D. G., Logie, R. H., & Gilhooly, K. J. (1999). Verbal repre-
sentations and spatial manipulations during mental synthesis.
Journal of Cognitive Psychology, 11, 295314.
Peters, M., Laeng, B., Latham, K., Jackson, M., Zaiyouna, R., &
Richardson, C. (1995). A redrawn Vandenberg and Kuse Mental
Rotations Test: Different versions and factors that affect perfor-
mance. Brain and Cognition, 28, 3958.
Poincar, H. (1913). The foundations of science. New York, NY:
Science House.
Quellmalz, E. (1985). Test review of Alternate Uses. From J. V. Mitchell,
Jr. (Ed.), The ninth mental measurements yearbook [Electronic
version]. Retrieved from Buros Institutes Test Reviews Online web-
site: www.unl.edu/buros
Rey, A., Goldstein, R. M., & Perruchet, P. (2009). Does unconscious
thought improve complex decision making? Psychological Research,
73, 372379.
Segal, E. (2004). Incubation in insight problem solving. Creativity
Research Journal, 16, 141148.
Seifert, C. M., Meyer, D. E., Davidson, N., Patalano, A. L., & Yaniv, I.
(1995). Demystification of cognitive insight: Opportunistic assimi-
lation and the prepared-mind perspective. In R. J. Sternberg & J. E.
Davidson (Eds.), The nature of insight (pp. 65124). Cambridge,
MA: MIT Press.
Simon, H. A. (1966). Scientific discovery and the psychology of problem
solving. In R. Colodny (Ed.), Mind and cosmos. Pittsburgh, PA:
University of Pittsburgh Press.
Simonton, D. K. (1995). Foresight in insight? A Darwinian answer. In
R. J. Sternberg & J. E. Davidson (Eds.), The nature of insight.
Cambridge, MA: MIT Press.
Sio, U. N., & Ormerod, T. C. (2009). Does incubation enhance problem
solving? A meta-analytic review. Psychological Bulletin, 135, 94120.
Snyder, A., Mitchell, J., Ellwood, S., Yates, A., & Pallier, G. (2004).
Nonconscious idea generation. Psychological Reports, 94, 13251330.
Wallas, G. (1926). The art of thought. New York, NY: Harcourt Brace.
Weisberg, R. W. (2006). Creativity: Understanding innovation in
problem solving, science, invention, and the arts. New York,
NY: Wiley.
Zhong, C.-B., Dijksterhuis, A., & Galinsky, A. D. (2008). The merits
of unconscious thought in creativity. Psychological Science, 19,
912918.
Social Cognition, Vol. 29, No. 6, 2011, pp. 683698
McMahon et al.
Driven to Distraction

DRIVEN TO DISTRACTION:
THE IMPACT OF DISTRACTER TYPE ON
UNCONSCIOUS DECISION MAKING
Kibby McMahon, Betsy Sparrow, Ljubica Chatman, and Travis Riddle
Columbia University

While there is a variety of wonderful ways to take a break from work, is

there a type of distraction that is actually more productive? Studies on the
deliberation-without-attention effect (Dijksterhuis, 2004; Dijksterhuis, Bos,
Nordgren, & Van Baaren, 2006) show that a period of distraction while
making complex decisions can actually lead to better decisions than a pe-
riod of conscious deliberation. Although there are a number of activities
that will distract participants from decision-making tasks, we investigated
if certain types of distracter tasks are better for demonstrating the deliber-
ation-without-attention effect. Since most people realistically take a break
with undemanding forms of activity, we hypothesized that undemanding
distracter tasks will yield the best results. In Experiment 1, participants
were given a choice between four different cars but before they made their
decision they instructed (1) to think consciously about the choices, (2) lis-
ten to their own music with a portable digital music player, (3) solve an
anagram puzzle, (4) solve a word search puzzle, or (5) just make a decision
immediately. In Experiment 2, participants were given a choice among four
different applicants for a hypothetical graduate school program and were
instructed (1) to think consciously about the applicants, (2) to engage in a
listening task with music, or (3) to just listen to the music. As predicted, the
participants who were distracted with the easier tasks (listening to music
and word search puzzles) made the best decision significantly more of-
ten than conscious thinkers and even outperformed participants distracted
with more difficult tasks.

Most of us have experienced mulling over a difficult decision or problem that ap-
peared impossible to solve when the answer just came to us only after putting
the matter aside for a while. Taking a break from tasks like complex decisions or
insight problems has shown to be actually productive for solving them (Beeftink,

Address correspondence to Kibby McMahon, 406 Schermerhorn Hall, 1190 Amsterdam Avenue
MC 5501, New York, NY 10027. Email: kibbymcmahon@gmail.com

2011 Guilford Publications, Inc.

683

48
684 MCMAHON ET AL.

Van Eerde, & Rutte, 2008; Dijksterhuis & Nordgren, 2006; Paulus, Nakui, Putman,
& Brown, 2006; Smith, 1995; Wallas, 1926), but are all forms of distraction created
equal? When you take a break for a few minutes, it feels very different to relax
with some music than to start busying yourself with something else on your to-do
list. In this article we examine types of distracters and their impact on decision
making.
Dijksterhuis and colleagues pioneered the work on what they call the deliber-
ation-without-attention effect (Dijksterhuis, 2004; Dijksterhuis, Bos, Nordgren, &
Van Baaren, 2006), which leads to improved decision making for complex deci-
sions. Specifically, they have shown that when participants are distracted from a
decision-making task involving a lot of pertinent information, they are more likely
to make the best decision than if they thought about it consciously. The common
paradigm Dijksterhuis et al. use consists of a pre-acquisition phase, followed by an
acquisition phase, an incubation period, and then the actual decision. In the pre-
acquisition phase, subjects are informed that they will read over the descriptions
of four different items and will have to choose the best one later in the study. After
reading over the descriptions of four different items, participants are assigned to a
conscious thought or an unconscious thought condition. Participants assigned
to the conscious thought condition are told to carefully consider the options before
making the choice. Those in the unconscious thought condition engage in an un-
related task before making their choice. Once the incubation period is finished, all
subjects indicate which item they think is best out of all four. Results consistently
show that participants in the unconscious thought condition make decisions or
judgments that favor the best choice more often than participants in the conscious
condition.
While there is evidence for the benefits of distraction, the quality of distraction
must be taken into account. Requiring participants to engage in another difficult
task in addition to the decision-making task adds cognitive demands (Hart, 1986;
Norman & Bobrow, 1975), which can lead to information overload that actually
hinders processing (Chewning & Harrell, 1990; Huber, 1985; Speier, Valacich, &
Vessey, 1999). Within a review of attention allocation and task demands, Kahne-
man (1973) mentions that when attending to different tasks, the type of one task
can influence the performance on the other. Working on two tasks can create at-
tentional conflict, which can be either structural interference if the content of the two
tasks are similar, or capacity interference if the task difficulties compete for cognitive
resources. The Distraction-Conflict Theory (Baron, 1986) argues that if a distract-
ing task is difficult enough to create such an attentional conflict, the arousal and
information overload will inhibit performance on complex tasks. Thus the type
and difficulty of a distracting task would have an influence on the performance
of a concurrent decision-making task. These theories do not directly address the
issue of a separate unconscious decision-making process. However, it has been
argued that conscious and unconscious processes are not completely independent
systems but instead have an interactive relationship (Evans, 2006; Kahneman &
Fredrick, 2002; Reber, 1993). Devoting conscious resources to an unrelated task al-
lows unconscious thought to process a different task, but it might not be so simple
to claim that all forms of distraction yield the same results; what we devote our
attention to might influence how the unconscious processes perform.
To induce unconscious thought, Dijksterhuis and colleagues (Dijksterhuis, 2004;
Dijksterhuis et al., 2006) have used a myriad of distracter tasks such as word search,

49
DRIVEN TO DISTRACTION 685

anagrams, and n-back tasks. All of these tasks seem to accomplish the goal of dis-
tracting participants from the decision-making task enough to let unconscious
thought processing arrive at the best decision. There has not been any report that
the differences in the tasks have an impact on the deliberation-without-attention
effect even though they require slightly different cognitive processes to solve. To
solve a word search puzzle, one must hunt through a matrix of letters to find a
target word. Priming the category of a target word-withina-word search has no
significant impact on performance, suggesting that word search puzzles are visual
perceptual tasks that do not require any semantic activation of the target word
(Karlin & Bower, 1976). However, anagram puzzles consist of a list of scrambled
words that participants must identify and spell correctly (i.e., the correct answer to
lisa is sail). There is some evidence that solving anagram puzzles activates a
deeper level of processing (Rajaram & Roediger, 1993; Srinivas & Roediger, 1990).
These two puzzles can both successfully occupy conscious resources, but it is un-
clear if the difference in their cognitive demands would influence the concurrent
unconscious processes.
In addition to a lack of research into the previously used distracter tasks, there
has not been any work demonstrating that the deliberation-without-attention ef-
fect will hold for more ecologically valid distracters. One of the most ubiquitous
distracters in daily life are portable music players. As twentieth-century English
conductor Sir Thomas Beecham astutely claimed, The function of music is to re-
lease us from the tyranny of conscious thought. Rauscher, Shaw, and Ky (1993)
came across an interesting finding in their studies that suggested classical music
(specifically Mozart pieces) temporarily enhance spatial reasoning performance.
This was dubbed the Mozart Effect and sparked new interest in how listening
to music can improve the performance of various cognitive processes (Fairfield &
Cornoldi, 2007; Lesiuk & Teresa, 2010; Mammarella, Fairfield, & Cornoldi, 2007;
Rauscher et al., 1993; Rideout, Dougherty, & Wernert, 1998). An interesting pilot
study by Caldwell and Riby (2007) explored how listening to a subjects preferred
music genre affects the processing of novel information. Rock and classical musi-
cians listened to either classical or rock music while completing an oddball task, an
attention exercise in which subjects have to pick out when a new stimulus appears
on a computer screen. Their event-related potentials suggested that conscious pro-
cesses like working memory were not as strongly deployed to the oddball task if
the subject listened to their preferred music genre. If listening to music has such
a notable influence on cognitive abilities and attention, it is worth exploring how
well it can apply to an unconscious decision-making paradigm, especially since it
is such a popular form of distraction.
In these experiments, we compared the efficacies of the different distracter tasks
to see if the type or difficulty of the distraction has any influence on unconscious
decision making. We predicted that we would replicate Dijksterhuis results in
the overall comparison of conscious thought and unconscious thought conditions
for difficult decisions. Furthermore, we examined the nuances of different uncon-
scious thought conditions in Experiment 1. Taking into account the previously
mentioned literature on task difficulty and attentional conflict, we predicted that
undemanding tasks such as listening to music and solving word search puzzles
would lead to higher performance on the complex decision-making task than the
more challenging anagram puzzles. In Experiment 2, we attempted to replicate the
deliberation-without-without attention effect with music, except within a social

50
686 MCMAHON ET AL.

decision-making task. We also explored the influence of different levels of diffi-

culty, or cognitive load, within the same type of distracter task. We predicted that
an easier and less challenging distracter task would lead to better unconscious
decision-making performance.

EXPERIMENT 1

METHOD

Participants and Design. Seventy-four Columbia University students were re-

cruited through the introductory psychology courses or through flyers posted
around campus and compensated with course credit or $5. All participants were
randomly assigned one of five conditions: an immediate decision condition (n =
15), a conscious thought condition (n = 15), a word search condition (n = 14), a
portable music or iPod condition (n = 15), and an anagram condition (n = 15). The
latter three conditions all fall under the unconscious thought category, but use dif-
ferent types of distraction.
Distracter Task Difficulty. Subjective ratings of cognitive load and working mem-
ory tasks have been used as a reliable measure of difficulty (Bratfisch, Borg, & Dor-
nic, 1972; Paas, Van Merrinboer & Adam, 1994). The distracter tasks were judged
for their relative difficulty with 17 separate participants from the same pool of
Columbia University students. The word search, anagram, and iPod tasks were
completed by each participant in random order and then rated for level of diffi-
culty on a scale of 1 (not difficult) to 7 (very difficult). A repeated measures ANOVA
determined a significant difference in difficulty among the distracter tasks, F(2,17)
= 42.08, p < .01. Post hoc pair-wise comparisons using Bonferroni correction re-
vealed that anagram puzzles (M = 5.18, SE = 1.43) were rated as significantly more
difficult than the word search task (M = 3.94, SE = 1.64) at p< .04 and the iPod-lis-
tening task (M = 1.35, SE = .86) at p < .01. The word search task was also considered
significantly more difficult than the iPod-listening at p < .001

PROCEDURE

Participants were seated at a lab computer that used the experimental program-
ming software Medialab (Jarvis, 2004). All participants were informed that they
would be presented with descriptions of four different cars and would have to
choose the car that is the best overall. As in Dijksterhuis (2004) study 2, partici-
pants were then presented with the description of the cars for 10 seconds each.
Each car had the same 12 attributes: handling, gas mileage, impact on the environ-
ment, legroom, quality of sound system, quality of service, difficulty to shift gears,
trunk size, age of car, choice in color, sunroof, and cup holders. These attributes
could either be positive (e.g.,The Hatsdun has good gas mileage) or negative
(e.g., The Nabusi has bad gas mileage). The best car, the Hatsdun, had 75% posi-
tive attributes, the medium cars had 50% positive attributes, and the worst car had
25% positive attributes. Even though this proportion of positive and negative at-
tributes makes the Hatsdun the overall best choice, we took into account the pos-
sibility that some car attributes are more important than others. The distribution of

51
DRIVEN TO DISTRACTION 687

TABLE 1. The Valence of the Attributes for Each Car

The Hatsdun The Nabusi The Kaiwa The Dasuka
Gas Mileage + - - -
Handling - - - +
Environment + + + -
Sound System + - - +
Service + - + -
Shifting Gears + - + +
Trunk Size + - + -
Legroom - + + +
New/Old + + - -
Colors + - + -
Sunroof - - - +
Cup Holders + - - +

positive and negative attributes of each car was based on Rey, Goldstein, and Per-
ruchets (2009) attribute weighting. His survey results provide a mean value of in-
fluence for each type attribute according to how important it is. When an attribute
is positive (e.g., The Hatsdun has good mileage) the mean value of influence
score is positive, and when the attribute is negative (e.g., The Nabusi has poor
handling) the mean value of influence is negative. A mean evaluation score for
each car was calculated from averaging the values of all the attributes. The higher
the mean evaluation score, the better the car is. Based on Reys attribute values,
our version of the Hatsdun had a mean evaluation score of 5.9, the Kaiwa had a
score of .5, the Dasuka had a score of -1.1, and the Nabusi had the lowest score of
-5.4. The positive and negative attribute distribution is displayed in Table 1.
After participants read the descriptions of the four cars, they were assigned to
the immediate, the conscious thought, the iPod, the anagram, or the word search
condition. In the immediate condition, participants chose the best car right after
they read the four descriptions. In the conscious thought condition, participants
were told to spend 3 minutes carefully considering the four different cars before
making their decision. The iPod, anagram, and word search conditions are all dif-
ferent forms of unconscious thought conditions. For the iPod condition, they were
told to listen to any song they liked on their portable music player with head-
phones for 3 minutes (all participants were asked to bring in a portable music
player in case they were randomized to this condition). In the word search and
anagram conditions, participants spent 3 minutes solving the assigned puzzle.
Once the 3-minute conscious thinking or distraction period was over, partici-
pants indicated which car they thought was the best overall out of all four in mul-
tiple choice format. They were then debriefed, compensated, and dismissed.

RESULTS AND DISCUSSION

The percentages of participants in each condition who chose the best car, the Hats-
dun, are displayed in Figure 1. We used a generalized linear model to analyze
this binary data. As we expected, we found an overall significant main effect of
thought, Wald chi-square = 13.29, df = 4, p < .02, indicating that some types of

52
688 MCMAHON ET AL.

FIGURE 1. Percentages shown are based on the proportion of subjects who chose the Hatsdun,
the best choice, to the total number of subjects in each condition.

thought condition performed better than others. Pair-wise comparisons revealed

that the word search condition significantly outperformed the conscious thought
condition (M = -.47, SE = .161, Z score= -2.9, df = 1, p < .01), as did the iPod con-
dition (M = -.45, SE = .166, Z score = -2.7, df = 1, p < .01). The participants in the
anagram condition did poorly and did not differ significantly from the conscious
condition (M = .07, SE = .154, Z score = .45, df =1, p =.67).
We also noted that participants in the immediate choice condition did signifi-
cantly worse than the word search condition (M = -.33, SE = .170, Z score = -1.9, df
= 1, p < .05), and marginally worse than the iPod condition (M = -.31, SE = .175, Z
score = -1.77, df = 1, p = .07). The immediate choice condition did not significantly
differ from the conscious thought (M = .13, SE = .170, Z score = .76, df = 1, p = .43) or
the anagram condition (M = .20, SE = .163, Z score = 1.23, df = 1, p = .22). This com-
parison provides additional evidence for the claim that the improved performance
of unconscious thought conditions is not due to using an on-line judgment made
during the acquisition phase. Since the immediate condition performance lies in
between the performances of unconscious and conscious thought, we are lead to
believe that it demonstrates the relative effectiveness of each kind of thought.
These results suggest that listening to music is an effective type of distraction
that can lead to improved decision making. Both the iPod and word search con-
ditions allowed subjects to take a break from the decision-making task with an
activity stimulating enough to capture their attention. We were surprised at how
poorly the anagram condition performed considering its success in previous work
(e.g., Dijksterhuis, 2004). While running the experiment, we noticed a difference
in students reactions to the distracter tasks. People in the word search and iPod
conditions looked engaged in their tasks and nothing seemed out of the ordinary.
After the incubation period in the anagram condition, however, most of the par-
ticipants appeared flustered and discouraged. Some of them hesitantly handed in
their anagram sheets, apologizing for not doing that well.
An interesting finding emerged when we explored the type of errors that con-
scious thinkers were making instead of choosing the Hatsdun. Further analyses
revealed that those in the underperforming conditions tended to select the first

53
DRIVEN TO DISTRACTION 689

FIGURE 2. Percentage of participants who chose the Kaiwa, the first car, out of all the
participants in each thought condition.

car presented, the Kaiwa. Percentages of car one (the Kaiwa) choices are shown in
Figure 2. The generalized linear model reveals no overall significance of thought
condition, Wald chi-square = 3.78, df =4, p = .44, but pair-wise comparisons reveal
that the conscious (M = -.40, SE = .126, Z score = -3.17, df = 1, p < .01), immediate
(M= -.33, SE = .122, Z score = 2.7, df = 1, p < .01), and anagram (M= -.33, SE = .122, Z
score = 2.7, df = 1, p < .01) conditions chose Kaiwa significantly more than the iPod
condition. The conscious condition also chose Kaiwa significantly more than the
word search condition (M= -.33, SE = .142, Z score = 2.32, df = 1, p < .01).
The Kaiwa does not have anything that would make it stand out from the other
cars, especially since it has negative attributes within the most important charac-
teristics. The only aspect worth noting of the Kaiwa is that it is the first car pre-
sented in the acquisition phase. Its popularity may be attributed to the primacy
effect, a serial order effect that biases judgment toward the first item presented in
a sequence (Deese & Kaufman, 1957). Heuristics reduce complex decision-making
and reasoning tasks into fast, automatic judgments based on previous experience
(Tversky & Kahneman, 1974), and streamline complex judgment and reasoning
processes by making use of a fewer, more select number of available cues (Shah &
Oppenheimer, 2008). In this way, deliberative conscious thought could exacerbate
the tendency to give some attributes too much weight within the decision, as Di-
jksterhuis and Nordgren (2006) describe in the unconscious thought theory. In Ex-
periment 2, we explored these heuristic-consistent errors as a secondary concern.

EXPERIMENT 2

The second experiment further replicates the deliberation-without-attention ef-

fect with music, but within a social decision-making task. Listening to portable
music proved to be an effective distracter that is not only ecologically valid, but
also leaves flexibility for different engagement and attention manipulations. We
compared conscious thought to both high and low cognitive load music listening
conditions to see if the levels of difficulty of the distracter task had any influence

54
690 MCMAHON ET AL.

on the deliberation-without-attention effect. In addition, we also investigated if

there were any consistent mistakes within incorrect responses.

METHOD

Participants and Design. Seventy-two Columbia University students participated

in this experiment for course credit or $5. All participants were taken from the
Columbia undergraduate population to increase the influence of in-group bias for
other Columbia students. Participants were randomly assigned to one of three
conditions: conscious thought condition (n = 24), easy iPod condition (n = 24), and
difficult iPod condition (n = 24).
Distracter Task Difficulty. As in Experiment 1, these two iPod (or unconscious
thought) tasks were rated on a 7-point difficulty scale by the same participants
who rated the other distracters, plus an additional Columbia University student
(for a total of 18 raters). A repeated measures ANOVA determined a significant
difference between among the tasks, F(1, 18) = 29.97, p < 0.01 (Mdifficult = 3.72, SE
= 1.87; Measy = 1.33, SE = .08). Subjects found that an added listening exercise in
the hard iPod condition provided a more difficult task than the condition in which
they could listen to music with no other demands in the easy iPod condition.
Procedure. Participants were seated at a computer that used the experimental
programming software Medialab (Jarvis, 2004). Participants were told that they
are in charge of admissions to a hypothetical graduate school that judges appli-
cants based on 12 attributes: Graduate Record Examination (GRE) percentile, un-
dergraduate university, grade point average (GPA), quality of recommendations,
prestige of the recommenders, interest in your particular institution, ability to meet
deadlines, honors, relevance of undergraduate major, ability to contribute ideas to
the program, email address, and amount of work experience. Participants were
then given full descriptions for four different applicants with all the attributes of
each applicant presented on the screen for 14 seconds each in randomized order.
Much like in the car context, these attributes could either be positive (e.g., Ra-
chel is good at meeting deadlines) or negative (e.g., Victoria is bad at meeting
deadlines). Rachel, the best applicant, had 75% positive attributes, Victoria was
considered the worst with 25% positive attributes, and the two medium applicants
both had 50% positive attributes. We based the distribution and relative attribute
weighting on the results of pretesting. These initial rankings were confirmed by
ranking scores the participants gave to each attribute (see Table 2).
Once the four applicant descriptions were presented, participants were random-
ly assigned to one of three conditions. In the conscious thought condition, they
were told to spend the following 3 minutes carefully considering the applicants.
The easy iPod condition was a kind of low load task in which the participants were
distracted with no performance or cognitive demands. The instructions they were
given were: you will take a break of three minutes. Please choose a song with
lyrics on your iPod by any band or artist you like. In the difficult iPod condition,
we wanted to use the same type of distraction except with an added motivation
to increase attention to the music. The participants were told, You will spend the
next three minutes listening carefully to a song with lyrics on your iPod by any
band or artist you like. First, write down the name of the song in the box provided

55
DRIVEN TO DISTRACTION 691

TABLE 2. The Valence of the Attributes for Each Applicant

Rachel Victoria James Peter
Grade Point Average + - - +
Recommendations + - + -
University - + + -
Graduate Record Examination - + + -
Work Experience + - - +
Ideas + + + -
Interest in Program - - + +
Honors + - - +
Prestige of Recommenders + - - +
Deadlines + - + -
Major + - - -
Email Address + - - +

and as you listen, remember and write down every word that begins or ends with
the letter T. Please list as many words as you can since your performance will be
scored.
After the 3-minute incubation period, participants were shown the applicant
descriptions for 1 second each as a quick review to prevent the participants from
forgetting which applicant was which. Then they had to choose the overall best
applicant in a randomized multiple-choice format. Afterward they filled out an
affect questionnaire and rated on a 7-point scale the levels of enthusiasm, stress,
interest, boredom, and alertness they felt during the experiment. They then ranked
the 12 application attributes in descending order of relative importance. Partici-
pants were then debriefed, compensated, and dismissed.

RESULTS AND DISCUSSION

For each condition, the percentages of participants who chose the best applicant
(Rachel) are displayed in Figure 3. We used a generalized linear model to assess
the probability of selecting Rachel as a function of the thought condition partici-
pants were randomly assigned to. We found a marginally significant main effect
of thought, Wald chi-square = 5.521, df = 2, p = .063, but pair-wise comparisons
revealed that the easy iPod condition performed significantly better than the con-
scious thought condition (M = -.33, SE = .136, Z score = -2.43, df =1, p < .01). The
difficult iPod condition marginally outperformed the conscious condition (M =
-.25, SE = .138, Z score = -1.81, df = 1, p = .06). As we hypothesized, both kinds of
distraction, easy and difficult, outperformed the conscious thought condition, but
only the easy iPod condition performed well enough to yield a significant differ-
ence. A comparison between the two iPod conditions did not yield a significant
difference (M = .08, SE = .142, Z score = .56, df = 1, p = .56). When participants take
a break with their own choice of music, their decisions reflect the performance of
unconscious thought when making complex decisions. This suggests that the de-
liberation-without-attention effect could be replicated in more ecologically valid

56
692 MCMAHON ET AL.

FIGURE 3. Percentage of participants who chose Rachel, the best applicant, of all the
participants in each thought condition.

methods of distraction. Although participants in both iPod conditions made better

decisions than conscious thinkers, a demand for more concentration on the dis-
tracter task in the hard iPod condition impaired the unconscious decision making.
The effect was not as strong as with the anagram condition in the previous experi-
ment, but the difficulty ratings suggest that the listening task was not as challeng-
ing as an anagram puzzle. However, as in Experiment 1, the participants who were
distracted with the easier task made the best decision most often.
Taking into consideration the pattern of incorrect decisions in Experiment 1, we
looked at which erroneous choice was the most popular in Experiment 2 and if
there was any implication of heuristic use. James was the less qualified applicant
and was only distinguished by his status as an alumnus of Columbia University.
Since all of the participants were members of the Columbia University community,
it can be argued that the decision making of the conscious thinkers were biased by
the in-group bias; that is, the tendency to favor members within ones own social
categories (Tajfel, 1970). An analysis of the number of people who chose James
as a function of type of thought condition revealed that there was a marginally
significant overall main effect of thought, Wald chi-square = 5.388, df = 2, p = .07,
but participants in the conscious thought condition chose James significantly more
often than those in the easy iPod condition (M = .33, SE = .131, Z score = 2.52, df =
1, p < .01). The percentages of participants who chose James and the percentages
of the participants who chose either of the other two incorrect responses within
each thought condition are displayed in Figure 4. These results are hard to inter-
pret considering the three conditions had unequal numbers of participants who
made an incorrect decision. However, within the incorrect decisions, there was a
trend toward choosing James more prominently in the conscious and difficult iPod
conditions. More exploration on heuristics and the deliberation-without-attention
effect would be necessary to draw any definite conclusions.

57
DRIVEN TO DISTRACTION 693

FIGURE 4. Percentage of participants who chose James, the Columbia applicant, compared
to the percentage who chose either Peter or Victoria, of all participants in each thought
condition.

AFFECT AND ATTRIBUTE RATINGS

The post-study questionnaires measured explicit reports of emotion and value

placed on the respective 12 applicant attributes. We ran one-way ANOVA analy-
ses to compare the effect of thought condition on each affect rating and only the
Alertness rating was significant, F (2)= 4.94, p < .02. A pair-wise comparison re-
vealed that the mean rating of Alertness in the difficult iPod condition (M = 1.75,
SE = .193) was significantly lower than in the conscious thought condition (M =
2.54, SE = .193). The participants in the easy iPod condition lay in between (M =
2.17, SE = .193) the two other conditions, feeling slightly less alert than conscious
thinkers but more than the difficult iPod participants. Although this is only based
on self-report, these results suggest that alertness during the experiment does not
definitively account for the differences in decision-making performance among
the conditions. It is possible that too much or too little attentiveness can impair
decision making, but the effect is too small to be determined as the mechanism of
deliberation-without-attention effect.
Each participant ranked each of the 12 attributes from 1 (most important) to 12
(least important) based on how influential the attributes were for assessing the
overall quality of the applicants. The means for university ratings are displayed
in Table 3; the lower the mean ranking score, the more important the attribute is
considered to be within an application. There was no significant effect of thought
condition on any of the rankings, suggesting that there are no major differences
in the explicit decision-making strategies the participants used to compare the ap-
plicants. However, we used these mean ranking scores to calculate the individual
applicants mean evaluation score. Rachel had a mean evaluation score of 2.84,
Victoria had a score of -2.74, Peter had a score of 1.01, and James had a score of .21.
Therefore, Rachel was confirmed as the best choice even taking into account that
some attributes of an application are considered more important than others.

58
694 MCMAHON ET AL.

TABLE 3. The Mean Ranking Scores of Each Application Attribute

Attribute Mean Standard Deviation
Grade Point Average 3.06 2.28
Recommendations 4.65 2.72
University 5.24 3.37
Graduate Record Examination 5.24 2.83
Work Experience 5.46 3.17
Ideas to Contribute 5.85 3.16
Interest in Program 6.44 3.08
Honors 7.01 2.19
Prestige of Recommenders 7.19 2.75
Ability to Meet Deadlines 7.46 2.82
Relevance of Major 8.75 2.73
Email Address 11.56 1.63

GENERAL DISCUSSION

We were interested in refining the idea of a distracter task within a deliberation-

without-attention context and determining which distractions are the most useful
in making complex decisions unconsciously. The first experiment was intended to
compare different types of distracters, especially to see if a popular form of distrac-
tion, listening to portable music, can yield the same unconscious thought perfor-
mance as other known tasks. The results of the experiment suggest that distracters
such as word searches and iPods lead to making better complex decisions than
thinking consciously about the decision. However, it was surprising how poorly
people in the anagram condition performed, as if the pressure the subjects felt to
do the anagram was enough to dampen unconscious thoughts abilities. We iso-
lated this factor of task difficulty in the second experiment by providing different
levels of difficulty to the participants while keeping the distracter task constant.
As we predicted, all the subjects who were distracted with music performed well
on the decision-making task, but only the subjects who listened to music with no
added cognitive demands made the best choice significantly more than conscious
thinkers. The type or level of difficulty that drives these effects is still unclear,
but the consistent high performance of participants with easy tasks suggests that
distractions with the least amount of cognitive demands are the optimal choice for
the deliberation-without-attention effect. Further work would have to determine
what type of distracter or the threshold of distracter difficulty would completely
obstruct unconscious processing.
If you recall a time when you took a break from a difficult situation, work as-
signment, or decision, most likely you engaged yourself in your favorite mindless
form of procrastination. In these studies, all the unconscious thought conditions
included an incubation period, but only the easy tasks of word searches and pas-
sive music listening really simulated how people take a real break. The feeling of
relaxation and low vigilance during these easy tasks could account for the high
decision-making performance. On the other hand, Yerkes & Dodson (1908) and
Zajoncs (1965) work on social facilitation both provide evidence that arousal over
a certain threshold can impede performance on challenging tasks. Therefore, the

59
DRIVEN TO DISTRACTION 695

added pressure that the anagram and difficult iPod conditions could have elevat-
ed subjects to a level of arousal that impaired their performance on the difficult
decision-making task. Conscious thought also has that added pressure of concen-
trating on the decision-making task, which would further impede performance.
The issue of affect in general has been raised as a confounding factor since people
listening to music may feel happier during the experiment. There is some evidence
that positive affect can improve performances on cognitive tasks such a decision
making and creative problem solving (Isen, Daubman, & Nowicki, 1987; Isen &
Means, 1983). To check for this, we had subjects report their levels of enthusiasm,
alertness, boredom, stress, and interest during the experiment. None of the mea-
sures showed a significant effect of positive affect among the thought conditions,
so we ruled this out as a possibility. The participants subjective ratings of alert-
ness do not support the argument that high arousal or attentiveness leads to lower
performance. Perhaps physiological measures can more clearly point to an effect
of arousal, but our lack of evidence leads us to believe that our data is less affected
by arousal and more a result of the different cognitive demands.
An exploratory look into the incorrect decisions revealed a trend toward the
heuristic-consistent choice. The Kaiwa car and James were choices that were oper-
ationally not the best option, but were appealing to those making decisions based
on shortcuts like the order of item presentation or the preference toward members
of your own social group. There is evidence that when people feel pressured to
complete a task and if the task is difficult, their attention resources are depleted
and they default to biased responses (Chaiken & Maheswaran, 1994; Kunda, 1990).
A series of studies by Pelham and Neter (1995) found an interaction with task dif-
ficulty and motivation on making heuristic-consistent errors. When subjects were
motivated to answer difficult versions of decision-making tasks, they were more
likely to rely on the heuristics and make inaccurate judgments than if they were
unmotivated within easy versions of the task. Although the heuristics data from
our studies are far from conclusive, it is interesting to note that the incorrect deci-
sions of the participants in the difficult distracter and conscious conditions tend to
be biased, a typical result of restricted attentional resources. Thinking consciously
about a complex decision or engaging in difficult distracters may tax attention
resources enough to reduce decision-making accuracy. These conditions lead to in-
formation overload, which negatively impacts task performance according to the
aforementioned capacity interference (Kahneman, 1973) and Distraction-Conflict
Theory (Baron, 1986). It questions the perspective that distractions activate only
unconscious thought and deliberation activates only conscious thought. Instead,
there is enough overlap within the two systems that the difficulty of the task dur-
ing incubation periods can influence the concurrent decision-making process.
Growing evidence for implicit processes within the field of working memory
further supports the interaction of unconscious and conscious processes. Given
that conscious attention is so limited, current working memory models claim that
consciousness is useful for processing stimuli input and activating the unconscious
networks to process this information (Baars, 2002, 2003; Baars & Ave, 1997). Recent
dual-process theories have proposed that we can make automatic judgments and
decisions, but conscious thought can monitor and modify their effect (Chaiken,
Liberman, & Eagly, 1989; Evans, 2006). These theories support the idea that ex-
plicit and implicit processes often work together instead of exist as two distinct
systems. The successful unconscious thought paradigms first require subjects to

60
696 MCMAHON ET AL.

attend to the necessary information on the four items before distraction can allow
for the unconscious working memory to process the decision. Attending to yet
another task might overwhelm the working memory and deplete both conscious
and unconscious resources. Our results suggest that the best way to facilitate the
deliberation-without-attention effect is to consciously attend to all of the available
information, then engage in a distracter task that is low in cognitive load and does
not require a lot of working memory or attention resources.

CONCLUSION AND IMPLICATIONS FOR FURTHER RESEARCH

Distracter tasks that are low in difficulty are shown to be the most successful in
demonstrating the deliberation-without-attention effect. Further replications of
this effect should take the quality of distracters into account, especially partici-
pants assessment of the tasks difficulty. Listening to portable music is an effec-
tive and accessible way to distract subjects without added pressure or too much
cognitive demand. Since it just provides auditory stimulation, it leaves room for
new unconscious manipulations that would need to present visual stimuli. For
example, subjects can listen to music while new applicant information is present-
ed or subliminally primed on the screen. Further studies can also more clearly
determine the mechanism for what makes the more difficult distracters like ana-
grams impede unconscious decision making. The performance of the hard iPod
participants showed that a more difficult distracter task can reduce unconscious
thoughts efficacy, but perhaps a task that requires a deeper level of processing
(complex memory tasks or other puzzles that require associative thinking) or adds
a higher level of cognitive load would eliminate unconscious thoughts advan-
tages in complex decision making.

REFERENCES

Baars, B. J. (2002). The conscious access hy-
pothesis: origins and recent evidence.
Trends in Cognitive Sciences, 6(1), 4752.
Baars, B. J. (2003). Working memory requires
conscious processes, not vice versa: A
global workspace account. In N. Osaka
(Ed.), Neural basis of consciousness. Phila-
delphia: John Benjamins.
Baars, B. J., & Ave, D. (1997). Global workspace
theory, a rigorous scientific theory of
consciousness. Journal of Consciousness
Studies, 4, 292309.
Baron, R. S. (1986). Distraction-conflict theory:
Progress and problems. Advances in Ex-
perimental Social Psychology, 19, 139.
Beeftink, F., Van Eerdea, W., & Rutteb, C.G.
(2008). The effect of interruptions and
breaks on insight and impasses: Do you
need a break right now? Creativity Re-
search Journal, 20, 358-364.
Bratfisch, O., Borg, G., & Dornic, S. (1972).
Perceived item-difficulty in three tests of
intellectual performance capacity (Rep. No.
29). Stockholm: Institute of Applied Psy-
chology.
Caldwell, G. N., & Riby, L. M. (2007). The ef-
fects of music exposure and own genre
preference on conscious and uncon-
scious cognitive processes: A pilot ERP
study. Consciousness and Cognition, 16,
992996.
Chaiken, S., Liberman, A., & Eagly, A. H.
(1989). Heuristic and systematic infor-
mation processing within and beyond
the persuasion context. In J. S. Uleman
& J. A. Bargh (Eds.), Unintended thought

61
DRIVEN TO DISTRACTION
(Vol. 16, pp. 212252). New York: Guil-
ford.
Chaiken, S., & Maheswaran, D. (1994). Heuris-
tic processing can bias systematic pro-
cessing: Effects of source credibility, ar-
gument ambiguity, and task importance
on attitude judgment. Journal of Persona-
lity and Social Psychology, 66, 460473.
Chewning, E. G., Jr., & Harrell, A. M. (1990).
The effect of information load on deci-
sion makers cue utilization levels and
decision quality in a financial distress
decision task. Accounting, Organizations,
and Society, 15, 527542.
Deese, J., & Kaufman, R. A. (1957). Serial effects
in recall of unorganized and sequential-
ly organized verbal material. Journal of
Experimental Psychology, 54, 180187.
Dijksterhuis, A. (2004). Think different: The
merits of unconscious thought in prefer-
ence development and decision making.
Journal of Personality and Social Psycholo-
gy, 87, 586598.
Dijksterhuis, A., Bos, M. W., Nordgren, L. F.,
& Van Baaren, R. B. (2006). On making
the right choice: The deliberation-with-
out-attention effect. Science, 311(5763),
10051007.
Dijksterhuis, A., & Nordgren, L. F. (2006). A
theory of unconscious thought. Perspec-
tives on Psychological Science, 1, 95109.
Evans, J. S. (2006). The heuristic-analytic theo-
ry of reasoning: Extension and evalua-
tion. Psychonomic Bulletin & Review, 13,
378395.
Huber, V. L. (1985). Effects of task difficulty,
goal setting, and strategy on perfor-
mance of a heuristic task. Journal of Ap-
plied Psychology, 70, 492504.
Isen, A. M., Daubman, K. A., & Nowicki, G. P.
(1987). Positive affect facilitates creative
problem solving. Journal of Personality
and Social Psychology, 52, 11221131.
Isen, A. M., & Means, B. (1983). The influence
of positive affect on decision-making
strategy. Social Cognition, 2, 1831.
Jarvis, B.G. (2004). MediaLab Research Soft-
ware, Version 2004 [Computer Pro-
gram]. New York: Empirisoft. (www.
empirisoft.com)
Kahneman, D. (1973). Attention and effort. En-
glewood Cliffs, NJ: Prentice-Hall.
Kahneman, D., & Frederick, S. (2002). Repre-
sentativeness revisited: Attribute sub-
stitution in intuitive judgement. In T.
62
697
Gilovich, D. Griffin, & D. Kahneman
(Eds.), Heuristics and biases: The psycholo-
gy of intuitive judgment (pp. 4981). Cam-
bridge: Cambridge University Press.
Karlin, M. B., & Bower, G. H. (1976). Semantic
category effects in visual word search.
Perception Psychophysics, 19, 417424.
Kunda, Z. (1990). The case for motivated
reasoning. Psychological Bulletin, 108,
480498.
Lesiuk, T. (2010). The effect of preferred music
on mood and performance in a high-
cognitive demand occupation. Journal of
Music Therapy, 47, 137154.
Mammarella, N., Fairfield, B., & Cornoldi, C.
(2007). Does music enhance cognitive
performance in healthy older adults?
The Vivaldi effect. Aging Clinical and Ex-
perimental Research, 19, 394399.
Norman, D. A., & Bobrow, D. G. (1975). On
data-limited and resource-limited pro-
cesses. Cognitive Psychology, 7, 44-64.
Paas, F., Van Merrinboer, J.J.G., & Adam, J. J.
(1994). Measurement of cognitive-load
in instructional research. Perceptual and
Motor Skills, 79, 419430.
Paulus, P. B., Nakui, T., Putman, V. L., & Brown,
V. R. (2006). Effects of task instructions
and brief breaks on brainstorming.
Group Dynamics: Theory, Research, and
Practice, 10, 206219.
Pelham, B. W., & Neter, E. (1995). The effect
of motivation of judgment depends on
the difficulty of the judgment. Journal
of Personality and Social Psychology, 68,
581594.
Rajaram, S., & Roediger III, H. L. (1993). Direct
comparison of four implicit memory
tests. Learning, Memory, and Cognition,
19, 765776.
Rauscher, F. H., Shaw, G. L., & Ky, K. N. (1993).
Music and spatial task performance. Na-
ture, 365(6447), 611.
Reber, A. S. (1993). Implicit learning and tacit
knowledge. Oxford: Oxford University
Press.
Rey, A., Goldstein, R. M., & Perruchet, P. (2009).
Does unconscious thought improve
complex decision making? Psychological
Research, 73, 372379.
Rideout, B.E., Dougherty, S., & Wernert, L.
(1998). Effect of music on spatial per-
formance: A test of generality. Perceptual
Motor Skills, 86, 512514.
698
Shah, A. K., & Oppenheimer, D. M. (2008).
Heuristics made easy: An effort-reduc-
tion framework. Psychological Bulletin,
134, 207222.
Smith, S. M. (1995). Fixation, incubation, and
insight in memory and creative think-
ing. In S. M. Smith, T. B. Ward, & R. A.
Finke (Eds.), The creative cognition ap-
proach (pp. 135156). Cambridge, MA:
The MIT Press.
Speier, C., Valacich, J. S., & Vessey, I. (1999).
The influence of task interruption on
individual decision making: An infor-
mation overload perspective. Decision
Sciences, 30, 337360.
Srinivas, K., & Roediger, H. L. (1990). Classify-
ing implicit memory tests: Category as-
sociation and anagram solution. Journal
of Memory and Language, 29, 389412.
63
MCMAHON ET AL.
Tajfel, H. (1970). Experiments in intergroup
discrimination. Scientific American, 223,
96102.
Tversky, A., & Kahneman, D. (1974). Judgment
under uncertainty: Heuristics and bias-
es. Science, 185(4157), 11241131.
Van Der Weiden, R.M.F. (2001). The Mozart ef-
fect. Journal of the Royal Society of Medici-
ne, 94, 553.
Wallas, G. (1926). The art of thought. New York:
Harcourt.
Yerkes, R. M., & Dodson, J. D. (1908). The re-
lation of strength of stimulus to rapid-
ity of habit-formation. Journal of Com-
parative Neurology and Psychology, 18,
459-482.
Zajonc, R. B. (1965). Social facilitation. Science,
149, 269-274.
 Psychological Science http://pss.sagepub.com/

Inspired by Distraction : Mind Wandering Facilitates Creative Incubation

Benjamin Baird, Jonathan Smallwood, Michael D. Mrazek, Julia W. Y. Kam, Michael S. Franklin and Jonathan W. Schooler
Psychological Science published online 31 August 2012
DOI: 10.1177/0956797612446024

The online version of this article can be found at:

http://pss.sagepub.com/content/early/2012/08/31/0956797612446024

Published by:

http://www.sagepublications.com

On behalf of:

Association for Psychological Science

Additional services and information for Psychological Science can be found at:

Email Alerts: http://pss.sagepub.com/cgi/alerts

Subscriptions: http://pss.sagepub.com/subscriptions

Reprints: http://www.sagepub.com/journalsReprints.nav

Permissions: http://www.sagepub.com/journalsPermissions.nav

>> OnlineFirst Version of Record - Aug 31, 2012

What is This?

Downloaded from pss.sagepub.com at UNIV CALIFORNIA SANTA BARBARA on September 26, 2012

64
 Psychological Science OnlineFirst, published on August 31, 2012 as doi:10.1177/0956797612446024

Research Report
Psychological Science

Inspired by Distraction: Mind Wandering XX(X) 16

The Author(s) 2012
Reprints and permission:
Facilitates Creative Incubation sagepub.com/journalsPermissions.nav
DOI: 10.1177/0956797612446024
http://pss.sagepub.com

Benjamin Baird1, Jonathan Smallwood2 , Michael D. Mrazek1,

Julia W. Y. Kam3, Michael S. Franklin1, and
Jonathan W. Schooler1
1
Department of Psychological and Brain Sciences, University of California, Santa Barbara; 2Department
for Social Neuroscience, Max Planck Institute for Human Cognitive Brain Sciences, Leipzig, Germany;
and 3Department of Psychology, University of British Columbia

Abstract
Although anecdotes that creative thoughts often arise when one is engaged in an unrelated train of thought date back
thousands of years, empirical research has not yet investigated this potentially critical source of inspiration. We used an
incubation paradigm to assess whether performance on validated creativity problems (the Unusual Uses Task, or UUT) can
be facilitated by engaging in either a demanding task or an undemanding task that maximizes mind wandering. Compared
with engaging in a demanding task, rest, or no break, engaging in an undemanding task during an incubation period led
to substantial improvements in performance on previously encountered problems. Critically, the context that improved
performance after the incubation period was associated with higher levels of mind wandering but not with a greater number
of explicitly directed thoughts about the UUT. These data suggest that engaging in simple external tasks that allow the mind
to wander may facilitate creative problem solving.

Keywords
creativity, consciousness, insight
Received 12/26/11; Revision accepted 3/29/12

Anecdotes of individuals solving problems after relinquishing
the effort to solve them date back millennia. Indeed, many
influential scientific thinkersincluding Newton, Poincar,
and Einsteinclaim to have had their moments of inspiration
while engaged in thoughts or activities not deliberately aimed
at solving the problem they were trying to solve. A key ques-
tion that arises from such examples is whether engaging in any
type of unrelated cognition increases the frequency of creative
solutions, or whether the thoughts that yield such insights have
specific features.
One common example of thinking that is unrelated to an
overt goal is the internally generated thought that occupies
ones attention during mind wandering (Smallwood & Schooler,
2006). Several lines of research suggest that mind wandering
could be linked to enhanced creativity, particularly for prob-
lems that have been previously encountered. First, individuals
with attention-deficit/hyperactivity disorder (which is known
to be associated with mind wandering; e.g., Shaw & Giambra,
1993) tend to score higher than individuals without ADHD on
laboratory measures of creativity (White & Shah, 2006) and on
questionnaire-based assessments of achievement in creative
areas (e.g., music, visual arts; White & Shah, 2011).
Second, focused deliberation on problems can undermine
creativity, whereas distraction can enhance creativity
(Dijksterhuis & Meurs, 2006). Third, a recent meta-analysis
of the conditions that maximize incubation effects (i.e.,
enhanced creative problem solving following a break) found
that the benefits of incubation intervals are greater when indi-
viduals are occupied by an undemanding task than when they
engage in either a demanding task or no task at all (Sio &
Ormerod, 2009). Given that mind wandering is more frequent in
undemanding tasks than in demanding tasks (e.g., Mason et al.,
2007; Smallwood, Nind, & OConnor, 2009), this finding sug-
gests that one feature that may characterize successful incuba-
tion intervals could be the opportunity for mind wandering.
Finally, a recent investigation found that when individuals
engaged in REM sleep during an incubation interval, they
showed enhanced integration of unassociated information
in the service of creative problem solving (Cai, Mednick,
Corresponding Author:
Benjamin Baird, Department of Psychological and Brain Sciences, Building
429, Room 102, University of California, Santa Barbara, CA 93106-9660
E-mail: baird@psych.ucsb.edu

Downloaded from pss.sagepub.com at UNIV CALIFORNIA SANTA BARBARA on September 26, 2012

65
2 Baird et al.
Harrison, Kanady, & Mednick, 2009). Although REM sleep is
very different from mind wandering, the fact that the formation
of associative networks during dreaming can lead to incubation
effects is certainly consistent with the prospect that the loose
associative processes of mind wandering (e.g., Smallwood,
Obonsawin, & Heim, 2003) might have similar effects.
However, caution must be taken in drawing firm conclu-
sions from the results of these studies; to date, no published
study has directly compared the effects of incubation intervals
of systematically varying difficulty within a single experi-
ment, nor has any study directly assessed the occurrence of
mind wandering during incubation. Furthermore, there are at
least two competing interpretations of the beneficial effects of
tasks with a light cognitive load: Easy tasks may simply allow
individuals a greater opportunity to explicitly think about pre-
vious problems, or easy tasks may encourage a global mental
set (e.g., Frster, Friedman, & Liberman, 2004) that might
facilitate creativity independently of any specific benefit of
incubation.
The study reported here used an incubation paradigm to
compare the effects of interpolated tasks that systematically
varied in their levels of attentional demand and thus in their
conduciveness to mind wandering. The tasks were interpo-
lated into the Unusual Uses Task (UUT), a classic and widely
used measure of divergent thinking (Guilford, 1967). The
UUT was selected because it yields particularly consistent and
robust incubation effects (Ellwood, Pallier, Snyder, & Gallate,
2009; Sio & Ormerod, 2009), unlike convergent-thinking
tasks (such as the Remote Associates Task), which have been
more prone to empirical inconsistencies (Vul & Pashler, 2007).
The UUT requires participants to generate as many unusual
uses as possible for a common object, such as a brick, in a set
amount of time. The originality of the responses is taken as an
index of creative thinking (e.g., Milgram & Milgram, 1976;
Torrance, 2008; Wallach & Kogan, 1965).
Following the procedure in Cai et al. (2009), we assessed
participants performance on UUT problems that were pre-
sented both before and after the incubation interval (repeated
exposure) and on UUT problems that were presented for the
first time after the incubation interval (new exposure). These
exposure conditions allowed us to distinguish between two
different types of improvements in problem solving: incuba-
tion effects (repeated-exposure condition), which correspond
to enhanced processing of previously encountered informa-
tion, and general increases in creative problem solving (new-
exposure condition), which could correspond to general
improvements in creative thinking or to other general facilita-
tive effects (e.g., arousal or fatigue).
We had four hypotheses for this study. First, we expected
that participants would exhibit more mind wandering in
an interpolated undemanding task than in an interpolated
demanding task, which would replicate previous findings that
attentional demand reduces mind wandering (Smallwood
et al., 2009). Given these anticipated differences in mind wan-
dering, we hypothesized, second, that the creative benefits of
Downloaded from pss.sagepub.com at UNIV CALIFORNIA SANTA BARBARA on September 26, 2012
66
incubation would be greater for participants who engaged in
the undemanding task than for participants who engaged in the
demanding task and, third, that this effect would not be attrib-
utable to a greater number of explicit thoughts about the previ-
ously encountered problems. Finally, we hypothesized that
performance would selectively improve on repeated-exposure
problems (i.e., not on new problems) following the undemand-
ing task, which would indicate that the performance improve-
ments resulted from an incubation process rather than a general
increase in creative problem solving.
Method
Participants
One hundred forty-five participants (35 males, 110 females)
completed the experiment (age range: 1932 years) as partial
fulfillment of a course requirement. Informed consent was
obtained from all participants, and ethical approval for the
study was obtained from the University of California, Santa
Barbara, institutional review board.
Procedure
Baseline UUT. Participants were randomly assigned to work
on two UUT problems (2 min per problem) in which they were
instructed to list as many unusual uses as possible for each
stimulus. Participants typed their responses on a computer,
directly into a text box that automatically expired after 2 min.
Incubation. After completing the baseline UUT, participants
were assigned to one of four between-subjects conditions, using
a counterbalanced design. In three of these conditions (demand-
ing task, undemanding task, and rest), the baseline UUT was
followed by an incubation period that lasted 12 min. Partici-
pants in the demanding-task condition performed a 1-back
working memory task that places a strong constraint on top-
down attention, whereas those in the undemanding-task condi-
tion performed a choice reaction time task (0-back) requiring
infrequent responses. Studies have shown that tasks without a
working memory load elicit more mind wandering than tasks
with a working memory load (e.g., Smallwood et al., 2009). In
the rest condition, participants were asked to sit quietly during
the incubation interval. Participants in the fourth condition (no
break) did not receive a break from the UUT.
Immediately following the incubation interval in the
demanding-task, undemanding-task, and rest conditions, we
administered a commonly used self-report measure of mind
wandering (e.g., Barron, Riby, Greer, & Smallwood, 2011;
Matthews et al., 1999) in order to confirm differences in mind-
wandering frequency between the two task conditions. (The
questionnaire was administered following the rest interval in
the rest condition in order to maintain consistency across incu-
bation conditions.) This questionnaire asks participants to rate
how often they engaged in different types of task-unrelated
thought, such as considering personal worries or future or past
Mind Wandering Facilitates Creative Incubation
events (rating scale from 1 to 5, with higher scores indicating
higher levels of mind wandering). To assess explicit thoughts
about the creativity task, we had participants in these three
conditions complete a separate questionnaire on the frequency
of their thoughts about the creativity problems during the incu-
bation interval.
Postincubation UUT. After the incubation interval (or fol-
lowing the baseline UUT, in the case of the no-break condi-
tion), participants were informed that they would work on the
UUT again. Four UUT problems (2 min per problem) were
presented in a random order: two repeat problems (repeated-
exposure condition) that were identical to the problems pre-
sented at baseline and two randomly assigned new problems
(new-exposure condition).
Assessing propensity to mind-wander. At the end of the
experiment, all participants completed the Daydreaming Fre-
quency subscale of the Imaginal Process Inventory (IPI),
which assesses individuals general propensity to mind-
wander (Singer & Antrobus, 1972).
Tasks
Interpolated tasks. Stimuli for the demanding and unde-
manding tasks were the digits from 1 through 9, which were
presented serially (in quasirandom order) in the center of a
computer screen for 1,000 ms each; each digit was followed
by a 1,500-ms fixation cross. In both of these tasks, nontargets
were black numbers that required no response, and nontargets
occurred frequently, whereas targets were infrequent. In the
undemanding task, targets were colored numbers, and partici-
pants had to determine whether each target stimulus was even
or odd. In the demanding task, targets were colored question
marks, and participants had to determine whether the stimulus
immediately preceding each target was even or odd. Partici-
pants in both conditions received a short practice session with
feedback.
UUT. Following the procedure used by Wallach and Kogan
(1965), we pooled responses to each UUT stimulus across the
sample, and points were assigned for statistically unique
responses.1 Percentage improvement on the UUT was calcu-
lated separately for each problem type (repeated exposure,
new exposure) and was compared across conditions (unde-
manding task, demanding task, rest, no break). This was calcu-
lated as [(postincubation UUT score baseline UUT score)/
(baseline UUT score)] 100 (see Cai et al., 2009, for a similar
analytic method). Percentage improvement was calculated at
the individual level and then averaged for each condition.
Although uniqueness scoring is the most standard method
of scoring divergent-thinking tasks (e.g., Milgram & Milgram,
1976; Torrance, 2008; Wallach & Kogan, 1965), it has been
criticized (Silvia et al., 2008) on the grounds that it may con-
found creativity with fluency (e.g., participants may receive
Downloaded from pss.sagepub.com at UNIV CALIFORNIA SANTA BARBARA on September 26, 2012
67
3
high creativity scores simply by virtue of generating a large
number of responses). Therefore, to assess fluency, we had
two independent raters blind to condition tabulate the number
of nonredundant responses each participant generated for each
UUT stimulus. The interrater classification of nonredundant
responses was highly reliable ( = .95). For each individual,
the two raters scores were averaged to yield a measure of
fluency.
Results
Mind wandering
Participants in the undemanding-task condition reported sig-
nificantly greater mind wandering (M = 2.47, SD = 0.66) in the
retrospective questionnaire than did participants in the
demanding-task condition (M = 2.15, SD = 0.67), F(1, 72) =
4.04, p < .05, 2 = .05. This result replicates previous findings
that working memory load decreases the frequency of mind
wandering.2 An analysis of the demanding-task, undemanding-
task, and rest conditions revealed no group differences in par-
ticipants retrospective reports about the degree to which they
had been explicitly thinking about the previous creativity task,
F(2, 106) = 0.09, p = .90, 2 = .002.
Incubation-task performance measures
No significant difference in accuracy was observed between
the undemanding task (M = .87, SD = .10) and the demanding
task (M = .88, SD = .20), F(1, 72) = 0.06, p = .80, 2 = .001.
Response time to targets was significantly faster in the
demanding task (M = 518.39 ms, SD = 117.55 ms) than in
the undemanding task (M = 648.97 ms, SD = 48.21 ms), F(1,
72) = 38.93, p < .001, 2 = .35. Faster response times were
expected in the demanding task because responses were based
on the previous (already-encoded) digit, whereas the unde-
manding task required participants to first encode the target
digit and then respond. This difference in response times
reflects the key difference in the structure of the two tasks: The
demanding task required that the identity of nontarget stimuli
be encoded, whereas the undemanding task did not require
that participants attend to nontarget stimuli.
UUT uniqueness scores
We first analyzed the UUT uniqueness scores using a mixed-
model analysis of variance (ANOVA) with exposure condition
(repeated exposure, new exposure) as a repeated measures fac-
tor and incubation condition (undemanding task, demanding
task, rest, no break) as a between-subjects factor. An Exposure
Condition Incubation Condition interaction emerged, F(1,
141) = 4.98, p < .01, 2 = .10. To further explore this effect, we
used univariate ANOVAs to analyze incubation-condition dif-
ferences in repeated-exposure and new-exposure UUT unique-
ness scores.
4 Baird et al.
60
Improvement on UUT (%)
40
20
0 between-condition differences in creativity as indexed by
20
40
Undemanding Demanding Rest
Incubation Condition
Fig. 1. Improvement in Unusual Uses Task (UUT) uniqueness scores (post-
incubation performance relative to baseline performance) for repeated-
exposure problems as a function of incubation condition. Error bars indicate
standard errors of the mean.
Repeated-exposure condition. There was a significant
effect of incubation condition in the repeated-exposure condi-
tion, F(1, 144) = 4.99, p < .01, 2 = .10. Participants who
engaged in an undemanding task during the incubation inter-
val displayed significantly greater improvement in UUT
uniqueness scores for repeated-exposure problems compared
with participants who engaged in a demanding task (p < .01),
a period of rest (p < .01), or no break (p < .01). No significant
difference in improvement was observed between participants
who received no break and those who engaged in either a
demanding task (p = .35) or a period of rest (p = .30); thus, no
incubation effect was observed in the latter two conditions
(see Fig. 1).
New-exposure condition. No incubation-condition differ-
ences were observed for improvement in uniqueness scores
for new problems, F(1, 144) = 1.01, p = .39, 2 = .02 (Fig. 2).
No significant difference was observed between participants
who received no break and those who engaged in an unde-
manding task (p = .21), a demanding task (p = .70), or rest
(p = .95). Thus, there was no significant incubation effect in
any incubation condition for the new-exposure problems.
Improvement on UUT (%)
60
40
20
0
20
40
Undemanding Demanding Rest
Incubation Condition
Fig. 2. Improvement in Unusual Uses Task (UUT) uniqueness scores
(postincubation performance relative to baseline performance) for new-
exposure problems as a function of incubation condition. Error bars indicate
standard errors of the mean.
Downloaded from pss.sagepub.com at UNIV CALIFORNIA SANTA BARBARA on September 26, 2012
UUT fluency
Fluency scores for the repeated-exposure problems did not dif-
fer significantly between incubation conditions, F(1, 144) =
1.15, p = .39, 2 = .02. This result rules out the possibility that
uniqueness scores were a result of confounding fluency and
creativity.
No Break
Individual differences in mind wandering and
UUT uniqueness scores
Scores on the Daydreaming Frequency subscale of the IPI
positively correlated with UUT uniqueness scores for both
repeated-exposure problems, r = .22, p < .05, and new-
exposure problems, r = .20, p < .05. This result provides pre-
liminary evidence that individuals who mind-wander more
frequently in their daily lives may be more creative in
general.
Discussion
Although research has suggested that taking a break can facili-
tate creativity, the mechanism of this incubation effect has
remained unclear and has been the source of considerable
empirical research and theoretical debate (e.g., Dijksterhuis &
Meurs, 2006; Smith & Blankenship, 1989; Yaniv & Meyer,
1987). The study reported here demonstrated that taking a
break involving an undemanding task improved performance
on a classic creativity task (the UUT) far more than did taking
a break involving a demanding task, resting, or taking no
break. Notably, this improvement was observed only for
repeated-exposure problems, which demonstrates that it
resulted from an incubation process rather than a general
increase in creative problem solving. Together, these data cor-
roborate, within a single experiment, the conclusion of a recent
meta-analysis (Sio & Ormerod, 2009) showing that incubation
effects were larger in studies in which individuals engaged in
an undemanding interpolated task than in studies that included
a demanding interpolated task or a rest period.
Our data support the notion that specific types of unrelated
thought facilitate creative problem solving. Even though the
act of encoding information in working memory was unrelated
to the solutions of the creativity problems, no incubation effect
was observed in the demanding-task condition. Moreover, the
undemanding-task condition was not associated with increased
frequency of thoughts explicitly about the creativity problems,
but was characterized by high levels of mind wandering. Thus,
our data indicate that creative problem solutions may be facili-
tated specifically by simple external tasks (i.e., tasks not
No Break related to the primary task) that maximize mind wandering.
The observation that performance selectively improved for
repeated-exposure problems (and not for new problems) indi-
cates that engaging in a task conducive to mind wandering
does not lead to general increases in creative problem-solving
ability. However, performance on both repeated-exposure and
68
Mind Wandering Facilitates Creative Incubation
new problems positively correlated with individuals general
propensity to mind-wander in everyday life (as assessed by the
IPI). This observation provides preliminary evidence that
there may be a relationship between individual differences in
mind wandering and creativity. Although this observation is
intriguing, it should be noted that this study lacked assess-
ments for a variety of other individual differences measures
(most notably, measures of inhibition) that could in principle
account for the association between propensity to mind-
wander and performance on the creativity task. An important
direction for future research will be to conduct a more thor-
ough assessment of the relationship between individual differ-
ences in mind wandering and creativity while controlling for
other factors that could contribute to this relationship.
Further research is needed to determine precisely why the
unrelated thoughts that occur during mind wandering uniquely
facilitate incubation. One possibility is that mind wandering
enhances creativity by increasing unconscious associative pro-
cessing, as predicted by the spreading-activation account of
incubation (e.g., Yaniv & Meyer, 1987; see also Dijksterhuis
& Meurs, 2006). A second possibility derives from recent neu-
roimaging work indicating that executive and default networks
interact during mind wandering (Christoff, Gordon, Small-
wood, Smith, & Schooler, 2009). Interactions between these
networks are observed relatively rarely in cognitive neurosci-
ence (although see Baird, Smallwood, & Schooler, 2011;
Gerlach, Spreng, Gilmore, & Schacter, 2011); considering that
activations in both networks are observed prior to successful
solution of insight problems (Kounios et al., 2008; Kounios
et al., 2006), engaging in tasks conducive to mind wandering
could contribute to incubation by creating a situation in which
default and executive systems mutually contribute to associa-
tive processing. Neurocognitive investigations of the brain
activations that occur during successful incubation intervals
might profitably explore this issue.
Anecdotal accounts of the inception of creative ideas have
long implicated mind wandering in the creative process. The
findings reported here provide arguably the most direct evi-
dence to date that conditions that favor mind wandering also
enhance creativity. From a theoretical perspective, this
research also helps to establish at least one benefit from
engaging in this otherwise seemingly dysfunctional mental
state. Although mind wandering may be linked to compro-
mised performance on an external task (Barron et al., 2011;
McVay & Kane, 2009) and may be a signature of unhappiness
(Killingsworth & Gilbert, 2010), it may also serve as a founda-
tion for creative inspiration.
Acknowledgments
We thank Steve Fiore for helpful discussion and James Schlegel,
Alex Weis, and Adam Haik for assistance in conducting the research.
Declaration of Conflicting Interests
The authors declared that they had no conflicts of interest with
respect to their authorship or the publication of this article.
Downloaded from pss.sagepub.com at UNIV CALIFORNIA SANTA BARBARA on September 26, 2012
69
5
Funding
This research was supported by the John Templeton Foundation
under Grant No. 24329, awarded to Jonathan W. Schooler. Benjamin
Baird is supported by a National Science Foundation Graduate
Research Fellowship under Grant No. DGE-0707430. Michael D.
Mrazek and Michael S. Franklin are supported through Office of
Education Grant No. R305H030235, awarded to Jonathan W.
Schooler and Jonathan Smallwood.
Notes
1. Following the procedure used by Wallach and Kogan (1965), we
categorically assigned statistically unique responses a score of 1.
Because problems were repeated in our incubation design, responses
appearing up to two times across the sample received points. An
alternative scoring method using a graded scale (from 1 to 5;
S. Fiore, personal communication, January 25, 2011) yielded nearly
identical results.
2. As noted, we also administered the mind-wandering questionnaire
following the rest interval. The score on the retrospective mind-
wandering scale in the rest condition (M = 2.35, SD = 0.57) was not
significantly different from the score on this scale in either the
undemanding-task condition (p = .44) or the demanding-task condi-
tion (p = .19), although this comparison is difficult to interpret
because the rest condition included no primary task to which internal
thoughts could fail to pertain.
References
Baird, B., Smallwood, S., & Schooler, J. W. (2011). Back to the
future: Autobiographical planning and the functionality of mind-
wandering. Consciousness and Cognition, 20, 16041611.
Barron, E., Riby, L. M., Greer, J., & Smallwood, J. (2011). Absorbed
in thought: The effect of mind wandering on the processing of
relevant and irrelevant events. Psychological Science, 22, 596
601.
Cai, D. J., Mednick, S. A., Harrison, E. M., Kanady, J. C., & Med-
nick, S. C. (2009). REM, not incubation, improves creativity by
priming associative networks. Proceedings of the National Acad-
emy of Sciences, USA, 106, 10130.
Christoff, K., Gordon, A. M., Smallwood, J., Smith, R., & Schooler,
J. W. (2009). Experience sampling during fMRI reveals default
network and executive system contributions to mind wandering.
Proceedings of the National Academy of Sciences, USA, 106,
87198724.
Dijksterhuis, A., & Meurs, T. (2006). Where creativity resides: The
generative power of unconscious thought. Consciousness and
Cognition, 15, 135146.
Ellwood, S., Pallier, G., Snyder, A., & Gallate, J. (2009). The incu-
bation effect: Hatching a solution? Creativity Research Journal,
21, 614.
Frster, J., Friedman, R. S., & Liberman, N. (2004). Temporal con-
strual effects on abstract and concrete thinking: Consequences for
insight and creative cognition. Journal of Personality and Social
Psychology, 87, 177189.
Gerlach, K. D., Spreng, R. N., Gilmore, A. W., & Schacter, D. L.
(2011). Solving future problems: Default network and executive
6 Baird et al.
activity associated with goal-directed mental simulations. Neuro-
Image, 55, 18161824.
Guilford, J. P. (1967). The nature of human intelligence. New York,
NY: McGraw-Hill.
Killingsworth, M. A., & Gilbert, D. T. (2010). A wandering mind is
an unhappy mind. Science, 330, 932.
Kounios, J., Fleck, J. I., Green, D. L., Payne, L., Stevenson, J. L.,
Bowden, E. M., & Jung-Beeman, M. (2008). The origins of
insight in resting-state brain activity. Neuropsychologia, 46, 281
291.
Kounios, J., Frymiare, J. L., Bowden, E. M., Fleck, J. I., Subra-
maniam, K., Parrish, T. B., & Jung-Beeman, M. (2006). The
prepared mind: Neural activity prior to problem presentation pre-
dicts subsequent solution by sudden insight. Psychological Sci-
ence, 17, 882890.
Mason, M. F., Norton, M. I., Van, H. J. D., Wegner, D. M., Grafton,
S. T., & Macrae, C. N. (2007). Wandering minds: The default
network and stimulus-independent thought. Science, 315, 393
395.
Matthews, G., Joyner, L., Gilliland, K., Campbell, S., Huggins, J., &
Falconer, S. (1999). Validation of a comprehensive stress state
questionnaire: Towards a state big three? In I. Mervielde, I. J.
Deary, F. DeFruyt, & F. Ostendorf (Eds.), Personality psychol-
ogy in Europe (Vol. 7, pp. 335350). Tilburg, The Netherlands:
Tilburg University Press.
McVay, J. C., & Kane, M. J. (2009). Conducting the train of thought:
Working memory capacity, goal neglect, and mind wandering in
an executive-control task. Journal of Experimental Psychology:
Learning, Memory, and Cognition, 35, 196204.
Milgram, R. M., & Milgram, N. A. (1976). Creative thinking and
creative performance in Israeli students. Journal of Educational
Psychology, 68, 255259.
Shaw, G. A., & Giambra, L. M. (1993). Task unrelated thoughts of
college students diagnosed as hyperactive in childhood. Develop-
mental Neuropsychology, 9, 1730.
Silvia, P. J., Barona, C. M., Cram, J. T., Hess, K. I., Martinez, J. L.,
Richard, C. A., . . . Willse, J. T. (2008). Assessing creativity with
divergent thinking tasks: Exploring the reliability and validity of
Downloaded from pss.sagepub.com at UNIV CALIFORNIA SANTA BARBARA on September 26, 2012
70
new subjective scoring methods. Psychology of Aesthetics, Cre-
ativity, and the Arts, 2, 6885.
Singer, J. L., & Antrobus, J. S. (1972). Daydreaming, imaginal pro-
cesses, and personality: A normative study. In P. W. Sheehan
(Ed.), The function and nature of imagery (pp. 175202). New
York, NY: Academic Press.
Sio, U. N., & Ormerod, T. C. (2009). Does incubation enhance prob-
lem solving? A meta-analytic review. Psychological Bulletin,
135, 94120.
Smallwood, J., Nind, L., & OConnor, R. C. (2009). When is your
head at? An exploration of the factors associated with the tempo-
ral focus of the wandering mind. Consciousness and Cognition,
18, 118125.
Smallwood, J., Obonsawin, M. C., & Heim, D. (2003). Task unre-
lated thought: The role of distributed processing. Consciousness
and Cognition, 12, 169189.
Smallwood, J., & Schooler, J. W. (2006). The restless mind. Psycho-
logical Bulletin, 132, 946958.
Smith, S. M., & Blankenship, S. E. (1989). Incubation effects. Bul-
letin of the Psychonomic Society, 27, 311314.
Torrance, E. P. (2008). Torrance Tests of Creative Thinking: Norms-
technical manual, verbal forms A and B. Bensenville, IL: Scho-
lastic Testing Service.
Vul, E., & Pashler, H. (2007). Incubation benefits only after people
have been misdirected. Memory & Cognition, 35, 701710.
Wallach, M. A., & Kogan, N. (1965). Modes of thinking in young
children: A study of the creativityintelligence distinction. New
York, NY: Holt, Rinehart, & Winston.
White, H. A., & Shah, P. (2006). Uninhibited imaginations: Creativ-
ity in adults with attention-deficit/hyperactivity disorder. Person-
ality and Individual Differences, 40, 11211131.
White, H. A., & Shah, P. (2011). Creative style and achievement in
adults with attention-deficit/hyperactivity disorder. Personality
and Individual Differences, 50, 673677.
Yaniv, I., & Meyer, D. E. (1987). Activation and metacognition of
inaccessible stored information: Potential bases for incubation
effects in problem solving. Journal of Experimental Psychology:
Learning, Memory, and Cognition, 13, 187205.
Cerebral Cortex August 2015;25:21602169
doi:10.1093/cercor/bhu022
Advance Access publication February 20, 2014
Neural Correlates of Subliminal Language Processing
Vadim Axelrod1,2, Moshe Bar1,4, Geraint Rees2,3 and Galit Yovel5,6
1
The Gonda Multidisciplinary Brain Research Center, Bar Ilan University, Ramat Gan, Israel, 2UCL Institute of Cognitive
Neuroscience, 3Wellcome Trust Centre for Neuroimaging, University College London, London, UK, 4Athinoula A. Martinos Center
for Biomedical Imaging, Massachusetts General Hospital and Harvard Medical School, Charlestown, MA, USA, 5School of
Psychological Sciences and 6Sagol School of Neuroscience, Tel Aviv University, Tel Aviv, Israel
Address correspondence to Vadim Axelrod. Email: vadim.axelrod@gmail.com
Language is a high-level cognitive function, so exploring the neural
correlates of unconscious language processing is essential for un-
derstanding the limits of unconscious processing in general. The
results of several functional magnetic resonance imaging studies
have suggested that unconscious lexical and semantic processing is
conned to the posterior temporal lobe, without involvement of the
frontal lobethe regions that are indispensable for conscious
language processing. However, previous studies employed a similarly
designed masked priming paradigm with briey presented single and
contextually unrelated words. It is thus possible, that the stimulation
level was insufciently strong to be detected in the high-level frontal
regions. Here, in a high-resolution fMRI and multivariate pattern
analysis study we explored the neural correlates of subliminal
language processing using a novel paradigm, where written mean-
ingful sentences were suppressed from awareness for extended dur-
ation using continuous ash suppression. We found that subjectively
and objectively invisible meaningful sentences and unpronounceable
nonwords could be discriminated not only in the left posterior
superior temporal sulcus (STS), but critically, also in the left middle
frontal gyrus. We conclude that frontal lobes play a role in uncon-
scious language processing and that activation of the frontal lobes
per se might not be sufcient for achieving conscious awareness.
Keywords: continuous ash suppression (CFS), decoding subliminal
content, fMRI imaging of unconscious processing, multivoxel pattern
classication analysis (MVPA), subliminal language processing
Introduction
What are the limits of unconscious language processing? This
question has been intensively researched during last 50 years
(for reviews see Kouider and Dehaene 2007; Lin and He 2009;
Van den Bussche et al. 2009). Though no consensus has been
reached, many behavioral experiments show that subliminally
presented text can be processed not only at a relatively low
orthographic level (e.g., Dehaene et al. 2001; Devlin et al.
2004), but also at a higher semantic level (e.g., Marcel 1983;
Jiang et al. 2007; Costello et al. 2009; Sklar et al. 2012; but see
Holender 1986; Abrams and Greenwald 2000). Neuroimaging
studies show that the visual word form area (VWFA) (Cohen
et al. 2000) is involved in unconscious orthographic word pro-
cessing (e.g., Dehaene et al. 2001; Kouider et al. 2007),
whereas unconscious semantic language processing is most
consistently observed along the left posterior STS (Devlin et al.
2004; Nakamura et al. 2007; see also Nakamura et al. 2005).
Whereas the central role of frontal lobes in various aspects of
conscious language processing is unquestionable (for reviews
The Author 2014. Published by Oxford University Press.
This is an Open Access article distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/by/3.0/), which permits unrestricted
reuse, distribution, and reproduction in any medium, provided the original work is properly cited.
71
see Vigneau et al. 2006; Price 2012), the role of the frontal
lobes in unconscious language processing remains elusive. In
particular, only one study reports activations in the inferior
frontal gyrus for subliminally presented words (Diaz and
McCarthy 2007). However, this study did not implement sub-
jective/objective awareness reports after each trial (Seth et al.
2008) and it is therefore difcult to condently determine
whether the activations indeed reect unconscious language
processing. It is noteworthy that while such limited empirical
evidence for unconscious activations in frontal lobes (e.g., Lau
and Passingham 2007; van Gaal et al. 2010) is in line with
some prominent theoretical models, such as Global Workspace
model (Dehaene et al. 1998), it is also possible that subliminal
sensory stimulation in the previous studies was too weak to ac-
tivate the frontal lobes (Haynes 2009). Specically, the earlier
imaging studies used subliminal priming masking paradigm
with a brief stimulus exposure, which could have resulted in
insufcient brain stimulation. In addition, a more general limit-
ation of previous studies was that none of them measured
awareness on each individual trial during the neuroimaging
experiment (Seth et al. 2008). Consequently, if on some trials
the primes were visible or at least partially visible (Kouider
et al. 2010) they would still be considered unaware and there-
fore their neural correlates may not reect only unconsciously
processed material.
In the current functional magnetic resonance imaging
(MRI) study we explored the neural correlates of subliminal
language processing, with a novel design that addresses the
concerns reviewed above. Observers were presented with
series of consecutively presented textual stimuli: meaningful
sentences or unpronounceable nonwords (Fedorenko et al.
2010), which were rendered invisible using continuous ash
suppression (CFS) (Tsuchiya and Koch 2005) for extended
period of time (10 s) (Fig. 1). Critically, as we sought for evi-
dence of unconscious language processing of any type, we
decided to use meaningful sentences that required not only
semantic, but also syntactic and structural processingthe
design which permitted to increase potential differences
between meaningful (sentences) and meaningless (non-
words) conditions. After each block of either sentences or
nonwords participants reported whether they had been
aware of even a single worda procedure which ensured that
data analyses were conducted only on blocks judged invisible
by participants. To discriminate between neural activity eli-
cited by the 2 conditions we used multivoxel pattern classi-
cation analyses (MVPA) focusing on the language network
(Fedorenko et al. 2010), which was localized on a per-
participant basis using the same stimuli while they were fully
visible. The principal goal of our research was to test whether

Figure 1. Schematic ow of one block with meaningful sentence in CFS invisible
experiment. The words are translated to English for illustrative purpose only while in
the experiment all the materials were in Russian. The structure of the blocks with
unpronounceable nonwords was the same as the blocks with meaningful sentences,
but substituting nonwords (random letter permutations) for words. In the functional
localizer (visible experiment) the block ow was similar to that depicted in this gure
while the words were visible via both eyes (no CFS mask) and the task was 1-back
word repetition (instead of awareness report at the end of the block).
the frontal lobes were involved in any unconscious proces-
sing of language. The secondary goal was to reveal whether
using a paradigm that is different from previous studies and
by measuring awareness report after each block, the neural
correlates of subliminal processing could still be found in the
left posterior temporal lobe. Given that different subliminal
paradigms do not always yield similar effects (Almeida et al.
2008, 2013; Kanai et al. 2010; Faivre et al. 2012), such a repli-
cation is important for establishing this general cognitive
phenomenon.
72
Materials and Methods
Apparatus
MRI data were collected using a 3T GE MRI scanner with an 8-channel
head coil. Echo planar imaging used a T2*-weighted sequence to
measure changes in blood oxygen level dependent (BOLD) signal. The
scanning parameters were as follows: repetition time (TR) = 2.5 s, time
echo (TE) = 30 ms, ip angle: 90, slice thickness: 3.4 mm no gap, eld
of view (FOV) 200 mm, 32 slices; data were acquired using 96 96
matrix (in plane resolution 2.08 2.08 mm), reconstruct into 128 128
matrix (in plane resolution 1.56 1.56 mm). Slice orientation was par-
allel to temporal lobe with full coverage of the cerebral cortex. An ana-
tomical SPRG scan with full brain coverage was collected with
1 1 1 mm resolution (TE = 3.52 ms, TR = 9.104 ms).
Participants
Seventeen healthy volunteers (age: 2343, 9 females, all right-handed)
with normal or corrected-to-normal vision participated in the exper-
iment. The mother-tongue of all participants was Russian (they were
born in the Soviet Union and lived there or in an ex-Soviet Union
country at least till the age of 14). The study was approved by the
ethics committee of the Tel Aviv Sourasky Medical Center. All partici-
pants signed informed consent to participate in the study. Data of 2
participants were excluded from the analysis due to inability to follow
the instructions (one participant) and excessive movements (>1 cm) in
the scanner (another participant).
Experimental Setup
Stimuli
Experiment textual material included series of words presented one
word at a time (Fig. 1). The stimuli were of 2 types: meaningful sen-
tences or series of unpronounceable nonwords sentences (random
permutation of the letters) (e.g., Fedorenko et al. 2010, 2011). The
meaningful sentences described neutral situations (e.g., weather,
nature description etc.). The textual material was in Russian (Cyrillic al-
phabet) written with the Arial font. The letters were presented in lower
case (including the rst letter of the sentence) and there was no period
sign at the end of the sentence. The words were shown at the center of
the screen; visual angle size varied between 8 5 (horizontal vertical)
and 30 5. The sentences in both conditions were 6 words in length.
In total, there were 40 different meaningful and 40 nonword sen-
tences.
Continuous Flash Suppression
A standard CFS procedure was used (Tsuchiya and Koch 2005). In the
scanner participants wore MR-compatible cardboard anaglyph cyan/
red glasses. Stimuli were projected with an LCD projector (NEC,
VT660K), positioned ahead of the participant and viewed through a
tilted mirror mounted on the MR head coil. Stimuli were projected
using the red RGB color channel (visible using red lter) and for the
Mondrian mask blue/green RGB channels were used. The red (target)
glass lter was always placed over the nondominant eye of each partici-
pant. Eye-dominance was assessed prior to the experiment by asking
the participants to view a distant object through a hole made by the
ngers of their 2 hands (Miles test) (Miles 1930; Mendola and Conner
2007). The screen luminance of the text was set to 40% ( percent of the
maximal screen luminance; dark gray) and of the background was 61%
(light gray). The Mondrian pattern was projected on the center of the
screen and its size was 34 6 of visual angle (to cover the longest
word). The pattern of the CFS mask consisted of unlled ellipses and
rectangles, which were similar in image pattern structure to letters (see
Fig. 1 for the examples of the pattern). The frequency rate at which the
Mondrian patterns were changed was 10 Hz (100 ms for each image).
Experimental Design
Participants underwent 2 separate fMRI experiments: the main exper-
iment using CFS masking and a functional localizer with fully visible
text. The sessions with visible text were always the last in order not to
Cerebral Cortex August 2015, V 25 N 8 2161
provide additional cues about the appearance of invisible stimuli. The
design and ow of both experiments were identical except for the visi-
bility level of the stimuli and the behavioral task (described below). In
the functional localizer experiment, no Mondrian mask was used and
the text was projected to both eyes using all 3 RGB channels (to elimin-
ate potential head movements in the scanner between 2 experiments,
participants still wore anaglyph glasses during this experiment as
well). Screen luminance of the text in the functional localizer exper-
iment was 0% (black) and the background was 61% (light gray).
The experiment used a block-design with each experimental block
lasting 10 s and interleaved with a xation block of 7.5-s duration. Fix-
ation block was an empty screen with + sign at the middle (0.2 0.2
of visual angle; background luminosity: 61% [light gray color], fore-
ground luminosity: 0% [black color]). Each session consisted of 12
experimental blocks (6 blocks of meaningful sentences and unpro-
nounceable nonwords, respectively). Each session started with a 10 s
xation cross. Total session duration was 3 min and 40 s. Number of
sessions per participant varied between participants: for the main
experiment (invisibility experiment) it was between 7 and 11 and for
the functional localizer experiment it was between 3 and 5. The use of
a larger number of short experimental sessions compared with a
smaller number of long experimental sessions improves classication
performance (Coutanche and Thompson-Schill 2012).
The schematic ow of one block (invisible CFS experiment) consist-
ing of meaningful sentences is shown in Figure 1. The ow of the con-
dition with unpronounceable nonwords was similar. For illustrative
purposes only, in the gure the words have been translated to English,
while all the experiment materials were in Russian. Duration of a
single word was 0.4 s; duration of a single sentence (6 words) was 2.4 s
(words appeared one after another without interstimulus interval).
Each experimental block consisted of 4 sentences (or 4 lists of non-
words). The sentences appeared back to back without delay between
the last word of the previous sentence and the rst word of a next sen-
tence. For some sentences of the block the last word of the previous
sentence and the rst word of the consecutive sentence was identical
(e.g., the queue was extremely long today, today the weather is very
nice). The number of repetitions of the last and rst word varied ran-
domly between blocks (minimum 0, maximum 2). The rst word of
each sentence appeared with a random horizontal position jitter (one
or 2 letters from centered position). This ensured that when the rst
and last word of the sentence repeated, there was no effect of word
freezing on the screen. The blocks of unpronounceable nonwords
were similar to blocks of meaningful sentences (Fig. 1) but substituting
nonwords (random letter permutations) for words. We decided to use
the random letter permutation and not the letter permutation of the
real words, since the later can be still occasionally recognized as real
words (e.g., Wentura et al. 2005). The 6-word sentence-like structure
and the repetition of the last and rst word were preserved for non-
words as well.
The behavioral tasks in the functional localizer and in the CFS sub-
liminal experiment were different. In the functional localizer partici-
pants were asked to press any button on the response box when they
detected a consecutive repetition of a word (1-back task). The repeated
words could only be the last and rst words of a sentence (see above).
This task ensured that participants were attentive to the stimuli.
In the CFS main experiment, participants were required to make 2
separate responses after every block (either 4 sentences or 4 nonword
lists, see Fig. 1). These responses were made during the xation block
(no instructions were presented). The participants were rst required
(objective response) whether they thought that a block consisted of
sentences (comprising meaningful words) or of nonwords. Partici-
pants were then required to make a second response (subjective
response) indicating whether their rst response was based on seeing
the stimuli or on a guess. Below are the instructions, which were
given to participants: You will be presented with blocks of either sen-
tences composed of meaningful words or series of nonwords. The
words or nonwords will be presented sequentially. The stimuli are pre-
sented in a way that makes it very hard and probably impossible to see
them. At the end of each block you need to make 2 responses. The rst
response asks you to indicate whether it was a block of sentences
(words) or nonwords. Because in each block we present only words or
2162 Neural Correlates of Subliminal Language Processing Axelrod et al.
nonwords, detecting one of the stimuli (single word or nonword)
during the block would allow you to indicate the correct response.
Even if you did not see anything, we ask you to guess. The second
response asks you to indicate whether your rst response was based
on seeing the text or guessing. It should be noted that as the condition
of sentences is comprised of various types of language processing
(sentence syntax and structure, words semantics etc.), it was crucial to
ensure that any part of language processing remained unconscious.
This was the reason that we made it clear to participants, that detecting
a single word in the block is sufcient for a correct answer. Participants
underwent a short training session outside the scanner as well as a
short training session inside the scanner at the beginning of the exper-
iment to ensure that they understood the instructions. Prior to starting
the experimental sessions all participants conrmed that the instruc-
tions are clear for them. The instructions were also repeated during the
experiment, between the sessions. At the end of the experiment,
during the informal debrieng, none of the participants indicated any
difculty with performing the task according to the instructions.
Data Analysis
Preprocessing
Data analysis used SPM5 (Wellcome Trust Centre for Neuroimaging,
London, UK; http://www.l.ion.ucl.ac.uk). The rst 4 volumes (4 TRs,
10 s) of each session were discarded to allow for T1 equilibration
effects. Preprocessing steps applied for functional (EPI) scans in-
cluded: realignment, slice-time correction, motion correction, normali-
zation to 2 2 3 voxel resolution using Montreal Neurological
Institute (MNI) template and spatial smoothing with a full-width at
half-maximum = 6 mm kernel. For the normalization we used a unied
segmentation procedure (Ashburner and Friston 2005).
Region of Interest Localization
For the language functional localizer (visible text) we estimated a GLM
model (HRF boxcar function) with 2 regressors: meaningful sentences
and nonwords. We used the contrast meaningful sentences >
nonwords to identify a network of language processing regions for
each participant (Fedorenko et al. 2010, 2011, 2012). To constraint
individual GLM-dened functional activations we used probabilistic
group-level functional masks (Fedorenko et al. 2010; http://web.mit.
edu/evelina9/www/funcloc/funcloc_parcels.html). Thus, for each
mask region/participant based on individual meaningful sentences >
nonwords GLM contrast we selected a contiguous cluster of most se-
lective voxels (number of voxels is specied below). The regions
dened by the masks are shown in Figure 2. There were 11 regions in
total: 5 regions in the left parieto-temporal lobe (angular gyrus, supra-
marginal gyrus, posterior STS, middle anterior temporal gyrus, and
anterior temporal gyrus), 2 regions in the right hemisphere of the tem-
poral lobe ( posterior STS, middle anterior temporal gyrus) and 4
regions in the left hemisphere of the frontal lobe (orbital inferior
frontal gyrus, inferior frontal gyrus, middle frontal gyrus, superior
frontal gyrus). Critically, as multivariate prediction is inuenced by
region of interest (ROI) size (e.g., Eger et al. 2008; Walther et al. 2009;
Said et al. 2010) we ensured the ROIs of different regions were of an
equal size of 100 voxels (1200 mm3). In additional analyses we also ex-
plored a range of different ROI sizes of 50 and 150 voxels. The ROI
size could not be increased further since the size of probabilistic
group-level functional masks (Fedorenko et al. 2010) of some of the
regions (e.g., left superior frontal gyrus) was <200 voxels. Dening
ROIs of equal size was undertaken using custom MATLAB code, where
for each region/participant the code selected the contiguous cluster of
voxels with the highest z-score values relating to the meaningful
sentences > nonwords contrast in the independent localizer with
visible stimuli (similar procedure had been previously applied for face-
selective voxels here [Axelrod and Yovel 2012]). The list of the ROIs
(100 voxels size) with their coordinates and average z-score values can
be found in Table 3.
73
Figure 2. Probabilistic group-level masks of language network (Fedorenko et al. 2010; http://web.mit.edu/evelina9/www/funcloc/funcloc_parcels.html) projected on a SPM
template T1 image. The names of the regions are: 1left angular gyrus, 2left supramarginal gyrus, 3left posterior STS, 4left middle anterior temporal, 5left anterior
temporal, 6left orbital inferior frontal gyrus, 7left inferior frontal gyrus, 8left middle frontal gyrus, 9left superior frontal gyrus, 10right posterior STS, 11right middle
anterior temporal. Regions IDs correspond to the IDs in Table 3.
Main Experiment: Multivariate Pattern Analysis
A multivariate pattern classication analysis was performed on raw
BOLD signal values. After a preprocessing stage (the step which was
identical for all EPI data), the data of the main (CFS) experiment were
detrended and normalized using the z-score MATLAB function. This
procedure was applied to the full scan voxel time course. The time
course was also shifted 3 TRs to account for hemodynamic lag. In the
multivariate analysis the global signal average for each of the 2 con-
ditions was subtracted (e.g., Serences et al. 2009; Misaki et al. 2010),
while this procedure was performed separately for each session in
order to prevent information leakage in course of cross-validation
procedure. Global signal average subtraction, which was applied as
part of per-condition normalization procedure, increases classication
performance (Raizada et al. 2010; see also Aksoy and Haralick 2001)
and may be benecial, as it prevents voxels with higher values and
range to dominate the classiers weights (Coutanche 2013). In
addition, subtraction of the global signal average from each condition
permits a more straightforward interpretation of the results, as even
minimal differences in global average between conditions might be a
result of different level of arousal or attention (Coutanche 2013). We
obtained qualitatively similar results when the analyses were con-
ducted without subtracting global signal average. The TRs within each
block were averaged, resulting in a single average data point value per
block that was used as the input to pattern classication analysis. The
LibSVM MATLAB implementation of a linear support vector machine
was used for classication (http://www.csie.ntu.edu.tw/~cjlin/libsvm/
) using a leave-one-session out cross-validation procedure, which was
repeated according to number of scans available for each participant.
Signicance of prediction rate was established using a one-tailed t-test
above the chance for the group classication rates (Eger et al. 2008;
Meyer et al. 2010; Nestor et al. 2011). Multiple comparison Bonferroni
correction was made based on the number of ROIs in posterior tem-
poral lobes (7 ROIs, P-value signicance threshold = 0.0071) and
frontal lobe: 4 ROIs, P-value signicance threshold = 0.0125). Of note,
the signicance of the results persists when the correction is made
based on total number of ROIs (11 ROIs). Supplementary analyses
tested whether successful prediction could also be achieved based on
signal global level and this analysis differed from the main analysis in 2
ways: 1) the average signal was not subtracted; 2) the timecourses of all
voxels in a ROI were averaged resulting in one timecourse (i.e., classi-
cation using one dimension). In both main and supplementary ana-
lyses only the blocks where the second (subjective) report of the
74
participants was guessed were used. To ensure that for each partici-
pant the equal number of data points (blocks) per condition (otherwise
the use of imbalanced data set might bias classication performance
[Japkowicz and Stephen 2002]) we randomly discarded the data points
from the condition with largest amount of data.
Results
Behavioral Results
To establish the level of stimulus awareness (during fMRI scan-
ning) after each block (duration of 10 s) participants were
asked to make 2 separate judgments: an objective response
(meaningful sentences/nonwords) and a subjective (con-
dence) response (knew/guessed). Most blocks of both con-
ditions were judged to be invisible: the percentage of blocks
on which participants responded with guessed for meaning-
ful sentences was 80.3% (standard error of mean [SEM] = 4.5%)
and for nonwords was 82.6% (SEM = 3.4%). Critically, objective
responses for these invisible (guessed) blocks were at
chance level: meaningful sentences = 51.7% (SEM = 2.1%;
P-value = 0.22, t < 1, one-tailed t-test vs. 50%) and nonwords =
51.7% (SEM = 2.4%; P-value = 0.24, t < 1, one-tailed t-test vs.
50%). The distribution of all subjective and objective responses
is shown in Tables 1 and 2. In addition, during the informal de-
brieng after the experiment, we asked the participants what
exactly they saw when they choose to answer guessed. Criti-
cally, all the participants indicated that when they responded
guessed then they could not see even a single letter within a
presented stimulus. Thus, taken together, we conclude that
invisibility manipulation was effective and that the trials,
which were reported as guessed, were genuinely invisible.
To explore unconscious processing, we restricted all sub-
sequent analyses to blocks ranked by participants as guessed
in their subjective response. The average number of blocks per
participant/condition was 38.53 (mean squared error [MSE] =
2.49). It was not feasible to investigate neural correlates of
Cerebral Cortex August 2015, V 25 N 8 2163
Table 1
Distribution of subjective responses (condence rating, second response)

Knew Guessed
Meaningful sentences 19.7% (SEM: 4.5%) 80.3% (SEM: 4.5%)
Nonwords 17.4% (SEM: 3.4%) 82.6% (SEM: 3.4%)

Table 2
Percent of correct objective (rst answer) responses per each category (numbers in the table
cells) binned for corresponding subjective (second answer) response (table columns)

Knew Guessed
Meaningful sentences 62.4% (SEM: 9.9%) 51.7% (SEM: 2.1%)]
Nonwords 86.5% (SEM: 4.9%) 51.7% (SEM: 2.4%)

Table 3
Average z-scores and average MNI coordinates (center of mass) of ROI used in the decoding
analysis of invisible stimuli

ID Region of interest Average z-score MNI coordinates

X Y Z
1 Left angular gyrus 1.46 (SEM: 0.47) 43 73 30
2 Left supramarginal gyrus 3.35 (SEM: 0.64) 54 59 15
3 Left posterior superior temporal sulcus (STS) 5.85 (SEM: 0.47) 57 42 4
4 Left middle anterior temporal 4.67 (SEM: 0.55) 57 18 9
5 Left anterior temporal 3.9 (SEM: 0.44) 54 2 17
6 Left orbital inferior frontal gyrus 3.5 (SEM: 0.51) 48 28 3 Figure 3. Prediction rate of discrimination between subliminal meaningful sentences
7 Left inferior frontal gyrus 4.94 (SEM: 0. 64) 50 17 23 and nonwords in language network regions in the temporal lobe (A) and the frontal
8 Left middle frontal gyrus 3.72 (SEM: 0.62) 43 1 52 lobe (B). Black line is a chance level = 50%; error bars denote standard error of the
9 Left superior frontal gyrus 0.13 (SEM: 0.47) 7 54 38 mean.
10 Right posterior superior temporal sulcus (STS) 3.24 (SEM: 0.31) 59 45 8
11 Right middle anterior temporal 3.08 (SEM: 0.47) 54 14 13

ID numbers in the rst column correspond to the numbers of anatomical masks in Figure 2. The
Z-score values are based on meaningful sentence > nonwords contrast in visible stimuli localizer.
Volume of all ROIs was 100 voxels (1200 mm3). Details of how the ROIs were generated are
described in the Materials and Methods section.
conscious processing (knew subjective, second response)
since there was not sufcient data: 45% of the participants had
<5 blocks per condition of this type and the average number of
blocks per participant/condition was 7.93 (MSE = 1.94).
Imaging Results
The goal of our analyses was to identify brain regions where
pattern signals were sufcient to discriminate subliminal
meaningful sentences from nonwords. Our multivoxel pattern
classication analysis (MVPA) (Norman et al. 2006) approach
focused on nodes in the language network (Fedorenko et al.
2010) identied by an independent localizer using visible
stimuli. Summary statistics (average selectivity z-score and co-
ordinates) of the ROIs is presented in the Table 3. It can be
seen that all the regions except for the left superior frontal
gyrus showed higher activation for meaningful sentences com-
pared with nonwords. Notably, the left superior frontal gyrus
was also among the less selective regions in the study of Fedor-
enko et al. (2010). It is noteworthy, that statistical contrast of
visible meaningful sentences versus nonwords identies only
the high-level language processing network and does not
include inferior temporal cortex (e.g., VWFA [Cohen et al.
2000]), which is implicated in more low-level orthographical
processing (Dehaene and Cohen 2011).
The performance of the support vector machine in dis-
tinguishing subliminal meaningful sentences from nonwords
in the parieto-temporal ROIs is shown in Figure 3A. Group-
level statistical signicance was assessed using one-tailed
t-tests against chance level of 50% (Bonferroni multiple com-
parison correction, see Materials and Methods). The only
parieto-temporal region, which showed prediction signi-
cantly above chance was left posterior STS: 56.2% (MSE: 2%,
t (14) = 3.01, P = 0.004). Prediction rate in the right posterior STS
was greater than chance (53.1% [MSE: 1.6%]), but it did not
reach statistical signicance after multiple comparison correc-
tion (t (14) = 1.93, P = 0.036)]. Performance in the other ROIs
did not differ from chance: left supramarginal gyrus: 53.1%
(MSE: 2.8%, t (14) = 1.1, P = 0.22), left angular gyrus: 51.6%
(MSE: 2.5%, t (14) < 1), left middle anterior temporal: 49.9%
(MSE: 2%, t (14) < 1), left anterior temporal: 50.7% (MSE: 2.7%,
t (14) < 1) and right middle anterior temporal: 49.1% (MSE:
2.8%, t (14) < 1). To compare the prediction rates between hemi-
spheres, for 2 regions which were localized in both hemi-
spheres (the posterior STS and the middle anterior temporal
region) we ran a 2-way repeated measures ANOVA with a
region and a hemisphere as factors. The results showed signi-
cant main effect of region [F1,14 = 7.781, P = 0.014], but no sig-
nicant effect of hemisphere [F1,14 < 1] and no signicant
interaction [F1,14 < 1] suggesting that higher prediction rate in
the posterior STS comparing to the middle anterior temporal
was a property of both hemispheres.
Decoding performance comparing subliminal meaningful
sentences and nonwords in frontal regions is shown in Figure 3B.
The only region which showed prediction rate signicantly
above chance was left middle frontal gyrus: 54.7% (MSE: 1.5%,

2164 Neural Correlates of Subliminal Language Processing Axelrod et al.

75
t(14) = 2.98, P < 0.004). In 3 other regions the prediction rate did
not differ from chance level: left orbital inferior frontal gyrus:
53.8% (MSE: 2.6%, t(14) = 1.47, P = 0.081), left inferior frontal
gyrus: 49.1% (MSE: 2.2%, t(14) < 1), left superior frontal gyrus:
51.4% (MSE: 2.3%, t(14) < 1]. To test whether the amount of
unconscious information differed between the highest classi-
cation rate region in the temporal and frontal lobes we compared
the prediction rates in the left posterior STS and left middle
frontal gyrus. No signicant difference was found (paired t-test,
t(14) < 1), suggesting that there is no evidence that one of the
regions contained more information than the other.
So far we have shown that it was possible to discriminate
between unconscious meaningful sentences and nonwords
based on multidimensional patterns of BOLD signals in the left
posterior STS and left middle frontal gyrus. Now, we asked
whether the 2 subliminal conditions could also be discrimi-
nated based on global signal level alonethe univariate ap-
proach, which is extensively used in fMRI research. We
therefore conducted additional analyses where the classi-
cation was done for only one dimension, which was the
average across all the voxels in the ROI. This analysis revealed
that across both parieto-temporal and frontal lobe regions only
the angular gyrus exhibited above chance prediction rate
(53.7%, MSE: 1.8%), but it did not reach signicance level after
multiple comparison correction [t (14) = 2.08, P = 0.028]. In all
other regions the prediction rate did not exceed 51.5% and did
not differ from chance [t (14) < 1]. We conclude that average
signal did not contain sufcient information for successful dis-
crimination between 2 conditions.
Finally, to ensure that the reported result was not idiosyn-
cratic for a specic ROI size, we repeated the multivariate ana-
lyses for the ROI size of 50 and 150 voxels. As in the main
analysis, the signicance was assessed based on Bonferroni
multiple comparison correction for each ROI size (see
Materials and Methods). The results of this analysis are shown
in Figure 4. Critically, for both left posterior STS and the left
middle frontal gyrus the prediction rate was always signi-
cantly above chance: left posterior STS [50 voxels: prediction
rate: 56.4%, MSE: 2.1%, t (14) = 3.02, P = 0.004; 150 voxels:
Figure 4. Discrimination between subliminal meaningful sentences and nonwords in language network regions for ROI size of 50, 100, and 150 voxels. Grey line represents a
chance level of performance (50%); error bars denote standard error of the mean.
76
prediction rate: 56.7%, MSE: 1.9%, t (14) = 3.6, P = 0.001)], left
middle frontal gyrus [50 voxels: prediction rate: 54.2%, MSE:
1.4%, t (14) = 2.99, P = 0.004; 150 voxels: prediction rate: 56.4%,
MSE: 1.6%, t (14) = 3.97, P < 0.001]. Consistent with our main
analysis the prediction rate was also above chance in the right
posterior STS, but statistical signicance was not reached after
multiple comparison correction [50 voxels: prediction rate:
53.7%, MSE: 1.9%, t (14) = 1.94, P = 0.04; 150 voxels: prediction
rate: 55.3%, MSE: 2.2%, t (14) = 2.4, P = 0.01]. Performance did
not differ from chance in left angular gyrus: [50 voxels: predic-
tion rate: 52.6%, MSE: 2.8%, t (14) < 1; 150 voxels: prediction
rate: 53.5%, MSE: 2.9%, t (14) = 1.19, P < 0.12], left supramargi-
nal gyrus: [50 voxels: prediction rate: 55%, MSE: 2.2%,
t (14) = 2.29, P = 0.018; 150 voxels: prediction rate: 52%, MSE:
3.1%, t (14) < 1], left orbital inferior frontal gyrus [50 voxels: pre-
diction rate: 53%, MSE: 2.5%, t (14) = 1.23, P = 0.12; 150 voxels:
prediction rate: 53.5%, MSE: 2.6%, t (14) = 1.37, P = 0.09], left
superior frontal gyrus [50 voxels: prediction rate: 53.8%, MSE:
2.2%, t (14) = 1.73, P = 0.05; 150 voxels: prediction rate: 50.2%,
MSE: 2.5%, t (14) < 1]. In left middle anterior temporal, left
anterior temporal, left inferior frontal gyrus and right middle
anterior temporal the prediction rate also did not differ from
chance and was <52.5% [t (14) < 1]. The results of this analysis
suggest that both the left posterior STS and the left middle
frontal gyrus contained the information, which permitted
reliable discrimination between meaningful sentences and
nonwords across different ROI sizes.
Discussion
The principal nding of this study was that subliminal meaning-
ful sentences and nonwords could be discriminated above
chance level based on BOLD signals in 2 regions: the left pos-
terior STS and the middle frontal gyrus. This supports the idea
that high-level language information can be processed in the
absence of awareness and critically provides important evidence
that unconscious language processing is not conned to
occipito-temporal lobes, but also involves the human frontal
lobes.
Cerebral Cortex August 2015, V 25 N 8 2165
 The language processing system is a widely distributed
brain network, which spans large regions of the left temporal
and frontal lobes (for reviews see Vigneau et al. 2006; Price
2012) as well as regions in the right temporal lobe (for reviews
see Jung-Beeman 2005; Vigneau et al. 2011). Strong activation
of the language network was observed in our functional locali-
zer experiment with visible stimuli, where all the regions
except for the superior frontal gyrus showed higher activation
to meaningful sentences compared with unpronounceable
nonwords (Table 3). Notably, the level of activations and the
amount of information is drastically attenuated with uncon-
scious stimulation. Even more relevant for the current discus-
sion, is that some regions like frontal lobes are usually not
activated at all during unconscious processing (for review see
Dehaene and Changeux 2011)evidence that is taken to
support theoretical models, such as Global Workspace Theory
(Dehaene et al. 1998). Yet, alternatively, it can be suggested
that weak sensory stimulation, which is an inevitable conse-
quence of rendering stimuli invisible, is simply not sufciently
strong to lead to activation in areas processing high-level infor-
mation such as the frontal lobes (Haynes 2009). In the current
study, to increase the probability of detecting the signals in the
frontal lobes we: 1) increased sensory stimulation by projecting
invisible stimuli for an extended period of time (10 s); 2) en-
hanced linguistic processing by showing meaningful sen-
tences, which in addition to semantics also contained syntax
and structure; 3) used multivariate ROI analysis approaches
that focused on language-selective regions, which were loca-
lized using independent experiment with visible stimuli. Criti-
cally, by implementing the awareness report procedure after
each block we ensured that the only blocks used in our analysis
were subjectively and objectively invisible on a per-participant
and per-block basis. We found that invisible meaningful sen-
tences and unpronounceable nonwords could be discriminated
beyond chance level in the left middle frontal gyrus (Fig. 3B).
The classication result was stable and did not depend on the
exact size of the ROI used (Fig. 4).
It is noteworthy, that because none of the previous studies
that have examined unconscious language processing could
reliably demonstrate unconscious language activity in the
frontal lobes, we deliberately designed the study to maximize
the difference between 2 conditions (meaningful sentences vs.
nonwords). Accordingly, the current design was not intended
to answer the question what type of language information (se-
mantic, syntactic, structural, and semantic context etc.) contrib-
uted to successful unconscious discrimination. Yet, the fact
that in the conscious language experiments the left middle
frontal gyrus has been shown to be most active in studies with
sentences (Bottini et al. 1994; Baumgaertner et al. 2002), text
processing (Vingerhoets et al. 2003), and complex language
material in general (for review see Vigneau et al. 2006) makes
it plausible that subliminal syntactic/structural information
might have contributed to successful discrimination between
meaningful sentences and nonwords. This interpretation is
also in line with a recent event-related potentials (ERP) study
(Batterink and Neville 2013), where using auditoryvisual at-
tentional blink paradigm (Raymond et al. 1992) the authors
showed that consciously undetected violations in written sen-
tence syntax processing elicited early frontal negativity 100
400 ms. Interestingly, we found no successful decoding in the
orbital inferior frontal gyrus and inferior frontal gyrus regions
(including Brocas area)the key regions of language
2166 Neural Correlates of Subliminal Language Processing Axelrod et al.
processing in general (Broca 1861). This result is apparently at
odds with the results of Diaz and McCarthy (2007) study, who
did report extended activations in the left inferior frontal gyrus
during a subliminal semantic task. Yet, as this study did not
assess awareness after each trial, it is possible that invisibility
was not complete and the participants were aware of the sub-
liminal stimuli during some of the trials.
From a broader conscious awareness theoretical perspec-
tive, the fact that unconscious information was successfully
decoded from a region in the frontal lobes suggests that infor-
mation can be processed by frontal lobes without automati-
cally triggering conscious awareness. This result is consistent
with rst order theories (Block 2005) and recurrent processing
view (Lamme 2006; van Gaal and Lamme 2011) which do not
attribute special role to the frontal lobes in achieving conscious
awareness. The present result does not support the original
formulation of Global Workspace Theory (Dehaene et al.
1998), which does not expect the frontal lobes to be activated
by unconscious stimulation. Yet, based on the recent elabor-
ation of this theory (Dehaene and Changeux 2011), our result
might not contradict it either since the activity we report was
localized in a specic region and did not span large portions of
the frontal lobe, as is the case for many conscious experiences
(Dehaene and Changeux 2011). Indeed, localized fMRI acti-
vations of the frontal lobes were previously shown in uncon-
scious cognitive control tasks (Lau and Passingham 2007; van
Gaal et al. 2008; see also EEG ndings: van Gaal et al. 2010,
2011). Yet, the neural systems responsible for cognitive control
and language processing are very different. Thus, the present
ndings complement current knowledge by showing that
unconscious language processing can also elicit localized
activity in the frontal lobes.
Interestingly, many previous studies using various stimuli,
such as faces (e.g., Sterzer et al. 2008; Schurger et al. 2010; Fah-
renfort et al. 2012) or words (Devlin et al. 2004; Nakamura et al.
2005, 2007) failed to nd unconscious information in the frontal
lobes. In the present study, we used a combination of exper-
imental design (long stimulation duration using CFS and mean-
ingful sentences) and data analysis (MVPA) procedures, which
has not been previously applied together. There was no way to
estimate a contribution of each one of these procedures, as our
experiment did not include direct, within experiment compari-
son, between different parameters (e.g., long stimulation dur-
ation using CFS vs. short backward masking stimulation).
Having said that, we were able to establish that the use of multi-
variate analysis (MVPA) played an important role in decoding
unconscious activity in the frontal lobes, as the successful de-
coding could not be achieved using univariate analysis. Similar
observation was made by another study, where invisible face/
scene stimuli could be decoded in the temporal lobe only by
using multivariate, but not using univariate approach (Sterzer
et al. 2008). Thus, future studies are needed to examine whether
the use of MVPA will permit to nd unconscious information in
the frontal lobes, also for the backward masking paradigms
(Devlin et al. 2004; Nakamura et al. 2005, 2007).
Additional nding of the current study was that subliminal
meaningful sentences and nonwords could be discriminated
beyond chance level from signals in the left posterior STS. This
result is consistent with previous studies that also reported
signals associated with unconscious processing of words in
this area (Devlin et al. 2004; Nakamura et al. 2005, 2007). Criti-
cally, only the present study implemented awareness report
77
after each block allowing characterization of subjective and ob-
jective invisibility. Compared with previous studies we used a
different methodological approach: experimental paradigm
(CFS vs. backward/forward mask priming), language stimuli
(sentences vs. words) and data analysis approach (MVPA)
(Norman et al. 2006) vs. fMR-adaptation (Grill-Spector et al.
1999)]. Importantly, despite the differences between the
studies the result in the left posterior STS was successfully re-
plicated, providing converging evidence and making a strong
case for the involvement of the left posterior STS in subliminal
text processing. It should be also noted, that discrimination
rate from signals in the right posterior STS was also relatively
high (Fig. 3B) and stable across ROIs of different sizes (Fig. 4),
thought it did not reach statistical signicance after correction
for multiple comparisons. An important role of bilateral pos-
terior STS in semantic processing with visible stimuli has been
shown using various tasks and paradigms (for reviews see
Jung-Beeman 2005; Vigneau et al. 2006, 2011). As this region
was also shown to be involved in syntactic processing (e.g.,
Ben-Shachar et al. 2004) it remains to be established what type
of information contributed the most to successful discrimi-
nation of the invisible stimuli. Interestingly, in our functional
localizer task with visible stimuli, while the activation of the
left posterior STS was the strongest among all the regions
(Table 3, average z-score column), the activation of the right
posterior STS was weaker than most other regions. Thus, suc-
cessful discrimination of invisible stimuli was not just a direct
consequence of a strong level of activation for visible stimuli
(Smith et al. 2011; Tong et al. 2012), but might also reect
regional specialization in the absence of aware processing.
Interestingly, while fMRI studies including the current
(Devlin et al. 2004; Nakamura et al. 2005, 2007; Diaz and
McCarthy 2007) consistently nd neural correlates of sublim-
inal meaningful text processing in the left posterior temporal
lobe, one recent ERP study, which also used CFS (Tsuchiya
and Koch 2005) failed to nd any modulation of the N400 se-
mantic component to invisible text content (Kang et al. 2011;
see also: Vogel et al. 1998; Heyman and Moors 2012). Though
differences in the signal measured by the 2 neuroimaging
methods can potentially explain these different results, based
on our current ndings we propose an additional interpret-
ation. The N400 is thought to originate from multiple sources
in the left temporal lobe (Kutas and Federmeier 2011). In the
present study we identied 5 regions in the left temporal lobe
(Fig. 2) that were all selective to meaningful text processing
when stimuli were visible (Table 3). Yet, when the text was
presented subliminally out of those 5 temporal regions only
one of them (left posterior STS) afforded successful discrimi-
nation of meaningful sentences from nonwords, while in the
other temporal ROIs performance did not differ from chance
level. Given that the N400 ERP component reects these mul-
tiple sources (Kutas and Federmeier 2011), it is possible that
the neural activity that originated in discriminative left pos-
terior STS was intermixed with the activity which originated in
neighboring, nondiscriminative regions (e.g., left middle
anterior temporal, left anterior temporal). As a result, the N400
component recorded on the scalp was not sensitive enough to
reect the subliminal semantic processing.
Finally, a novel methodological aspect of our work was the
presentation of long invisible sentences. While the CFS para-
digm was previously used for presenting subliminal words
(Costello et al. 2009; Kang et al. 2011; Yang and Yeh 2011) or 3
78
words sentence on one screen (Sklar et al. 2012) here we
propose a method of presenting invisible sentences of the un-
limited length. The presentation of invisible sentences for pro-
longed duration (e.g., 10 s in our case) gives an opportunity to
present not only subliminal text with much richer semantic
information, but also permits exploration of unconscious
neural correlates of complex language processing (e.g., syntax
processing). We suggest that this paradigm can be useful for
future studies in the eld.
In conclusion, in the current study we demonstrated that
based on the activity in the human left posterior STS and left
middle frontal gyrus it was possible to discriminate between
subliminally presented meaningful sentences and nonwords.
This result supports the notion that high-level language func-
tions might be processed subliminally and provides important
evidence that frontal regions might be involved in unconscious
language processing.
Funding
This study was supported by the Israeli Center of Research Ex-
cellence in Cognitive Sciences (V.A., M.B., and G.Y.), Daniel
Turnberg Travel Fellowship (V.A.) and the Wellcome Trust
(G.R.). Funding to pay the Open Access publication charges
for this article was provided by Wellcome Trust grant 100227/
Z/12/Z and Wellcome Trust strategic award 091593/Z/10/Z.
Notes
We thank Ev Fedorenko for assistance with establishing experimental
paradigm and for the permission to use anatomical masks. We would
also like to thank 2 anonymous reviewers for constructive comments.
Conict of Interest: None declared
References
Abrams RL, Greenwald AG. 2000. Parts outweigh the whole (word) in
unconscious analysis of meaning. Psychol Sci. 11:118124.
Aksoy S, Haralick RM. 2001. Feature normalization and likelihood-
based similarity measures for image retrieval. Pattern Recognit Lett.
22:563582.
Almeida J, Mahon BZ, Nakayama K, Caramazza A. 2008. Unconscious
processing dissociates along categorical lines. Proc Natl Acad Sci
USA. 105:1521415218.
Almeida J, Pajtas P, Mahon B, Nakayama K, Caramazza A. 2013. Affect
of the unconscious: Visually suppressed angry faces modulate our
decisions. Cogn Affect Behav Neurosci. 13:94101.
Ashburner J, Friston KJ. 2005. Unied segmentation. Neuroimage.
26:839851.
Axelrod V, Yovel G. 2012. Hierarchical processing of face viewpoint in
human visual cortex. J Neurosci. 32:24422452.
Batterink L, Neville HJ. 2013. The human brain processes syntax in the
absence of conscious awareness. J Neurosci. 33:85288533.
Baumgaertner A, Weiller C, Bchel C. 2002. Event-related fMRI reveals
cortical sites involved in contextual sentence integration. Neuro-
image. 16:736745.
Ben-Shachar M, Palti D, Grodzinsky Y. 2004. Neural correlates of syn-
tactic movement: converging evidence from two fMRI experiments.
Neuroimage. 21:13201336.
Block N. 2005. Two neural correlates of consciousness. Trends Cogn
Sci. 9:4652.
Bottini G, Corcoran R, Sterzi R, Paulesu E, Schenone P, Scarpa P, Frack-
owiak RS, Frith D. 1994. The role of the right hemisphere in the
interpretation of gurative aspects of language A positron emission
tomography activation study. Brain. 117:12411253.
Cerebral Cortex August 2015, V 25 N 8 2167
Broca P. 1861. Remarques sur le sige de la facult du langage articul,
suivies dune observation daphmie ( perte de la parole). Bull Soc
Anat. 6:330357.
Cohen L, Dehaene S, Naccache L, Lehricy S, Dehaene-Lambertz G,
Hnaff M-A, Michel F. 2000. The visual word form area. Brain.
123:291307.
Costello P, Jiang Y, Baartman B, McGlennen K, He S. 2009. Semantic
and subword priming during binocular suppression. Conscious
Cogn. 18:375382.
Coutanche MN. 2013. Distinguishing multi-voxel patterns and mean
activation: why, how, and what does it tell us? Cogn Affect Behav
Neurosci. 13:667673.
Coutanche MN, Thompson-Schill SL. 2012. The advantage of brief
fMRI acquisition runs for multi-voxel pattern detection across runs.
Neuroimage. 61:11131119.
Dehaene S, Changeux J-P. 2011. Experimental and theoretical ap-
proaches to conscious processing. Neuron. 70:200227.
Dehaene S, Cohen L. 2011. The unique role of the visual word form
area in reading. Trends Cogn Sci. 15:254262.
Dehaene S, Kerszberg M, Changeux J-P. 1998. A neuronal model of a
global workspace in effortful cognitive tasks. Proc Natl Acad Sci
USA. 95:1452914534.
Dehaene S, Naccache L, Cohen L, Bihan DL, Mangin J-F, Poline J-B,
Riviere D. 2001. Cerebral mechanisms of word masking and uncon-
scious repetition priming. Nat Neurosci. 4:752758.
Devlin JT, Jamison HL, Matthews PM, Gonnerman LM. 2004. Mor-
phology and the internal structure of words. Proc Natl Acad Sci
USA. 101:1498414988.
Diaz MT, McCarthy G. 2007. Unconscious word processing engages a
distributed network of brain regions. J Cogn Neurosci. 19:
17681775.
Eger E, Ashburner J, Haynes JD, Dolan RJ, Rees G. 2008. fMRI activity
patterns in human LOC carry information about object exemplars
within category. J Cogn Neurosci. 20:356370.
Fahrenfort JJ, Snijders TM, Heinen K, van Gaal S, Scholte HS, Lamme
VAF. 2012. Neuronal integration in visual cortex elevates face cat-
egory tuning to conscious face perception. Proc Natl Acad Sci USA.
109:2150421509.
Faivre N, Berthet V, Kouider S. 2012. Nonconscious inuences from
emotional faces: a comparison of visual crowding, masking, and
continuous ash suppression. Front Psychol. 3:129.
Fedorenko E, Behr MK, Kanwisher N. 2011. Functional specicity for
high-level linguistic processing in the human brain. Proc Natl Acad
Sci USA. 108:1642816433.
Fedorenko E, Hsieh P-J, Nieto-Castan A, Whiteld-Gabrieli S, Kanw-
isher N. 2010. New method for fMRI investigations of language: de-
ning ROIs functionally in individual subjects. J Neurophysiol.
104:11771194.
Fedorenko E, Nieto-Castaon A, Kanwisher N. 2012. Lexical and syn-
tactic representations in the brain: an fMRI investigation with multi-
voxel pattern analyses. Neuropsychologia. 50:499513.
Grill-Spector K, Kushnir T, Edelman S, Avidan G, Itzchak Y, Malach R.
1999. Differential processing of objects under various viewing con-
ditions in the human lateral occipital complex. Neuron. 24:187203.
Haynes J-D. 2009. Decoding visual consciousness from human brain
signals. Trends Cogn Sci. 13:194202.
Heyman T, Moors P. 2012. Using interocular suppression and EEG to
study semantic processing. J Neurosci. 32:15151516.
Holender D. 1986. Semantic activation without conscious identication
in dichotic listening, parafoveal vision, and visual masking: a
survey and appraisal. Behav Brain Sci. 9:123.
Japkowicz N, Stephen S. 2002. The class imbalance problem: a sys-
tematic study. Intell Data Anal. 6:429449.
Jiang Y, Costello P, He S. 2007. Processing of invisible stimuli: advan-
tage of upright faces and recognizable words in overcoming intero-
cular suppression. Psychol Sci. 18:349355.
Jung-Beeman M. 2005. Bilateral brain processes for comprehending
natural language. Trends Cogn Sci. 9:512518.
Kanai R, Walsh V, Tseng C-h. 2010. Subjective discriminability of invisi-
bility: a framework for distinguishing perceptual and attentional
failures of awareness. Conscious Cogn. 19:10451057.
2168 Neural Correlates of Subliminal Language Processing Axelrod et al.
Kang M-S, Blake R, Woodman GF. 2011. Semantic analysis does not
occur in the absence of awareness induced by interocular suppres-
sion. J Neurosci. 31:1353513545.
Kouider S, de Gardelle V, Sackur J, Dupoux E. 2010. How rich is con-
sciousness? The partial awareness hypothesis. Trends Cogn Sci.
14:301307.
Kouider S, Dehaene S. 2007. Levels of processing during non-
conscious perception: a critical review of visual masking. Philos
Trans R Soc B Biol Sci. 362:857875.
Kouider S, Dehaene S, Jobert A, Le Bihan D. 2007. Cerebral bases of
subliminal and supraliminal priming during reading. Cereb Cortex.
17:20192029.
Kutas M, Federmeier KD. 2011. Thirty years and counting: nding
meaning in the N400 component of the event-related brain poten-
tial (ERP). Annu Rev Psychol. 62:621647.
Lamme VAF. 2006. Towards a true neural stance on consciousness.
Trends Cogn Sci. 10:494501.
Lau HC, Passingham RE. 2007. Unconscious activation of the cognitive
control system in the human prefrontal cortex. J Neurosci.
27:58055811.
Lin Z, He S. 2009. Seeing the invisible: the scope and limits of uncon-
scious processing in binocular rivalry. Prog Neurobiol. 87:195211.
Marcel AJ. 1983. Conscious and unconscious perception: experiments
on visual masking and word recognition. Cogn Psychol. 15:197237.
Mendola JD, Conner IP. 2007. Eye dominance predicts fMRI signals in
human retinotopic cortex. Neurosci Lett. 414:3034.
Meyer K, Kaplan JT, Essex R, Webber C, Damasio H, Damasio A. 2010.
Predicting visual stimuli on the basis of activity in auditory cortices.
Nat Neurosci. 13:667668.
Miles WR. 1930. Ocular dominance in human adults. J Gen Psychol.
3:412430.
Misaki M, Kim Y, Bandettini PA, Kriegeskorte N. 2010. Comparison of
multivariate classiers and response normalizations for pattern-
information fMRI. Neuroimage. 53:103118.
Nakamura K, Dehaene S, Jobert A, Bihan DL, Kouider S. 2005. Sublim-
inal convergence of Kanji and Kana words: further evidence for
functional parcellation of the posterior temporal cortex in visual
word perception. J Cogn Neurosci. 17:954968.
Nakamura K, Dehaene S, Jobert A, Le Bihan D, Kouider S. 2007. Task-
specic change of unconscious neural priming in the cerebral
language network. Proc Natl Acad Sci USA. 104:1964319648.
Nestor A, Plaut DC, Behrmann M. 2011. Unraveling the distributed
neural code of facial identity through spatiotemporal pattern analy-
sis. Proc Natl Acad Sci USA. 108:999810003.
Norman KA, Polyn SM, Detre GJ, Haxby JV. 2006. Beyond mind-
reading: multi-voxel pattern analysis of fMRI data. Trends Cogn Sci.
10:424430.
Price CJ. 2012. A review and synthesis of the rst 20 years of PET and
fMRI studies of heard speech, spoken language and reading. Neu-
roimage. 62:816847.
Raizada RD, Tsao F-M, Liu H-M, Holloway ID, Ansari D, Kuhl PK.
2010. Linking brain-wide multivoxel activation patterns to behav-
iour: examples from language and math. Neuroimage. 51:462471.
Raymond JE, Shapiro KL, Arnell KM. 1992. Temporary suppression of
visual processing in an RSVP task: an attentional blink. J Exp
Psychol Hum Percept Perform. 18:849860.
Said CP, Moore CD, Engell AD, Todorov A, Haxby JV. 2010. Distributed
representations of dynamic facial expressions in the superior tem-
poral sulcus. J Vis. 10:112.
Schurger A, Pereira F, Treisman A, Cohen JD. 2010. Reproducibility dis-
tinguishes conscious from nonconscious neural representations.
Science. 327:9799.
Serences JT, Saproo S, Scolari M, Ho T, Muftuler LT. 2009. Estimating the
inuence of attention on population codes in human visual cortex
using voxel-based tuning functions. Neuroimage. 44:223231.
Seth AK, Dienes Z, Cleeremans A, Overgaard M, Pessoa L. 2008.
Measuring consciousness: relating behavioural and neurophysiolo-
gical approaches. Trends Cogn Sci. 12:314321.
Sklar AY, Levy N, Goldstein A, Mandel R, Maril A, Hassin RR. 2012.
Reading and doing arithmetic nonconsciously. Proc Natl Acad Sci
USA. 109:1961419619.
79
Smith AT, Kosillo P, Williams AL. 2011. The confounding effect of
response amplitude on MVPA performance measures. Neuroimage.
56:525530.
Sterzer P, Haynes J-D, Rees G. 2008. Fine-scale activity patterns in high-
level visual areas encode the category of invisible objects. J Vis. 8:112.
Tong F, Harrison SA, Dewey JA, Kamitani Y. 2012. Relationship between
BOLD amplitude and pattern classication of orientation-selective
activity in the human visual cortex. Neuroimage. 63:12121222.
Tsuchiya N, Koch C. 2005. Continuous ash suppression reduces nega-
tive afterimages. Nat Neurosci. 8:10961101.
Van den Bussche E, Van den Noortgate W, Reynvoet B. 2009. Mechan-
isms of masked priming: a meta-analysis. Psychol Bull. 135:452477.
van Gaal S, Lamme VA, Fahrenfort JJ, Ridderinkhof KR. 2011. Dissoci-
able brain mechanisms underlying the conscious and unconscious
control of behavior. J Cogn Neurosci. 23:91105.
van Gaal S, Lamme VAF. 2011. Unconscious high-level information pro-
cessing: implication for neurobiological theories of consciousness.
The Neuroscientist. 18:287301.
van Gaal S, Ridderinkhof KR, Fahrenfort JJ, Scholte HS, Lamme VA.
2008. Frontal cortex mediates unconsciously triggered inhibitory
control. J Neurosci. 28:80538062.
van Gaal S, Ridderinkhof KR, Scholte HS, Lamme VA. 2010. Uncon-
scious activation of the prefrontal no-go network. J Neurosci.
30:41434150.
80
Vigneau M, Beaucousin V, Herv PY, Duffau H, Crivello F, Houd O,
Mazoyer B, Tzourio-Mazoyer N. 2006. Meta-analyzing left hemi-
sphere language areas: Phonology, semantics, and sentence proces-
sing. Neuroimage. 30:14141432.
Vigneau M, Beaucousin V, Herv P-Y, Jobard G, Petit L, Crivello F,
Mellet E, Zago L, Mazoyer B, Tzourio-Mazoyer N. 2011. What is
right-hemisphere contribution to phonological, lexico-semantic,
and sentence processing? Insights from a meta-analysis. Neuro-
image. 54:577593.
Vingerhoets G, Borsel JV, Tesink C, van den Noort M, Deblaere K, Seur-
inck R, Vandemaele P, Achten E. 2003. Multilingualism: an fMRI
study. Neuroimage. 20:21812196.
Vogel EK, Luck SJ, Shapiro KL. 1998. Electrophysiological evi-
dence for a postperceptual locus of suppression during the at-
tentional blink. J Exp Psychol Hum Percept Perform.
24:16561674.
Walther DB, Caddigan E, Fei-Fei L, Beck DM. 2009. Natural scene cat-
egories revealed in distributed patterns of activity in the human
brain. J Neurosci. 29:1057310581.
Wentura D, Kulfanek M, Greve W. 2005. Masked affective priming by
name letters: evidence for a correspondence of explicit and implicit
self-esteem. J Exp Soc Psychol. 41:654663.
Yang Y-H, Yeh S-L. 2011. Accessing the meaning of invisible words.
Conscious Cogn. 20:223233.
Cerebral Cortex August 2015, V 25 N 8 2169
 On Making the Right Choice: The
Deliberation-Without-Attention Effect
Ap Dijksterhuis, et al.
Science 311, 1005 (2006);
DOI: 10.1126/science.1121629

The following resources related to this article are available online at

www.sciencemag.org (this information is current as of May 27, 2007 ):

Updated information and services, including high-resolution figures, can be found in the online
version of this article at:
http://www.sciencemag.org/cgi/content/full/311/5763/1005

Supporting Online Material can be found at:

http://www.sciencemag.org/cgi/content/full/311/5763/1005/DC1

Downloaded from www.sciencemag.org on May 27, 2007

A list of selected additional articles on the Science Web sites related to this article can be
found at:
http://www.sciencemag.org/cgi/content/full/311/5763/1005#related-content
This article cites 13 articles, 2 of which can be accessed for free:
http://www.sciencemag.org/cgi/content/full/311/5763/1005#otherarticles

This article has been cited by 11 article(s) on the ISI Web of Science.

This article has been cited by 3 articles hosted by HighWire Press; see:
http://www.sciencemag.org/cgi/content/full/311/5763/1005#otherarticles

This article appears in the following subject collections:

Psychology
http://www.sciencemag.org/cgi/collection/psychology

Information about obtaining reprints of this article or about obtaining permission to reproduce
this article in whole or in part can be found at:
http://www.sciencemag.org/about/permissions.dtl

Science (print ISSN 0036-8075; online ISSN 1095-9203) is published weekly, except the last week in December, by the
American Association for the Advancement of Science, 1200 New York Avenue NW, Washington, DC 20005. Copyright
c 2006 by the American Association for the Advancement of Science; all rights reserved. The title SCIENCE is a
registered trademark of AAAS.
81

13. E. J. Eide, H. Kang, S. Crapo, M. Gallego, D. M. Virshup,
Methods Enzymol. 393, 408 (2005).
14. S. Martinek, S. Inonog, A. S. Manoukian, M. W. Young,
Cell 105, 769 (2001).
15. M. Abe, E. D. Herzog, G. D. Block, Neuroreport 11, 3261
(2000).
16. C. Iitaka, K. Miyazaki, T. Akaike, N. Ishida, J. Biol. Chem.
280, 29397 (2005).
17. P. S. Klein, D. A. Melton, Proc. Natl. Acad. Sci. U.S.A. 93,
8455 (1996).
18. G. Adelmant, A. Begue, D. Stehelin, V. Laudet, Proc. Natl.
Acad. Sci. U.S.A. 93, 3553 (1996).
19. B. W. Doble, J. R. Woodgett, J. Cell Sci. 116, 1175
(2003).
20. A. Balsalobre, F. Damiola, U. Schibler, Cell 93, 929 (1998).
21. F. Zhang, C. J. Phiel, L. Spece, N. Gurvich, P. S. Klein,
J. Biol. Chem. 278, 33067 (2003).
22. A. J. Gelenberg et al., J. Med. 321, 1489 (1989).
23. J. A. Ripperger, L. P. Shearman, S. M. Reppert,
U. Schibler, Genes Dev. 14, 679 (2000).
24. E. Iwahana et al., Eur. J. Neurosci. 19, 2281 (2004).
25. F. Benedetti et al., Neurosci. Lett. 368, 123 (2004).
26. R. H. Lenox, T. D. Gould, H. K. Manji, Am. J. Med. Genet.
114, 391 (2002).
27. M. Gottlicher et al., EMBO J. 20, 6969 (2001).
28. C. J. Phiel et al., J. Biol. Chem. 276, 36734 (2001).
29. M. E. Dokucu, L. Yu, P. H. Taghert, Neuropsychopharma-
cology 30, 2216 (2005).
REPORTS
30. We thank the University of Pennsylvania Diabetes and
Endocrinology Research Center Vector Core (NIH DK
19525) and M. J. Birnbaum for GSK3b adenoviruses,
and J. D. Alvarez and A. Sehgal for helpful discussions.
Supported by NIH DK45586 (to M.A.L.) and NIH
MH058324 (to P.S.K.).
Supporting Online Material
www.sciencemag.org/cgi/content/full/311/5763/1002/DC1
Materials and Methods
Figs. S1 to S7
20 October 2005; accepted 14 January 2006
10.1126/science.1121613

On Making the Right Choice: The hardly developed beyond the status of Bfolk
wisdom.[ It has been postulated or investigated
by scientists infrequently Ebut see (1013)^. The
Deliberation-Without-Attention Effect question addressed here is whether this view is
justified. We hypothesize that it is not.

Downloaded from www.sciencemag.org on May 27, 2007

Ap Dijksterhuis,* Maarten W. Bos, Loran F. Nordgren, Rick B. van Baaren First, conscious thought does not always
lead to sound choices. For example, participants
Contrary to conventional wisdom, it is not always advantageous to engage in thorough conscious who chose their favorite poster among a set of
deliberation before choosing. On the basis of recent insights into the characteristics of conscious five after thorough contemplation showed less
and unconscious thought, we tested the hypothesis that simple choices (such as between different postchoice satisfaction than participants who
towels or different sets of oven mitts) indeed produce better results after conscious thought, but only looked at them briefly (14, 15). Further-
that choices in complex matters (such as between different houses or different cars) should be left more, conscious deliberation can make multiple
to unconscious thought. Named the deliberation-without-attention hypothesis, it was confirmed evaluations of the same object less consistent
in four studies on consumer choice, both in the laboratory as well as among actual shoppers, that over time (16). Two reasons why conscious
purchases of complex products were viewed more favorably when decisions had been made in the deliberation sometimes leads to poor judgments
absence of attentive deliberation. have been identified. First, consciousness has
a low capacity (17, 18), causing choosers to take

C
ommon knowledge holds that thor-
ough conscious thought leads to good
decisions and satisfactory choices.
Whether purchasing a new car, a desktop com-
puter, or a pair of shoes, people generally
believe that serious conscious deliberation in-
creases the probability that they will make the
Bright[ choice. This idea applies especially to
choices between products that are complex,
multifaceted, and expensive. Whereas most
people are willing to buy a new set of towels
without much thought, they are unlikely to buy
a new car or outfit a new kitchen without
deliberation.
A second pervasive idea is that the quality
of a choice benefits from Bsleeping on it.[
Rather than (or in addition to) thinking con-
sciously, people usually feel that Bunconscious
thought[ is useful for making sound decisions.
Whereas conscious thought refers to thought or
deliberation while conscious attention is di-
rected at the problem at hand, unconscious
thought can be defined as thought or delibera-
tion in the absence of conscious attention di-
rected at the problem (1). An example of
unconscious thought is the following: One
compares two holiday destinations (say the
Costa Brava and Tuscany) and does not know
what to decide. One puts the problem aside and
Department of Psychology, University of Amsterdam,
Roetersstraat 15, 1018 WB, Amsterdam, the Netherlands.
*To whom correspondence should be addressed. E-mail:
a.j.dijksterhuis@uva.nl
after 48 hours of not thinking about it con-
sciously, suddenly the thought BIt_s going to be
Tuscany![ pops into consciousness. This thought
itself is conscious, but the transition from
indecision to a preference 2 days later is the
result of unconscious thought, or of deliberation
without attention.
The scientific literature has emphasized the
benefits of conscious deliberation in decision
making for hundreds of years (2, 3). The idea
that conscious deliberation is the ideal (if not
always attainable) way to approach a decision
forms the backbone of classic (4, 5) as well as
contemporary perspectives on decision making
(6, 7) and attitude formation (8, 9). In contrast,
the notion that unconscious thought is fruitful
Fig. 1. Percentage of participants who chose the
most desirable car as a function of complexity of
decision and of mode of thought (n 0 18 to 22 in
each condition). Error bars represent the stan-
dard error.
into account only a subset of the relevant
information when they decide (13, 19). Second,
conscious thought can lead to suboptimal weight-
ing of the importance of attributes (1316):
We tend to inflate the importance of some at-
tributes at the expense of others, leading to
worse choices.
Conversely, unconscious thought, or thought
without attention, can lead to good choices
(13, 14). In a recent experiment, participants
read information about four apartments of dif-
ferent desirability (20). They were either asked
to choose their favorite immediately, or given
the opportunity to choose after a period of
conscious thought, or distracted for some time
Fig. 2. Difference in attitude (on a scale of 25 to
25) toward the desirable and undesirable car as a
function of complexity of decision and of mode of
thought (n 0 12 to 14 in each condition). Error
bars represent the standard error.

www.sciencemag.org 82 VOL 311

SCIENCE 17 FEBRUARY 2006 1005
 REPORTS
before they chose. In the third of these con-
ditions, participants could only engage in
unconscious deliberation: They knew they
would have to choose later, but the distraction
task prevented them from devoting conscious
attention to the choice. Interestingly, un-
conscious thinkers made better decisions than
conscious thinkers or than immediate choosers
(13, 14).
Recently, we formulated the Unconscious
Thought Theory (UTT) (21) about the strengths
and weaknesses of conscious thought and un-
conscious thought, that is, of deliberation with
and without attention. Two characteristics of
conscious and unconscious thought are impor-
tant in the current context. First, conscious
thought is rule-based and very precise (22, 23).
Unconscious thought can conform to rules in
that it detects recurring patterns, as the literature
on implicit learning shows (24). However, in
order to actively follow strict rules, conscious
attention is necessary. For example, one cannot
do arithmetic without conscious attention. This
capacity to follow rules makes conscious thought
more precise in decision making, because it
can strictly follow self-generated rules such as
not exceeding a maximum price. Second, as
alluded to earlier, conscious thought suffers
from the low capacity of consciousness, mak-
ing it less suitable for very complex issues.
Unconscious thought does not suffer from low
capacity. Indeed, it has been shown that during
unconscious thought, large amounts of infor-
mation can be integrated into an evaluative
summary judgment (13).
These characteristics of conscious and un-
conscious thought led us to postulate the
Bdeliberation-without-attention[ hypothesis, on
the relation between mode of thought or
deliberation (conscious versus unconscious)
and the complexity and quality of choice.
Complexity is defined as the amount of in-
formation a choice involves. A choice between
objects for which one or two attributes are
important (such as oven mitts or toothpaste) is
simple, whereas a choice between objects for
which many attributes are important (cars or
Fig. 3. The relation between mode of
thought and postchoice satisfaction
(on a scale of 1 to 7) for the six
products most frequently chosen in
study 3. Higher bars indicate more
satisfaction. The more complex the
product (on a scale of 1 to 5), the fur-
ther to the right it is shown. The
complexity score is given in parenthe-
ses. Participants were divided into
conscious and unconscious thinkers
on the basis of a median-split for
each product individually. Each bar
represents between two and five
participants.
1006 17 FEBRUARY 2006
houses) is complex. Conscious thought is hy-
pothesized, due to its precision, to lead to good
choices in simple matters. However, because of
its low capacity, conscious thought leads to
progressively worse choices with more com-
plex issues. Unconscious thought (i.e., deliber-
ation without attention) is expected, because of
its relative lack of precision, to lead to choices
of lower quality. However, the quality of choice
does not deteriorate with increased complexity,
allowing unconscious thought to lead to better
choices than conscious thought under com-
plex circumstances, this latter idea being the
kernel of the deliberation-without-attention
hypothesis. Quality of choice was operation-
alized both normatively (studies 1 and 2) as
well as subjectively (as postchoice satisfaction,
in studies 3 and 4).
Study 1. Participants were subjected to a 2
(mode of thought: conscious versus uncon-
scious) ! 2 (complexity of choice problem: sim-
ple versus complex) factorial design (25). All
participants read information about four hypo-
thetical cars. Depending on the condition, each
car was characterized by 4 attributes (simple) or
by 12 attributes (complex). The attributes were
either positive or negative. One car was char-
acterized by 75% positive attributes, two by
50% positive attributes, and one by 25% posi-
tive attributes (supporting online text). After
reading the information about the four cars,
participants were assigned either to a conscious
thought condition or to an unconscious thought
condition. In the conscious thought condition,
participants were asked to think about the cars
for 4 min before they chose their favorite car.
In the unconscious thought condition, partic-
ipants were distracted for 4 min (they solved
anagrams) and were told that after the period of
distraction they would be asked to choose the
best car.
The percentages of participants who chose
the best car are shown in Fig. 1. The crucial two-
way interaction supporting the deliberation-
without-attention hypothesis was significant
EF(1,76) 0 4.85, P G 0.04^. Unconscious think-
ers fared relatively well and showed no differ-
VOL 311 83
SCIENCE www.sciencemag.org
ences between conditions (F G 1, not significant).
Conscious thinkers generally made the proper
choice under simple conditions, but performed
poorly under complex circumstances EF(1,40) 0
4.95, P G 0.04^.
Study 2. For the second study we made one
change (25). Rather than asking for a choice,
we asked participants about their attitudes
toward each of the four cars. As the dependent
variable, we used the difference in attitude
toward the best car and the worst car. Again,
conscious thinkers were better able to differen-
tiate the quality of the cars under simple
conditions, whereas unconscious thinkers were
better able to differentiate the quality of the cars
under complex conditions EF(1,47) 0 5.63, P G
0.03^. The means are shown in Fig. 2.
Study 3. In a pilot study, undergraduate
students were asked how many aspects of a
Downloaded from www.sciencemag.org on May 27, 2007
product they would take into account in the
purchase of 40 different products. In this way,
we obtained an average Bcomplexity score[
for 40 different products (supporting online
text).
For the actual study, other students were
presented with this list of 40 products. From the
list, they were asked to choose a product that
they had recently purchased and were asked the
following questions: Which product did you
purchase? Did you know the product before
you went on the shopping trip? How much did
you think about the product between seeing it
for the first time and buying it? How satisfied
are you with the product?
To test our hypothesis, we distinguished
participants who thought (either consciously or
unconsciously) about their purchase from im-
pulse buyers who did not think much at all.
Hence, participants who indicated that they
bought a product they had never come across
before the shopping trip were not included,
leaving only participants who knew the product
beforehand (n 0 49).
It is impossible to know whether people are
engaged in unconscious thought by asking
them, so strictly speaking, we can only test
the relationship between conscious thought,
Fig. 4. Postchoice satisfaction of IKEA (n 0 27)
and Bijenkorf (n 0 27) shoppers as a function
of mode of thought. Error bars represent the
standard error.

complexity, and quality. However, it follows
from our definition of conscious and uncon-
scious thought (according to which attention
to the problem at hand is the crucial distin-
guishing factor) that they are at least partly
dependent. At any one point in time, attention
is either directed at the decision under con-
sideration, or it is not; that is, at any partic-
ular point in time, either you are attending to
buying a car, or you are not. The more you
think about a decision consciously (that is,
with attention), the less time remains to think
about the same decision unconsciously (that
is, without attention).
We regressed the amount of thought and the
average number of aspects on postchoice satis-
faction. As expected, thinking does not make
people more satisfied, nor does complexity
(t_s G 1). However, the interaction of the two
parameters significantly predicted postchoice
satisfaction Et(48) 0 2.13, P G 0.04^. Correla-
tions were calculated between amount of
thought and postchoice satisfaction for three
categories of products: complex, medium, and
simple. For products of medium complexity, no
correlation was found Er(18) 0 0.03^; for
simple products, a positive correlation was
found Er(15) 0 0.57, P G 0.03^; and for complex
products, a negative correlation was found
Er(16) 0 0.56, P G 0.03^. As expected, the
more people thought consciously about simple
products, the more satisfied they were with
their purchase. Conversely, the more people
thought consciously about complex products,
the less satisfied they were with their purchase.
Figure 3 depicts satisfaction as a function of
mode of thought for the six most frequently
chosen products (26).
Study 4. On the basis of the pilot study to
study 3, two shops were selected: one where
people generally buy complex products (IKEA,
which sells mainly furniture) and one where
people generally buy simple products (Bijenkorf,
a department store like Macy_s that sells
clothes, clothing accessories, and kitchen ac-
cessories). At the exit, shoppers were asked the
following questions: What did you buy? How
expensive was it? Did you know the product
before you went on the shopping trip? and
How much did you think about the product
between seeing it for the first time and buying
it? A few weeks later, the shoppers were asked
(over the phone) how satisfied they were with
their purchases. As in study 3, participants who
indicated that they bought a product they had
never come across before the shopping trip
were not included.
We divided participants (Bthinkers[) on the
basis of a median-split procedure into those
who engaged in much conscious thought
(conscious thinkers) and those who engaged
in little conscious thought (unconscious think-
ers). As expected, conscious thinkers reported
more postchoice satisfaction than uncon-
scious thinkers for Bijenkorf products (simple
www.sciencemag.org
products) EF(1,25) 0 6.52, P G 0.02^. The oppo-
site was true for the IKEA customers (complex
products), in which case unconscious thinkers
showed more postchoice satisfaction than con-
scious thinkers EF(1,25) 0 6.12, P G 0.02^
(Fig. 4).
In sum, in four studies we demonstrated the
deliberation-without-attention effect. Conscious
thinkers were better able to make the best
choice among simple products, whereas un-
conscious thinkers were better able to make the
best choice among complex products. Among
people who knew the product they purchased
before they went on a shopping trip, the amount
of conscious thought was positively related to
postchoice satisfaction for simple products and
negatively related to postchoice satisfaction for
complex products.
Our aim was to test the Bdeliberation-
without-attention[ hypothesis both in the labo-
ratory and among shoppers. In that sense, it is
important to view our set of studies as a whole
rather than as a series of individual studies.
Study 4 has unavoidable disadvantages such as
that the IKEA and Bijenkorf samples may have
differed (after all, different shops attract a
different clientele), which naturally opens the
potential for alternative explanations. There-
fore, study 3 was done in order to Bbridge[ the
laboratory studies with study 4. It has many of
the assets of study 4 (real choices between real
products with satisfaction as the dependent
variable), except that all participants were
students.
Although we investigated choices among
consumer products in our studies, there is no a
priori reason to assume that the deliberation-
without-attention effect does not generalize to
other types of choicespolitical, managerial,
or otherwise. In such cases, it should benefit
the individual to think consciously about
simple matters and to delegate thinking about
more complex matters to the unconscious.
References and Notes
1. It is important to note that attention to the problem at
hand is the crucial distinction in our definitions of conscious
and unconscious thought. Thinking about buying a new
car while attention is directed at possible new cars is
conscious thought. Thinking about buying a new car
while attention is temporarily directed elsewhere is
unconscious thought. This distinction does not mean that
conscious thought only comprises conscious processes.
One can compare it to speech. Speech is a conscious
process (i.e., attention is directed at it while one speaks),
but it is in part dependent on accompanying unconscious
processes (such as processes responsible for syntax or
word choice).
2. R. Descartes, The Philosophical Writings of Descartes
(Cambridge Univ. Press, Cambridge, 1984).
3. J. Locke, An Essay Concerning Human Understanding
(Penguin, London, 1689/1997).
4. I. L. Janis, L. Mann, Decision Making: A Psychological
Analysis of Conflict, Choice, and Commitment (Free
Press, New York, 1977).
5. H. A. Simon, Q. J. Econ. 69, 99 (1955).
6. J. R. Bettman, M. F. Luce, J. W. Payne, J. Consum. Res. 25,
187 (1998).
7. D. Kahneman, Am. Psychol. 58, 697 (2003).
8. S. Chaiken, J. Pers. Soc. Psychol. 39, 752 (1980).
84 VOL 311
SCIENCE 17 FEBRUARY 2006
REPORTS
9. R. E. Petty, J. T. Cacioppo, in Advances in Experimental
Social Psychology, L. Berkowitz, Ed. (Academic Press,
New York, 1986), vol. 19, pp. 123205.
10. As far as we know, the possibility of unconscious thought
(as well as the term unconscious thought) was explicitly
used for the first time by Schopenhauer, who remarked
that One might almost believe that half of our thinking
takes place unconsciously (27).
11. G. Claxton, Hare Brain, Tortoise Mind: How Intelligence
Increases When You Think Less (Harper Collins, New York,
1997).
12. The idea of unconscious thought was also studied by
various researchers interested in incubation processes in
creativity. An example is K. S. Bowers et al. (28).
13. A. Dijksterhuis, J. Pers. Soc. Psychol. 87, 586 (2004).
14. A. Dijksterhuis, Z. van Olden, J. Exp. Soc. Psychol.,
in press.
15. T. D. Wilson et al., Pers. Soc. Psychol. Bull. 19, 331
(1993).
16. G. M. Levine, J. B. Halberstadt, R. L. Goldstone, J. Pers.
Soc. Psychol. 70, 230 (1996).
17. G. A. Miller, Psych. Rev. 63, 81 (1956).
18. T. Nrretranders, The User Illusion: Cutting Consciousness
Downloaded from www.sciencemag.org on May 27, 2007
Down to Size (Viking, New York, 1998).
19. T. D. Wilson, J. W. Schooler, J. Pers. Soc. Psychol. 60, 181
(1991).
20. Quality of decision was operationalized from a
normative perspective. One of the choice options was
made more desirable than the others because it
had been assigned more positive aspects than the
other apartments.
21. A. Dijksterhuis, L. F. Nordgren, Persp. Psychol. Sci.,
in press.
22. The important distinction between following rules and
merely conforming to them (and the need for conscious
attention in the former) was made by S. A. Sloman
(29).
23. E. R. Smith, J. DeCoster, Dual-Process Theories in
Social Psychology, S. Chaiken, Y. Trope, Eds. (Guilford,
New York, 1999), pp. 323336.
24. Unconscious thought is reminiscent of implicit learning,
but there is an important difference. Implicit learning
refers to aspects of a task that are learned while
working on the task (and that are inaccessible to
consciousness). Unconscious thought refers to thought
processes that take place after the encoding of relevant
information. A good example of this definition of
unconscious thought is the groundbreaking work by
Stickgold and colleagues on learning during sleep.
See, e.g., (30, 31).
25. Materials and methods are available as supporting
material on Science Online.
26. We found a correlation between number of aspects and
amount of thought (r 0 0.54, P G 0.001): The more
complex a product is, the more people think consciously
when deciding to purchase it. Understandable as this
may be, our analysis suggests that people should do the
opposite, i.e., think unconsciously when deciding to
purchase a complex product. The correlation between
number of aspects and price was also significant
(r 0 0.45, P G 0.001): Expensive products were more
complex than inexpensive ones.
27. A. Schopenhauer, in Essays and Aphorisms (Penguin,
London, 1851/1970), p. 123.
28. K. S. Bowers, G. Regehr, C. Balthazard, K. Parker,
Cogn. Psychol. 22, 72 (1990).
29. S. A. Sloman, Psychol. Bull. 119, 3 (1996).
30. R. Stickgold, M. Walker, Trends Cogn. Sci. 8, 191 (2004).
31. R. Stickgold et al., Science 294, 1052 (2001).
32. We thank E. Neimeijer, L. Schreers, and R. Wassenberg
for help with conducting study 4. This research was
supported by a grant from Nederlandse Organisatie voor
Wetenschappelijk Onderzoek (016.025.030).
Supporting Online Material
www.sciencemag.org/cgi/content/full/311/5763/1005/DC1
Materials and Methods
20 October 2005; accepted 9 January 2006
10.1126/science.1121629
1007
 ARTICLES

Probabilistic word pre-activation during language

2005 Nature Publishing Group http://www.nature.com/natureneuroscience

comprehension inferred from electrical brain activity

Katherine A DeLong1, Thomas P Urbach1 & Marta Kutas1,2

Despite the numerous examples of anticipatory cognitive processes at micro and macro levels in many animal species, the idea
that anticipation of specific words plays an integral role in real-time language processing has been contentious. Here we exploited
a phonological regularity of English indefinite articles (an precedes nouns beginning with vowel sounds, whereas a precedes
nouns beginning with consonant sounds) in combination with event-related brain potential recordings from the human scalp to
show that readers brains can pre-activate individual words in a graded fashion to a degree that can be estimated from the
probability that each word is given as a continuation for a sentence fragment offline. These findings are evidence that readers
use the words in a sentence (as cues to their world knowledge) to estimate relative likelihoods for upcoming words.

Despite the variety of real-time processing domains across the phylo-
genetic spectrum in which anticipatory processing has been observed (at
both micro and macro levels), the concept of anticipation has played a
relatively minor role in language processing theories. Human languages
offer unlimited possibilities not only for saying new things but also for
saying old things in new waysfar too many ways, some have argued,
to make prediction of words a viable and effective strategy except when
contextual constraint is unusually high1. Accordingly, early language
processing models often included some form of memory buffer wherein
sentential elements were temporarily stored for later integration at
phrasal, clausal or sentence boundaries26. Since the 1970s, however, the
consensus view has been that sentence processing is continuous and
incremental, with provisional commitments made that at least tem-
porarily resolve linguistic ambiguities as each word is processed upon its
occurrence and rapidly integrated into the sentence representation716.
More recently, a few researchers have argued for the predictive power
of context in generating expectancies during sentence processing, but it
has proven difficult to distinguish prediction from integration. What
some researchers take as evidence for neural pre-activation (prediction,
at a psychological level), others take as a sign of the ease or difficulty in
integrating words into message-level representations upon, but not
before, their occurrence. A case in point is the N400 component
observed in event-related brain potential (ERP) studies, in which
cortical neuronal ensembles generate potentials measurable at the scalp.
The N400 (B200500 ms postitem onset) is the brains neural
response to any potentially meaningful item. Its amplitude is sensitive
to word frequency, repetition and concreteness, among other factors.
The N400 is especially large to nouns that do not meaningfully fit with
their preceding contexts17. However, N400s also characterize responses
to all but the most highly expected nouns, even when they fit
contextually, with amplitudes inversely related (r E !0.9) to their
offline cloze probabilities18. An items cloze probability is the percen-
tage of individuals that continue a sentence fragment with that item in
an offline sentence completion task. Despite the sensitivity of the N400
in response to offline semantic expectancy, it is impossible to determine
whether variation in N400 amplitude in response to eliciting words
during online, real-time sentence processing means that comprehen-
ders are using context to generate expectancies for upcoming items
(prediction view) or whether they are forced by the words themselves to
devote more or fewer resources to integrating words into sentence
representations (integration view).
Tracking prediction in sentence processing requires a measure that
has high temporal resolution (such as ERPs, magnetoencephalogram or
eye movements) and does not alter the comprehension process under
study. Additionally, it calls for a design that precludes interpretation in
terms of integrative difficulty. A few recent ERP and eye-tracking studies
have demonstrated contextually generated expectancies for semantic or
syntactic features of upcoming words1927. None, however, has demon-
strated contextual generation of expectancies for specific word forms in
semantically meaningful, syntactically well-formed sentences.
To this end, we designed an experiment which capitalized on the
phonological regularity in English whereby the singular indefinite
article meaning some one thing is phonologically realized as an
before words beginning with a vowel sound and a before words
beginning with a consonant sound (for example, an airplane and a
kite). To determine whether comprehenders pre-activate specific
articles and nouns before their occurrence, we used sentences of varying
constraint that led to expectations for particular consonant- or vowel-
initial nouns. Across sentences, target nouns ranged from highly
probable to unlikely, based on offline cloze probability norming. For
instance, given The day was breezy so the boy went outside to flyy,
the most likely continuation was a kite (cloze of a 86%, kite

1Department of Cognitive Science and 2Department of Neurosciences, University of California San Diego, La Jolla, California 92093, USA. Correspondence should be

addressed to K.A.D. (kadelong@cogsci.ucsd.edu).

Published online 10 July 2005; doi:10.1038/nn1504

NATURE NEUROSCIENCE VOLUME 8 [ NUMBER 8 [ AUGUST 2005 1117

85
 ARTICLES

89%). However, the sentence could continue with a plausible, though
less likely, alternative such as an airplane. Based on previous studies we
knew that the N400 in response to kite would be smaller than that to
airplane, and more generally that noun N400 amplitude would be
highly inversely correlated with cloze probability. However, as pre-
viously mentioned, the pattern of noun effects may be a consequence
either of the brains surprise at encountering an item different than
what it expects (prediction view), or greater difficulty integrating the
received word into the sentence representation (integration view).
2005 Nature Publishing Group http://www.nature.com/natureneuroscience
RESULTS
We obtained offline probabilities for all article and noun targets.
Participants were asked to provide the best continuations for sentences
truncated before the article or noun. Article cloze ranged from 096%;
noun cloze ranged from 0100%. These broad ranges of expectancy
allowed for analysis of the correlations between the ERP effects and the
offline probabilities of the relevant continuations.
In the ERP experiment, different participants read sentences of
varying contextual constraint that included target articles and nouns

Indeed, on the basis of our experiences, kite may be easier to
integrate into the developing sentence representation than airplane.
Given the difference in their meanings, it is likely that kite and
airplane also differ in how well they fit with event schemas that the
sentence brings to mind through semantic memory processes. How-
ever, whereas kite and airplane differ in meaning, a and an do not,
being distinguished only by their phonological forms. Since their
semantics are identical and they differ only in frequency of usage and
length, there is no reason for the articles to be differentially difficult to
integrate into a given sentence representation unless (i) a is always
easier to integrate, because it is shorter and/or more frequent than an
in everyday usage, or, as we will maintain, (ii) comprehenders have
already (unconsciously) formed a higher, non-trivial expectation for
kite than for airplane.
If anticipation is an integral part of language processing, then it
should be reflected in the brain activity probed by the more and less
expected indefinite articles. If the amount of pre-activation is driven
strictly by word length or frequency, then whatever the ERP effect, it
would be context independent, with all examples of a (versus an)
patterning together. Even if pre-activation is context dependent, the
brain may react to the anticipated article with one response and to
anything else with a different response, in a binary rather than a graded
fashion. Finally, we hypothesize that the language processor exploits
sentence context to probabilistically pre-activate possible continua-
tions, consistent with constraint-based models; if so, the N400 should
respond to a degree that can be estimated from the articles offline cloze
probability. In sum, no observable difference in the brains response to
more- versus less-expected articles would be a sharp blow to predictive
processing accounts, whereas a graded N400 effect correlated with the
articles offline cloze probability would support incremental, predictive
processing that is sensitive to meaning-based constraints.
with large ranges of cloze. Across participants the same sentence
context appeared with both higher- and lower-probability articles
and nouns. Notably, although some continuations were more probable
than others, none was nonsensical, barring participants from develop-
ing a strategy (conscious or unconscious) whereby an improbable
article was taken to signal an impending semantic anomaly.
ERP recording and analyses
To measure brain activity associated with prediction effects, we
recorded electroencephalograms at 26 scalp locations as 32 partici-
pants read sentences word-by-word from a CRT at a rate of 2 words/s
(200-ms duration each). ERPs were analyzed for target articles and
nouns in 160 sentences, with 16 participants viewing each item. The
160 articles and nouns were sorted into ten equal-width bins as a
function of each items cloze probability, from highest (90100%) to
lowest (010%). ERPs for each 10% bin were averaged first within, then
across, participants. The average numerical cloze probability of each
Vertex ERPs by median split on cloze probability,
a e.g., 'The day was breezy so the boy went outside to fly ...'
Articles Nouns
5 V
5 V
'airplane'
'an'
0 500 ms 0 500 ms
'a'
'kite'
< 50% Article cloze < 50% Noun cloze
50% Article cloze 50% Noun cloze

N400 x cloze probability correlations at vertex

b
Figure 1 ERP waveforms and correlations between N400 amplitude and 2.5 Articles 2 Nouns
cloze probability showing that specific words were predicted during language
2.0 1
200500 ms (V)

200500 ms (V)
Mean amplitude

Mean amplitude

comprehension. Articles and nouns were analyzed separately. (a) Illustrative

ERPs at the midline central (vertex) recording site according to median splits 1.5 0
on cloze probabilities. Negative amplitudes are plotted upwards. Both articles
and nouns with o50% cloze elicited greater negativity between 200500 ms 1.0 1

post-stimulus onset (N400) than those with Z50% cloze. Although the 0.5 r = 0.68 2 r = 0.79
ERPs include responses to both article types (a, an) and both noun types
(kite, airplane), a single sample sentence is provided for simplicity. 0 3
0 0.2 0.4 0.6 0.8 1.0 0 0.2 0.4 0.6 0.8 1.0
(b) Mean N400 amplitudes were inversely correlated with items cloze Article cloze probability Noun cloze probability
probabilities. Scatter diagrams show strong inverse relations between cloze
and N400 amplitude at the vertex for both articles (r 0.68, P o 0.05) c Articles
r-values at all recording sites
Nouns
and nouns (r 0.79, P o 0.01). Best-fitting regression lines are also 1.00
plotted. (c) The r-values for all 26 electrode sites plotted on an idealized 0.34 0.25
0.41 0.53 0.36 0.36
head, looking down, nose at the top. Darker shading indicates larger negative 0.34 0.54 0.67 0.32 0.41
correlations, with r-values between sites estimated by spherical spline 0.30 0.62 0.33 0.44 0.55
0.19 0.43 0.60 0.63 0.21 0.63 0.66 0.05
interpolation. The dotted circle demarcates the vertex. Although ERP
0.49 0.68 0.68 0.00 0.71 0.79 0.65
prediction effects were slightly larger over the right hemisphere for both 0.59 0.69 0.83 0.79
articles and nouns and over posterior sites for nouns, the correlations for 0.27 0.63 0.67 0.33 0.08 0.82 0.70
0.51 0.72 0.81 0.76
both articles and nouns showed maximal values over centroparietal sites,
0.53 0.61 0.67 0.77 0.75
with a right hemisphere bias only for the articles (see Supplementary Note 0.27 0.52
0.75 0.75
0.60 1.00 0.84
for detailed distributional analyses).

1118 VOLUME 8 [ NUMBER 8 [ AUGUST 2005 NATURE NEUROSCIENCE

86

bin was then calculated and correlated with mean ERP amplitude in the
N400 time window (200500 ms) for articles and nouns separately.
Correlation coefficients (r-values) and percentage of variance explained
by offline probability (r2) were then calculated separately for all 26
electrode sites.
N400 effects and correlations
As expected, N400 amplitude decreased (became less negative) as noun
cloze probabilities increased (Fig. 1a). We replicated the well-known
2005 Nature Publishing Group http://www.nature.com/natureneuroscience
correlation between N400 amplitude and offline cloze of the target
nouns (Fig. 1b), with correlation coefficients ranging from r 0.36
(not significant) to r 0.84 (P o 0.01) at various scalp sites (Fig. 1c).
Noun cloze probability thus accounted for up to 71% of variance in
brain activity between 200500 ms after a nouns appearance. More-
over, correlations peaked over posterior sites, where N400 amplitudes
are typically largest, whereas anterior sites (where visual N400s are
usually less prevalent) showed little if any evidence of similarly
correlated brain activity (Fig. 1c). These results were an important
precondition for analysis of the articles because they demonstrated that
the different degrees of constraint in these materials were reflected
in offline expectancies and N400 amplitude modulations in the usual
way. However, as previously noted, the noun correlation pattern does
not settle the question of prediction because high correlations could
reflect either the degree of pre-activation or the variance in the
integrability of the noun with the mental representation of the sentence
up to that moment.
To address the question of prediction directly we turned to the
correlation pattern of the target articles. Although article ERP wave-
forms were significantly smaller than those elicited by nouns, the
amplitude of the negativity in the N400 time window did indeed
vary as a function of article expectancy (Fig. 1a). Just as for the nouns,
the higher the articles cloze probability, the smaller the ERP negativity
between 200500 ms post-onset (Fig. 1b), with correlation coefficients
ranging from r "0.19 (not significant) to r "0.72 (P o 0.05) at
various recording sites (Fig. 1c). Moreover, maximum correlations
clustered over centroparietal sites, similar to nouns, albeit more
lateralized to the right (Fig. 1c and Supplementary Note online). So
at least over certain scalp areas, up to 52% of variance in article N400
amplitude was accounted for by the average probability that individuals
would continue the sentence context with that particular article offline.
DISCUSSION
By constructing sentence contexts that led to varying offline expecta-
tions for nouns beginning with vowel or consonant sounds, we could
assess the extent to which such expectations were formed online by
preceding the noun with the phonologically appropriate indefinite
article or the other semantically identical, sententially congruent but
phonologically inappropriate one. Similar to the nouns, the more
contextually unexpected an indefinite article was, the more negative the
ERP mean amplitude between 200500 ms post-word onset (N400). In
other words, the brains response to the articles differed in a graded
fashion as a function of contextual constraint. Our results thus
demonstrate not only that readers can rapidly, incrementally integrate
incoming words into evolving mental sentence representations, but
that they do so in part by exploiting various constraining forces to form
probabilistic predictions of which specific words will come next. Here,
we clearly showed this for the target articles and nouns, although we
have no reason to assume the same would not hold for every word in a
sentence throughout the range of normal reading rates. Notably,
maximum correlations for both nouns and articles were not randomly
distributed across the scalp but rather clustered over centroparietal
NATURE NEUROSCIENCE VOLUME 8 [ NUMBER 8 [ AUGUST 2005
87
ARTICLES
scalp sites (Fig. 1c) where previous reading studies have shown the
largest N400 effects. This topographical pattern indicated that the
values were not simply the spurious outcome of multiple testing at the
26 electrode sites.
In conducting the correlational analyses, we established the func-
tional relationship of the negativity between 200500 ms post-article
onset to the canonical N400 typically elicited by nouns and verbs.
Notably, we also demonstrated that this negativity indexed expectancy
for the eliciting article (and upcoming noun). Given that all articles
were grammatically and semantically congruent within their contexts
and that a and an have identical semantics, there was no reason for
either article type to have been any more or less difficult to integrate
into the sentence representation. Systematic variation in amplitude
of the ERP negativity in relation to offline article cloze probability
thus constitutes strong evidence that participants were indeed antici-
pating the phonological form of a particular noun and therefore had
formed expectations for one article type relative to the other and
seemed to experience some processing difficulty when the less-expected
article appeared.
Articles are relatively short; highly frequent; highly predictable as a
word class; not as semantically rich as nouns, verbs, adjectives or
adverbs and are often skipped over during natural reading28. In
addition to providing unequivocal evidence for lexically specific pre-
diction, the article correlations are compelling evidence that articles,
too, are predicted and integrated with context in qualitatively similar
ways as nouns. For reasons not yet known, the correlations with offline
probability were on average lower for articles than for nouns (although
at some electrode sites, the two were statistically similar). Nonetheless,
the article correlations clearly demonstrated that prediction is not
limited to highly constraining contexts. We believe that this sort of
anticipation is an integral (perhaps inevitable) part of real-time
language processing and is likely to have a functional role, although
this has yet to be demonstrated.
Our findings thus suggest that individuals can use linguistic input to
pre-activate representations of upcoming words in advance of their
appearance. Exactly what informs these predictions, as well as the
neural mechanism for predictive language processing, are matters for
empirical and computational investigations. An open question, for
instance, is how the human sentence comprehension system handles
variation in natural input rates (for example, 2 to 3.5 words/s), and in
particular, whether the same (or different) mechanisms are engaged. In
line with most studies of comprehension that draw general conclusions
without systematically varying input rate, we assume that basic
language processing mechanisms do not vary fundamentally across
the range of normal input rates. This parsimonious assumption is
bolstered by results of N400 studies in which variation in presentation
rate showed no evidence for the engagement of qualitatively different
neural mechanisms29,30. Although the current study demonstrated
graded prediction only at the slower end of natural input rates, we
suggest that this conclusion may generalize to faster rates, given the
aforementioned arguments and ERP evidence for binary prediction
(expected versus unexpected) in natural speech26,27. Subsequent
experimentation will undoubtedly shed more light on these issues.
We propose that single words and combinations of words in a
sentence tap into and differentially activate information via semantic
memory, going beyond the immediate physical input. Semantic mem-
ory is presumed to include information about individual words as well
as world knowledge built up from experience of people, places, things
and events. We maintain that probabilistic pre-activation of particular
word forms follows the access of this experiential knowledge as a result
of linguistic input. Our observation of an ERP expectancy effect at the
1119
 ARTICLES
article leads us to conclude that predictions can be for specific
phonological formswords beginning with either vowels or conso-
nants. In this sense, we propose that prediction can be highly specific, at
least under some circumstances.
Our results are in line with a growing list of empirical studies
demonstrating that the brains language parser projects probabilistic
expectations about various aspects of linguistic processing during
online sentence comprehension tasks. Several studies suggest that the
parser uses constraints accruing as a sentence is analyzed word-by-word
2005 Nature Publishing Group http://www.nature.com/natureneuroscience
to (i) compute likely relationships among referents in linguistic and
visual contexts (for example, upon hearing the word eat, a person is
likely to scan the environment for something edible)19, (ii) pre-activate
semantic features of categories (for example, expecting a particular
kind of tree pre-activates features of trees, even when not all trees would
be plausible in the sentence context)24 or (iii) anticipate various
syntactic aspects of to-be-presented material (for example, expecting
the grammatical gender of upcoming items in gender-marked lan-
guages such as Spanish or Dutch) during word-by-word reading and
natural speech2527.
In particular, our study expands on findings from the aforemen-
tioned grammatical gender studies25,27 in which nouns were preceded
by words whose syntactic gender marking was inconsistent with that of
the expected noun. Although both the gender studies and our study
used the same general logic, the differences in experimental manipula-
tions, design and analyses lead to substantive differences in the
justifiable conclusions. Whereas the Spanish and Dutch studies utilized
pre-nominal gender marking on determiners and adjectives, respec-
tively, our study relied upon a purely phonological (sound representa-
tion) relation between probed articles and upcoming nouns. Thus, our
observation of article ERP variation provides a strong test of whether
the language system predicts word forms with specific phonological
content (lexemes), instead of simply representations specifying words
semantic and syntactic properties (lemmas). In addition, having tested
prediction with semantically identical a/an articles (function words)
instead of words richer in meaning (content words) such as adjectives,
we effectively counter the argument that the observed difference
between more- versus less-predictable articles reflects difficulty inter-
preting them. And perhaps most notably, only our study compares
brain activity elicited by a range of more- or less-predictable articles,
not simply most- versus least-expected. The article correlation findings
thus show for the first time that the language system does not simply
pre-activate a single word when its representation exceeds some thresh-
old given a highly constraining context. Instead, a gradient of pre-
activation shows that the system makes graded predictions.
Our electrophysiological results extend previous prediction findings
in several other notable ways. First, they demonstrate that a candidate
entity (or its depiction) need not be physically present in order for the
brain to narrow the possibilities for likely continuations; rather, pre-
dictions can emerge on the basis of associations that form as sentential
context accrues. Second, our results illustrate that at least one subclass
of function words (which generally provide more grammatical struc-
ture than lexical meaning), indefinite articles, can be important in
building context and facilitating linguistic processing. This finding is
particularly relevant given the paucity of evidence in the comprehen-
sion literature about semantic context effects on function words31,32.
And finally, our findings unambiguously show that anticipatory pro-
cessing can happen not only for conceptual or semantic features but
also for specific phonological word forms. In sum, although natural
language comprehension must occur over a range of input rates with a
nearly infinite number of possible word combinations, these factors do
not seem to prevent the brain from anticipating the most probable
1120
88
continuations for sentences in advance of the actual input. In this
regard, language comprehension appears to involve a special case of the
anticipatory behavior observed in other biological systems.
METHODS
Experimental design and materials. Stimuli consisted of 80 sentence contexts
with two possible target types: relatively expected and unexpected indefinite
article/noun pairs. Each article/noun pair served as a more- and less-expected
target in different contexts. Targets were sentence medial and congruent (that
is, no agreement violations such as a airplane). The 160 stimuli were divided
into two lists of 80 sentences, each participant viewing one list. Sentence
contexts and article/noun targets were used only once per list. Each list
contained equal numbers of relatively expected and unexpected targets as well
as a and an targets. One-quarter of sentences were followed by yes/no
comprehension questions, 94% of which were answered correctly by partici-
pants, on average (range, 88100%).
Norming for articles and nouns was done with different groups of student
volunteers. Informed written consent was obtained from all norming (and
ERP) participants. For articles, cloze ratings were obtained from 30 participants
for 80 sentence contexts truncated before the target article. For nouns,
sentences were truncated after target articles, with two versions of each context
(160 sentences total): one with the more-probable article supplied, the other
with the less-probable article. Individual participants saw only one version,
with each normed by 30 participants.
Participants. Thirty-two volunteers (23 women) participated in the ERP
experiment for course credit or for cash. All were right-handed, native English
speakers with normal or corrected-to-normal vision, between 1837 years
(mean, 21 years). Seven participants reported a left-handed parent or sibling.
Procedure. Volunteers were tested in a single session, with visual sentences
presented centrally, one word at a time (200-ms duration, 500-ms stimulus
onset asynchrony). The instructions were to read sentences for comprehension
and answer yes/no comprehension questions by pressing hand-held buttons.
The electroencephalogram (EEG) was recorded from 26 electrodes arranged
geodesically in an Electro-cap, each referenced online to the left mastoid. Blinks
and eye movements were monitored from electrodes placed on the outer canthi
and under each eye, referenced to the left mastoid. Electrode impedances were
kept below 5 KO. The EEG was amplified with Grass amplifiers with a band-
pass of 0.01 to 100 Hz, continuously digitized at a sampling rate of 250 Hz.
Data analysis. Trials contaminated by eye movements, excessive muscle activity
or amplifier blocking were rejected offline before averaging; on average, 10.7%
of article trials and 11.4% of noun trials were rejected. Data with excessive
blinks were corrected using a spatial filter algorithm. A digital band-pass filter
set from 0.2 to 15 Hz was used on all data to reduce high-frequency noise. Data
were re-referenced offline to the algebraic sum of left and right mastoids and
averaged for each experimental condition, time-locked to the target article and
noun onsets.
Note: Supplementary information is available on the Nature Neuroscience website.
ACKNOWLEDGMENTS
Many thanks to E. De Ochoa and J. Cagle for help with data collection and to
J. Elman for helpful comments on this manuscript. This research was supported
by US National Institute of Child Health and Human Development grant
HD22614 and National Institute on Aging grant AG08313 to M.K. and by
a Center for Research in Language Training Fellowship to K.A.D.
COMPETING INTERESTS STATEMENT
The authors declare that they have no competing financial interests.
Received 21 April; accepted 21 June 2005
Published online at http://www.nature.com/natureneuroscience/
1. Stanovich, K.E. & West, R.F. Mechanisms of sentence context effects in reading:
automatic activation and conscious attention. Mem. Cognit. 7, 7785 (1979).
2. Kintsch, W. & van Dijk, T.A. Towards a model of text comprehension and production.
Psychol. Rev. 85, 363394 (1978).
VOLUME 8 [ NUMBER 8 [ AUGUST 2005 NATURE NEUROSCIENCE

3. Carroll, J. & Bever, T. Sentence comprehension: a case study in the relation of knowledge
and perception. in Handbook of Perception, Vol. 7 (eds. Carterette, E. & Friedman, M.)
299317 (Academic, New York, 1978).
4. Just, M.A. & Carpenter, P.A. A theory of reading: From eye fixations to comprehension.
Psychol. Rev. 87, 329354 (1980).
5. Daneman, M. & Carpenter, P.A. Individual differences in integrating information between
and within sentences. J. Exp. Psychol. Learn. Mem. Cogn. 9, 561584 (1983).
6. Mitchell, D.C. & Green, D.W. The effects of context and content on immediate processing
in reading. Q. J. Exp. Psychol. 30, 609636 (1978).
7. Marslen-Wilson, W.D. Sentence perception as an interactive parallel process. Science
189, 226228 (1975).
8. Altmann, G.T.M. & Steedman, M.J. Interaction with context during human sentence
2005 Nature Publishing Group http://www.nature.com/natureneuroscience
processing. Cognition 30, 191238 (1988).
9. Pickering, M.J. Processing local and unbounded dependencies: a unified account.
J. Psycholinguist. Res. 23, 323352 (1994).
10. Tyler, L.K. & Marslen-Wilson, W.D. The on-line effects of semantic context on syntactic
processing. J. Verb. Learn. Verb. Behav. 16, 683692 (1977).
11. Traxler, M., Bybee, M. & Pickering, M. Influence of connectives on language comprehen-
sion: eye-tracking evidence for incremental interpretation. Q. J. Exp. Psychol. A 50A,
481497 (1997).
12. Marslen-Wilson, W.D. & Tyler, L.K. The temporal structure of spoken language under-
standing. Cognition 8, 171 (1980).
13. Steedman, M. Grammar, interpretation, and processing from the lexicon. in Lexical
Representation and Process (ed. Marslen-Wilson, W.D.) 463504 (MIT Press, Cam-
bridge, Massachusetts, 1989).
14. Boland, J.E., Tanenhaus, M.K., Garnsey, S.M. & Carlson, G.N. Verb argument structure
in parsing and interpretation: Evidence from wh-questions. J. Mem. Lang. 34, 774806
(1995).
15. Tanenhaus, M., Spivey-Knowlton, M., Eberhard, K. & Sedivy, J. Integration of visual
and linguistic information during spoken language comprehension. Science 268,
16321634 (1995).
16. Kutas, M. & Hillyard, S.A. Event-related brain potentials to grammatical errors and
semantic anomalies. Mem. Cognit. 11, 539550 (1983).
17. Kutas, M. & Hillyard, S.A. Reading senseless sentences: brain potentials reflect
semantic incongruity. Science 207, 203204 (1980).
18. Kutas, M. & Hillyard, S.A. Brain potentials during reading reflect word expectancy and
semantic association. Nature 307, 161163 (1984).
NATURE NEUROSCIENCE VOLUME 8 [ NUMBER 8 [ AUGUST 2005
89
ARTICLES
19. Altmann, G.T.M. & Kamide, Y. Incremental interpretation at verbs: restricting the
domain of subsequent reference. Cognition 73, 247264 (1999).
20. Tanenhaus, M.K., Magnuson, J.S., Dahan, D. & Chambers, C. Eye movements and
lexical access in spoken-language comprehension: evaluating a linking hypothesis
between fixations and linguistic processing. J. Psycholinguist. Res. 29, 557580
(2000).
21. Kamide, Y., Scheepers, C. & Altmann, G.T.M. Integration of syntactic and semantic
information in predictive processing: cross-linguistic evidence from German and
English. J. Psycholinguist. Res. 32, 3755 (2003).
22. Chambers, C.G. & Smyth, R. Structural parallelism and discourse coherence: a test of
centering theory. J. Mem. Lang. 39, 593608 (1998).
23. Sedivy, J.C., Tanenhaus, M.K., Chambers, C.G. & Carlson, G.N. Achieving incremental
semantic interpretation through contextual representation. Cognition 71, 109148
(1999).
24. Federmeier, K.D. & Kutas, M. A rose by any other name: long-term memory structure and
sentence processing. J. Mem. Lang. 41, 469495 (1999).
25. Wicha, N.Y.Y., Moreno, E.M. & Kutas, M. Anticipating words and their gender: an
event-related brain potential study of semantic integration, gender expectancy, and
gender agreement in Spanish sentence reading. J. Cogn. Neurosci. 16, 12721288
(2004).
26. Wicha, N.Y.Y., Bates, E.A., Moreno, E.M. & Kutas, M. Potato not Pope: human brain
potentials to gender expectation and agreement in Spanish spoken sentences. Neurosci.
Lett. 346, 165168 (2003).
27. Van Berkum, J.J.A., Brown, C.M., Zwitserlood, P., Kooijman, V. & Hagoort, P. Anticipat-
ing upcoming words in discourse: evidence from ERPs and reading times. J. Exp.
Psychol. Learn. Mem. Cogn. 31, 443467 (2005).
28. ORegan, K. Saccade size control in reading: evidence for the linguistic control
hypothesis. Percept. Psychophys. 25, 501509 (1979).
29. Kutas, M. Event-related brain potentials (ERPs) elicited during rapid serial visual
presentation of congruous and incongruous sentences. Electroencephalogr. Clin. Neu-
rophysiol. 40 (suppl.), 406411 (1987).
30. Gunter, T.C., Jackson, J.L. & Mulder, G. An electrophysiological study of semantic
processing in young and middle-aged academics. Psychophysiology 29, 3854 (1992).
31. King, J.W. & Kutas, M. A brain potential whose latency indexes the length and frequency
of words. CRL Newsletter 10, 19 (1995).
32. Kluender, R. & Kutas, M. Subjacency as a processing phenomenon. Lang. Cognitive
Proc. 8, 573633 (1993).
1121
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
Journal of Personality and Social Psychology Copyright 2004 by the American Psychological Association
2004, Vol. 87, No. 5, 586 598 0022-3514/04/$12.00 DOI: 10.1037/0022-3514.87.5.586

Think Different: The Merits of Unconscious Thought in Preference

Development and Decision Making
Ap Dijksterhuis
University of Amsterdam

The role of unconscious and conscious thought in decision making was investigated in 5 experiments.
Because of the low processing capacity of consciousness, conscious thought was hypothesized to be
maladaptive when making complex decisions. Conversely, unconscious thought was expected to be
highly effective. In Experiments 13, participants were presented with a complex decision problem in
which they had to choose between various alternatives, each with multiple attributes. Some participants
had to make a decision immediately after being presented with the options. In the conscious thought
condition, participants could think about the decision for a few minutes. In the unconscious thought
condition, participants were distracted for a few minutes and then indicated their decision. Throughout
the experiments, unconscious thinkers made the best decisions. Additional evidence obtained in Exper-
iments 4 and 5 suggests that unconscious thought leads to clearer, more polarized, and more integrated
representations in memory.

When making a decision of minor importance, I have always found it
advantageous to consider all the pros and cons. In vital matters
however . . . the decision should come from the unconscious, from
somewhere within ourselves.
Sigmund Freud
A few years ago, I accepted my current job at the University of
Amsterdam and shortly afterward started my quest for a place to
live. My timing was awful. The housing market was incredibly
tight. Apartments were very expensive and very hard to get.
During those days, brokers adopted a procedure that was rather
demanding for the blood pressure of buyers. A house on sale was
widely advertised, and all potential buyers were invited to visit at
the same time. The first person who made a bid (provided it was
reasonable) would get the house. When I finally visited an apart-
ment I thought I liked, I knew I had to make a bid quickly. Very
quickly. Between the moment I first passed through the doorway
of the apartment and the moment I essentially became the owner of
the place, all of five minutes passed. As far as I remember, I had
glanced at the bathroom for about three seconds before I commit-
ted myself to a mortgage of alarming proportions. I did not sleep
much that night.
This research was supported by Nederlandse Organisatie voor Weten-
schappelijk OnderzoekVernieuwingsimpuls Grant 016.025.030. I thank
Teun Meurs for his great help throughout this project and Odin de Vries
and Bregje Zwaan for the collection of the data of Experiment 3. I
benefited greatly from comments on an earlier version by Henk Aarts,
Jonathan Schooler, Pamela Smith, Tim Wilson, Helma van den Berg, Karin
Bongers, Yael de Liver, Kirsten Ruys, and Daniel Wigboldus. Thanks.
The title of this article was inspired by the well-known Think Differ-
ent Apple ad campaign.
Correspondence concerning this article should be addressed to Ap
Dijksterhuis, Social Psychology Program, University of Amsterdam,
Roetersstaat 15, 1018 WB, Amsterdam, the Netherlands. E-mail:
a.j.dijksterhuis@uva.nl
586
The problem is that it feels wrong to make such an important
decision so quickly. It was not necessarily a bad decisionI can
now say it was a good one but it felt like it was poorly made.
After all, psychologists, teachers, and parents have all taught us
that we need to think hard before we make an important decision.
Their advice is both well meant and intuitively compelling. How-
ever, whether thorough conscious thinking always leads to good
decisions is not clear. Freud questioned it, as did his contemporary
Whitehead (1911), who said, It is a profoundly erroneous truism,
repeated by all copybooks and by eminent people making
speeches, that we should cultivate the habit of thinking about what
were doing. The precise opposite is the case (p. 143; also quoted
in Bargh, 1997, p. 10, and in Claxton, 1997, p. 15).
It is interesting that these same parents also advised us to take
our time when faced with an important decision by telling us to
sleep on it. This suggests another approach to making a decision.
Here, people intuitively sense that letting the unconscious mull
over the problem may lead to better decisions. However, whether
what we may call unconscious thought contributes to good
decisions is also not clear. Some recent treatments of the uncon-
scious have been optimistic about its powers (Bargh & Chartrand,
1999; Claxton, 1997; Wegner & Smart, 1997; Wilson, 2002), but
it has become so fashionable to view the unconscious as a rela-
tively stupid system that many people will surely not expect
unconscious thought to be beneficial at all.
The phrase unconscious thought may raise eyebrows. In the
present context, conscious thought and unconscious thought are
defined in the following way. Conscious thought refers to the
cognitive and/or affective task-relevant processes one is con-
sciously aware of while attending to a task. For instance, one may
compare two holiday destinations and consciously think, The
Spanish coast is cheap but I do not want to go there because it is
way too crowded. Unconscious thought, on the other hand, refers
to cognitive and/or affective task-relevant processes that take place
outside conscious awareness. One may compare two holiday des-
tinations and not know which one to choose. Subsequently, one

136
 MERITS OF UNCONSCIOUS THOUGHT
does not consciously attend to the problem for a few days, and
suddenly the thought, Its going to be Tuscany! pops into mind. prevent it from taking all this information into account simulta-
This thought itself is conscious, but the transition from indecision
to a preference a few days later is the result of unconscious
thought.
In this article, conscious and unconscious thought are pitted
against each other to answer the question, Who is the better
decision maker of the two? This question is investigated with
complex decision problems. When choosing between jobs, be-
tween houses, between different holiday destinations, between
roommates, people are faced with various alternatives, each char-
acterized by many pros and cons. Making sound decisions requires
integration of large amounts of information into impressions and a
comparison between these impressions to arrive at a preference. To
do this, two things are needed. One needs enough processing
capacity to deal with large amounts of information, and one needs
skills sophisticated enough to integrate information in a meaning-
ful and accurate way.
Processing Capacity
A little introspection reveals that the processing capacity of
consciousness is limited. People are not able to concentrate con-
sciously on two different things simultaneously. About a half a
century ago, researchers started to try to quantify the processing
capacity of consciousness and the unconscious. Miller (1956)
demonstrated that the maximum amount of information that can be
kept under conscious scrutiny at any given time is about seven
units. That is not a lot. Others who investigated the processing
capacity of consciousness also drew rather sobering conclusions.
Consciousness can process about 40 60 bits per second (for brief
reviews of this research, see Nrretranders, 1998; Wilson, 2002).
Take reading, for example. Each letter of the alphabet requires
about 5 bits. This means that a random letter string of eight letters
(noahlief) contains 40 bits. However, in real language, letters are
not random (e.g., zzzgh never appears in English), so a real
eight-letter word has far less than 40 bits. Taking this into account,
when people read, they process about 45 bits per second, equiva-
lent to a fairly short sentence. Although the number of bits con-
sciousness can process is dependent on both the task and the
experience (a skilled reader can process more bits while reading
than a child who is still mastering it), the processing capacity of
consciousness is low.
In contrast, the processing capacity of the entire human system,
or, in other words, of conscious and unconscious processes com-
bined, is enormous. The capacity of the entire system is about
11,200,000 bits. The visual system alone can deal with 10 million
bits per second. It follows that the unconscious does not have a
capacity problem. If the unconscious is a modern computer, con-
sciousness is nothing more than an old abacus.
The low capacity of consciousness suggests that it may not be
up to the task of making complex decisions: The sheer amount of
information involved may be too much for it to tackle. Take the
example of choosing a graduate school. Different programs have
different professors with different research interests, teaching eval-
uations, and reputations as advisors. Some programs have a higher
status than others, but this status is not always perfectly correlated
with the status of the university as a whole. Different schools are
also located in different cities with their own pros and cons on
137
587
several dimensions. The low capacity of consciousness will likely
neously, the consequence being that consciousness will only deal
with a subset of information. This may come at the expense of the
final decision.
Indeed, it has been shown that when people can devote only a
limited amount of information processing capacity to making a
decision (e.g., when they are under time pressure), normative,
consciously driven processes can lead to worse decisions than
more heuristic strategies (Payne, Bettman, & Johnson, 1988). This
is because more elaborate, normative strategies only work well
when all information is taken into account.
The work by Wilson, Schooler, and colleagues provides direct
evidence for consciousness as a poor decision maker because of its
limited capacity. Wilson and Schooler (1991; see also Schooler &
Melcher, 1995; Schooler, Ohlsson, & Brooks, 1993; Wilson et al.,
1993) had participants evaluate objects, such as different college
courses. In one condition, participants were simply asked to eval-
uate the different objects. They most likely engaged in little
conscious thought. In another condition, participants were pressed
to carefully analyze the reasons for their evaluations before mak-
ing them and to write down these reasons. They engaged in more
thorough conscious thinking. This did not help them. In fact, they
did a poorer job than participants who thought less. As expected,
additional evidence indicated that conscious thought led people to
focus on a limited number of attributes at the expense of taking
into account other relevant attributes.
Other evidence comes from Pelham and Neter (1995). They
asked participants to solve various problems. Some problems were
transparent and easy to solve, whereas others were difficult to
solve, and participants had to avoid pitfalls (they ran the risk of
heuristics leading them astray). Some people were simply asked to
solve the problems, whereas others were strongly motivated to
solve the problems accurately. This increased motivation helped
participants to be more accurate on the easy problems, but it
hindered solving the complex problems. If one is willing to assume
that the motivated participants engaged in more conscious think-
ing, the results support the notion of consciousness as a low-
capacity system. Ironically, more conscious thought reduced the
chance that people took crucial information into account.
The Skills to Think
Another important requirement for making a good decision is
the integration of information in a meaningful way. Simply put, it
is important to be able to think. Are both consciousness and the
unconscious able to associate, integrate, elaborate, weigh?
Consciousness may suffer from a power cut when too much
pressure is put on its limited capacity, but as long as its capacity is
enough to deal with a particular problem, it is likely to be a good
thinker. For instance, consciousness can effectively moderate im-
pressions that were initially made unconsciously (e.g., Gilbert,
1989). When we see someone shout, we come to the impression
that this person is aggressive, but conscious intervention can alter
this impression if warranted (wait, he was provoked). With
decisions, such conscious interventions can be highly effective.
Faced with the opportunity to buy an apartment with many wildly
positive attributes and a single critical negative one (it is much too
588
expensive), consciousness will be good at quickly deciding
against it.
The question is whether the unconscious is a good thinker. If it
can devote a relatively small part of its enormous processing
capacity to thought, it can potentially be very useful. However, a
vast processing capacity does not necessarily imply useful work. If
you shut down your computer, it cannot do anything. Or, to stick
to the computer metaphor, some computers are incredibly power-
ful but do only one simple thing, like sorting mail by zip code. So
what can the unconscious contribute to making decisions?
An area where people have devoted time and effort to investi-
gating the fruits of the unconscious is the domain of creativity.
Researchers have long recognized the importance of incubation,
the process whereby a problem is consciously ignored for a while,
after which the unconscious offers a solution. The amount of
anecdotal evidence for incubation is enormous. Nobel laureates
and famous artists often refer to this process as the true key to
insight (see, e.g., Gardner, 1993; Ghiselin, 1952). George Spencer
Brown has famously said about Sir Isaac Newton that
to arrive at the simplest truth, as Newton knew and practiced, requires
years of contemplation. Not activity. Not reasoning. Not calculating.
Not busy behavior of any kind. Not reading. Not talking. Not making
an effort. Not thinking. Simply bearing in mind what it is that one
needs to know. (Quoted in Claxton, 1997, p. 58)
Many of us recognize processes of incubation in ourselves. We
think about a paper we want to write and mull over the order in
which we are going to develop our argument in the introduction.
We put things to rest for a while and then suddenly, BING, we
feel we know it.
Strong as the anecdotal evidence may be, for a long time proof
of incubation was notoriously hard to establish in a research lab
(e.g., Olton, 1979). In addition, the empirical evidence for incu-
bation available these days (e.g., Schooler & Melcher, 1995;
Smith, 1995) is usually not explained by unconscious thought.
Instead, incubation is seen as fruitful because one is distracted
from the problem at hand. Not thinking about a problem for a
while may lead people to forget wrong heuristics or inappropriate
strategies in general. Distraction, then, allows people to give the
problem a fresh look. The distinction between distraction, whereby
the role of the unconscious is passive, and true active, unconscious
thought, is discussed in the introduction to Experiments 4 and 5
below.
Whereas the conclusions that can be drawn from the literature
on incubation are rather sobering regarding unconscious thought,
recent evidence by Betsch, Plessner, Schwieren, and Gutig (2001)
shows that the unconscious can integrate large amounts of infor-
mation. Betsch et al.s participants looked at ads shown on a
computer screen. They were asked to carefully look at the ads
because their memory for the ads would be probed. At the same
time, the numerical increases and decreases of five hypothetical
shares were shown. Participants were presented with 75 units of
information, all only briefly presented on the computer screen.
Afterward, they were asked specific questions about each of five
shares, such as what the average money returns were. They were
clueless. However, when they were only asked to give their atti-
tudes, they somehow knew what the best and worst shares were.
Participants had developed a gut feeling toward the shares, indi-
cating that they had integrated the information. If participants had
DIJKSTERHUIS
failed to take into account even a small portion of the 75 units of
information, this would have been impossible.
Other evidence for unconscious thought processes comes from
research by Bowers, Regehr, Balthazard, and Parker (1990). Their
participants were asked to identify words while from time to time
they were given a hint, such as an associated word. After each hint,
they were pressed to guess. When people solve such problems,
they feel as if they suddenly know the answer. Indeed, the
answer suddenly pops up in consciousness (red . . . bowl . . . fresh
. . . of course, they mean fruit!). However, peoples successive
guesses indicated that the process is not quite as sudden if seen
from the perspective of the unconscious. Successive guesses con-
verged, and the unconscious seemed to be closing in on the right
answer quite a while before the answer was accessible to con-
sciousness. Related findings come from research on tip-of-the-
tongue phenomena. Yaniv and Meyer (1987) offered participants
definitions of rare words they could not recall but felt they knew.
In a lexical decision task, the target tip-of-the-tongue words were
highly accessible. Although the words were inaccessible to con-
sciousness, the unconscious had found and activated them.
To summarize, there is enough evidence to at least assume that
the unconscious continues to think about pressing matters in the
absence of any conscious attention.
Overview of Experiments
Two hypotheses were tested in the current research. First, when
making complex decisions, a brief period of unconscious thought
will lead to a better decision relative to conditions under which
unconscious thought is prevented. Second, when making complex
decisions, conscious thought is inferior relative to unconscious
thought. The first three experiments directly tested these two
hypotheses. The fourth and fifth experiments were designed to
shed light on the nature of unconscious thought.
In the first three experiments, the same paradigm was used.
Participants were presented with information about various alter-
natives (apartments in Experiments 1 and 2, roommates in Exper-
iment 3). Alternatives were described by both positive and nega-
tive attributes, and one was made rather desirable and another one
rather undesirable. Hence, the quality of the decision is judged
from a normative perspective. In the experiments, filler alterna-
tives were included in order to increase the complexity of the
decision problem. These fillers were constructed to be neutral.
Participants indicated their preference either directly, by choos-
ing an alternative (Experiment 2), or indirectly, by indicating their
evaluation of each alternative (Experiments 1 and 3). In Experi-
ments 1 and 3, the dependent variable was the difference in attitude
toward the desirable and the undesirable alternatives. There are
two reasons for this choice. First, when choosing between alter-
natives, recognizing the best alternative is obviously important.
However, recognizing and rejecting a particularly unattractive
alternative is in many cases just as important. The second reason
is practical. A difference score between two opposite attitudes
provides more statistical power than a single attitude score.
In all experiments, there were three conditions. Some partici-
pants were provided with the relevant information and had to
decide (or evaluate) immediately afterward. This condition is the
immediate decision condition. Participants in the conscious
thought condition were given a few minutes to think about the
138
 MERITS OF UNCONSCIOUS THOUGHT
information before they decided. Finally, participants in the un-
conscious thought condition were distracted for a few minutes
before they decided, thereby enabling them to think unconsciously
while at the same time preventing conscious thought. The imme-
diate decision condition can be seen as a baseline where little or no
thought takes place after participants read the information about
the various alternatives. The unconscious thought condition is
devised in such a way as to prevent conscious thought. The first
hypothesis, that unconscious thought helps to make decisions,
should lead to better performance in the unconscious thought
condition relative to the immediate decision condition.
The conscious thought condition is a little more complex be-
cause it is not purely a conscious thought condition. Making
people think consciously does not stop them from thinking uncon-
sciously. The second hypothesis about the relative inferiority of
conscious thought should not necessarily lead to worse perfor-
mance under conscious thought conditions than under immediate
decision conditions, because the former group benefits from un-
conscious thought. Instead, the second hypothesis implies that the
conscious thought condition should show inferior performance
relative to the unconscious thought condition.
Experiment 1
Method
Participants and design. Sixty-three undergraduate students (48
women and 15 men) from the University of Amsterdam were randomly
assigned to one of three conditions: an immediate decision condition, a
conscious thought condition, and an unconscious thought condition. They
either received course credits or money (5; approximately US$6) for their
participation.
Procedure and materials. The experiment was the last in a longer
session with various unrelated experiments. Participants worked in separate
cubicles. The experiment was described as an experiment on decision
making. Participants were told that they would be presented with informa-
tion about four hypothetical apartments in Amsterdam (labeled Apartments
1 4). All apartments were described by positive (e.g., a very nice area)
and negative (e.g., rather noisy) attributes. Participants were asked to
form an impression of the apartments, and they were told that they would
be asked to choose one of the apartments at a later stage.
Information was then presented about the four apartments. Each apart-
ment was described by 12 attributes, for a total of 48 pieces of information.
These 48 attributes were presented in random order. Each attribute was
presented for 4 s in the center of the screen, automatically followed by the
next attribute. It is important to note that Apartment 2 was the most
attractive apartment, with 8 positive (e.g., Apartment 2 is in the city
center) and 4 negative (e.g., Apartment 2 has an unfriendly landlord)
attributes. Apartment 4 was the worst apartment, with 4 positive and 8
negative attributes. Apartments 1 and 3 were of medium attractiveness,
with 6 positive and 6 negative attributes. The last two apartments can be
seen as fillers, used so participants were presented with a large amount of
information.
The stimulus information was assembled as follows. A large number of
attributes were pretested, and all extremely negative or extremely positive
attributes were excluded. When students were asked which attributes they
found important, two stood out: size of the apartment and cost. These
dimensions were used with care in that items were phrased to make them
not too extreme (e.g., Apartment 2 is fairly large rather than Apartment
2 is enormous). No attributes were used more than once, although some
attribute dimensions were used twice (e.g., one was in a nice area, another
in a troublesome area).
139
589
After participants read all the information, they were randomly allocated
to one of three conditions. In the immediate decision condition, they were
immediately asked to give their attitude toward each of the four apartments.
The questions were phrased How would you judge Apartment . . . ?
Participants were asked to indicate their answer on a 10-point scale ranging
from 1 (extremely negative) to 10 (extremely positive). All participants
rated the apartments in numerical order, starting with Apartment 1.
In the conscious thought condition, participants were first asked to very
carefully think about what you think of each of the four apartments.1 They
were given 3 min. During this time, the computer screen was blank except
for a clock indicating how much time they had left. After 3 min, partici-
pants answered the attitude questions.
In the unconscious thought condition, participants performed a distractor
task aimed at preventing conscious thought: the n-back task (e.g., Jonides
et al., 1997). In this task, participants are presented with a series of digits,
and for each digit they have to decide whether it matches the digit that
preceded it by n places. Here participants completed a 2-back task. This
demanding task affects executive functioning quite severely and can there-
fore be expected to successfully eliminate conscious thought. Participants
performed the 2-back task for 3 min (including a 20-s instruction screen).
A number between 1 and 9 appeared on the screen every second, and
participants had to indicate a match by pressing the space bar. After 3 min,
they were asked to complete the attitude questions. In general, participants
did well on the n-back task, except for 3 participants who erred on over
10% of the trials. These participants were not taken into account in further
analyses. After completing the attitude questions, all participants were
debriefed, thanked, and dismissed.
Results
It was first confirmed that overall, the attractive apartment was
judged as more attractive than the unattractive apartment. Indeed,
the overall attitude toward the attractive apartment was higher
(M ! 6.18) than the attitude toward the unattractive apartment
(M ! 5.38), with the attitude toward the two fillers falling in
between (Ms ! 5.74 and 5.69).
The measure of interest is how well participants could differ-
entiate between the attractive apartment and the unattractive apart-
ment. Hence, difference scores were calculated by subtracting the
attitude toward the unattractive apartment from the attitude toward
the attractive apartment. Both participants in the immediate deci-
sion condition (M ! 0.47, SD ! 1.71) and in the conscious thought
condition (M ! 0.44, SD ! 1.48) performed poorly. Their scores
did not significantly differ from zero (ts " 1.15), indicating no
clear preference for the attractive apartment. Participants in the
unconscious thought condition did better (M ! 1.23, SD ! 2.05).
Their score was higher than zero, t(21) ! 2.75, p " .02.
It was expected that unconscious thought would outperform
participants in the immediate decision condition. The second hy-
pothesis was that conscious thought would be maladaptive, which
should lead to underperformance of conscious thinkers relative to
unconscious thinkers. It became clear after an inspection of the cell
means that female and male participants may have responded
differentially to our manipulation, so this factor was included in
the analyses. A 3 (experimental condition) # 2 (sex of participant)
analysis of variance (ANOVA) yielded a significant effect of
condition, F(2, 54) ! 3.40, p " .05, and a nonsignificant two-way
interaction, F(2, 54) ! 2.79, p " .08. The three conditions were
compared via three separate ANOVAs. As expected, a comparison
1
Original instructions were in Dutch.
590 DIJKSTERHUIS
between the immediate decision condition and the conscious
thought condition did not yield any significant results. Participants
in the unconscious thought condition performed better than par-
ticipants in the immediate decision condition, F(1, 37) ! 4.96, p "
.04. This effect was qualified by a two-way interaction of equal
magnitude, F(1, 37) ! 4.96, p " .04, showing that male partici-
pants were especially sensitive to our manipulation. They scored
very poorly in the immediate decision condition (M ! #1.00,
SD ! 2.16) and exceptionally well in the unconscious thought
condition (M ! 2.00, SD ! 1.83). This difference was absent for
female participants (both Ms ! 0.87). Finally, a comparison be-
tween the conscious and unconscious thought conditions showed
the predicted main effect of condition. Participants in the uncon-
scious thought condition outperformed those in the conscious
thought condition, although this effect was not significant, F(1,
37) ! 3.47, p " .08. No other effects were significant (Fs " 1).
To summarize, participants who could only engage in uncon-
scious thought were able to differentiate between the attractive
apartment and the unattractive apartment, whereas participants
who either were allowed to think consciously or were not allowed
to think at all could not, thereby supporting the two hypotheses.
Experiment 2
Experiment 2 served various purposes. One goal was to replicate
the effects of Experiment 1. However, rather than asking partici-
pants to evaluate each apartment separately, they were now asked
to choose one of the apartments. In addition, it is shown above that
conscious thought can be maladaptive because of the low capacity
of consciousness. Therefore, consciousness is not able to deal with
a large amount of information and must resort to focusing on a
limited subset. In concrete terms, when choosing between four
apartments, each described with 12 attributes, consciousness
would not be able to take all the information into account and
would, by necessity, focus on only a few attributes. Unconscious
thought, however, was expected not to suffer from capacity prob-
lems. It should therefore be easier for unconscious thought to form
a more global judgment based on all (or almost all) information.
This possibility was investigated by asking participants, after they
chose an apartment, to indicate whether their choice was based on
a global impression or on only one or two specific attributes.
The paradigm changed a little from Experiment 1, in which the
decision-making task was very taxing, because participants in two
out of three conditions could not discriminate between the best and
worst apartments. One reason for this experienced difficulty is that
the information about the apartments was presented in random
order. It was decided to present the information in Experiment 2
about each apartment individually in a fixed order.
Method
Participants and design. Ninety-four undergraduate students (80
women and 14 men) from the University of Amsterdam were randomly
assigned to one of three conditions: an immediate decision condition, a
conscious thought condition, and an unconscious thought condition. They
either received course credits or money (5) for their participation.
Procedure and materials. Participants were seated in a cubicle. The
experiment was described as an experiment on decision making. Partici-
pants were told that they would be presented with information about four
hypothetical apartments in Amsterdam (called Apartments 1 4). All apart-
140
ments were described by both positive and negative attributes. They were
asked to form an impression of the four apartments, and they were told that
they would be asked to choose one of the apartments at a later stage.
Information was then presented about the four apartments. Each apart-
ment was described by 15 attributes, for a total of 60 pieces of information.
As opposed to Experiment 1, the attributes were not presented in random
order. Participants were first given the 15 attributes describing Apartment
1. All 15 attributes appeared at once and were presented as a list. After 12 s,
the list for Apartment 2 appeared to the right of the list for Apartment 1.
Again after 12 s, a third list describing Apartment 3 was added, and after
a further 12 s, the list for Apartment 4 was added. The four lists remained
on the screen for another 12 s, after which all the information disappeared.
Apartment 3 was the most attractive apartment, with 8 positive, 4 negative,
and 3 neutral attributes. Because participants were asked to choose an
apartment rather than indicating their attitude toward each apartment, no
particularly unattractive apartment was included. The three remaining
apartments (i.e., Apartments 1, 2, and 4) were all characterized by 5
positive, 6 negative, and 4 neutral attributes.
After participants read all the information, they were randomly allocated
to one of three conditions. The conditions were the same as in Experiment 1.
Afterward, participants answered two questions. They were first asked,
If you had to choose one of the apartments, which one would you
choose? They indicated their answer by typing the corresponding number.
The second question pertained to the way they reached their choice. They
were asked, Is your choice based on a more global judgment, or is your
choice based on only one or two specific attributes? Participants answered
by clicking on one of two boxes, labeled global and specific. After
completing the attitude questions, participants were debriefed and
dismissed.
Results
Because of the extremely low number of male participants (only
3 in the unconscious thought condition), sex of participant was not
included in the analyses. The percentages of participants choosing
the attractive apartment were compared. As expected, participants
in the unconscious thought condition most often made the right
choice (59.3%). Participants in the conscious thought condition
and the immediate decision condition did not perform as well
(47.1% and 36.4% made correct choices, respectively). Chi-square
tests demonstrated that the goal to make the decision problem
easier was successful. In all conditions, participants performed
better than chance, all !2s(92, N ! 93) $ 5.19, ps " .03. Further-
more, the difference between the unconscious thought condition
and the immediate decision condition was significant, !2(59, N !
60) ! 3.13, p " .04, one-tailed.
The way participants reached their choice differed between
conditions. In the immediate decision condition, 42.4% of the
participants indicated they made a global judgment. This percent-
age was lower for participants in the conscious thought condition
(26.5%) and higher in the unconscious thought condition (55.6%).
These latter two percentages differed significantly, !2(60, N !
61) ! 6.69, p " .01, one-tailed. The way participants reached their
choice was related to its quality. Across participants, the correla-
tion between the answers on the two questions was .22 ( p " .04).
Participants who made global judgments more often chose the
attractive apartment.
Experiment 3
Experiment 3 was designed to replicate the finding of superior
unconscious thought with different stimulus materials. In Experi-
 MERITS OF UNCONSCIOUS THOUGHT
ments 1 and 2, apartments were chosen as objects in the hope that
students would find them both relevant and, to some extent,
appealing. In Experiment 3, however, potential roommates were
used as objects. This is as relevant, but possibly more appealing,
than apartments because participants would be thinking about
people.
In the first two experiments, quality of judgment was evaluated
from a normative perspective. One apartment was better than
another because it was characterized by more positive attributes
and fewer negative attributes. However, people differ as to which
attributes they find most important. Some people may be willing to
live in a very small apartment as long as it is in the city center,
whereas others need more space and prefer to move to the suburbs.
Likewise, it is easier for most people to list the appealing aspects
of a holiday in Tuscany (great cities and towns, lots of art, a
beautiful countryside, great food, good wine, etc.) than a holiday
on a Spanish costa (one can swim and choose between dozens of
bars each night to get a drink). Still, many people prefer the
Spanish costas because they are not interested in the highlights of
Tuscany. In other words, to reach a sound decision people should
give different attributes different, idiosyncratic weights.
One may argue that this weighing of attributes of unequal
importance is a task at which consciousness excels. Although
intuitively logical, this remains to be seen. There are two separate
assumptions behind this idea: (a) Consciousness is good at weigh-
ing attributes, and (b) the unconscious is not very good at it. There
is reason to disagree with both.
First, is consciousness good at assigning appropriate weights to
attributes? When the decision problem is simple, consciousness is
indeed likely to be good. As argued in the introduction, if one is
faced with an apartment with many wildly positive attributes and
a single critical negative one (it is much too expensive), conscious-
ness will likely be good at quickly deciding against it. However,
when the situation is much more complex, the low capacity of
consciousness should obstruct this weighing process. What if one
finds three attributes very important, four attributes moderately
important, four attributes rather unimportant, and two attributes
not important at all?
A telling example comes from work by Wilson et al. (1993).
Participants were presented with five posters and asked to choose
one to take home. Later, they were called and asked how satisfied
they were with their choicea wonderful measure of whether they
made the right decision from a subjective point of view. Before
choosing, some participants thought about the posters for a little
while, whereas others were asked to carefully analyze the pros and
cons of each poster. It turned out that people who carefully
analyzed were less satisfied than people who merely thought about
them. That is, people who pressed consciousness to carefully
weigh the various attributes made relatively poor decisions. Wil-
son and colleagues (see, e.g., Wilson, Dunn, Kraft, & Lisle, 1989)
explained these and other findings by claiming that too much
conscious reasoning increases the weight that people attach to
reasons that are very accessible and easy to verbalize. These
reasons are not always the ones that should receive more weight.
In essence, conscious reasoning leads to a weighing process, but
the weights are wrong.
Levine, Halberstadt, and Goldstone (1996) obtained evidence
further supporting this claim. In their work, participants were
presented with 60 faces that differed on a number of dimensions.
141
591
They were asked to indicate their liking for each face, either after
receiving the instruction to merely rate each face or after being
instructed to carefully think about the reasons for their liking. As
one might expect, peoples weighing of the different dimensions in
determining their liking of the faces was more variable after
thinking. However, they were also decidedly more inconsistent. In
light of these findings, it is not appropriate to conclude that
conscious thought is necessarily good at weighing the importance
of attributes.
The second widely held assumption is that the unconscious is
not good at weighing different attributes. However, intuitive as this
assumption is, there is no evidence that the unconscious is not able
to deal with subtleties such as weighing the importance of at-
tributes. Conversely, there are reasons to believe it can. The
evidence by Betsch et al. (2001) discussed above strongly suggests
that the unconscious can weigh information. In addition, there is
more general evidence that the unconscious can deal with many
subtle processes quite wellindeed, often better than conscious-
ness can. The unconscious can sense minor differences between
stimuli that consciousness cannot (Dijksterhuis & Aarts, 2003;
Marcel, 1983; Pierce & Jastrow, 1884). The unconscious can
develop preferences for people and objects, whereas consciousness
is not even aware of these people or objects, as shown by research
on mere exposure (e.g., Kunst-Wilson & Zajonc, 1980) and sub-
liminal evaluative conditioning (Dijksterhuis, 2004; Krosnick,
Betz, Jussim, & Lynn, 1992). The unconscious can master wildly
difficult tasks that consciousness cannot master at all, even if
pressed (e.g., Hassin & Bargh, 2003; Lewicki, Hill, & Czyzewska,
1992). The unconscious can decide to behave more intelligently
without any conscious mediation (Bargh, Gollwitzer, Lee-Chai,
Barndollar, & Trotschel, 2001; Dijksterhuis & van Knippenberg,
1998). Finally, recent research has shown that the unconscious can
force a person to behave intelligently when consciousness is trying
to obstruct this (Wegner, Fuller, & Sparrow, 2003).
In Experiment 3, peoples decisions were investigated from both
a normative perspective (as in Experiments 1 and 2) and a sub-
jective perspective. The information about the alternatives was
again presented in random order, as in Experiment 1. However, to
avoid making the task too difficult, only one filler alternative was
presented, rather than two as in Experiment 1.
Method
Participants and design. One hundred forty-five undergraduate stu-
dents (107 women and 38 men) from the University of Amsterdam par-
ticipated in the experiment. They were randomly assigned to one of three
conditions: an immediate decision condition, a conscious thought condi-
tion, and an unconscious thought condition. They either received course
credits or money (10; approximately US$12) for their participation.
Procedure and materials. The experiment was the last experiment in a
longer session with multiple, unrelated experiments. Participants worked in
separate cubicles. The experiment was announced as an experiment on
decision making. Participants were asked to imagine they would have to
find a roommate to share an apartment with (it is very common for
Amsterdam undergraduates to have one or more roommates). They were
told they would be presented with information about three hypothetical
roommates that were all described by various positive and negative at-
tributes. They were asked to form an impression of the three roommates
and to choose between the roommates at a later stage.
Each roommate was described with 12 attributes, for a total of 36 pieces
of information. These 36 attributes were presented in random order. Each
592 DIJKSTERHUIS
attribute was presented for 2 s in the center of the screen, followed, after
a 0.5-s pause, by the next attribute. Roommate A was the most attractive
person, with 8 positive and 4 negative attributes, whereas Roommate C was
the least attractive, with 4 positive and 8 negative attributes. Roommate B
was of medium attractiveness, with 6 positive and 6 negative attributes. All
roommates were described on the same 12 dimensions (e.g., how friendly,
how neat). Roommate A and C were exact opposites. On the 8 dimensions
where Roommate A was described as positive (e.g., has fun friends, is
neat), Roommate C was described as negative (e.g., has annoying
friends, is very messy), and vice versa. Roommate B partly overlapped
with both of them. The choice for the attributes used was based on pilot
testing aimed at assessing which attributes people find important. Only
attributes that students indicated were at least moderately important were
usedat least 4.6 on a 7-point scale ranging from 1 (extremely unimpor-
tant) to 7 (extremely important) but they were phrased in ways that made
them not too extreme.
After participants read all information, they were randomly allocated to
one of three conditions. In the immediate decision condition, they were
immediately asked to give their attitude toward each of the three room-
mates. Participants were asked to indicate their answers on 9-point scales
ranging from 1 (extremely negative) to 9 (extremely positive). All partici-
pants answered the questions in the same order, starting with Roommate A
and ending with Roommate C.
The conditions were the same as in Experiments 1 and 2, except that the
time people were given to think (in the conscious thought condition) or the
time people were distracted (in the unconscious thought condition) was
now 4 min. In the unconscious thought condition, participants were dis-
tracted with an anagram task. They were presented with one anagram at a
time (e.g., ecipbs); the next anagram appeared after participants had solved
the anagram (biceps), or after 45 s had passed. The task was interrupted
after 4 min, and participants were then asked to complete the attitude
questions.
After a filler task that lasted about 4 min, all participants were asked to
indicate how important they found various attributes of people when trying
to find a roommate. They were asked to indicate the importance of each of
the 12 dimensions used in the experiment (e.g., How important is it for
you that a roommate is neat?). Participants indicated their answers on
7-point scales ranging from 1 (very unimportant) to 7 (very important). The
12 questions were presented in random order.
Results
It was first confirmed that the desirable roommate was judged to
be more positive than the undesirable roommate. Indeed, the
overall attitude toward the desirable roommate was higher (M !
6.85) than that toward the undesirable (M ! 4.20), with the attitude
toward the filler roommate falling in between (M ! 5.45).
Difference scores were computed by subtracting the attitude
toward the least attractive roommate from the attitude toward the
most attractive one. As expected, participants in the unconscious
thought condition scored highest (M ! 3.15, SD ! 1.92). Partic-
ipants in the immediate decision condition scored lowest (M !
2.08, SD ! 1.80), with the participants in the conscious thought
condition falling in between (M ! 2.72, SD ! 1.60). These scores
were reliably higher than zero, as confirmed by t tests, all ts(50, 47,
45) " 7, ps # .01, indicating that in all conditions participants
preferred the attractive roommate.
A 3 (experimental condition) $ 2 (sex of participant) ANOVA
revealed a main effect of condition, F(2, 133) ! 4.69, p # .02, and
a nonsignificant two-way interaction, F(2, 133) ! 2.34, p # .13.
Participants in the unconscious thought condition did better than
participants in the immediate decision condition, F(1, 88) ! 8.07,
142
p # .01, and than participants in the conscious thought condition,
F(1, 84) ! 4.60, p # .04. The comparison between the uncon-
scious thought condition and the conscious thought condition also
revealed a significant two-way interaction, F(1, 84) ! 4.03, p #
.05. Male participants performed poorly in the conscious thought
condition (M ! 1.56) and very well in the unconscious thought
condition (M ! 3.44), whereas this difference was almost absent
for female participants (Ms ! 3.00 and 3.06, respectively). The
conscious thought condition and the immediate decision condition
did not differ from each other, F(1, 94) ! 0.55, but the comparison
did reveal a main effect of sex of participant, F(1, 94) ! 6.42, p #
.02. Female participants (M ! 2.69) outperformed male partici-
pants (M ! 1.65).
In all experiments so far, preferences were evaluated from a
normative perspective. In this experiment, ratings people gave
about the importance of the 12 dimensions used to describe the
three roommates were used to look at subjective preferences. For
each participant, the scores on the 8 dimensions on which the
attractive roommate was described as positive and the unattractive
roommate was described as negative were added. The scores on
the 4 dimensions on which the attractive roommate was described
as negative and the unattractive roommate was described as pos-
itive were then subtracted. The higher the resulting score, the more
favorable a participant should be about the attractive roommate
and the more unfavorable a participant should be about the unat-
tractive roommate. The scores were correlated with the difference
score of the attitudes toward the attractive roommate and the
unattractive roommate. The higher the correlation, the better the
participants preference is from a subjective perspective. That is,
the higher the correlation, the more a participant chose the one he
or she should have chosen according to his or her own weighing of
the attributes. The correlation in the immediate decision condition
was significant, r(51) ! .39, p # .005. The correlation in the
conscious condition was nonsignificant, r(47) ! .21, p # .17,
whereas the correlation in the unconscious thought condition was
the largest of the three, r(41) ! .48, p # .002. On the basis of these
data, there is no reason to assume that conscious thought helps to
make better decisions from a subjective point of view, nor is there
any reason to assume that unconscious thought hampers this pro-
cess. The data point in the opposite direction, although it should be
noted that the correlations do not differ from each other signifi-
cantly. The comparison between the unconscious and conscious
thought conditions failed to reach significance ( p ! .16).2
What Is Unconscious Thought? Experiments 4 and 5
The first three experiments show that participants in the uncon-
scious thought condition generally outperform participants in the
remaining two conditions. However, it is not yet clear what exactly
happens during the unconscious thought period. Experiments 4 and
5 were designed to shed more light on this process.
A number of different processes can take place during distrac-
tion. One should distinguish between processes whereby the role
of the unconscious is passive and a process whereby the uncon-
2
Another reason for being careful is that the ratings of importance of
dimensions were administered after rather than before people gave their
attitudes toward the different roommates.
 MERITS OF UNCONSCIOUS THOUGHT
scious engages in active thought. Most findings in the domain of
incubation are explained by processes whereby the role of the
unconscious is passive.
Schooler and Melcher (1995) reviewed findings showing that
distraction can lead to the change of a mental set. Here, the role
of the unconscious is proposed to be passive. People often ap-
proach a problem with wrong cues, wrong heuristics, and/or wrong
information. Following a period of distraction, wrong approaches
become less accessible or are forgotten altogether. The effects of
distraction on a change of mental set can be either fairly strong
(such as when one tries to solve a chess problem and initially gets
truly fixed in thinking along a wrong path) or relatively subtle
(such as when distraction merely attenuates the biasing influence
of primacy or recency effects). One could lump these processes
together under the umbrella of the fresh look explanation: Putting
a problem aside for a while allows for a fresh, unbiased new start.
These effects notwithstanding, here it is proposed that the un-
conscious also actively thinks. In the first three experiments,
participants in the unconscious thought conditions were better able
to distinguish an attractive alternative from an unattractive alter-
native. A reasonable assumption is that the superior judgments of
the unconscious thinkers were based on their representations of the
various alternatives in memory. These participants were distracted
for a while, and when they were asked to judge the various
alternatives, they must have relied on the representations they
retrieved from memory. Because their judgments were better than
the judgments made by participants in the other conditions, their
representations were also somehow better (or at least different).
Their representations must have changed during distraction.
The hypothesis that representations change over time is akin to
the idea that people engage in thought. The term to think is derived
from the Latin verb cogitare, which literally means to shake
together (Koestler, 1964). This meaning reflects the process of
unconscious thought proposed here quite well. If people are pre-
sented with a lot of information in a relatively short period of time,
the resulting representation in memory is likely to be disorganized.
Individual pieces of information still have to be associated and
integrated. Unconscious thought, it is proposed, does exactly this.
That is, unconscious thought is expected to turn an initial, disor-
ganized set of information into a clearer and more integrated
representation of information in memory. It is quite possible that
the representations can change in various ways. Given the results
of the experiments above, it is likely that unconscious thought
leads to representations that become more polarizedthat is, the
representations of moderately positive alternatives become more
dominated by positive aspects, whereas the representations of
negative alternatives become more negative over time. A second
possibility is that unconscious thought, by a process of continued
associative activity, results in a more organized representation. It
could lead to greater clustering, where pieces of information that
load on the same dimension or pertain to the same aspect become
clustered.
Experiments 4 and 5 were designed to test the hypothesis that
people engage in active unconscious thought and to shed light on
the nature of this process. Experiment 4 was designed to test the
polarization hypothesis, and Experiment 5 was designed to test the
clustering hypothesis. In Experiment 4, the same paradigm was
used as in Experiment 3, but rather than measuring peoples
attitudes toward the different alternatives, recognition of the var-
143
593
ious aspects was assessed. This was done to test the hypothesis of
greater polarization. Is it true that unconscious thought leads to
better judgment because representations become more polarized?
This should mean that positive aspects of the attractive alternative
and negative aspects of the unattractive alternative should come to
dominate the representation. Both accuracy and speed of recogni-
tion were assessed to answer these questions. Accuracy is indica-
tive of availability of information. Polarization may lead to rela-
tively accurate recognition of positive aspects of the attractive
alternative and negative aspects of the unattractive alternative.
Speed is indicative of the accessibility of information. Polarization
may lead to higher accessibility of positive aspects of the attractive
alternative and negative aspects of the unattractive alternative
relative to other information.
Experiment 5 was designed to test the clustering hypothesis. Is
it true that unconscious thought leads to better organization in
memory whereby information is more meaningfully clustered?
The paradigm differs from the one used in Experiment 3. Partic-
ipants were given behavioral information about a person. In this
information, three trait dimensions were hidden. Some informa-
tion suggested that the stimulus person was intelligent, some that
the person was extroverted, and some that the person was politi-
cally left-wing. If the unconscious is a better integrator of infor-
mation, unconscious thought should lead participants to organize
their memory more along these three trait dimensions than partic-
ipants in the remaining two conditions.
Experiment 4
Method
Participants and design. One hundred fourteen undergraduate students
(88 women and 26 men) from the University of Amsterdam participated in
the experiment. They were randomly assigned to one of three conditions:
an immediate decision condition, a conscious thought condition, and an
unconscious thought condition. They either received course credits or
money (10; approximately US$12) for their participation.
Procedure and materials. Experiment 4 was exactly the same as Ex-
periment 3, with one exception. Rather than measuring attitudes, we
assessed recognition. After participants read the information about the
roommates (in the immediate condition), or after either thinking for 4 min
(in the conscious thought condition) or after being distracted for 4 min (in
the unconscious thought condition), they were presented with a recognition
task. One by one, in random order, all 36 aspects of the roommates were
presented to them again. This time the aspects were presented without
roommate labels. Participants task was to quickly decide whether an
aspect belonged to Roommate A, B, or C by pressing a corresponding
button.
Results
No effects or even trends of sex of participants were obtained
(all Fs ! .43), so this factor was not further investigated.
Recognition accuracy. First, the proportion of correct recog-
nition of positive and negative aspects for the three roommates
were calculated separately. These proportions were subjected to a
3 (condition: immediate vs. conscious thought vs. unconscious
thought) between-subjects " 3 (roommate: attractive vs. unattrac-
tive vs. neutral filler) " 2 (valence of aspect: positive vs. negative)
within-subjects ANOVA. A main effect of condition was found,
F(2, 111) # 5.29, p ! .01, indicating that participants in the
594 DIJKSTERHUIS
immediate condition had higher proportions of correct recognition.
This is not surprising, because these participants did not pause
between information acquisition and recognition. More interesting,
this main effect was qualified by a two-way interaction of Condi-
tion ! Roommate, F(2, 111) " 2.91, p # .03. Superior recognition
in the immediate condition accurately recognized information of
all roommates, whereas participants in the other conditions showed
impaired recognition of information of the neutral filler roommate.
It seems that both conscious and unconscious thinkers focused on
the attractive and unattractive roommate and forgot the irrele-
vant filler.
To investigate the possibility of a more polarized memory
representation, the same analyses were done on the proportions of
recognition for the attractive and unattractive roommate only. A 3
(condition: immediate vs. conscious thought vs. unconscious
thought) between-subjects ! 2 (roommate: attractive vs. unattrac-
tive) ! 2 (valence of aspect: positive vs. negative) within-subjects
ANOVA revealed a two-way interaction of Roommate ! Valence
of Aspect, F(1, 111) " 63.37, p # .0001, indicating that people
polarized. Recognition of positive aspects of the attractive room-
mate and negative aspects of the unattractive roommate was su-
perior to memory for negative aspects of the attractive roommate
and positive aspects of the unattractive roommate. It was predicted
that the unconscious thinkers especially would polarize, but the
three-way interaction was not significant, F(2, 111) " 2.23, p # .12.
As can be seen in Figure 1, people generally polarized, and
participants in the unconscious thought condition tended to do this
to an even greater degree. Pairwise contrasts were calculated to
compare recognition accuracy for positive and negative aspects for
each roommate in each condition. Unconscious thinkers had better
recognition for positive aspects of the attractive roommate than for
negative aspects of the attractive roommate ( p # .001), whereas
the reverse was true for the unattractive roommate ( p # .003). In
both other conditions, one of the two contrasts failed to reach
significance ( ps $ .05).
Recognition speed. In the analyses on speed of recognition,
only data for aspects that were recognized accurately were used.
The analyses performed on the speed data were the same as those
Figure 1. Experiment 4: Proportions of accurate recognition of positive (Pos) and negative (Neg) aspects of the
attractive (Attr) and unattractive (Unattr) roommate per condition.
144
on the accuracy data. They were first analyzed with a 3 (condition:
immediate vs. conscious thought vs. unconscious thought) be-
tween-subjects ! 3 (roommate: attractive vs. unattractive vs. neu-
tral filler) ! 2 (valence of aspect: positive vs. negative) within-
subjects ANOVA. The only reliable effect was the two-way
interaction of Condition ! Roommate, F(2, 111) " 3.10, p # .02.
Participants in the immediate condition recognized information
about all three roommates with the same speed, whereas partici-
pants in the other two conditions showed slower recognition of
information about the neutral roommate.
The same analysis was done on the speed of recognition for the
attractive and unattractive roommates only. It revealed a two-way
interaction of Roommate ! Valence of Aspect, F(1, 111) " 13.31,
p # .001, which indeed indicated that people polarized. Recogni-
tion of positive aspects of the attractive roommate and negative
aspects of the unattractive roommate was clearly faster than rec-
ognition of negative aspects of the attractive roommate and posi-
tive aspects of the unattractive roommate. However, here the
predicted three-way interaction was significant, F(2, 111) " 4.19,
p # .02. As can be seen in Figure 2, evidence for polarization was
only obtained for unconscious thinkers.
The evidence for polarization of representations in memory is
fairly strong. Although the evidence based on recognition accuracy
is suggestive, the evidence based on speed is unequivocal. There is
one caveat. Throughout the presentation of this experiment, the
term recognition was used even though it was essentially an
allocation task. Participants did not have to recognize the aspects,
but they were asked to determine (or recognize) the source (Room-
mate A, B, or C). Therefore, allocation biases played a role as well.
Presented with a positive aspect, people may have allocated it to
the roommate they thought to be the most positive (Friendly?
This must have been Roommate A, because I like him most).
However, an allocation bias is based on expectations (I like A, so
he must have this positive attribute), and these expectations are
based on underlying representations. So although pure recognition
and allocation are different processes (one is mediated by expec-
tations, whereas the other is not necessarily), effects in both reflect
effects of the same underlying representation. If one wants to
 MERITS OF UNCONSCIOUS THOUGHT
Figure 2. Experiment 4: Speed of recognition (in seconds) of positive (Pos) and negative (Neg) aspects of the
attractive (Attr) and unattractive (Unattr) roommate per condition .
explain the differences between conditions with differences in
allocation bias, one still has to first assume differences in under-
lying representation.
Experiment 5
Experiment 5 was designed to investigate whether unconscious
thought leads to more integrated representations of information in
memory. To test this, a paradigm was used that was used before by
Hamilton, Katz, and Leirer (1980; see also Chartrand & Bargh,
1996). In their work, participants were presented with information
about a stimulus person and instructions to either form an impres-
sion or to memorize the information. At a later stage, participants
recall was assessed. It is interesting to note that participants with
an impression instruction recalled more information than people
who were specifically asked to memorize the information. In
addition, people with an impression instruction also showed a
more integrated organization of information in memory, as as-
sessed by the order of recall.3
The present experiment examined whether unconscious thought
contributes to this improved organization of information in mem-
ory. As in Hamilton et al. (1980), participants were presented with
behavioral information about a stimulus person. Each behavioral
description was indicative of one of three trait concepts. After-
ward, participants were either asked to recall the information
immediately, after some conscious thought, or after unconscious
thought. Our hypothesis was that unconscious thinkers would have
the highest clustering scores, indicative of more integrative mem-
ory organization. The two different instructions (impression vs.
memory) were included. These are not directly related to the
current hypothesis; they were kept as a tribute to the original work.
Method
Participants and design. Sixty-nine undergraduate students (58
women and 11 men) from the University of Amsterdam participated in the
experiment. They were randomly assigned to the cells of a 2 (instruction:
impression vs. memory) ! 3 (condition: immediate vs. conscious thought
145
595
vs. unconscious thought) design. They either received course credits or
money (7; approximately US$9) for their participation.
Procedure and materials. The experiment was the last experiment in a
longer session with multiple, unrelated experiments. Participants worked in
separate cubicles. The experiment was announced as an experiment on
person perception. Participants were told they would be presented with
information about a person named Jeroen. They were told that they
would have to read various behavioral descriptions about Jeroen. Half the
participants were asked to form an impression of him, whereas the other
half were asked to memorize the information.
Subsequently, 18 short sentences were presented one by one on the
screen in random order. A sentence stayed on the screen for 5 s, with the
next sentence appearing after a pause of 0.5 s. All sentences were pretested
to load on one of three trait categories. Six of the sentences indicated
intelligence, 6 others indicated extroversion, and the remaining 6 were
indicative of Jeroen being politically left-wing.
After presentation of the sentences, participants in the immediate recall
condition were given the recall task, during which they had 4 min to list as
many of the behavioral descriptions as possible. Participants in the con-
scious thought condition were asked to think about Jeroen for 4 min prior
to the recall task, whereas participants in the unconscious thought condi-
tions were distracted for 4 min. The distractor task was the same as in
Experiments 3 and 4.
Results
The number of descriptions recalled correctly was assessed
using a gist criterion (Hamilton et al., 1980). The recall scores
3
The reason for using this paradigm rather than the paradigms used in
Experiments 1 4 is that the dimension that people can use to cluster should
not be too obvious. Underlying trait constructs are excellent, but offering
people information about three individuals named A, B, and C presumably
leads to very strongand intentional clustering on the basis of these
three individuals (lets begin with listing what I remember of person A).
Adding a second, underlying dimension is a possibility, but this would
require much more stimulus material. More stimulus material, however,
leads to poorer relative recall, which in turn makes clustering scores less
reliable.
596 DIJKSTERHUIS
were subjected to a 2 (instruction: impression vs. memory) ! 3
(condition: immediate vs. conscious thought vs. unconscious
thought) between-participants ANOVA. No significant effects
emerged. The absence of an effect of instruction may have been
caused by the fact that compared with the experiments by Hamil-
ton et al. (1980) and Chartrand and Bargh (1996), our task was
easy. People were given few behavioral descriptions, and overall
recall was high (over 50%). However, the immediate decision
condition comes closest to the procedure in the original Hamilton
et al. (1980) studies. Under these conditions, participants in the
impression set condition indeed recalled more information than
participants who received a memory set. This difference was not
significant, F(1, 20) " 2.95, p # .11.
To assess organization in memory, conditional recall probabil-
ities were computed by looking at the order of recall. The number
of same-trait sequences (e.g., an intelligent behavior recalled after
another intelligent behavior) was divided by the total number of
behaviors recalled minus one. The resulting probabilities are listed
in Table 1. A 2 (instruction: impression vs. memory) ! 3 (condi-
tion: immediate vs. conscious thought vs. unconscious thought)
between-participants ANOVA revealed both predicted main ef-
fects. First, in keeping with Hamilton et al. (1980) and Chartrand
and Bargh (1996), there was a main effect of instruction showing
that participants with an impression instruction (M " .44, SD "
.13) had higher clustering scores than participants with a memory
instruction (M " .35, SD " .14), F(1, 63) " 6.56, p # .02. In
addition, the predicted main effect of condition was obtained, F(2,
63) " 4.32, p # .02. Participants in the unconscious thought
condition had higher clustering scores (M " .46, SD " .11) than
participants in the conscious thought condition (M " .37, SD "
18), F(1, 43) " 5.50, p # .03, and than participants in the
immediate recall condition (M " .35, SD " .13), F(1, 42) " 9.58,
p # .005. These latter two conditions did not differ. It is interesting
that recall in random order (without meaningful clustering) would
lead to a clustering score of about .30. Only participants in the
unconscious thought condition had scores considerably higher than
.30. In sum, only unconscious thought led to more integrated
representations in memory.
General Discussion
Unconscious thought improved the quality of decisions. When
people were faced with complex decisions, a few minutes of
distraction during which people could engage in unconscious
thought but not in conscious thoughtled to superior decisions
compared with circumstances under which people could not en-
gage in unconscious thought or to circumstances under which
people engaged in conscious thought. Moreover, a few minutes of
conscious thought generally did not lead to better decisions com-
Table 1
Experiment 5: Clustering Scores as a Function of Condition
Condition
Instruction Immediate Conscious Unconscious
Memory .29 .35 .41
Impression .42 .38 .52
146
pared with conditions where people did not consciously think. The
relative inferiority of conscious thought was expected to be the
consequence of the low processing capacity of consciousness.
Some supporting evidence for this idea comes from Experiment 2.
Here, conscious thinkers reported that their decisions were often
based on a few specific relevant attributes, whereas unconscious
thinkers reported forming a more global judgment based on much
more information.
Experiments 4 and 5 were designed to shed more light on the
processes underlying unconscious thought. It was shown that un-
conscious thought led people to develop more polarized, clearer,
and more integrated representations in memory, supporting the
proposed process underlying superior unconscious decision mak-
ing. It is interesting to note with respect to Experiment 5 that
conscious thought almost completely prevented meaningful clus-
tering. Clustering is an associative process, and it is likely that this
process was disrupted because, as Wilson and others have shown
(e.g., Wilson et al., 1989; Wilson & Schooler, 1991; see also
Experiment 2 of the current article), conscious thought biases
people toward paying a lot of attention to very few attributes at the
expense of others. This may lead to meaningful associations be-
tween the few attributes that received attention, but it may hinder
associative processes incorporating the other attributes.
One avenue for further research concerns the goal directedness
of unconscious thought. In all experiments, participants were told
that they had to decide between various alternatives before they
engaged in unconscious thought. Would the unconscious have
engaged in relevant thinking had it not been instructed to do so?
This is likely not the case. The literature on creativity and incu-
bation (e.g., Bowers et al., 1990; Smith & Blankenship, 1989) has
suggested that unconscious thought can be goal directed. The
quote about Newton in the introduction also points in this direc-
tion. He knew that he had to bear in mind what he needed to know.
One could say that he gave his unconscious a clear goal. In
addition, recent evidence has shown that goal activation and goal
pursuit can be achieved unconsciously (Bargh et al., 2001; Char-
trand & Bargh, 1996). This also implies that unconscious thought
processes can be goal directed. However, this has not yet been
shown.
A Note on Incubation
As said above, empirical evidence for incubation is limited.
People often have failed to find it, and the little evidence available
is hard to replicate (Olton, 1979). Furthermore, most effects that
were obtained could be explained by processes whereby the role of
the unconscious was passive. It is possible that the nature of the
problems people had to solve was responsible for this state of
affairs. Researchers used insight problemsthe kind of problems
where the solution to a problem is very specific and often coun-
terintuitive, resulting in a eureka experience. This choice is
understandable, because incubation is studied in the context of
creativity, and creativity is often (but not always; see Weisberg,
1995) the consequence of such insights. However, the unconscious
may not have been given a fair chance. Solving insight problems
is often like trying to find a needle in a haystack, and the uncon-
scious may need more time than the amount people are given in a
lab experiment. The current research points to the possibility that
 MERITS OF UNCONSCIOUS THOUGHT
incubation can be obtained even within minutes if more mundane,
analytical problems are used rather than insight problems.
Should We Stop Thinking? Some Potential Moderators
One may infer from the present work that it constitutes good
news for people who dislike the hard labor of conscious thinking.
The unconscious can be left to deal with making decisions while
consciousness can be directed elsewhere, such as at fun things.
This conclusion is clearly too bold. First of all, the findings
reported here should be placed in the context of the paradigm used.
Both consciousness and the unconscious had to work within the
confines of this paradigm. Perhaps the circumstances under which
consciousness had to work diverged too strongly from how con-
sciousness often deals with decision problems in real life. For one
thing, consciousness (but also the unconscious) was only given a
few minutes to think. Maybe it performs better with more time. In
addition, conscious thought took place after participants had read
the information. Maybe conscious thought would have led to better
results if the information had been visually available during
thought. Hence, it is important to pit conscious and unconscious
thought against each other in different paradigms before bold
conclusions can be drawn. For now, it is more interesting to
speculate about when conscious versus unconscious thought may
be more fruitful.
As to when one mode of thought may be better than the other,
some educated hypotheses can be formulated. First, there are
various reasons to propose the somewhat counterintuitive idea that
the more complex a problem is, the less likely it becomes that
conscious thought can contribute much. If a problem is complex,
it by definition means that a lot of information has to be taken into
account. Conscious thought is not very good at this. One could say
that conscious thought is very focused but not very encompassing
or inclusive. There are various (related) reasons for this. Con-
sciousness has low capacity, and conscious reasoning biases the
weights people attach to different pieces of information (e.g.,
Wilson et al., 1989). In addition, verbalizing information (an act of
consciousness) makes these biases even worse, as the work on
verbal overshadowing (Schooler et al., 1993) has demonstrated. If
we are willing to assume that mundane decisions are often rela-
tively simple, whereas vital matters are usually complex (which is
not always the case), we may conclude that Freud was right: Use
consciousness for relatively simple and mundane decisions, but
refrain from using it too much for more complex matters.
The observation that on some occasions weighing the impor-
tance of attributes is easier than in other cases may also lead to a
hypothesis. Sometimes weighing is easy or even unnecessary.
When someone from the support staff comes to fix a software
problem, all we care about is whether she is able to do the job
properly in a reasonable amount of time. On the other hand,
sometimes weighing attributes is mind-boggling, such as when
choosing a doctoral program. When weighing is easy, one can
make decisions by first comparing how well various alternatives
score on the one or two attributes that matter. When weighing is
hard, one could start out by forming global impressions of the
different alternatives and then start to compare. It is likely that
conscious thought is more proficient in the former case, whereas
unconscious thought is better able to do the latter. If this is true,
conscious thought may be fruitful when the decision problem is
147
597
well laid out with few attributes that do allow for meaningful
comparisons. If not, unconscious thought should be used to work
on more global or holistic impressions first.
One should not infer from the present work that consciousness
is rather stupid and the unconscious is smart. Rather than making
such categorizations, it seems fruitful to more closely examine the
strong and weak points of both systems (see Claxton, 1997; Kihl-
strom, Barnhardt, & Tataryn, 1992). General claims such as the
unconscious is fast and consciousness is slow or the unconscious
is stupid and consciousness is smart do not make much sense. The
unconscious automatically evaluates stimuli within milliseconds
(Bargh, Chaiken, Govender, & Pratto, 1992; Fazio, Sanbonmatsu,
Powell, & Kardes, 1986), but Newton, who contemplated for
years, may not have agreed that the unconscious is fast. Likewise,
in this article, much evidence has been described implying that the
unconscious can be very smart, but when asked, What is the
square root of 625? your unconscious is not going to solve it. You
could be given a distractor task lasting for months, but this ques-
tion can only be answered (assuming no calculator is used) after
conscious work. Likewise, consciousness is neither always smart
(as the current work shows) nor always very slow (it finds the
square root of 625 in seconds). The bottom line is that both
systems can be fast, slow, smart, or stupid. It all depends on what
they are asked to do.
Nevertheless, the current work demonstrates one thing the un-
conscious is good at: making complex decisions. When faced with
complex decisions such as where to work or where to live, do not
think too much consciously. Instead, after a little initial conscious
information acquisition, avoid thinking about it consciously. Take
your time and let the unconscious deal with it.
References
Bargh, J. A. (1997). The automaticity of everyday life. In R. S. Wyer Jr.
(Ed.), Advances in social cognition (Vol. 10, pp 1 61). Mahwah, NJ:
Erlbaum.
Bargh, J. A., Chaiken, S., Govender, R., & Pratto, F. (1992). The generality
of the automatic evaluation effect. Journal of Personality and Social
Psychology, 62, 893912.
Bargh, J. A., & Chartrand, T. L. (1999). The unbearable automaticity of
being. American Psychologist, 54, 462 476.
Bargh, J. A., Gollwitzer, P. M., Lee-Chai, A., Barndollar, K., & Trotschel,
R. (2001). The automated will: Nonconscious activation and pursuit of
behavioral goals. Journal of Personality and Social Psychology, 81,
1014 1027.
Betsch, T., Plessner, H., Schwieren, C., & Gutig, R. (2001). I like it but I
dont know why: A value-account approach to implicit attitude forma-
tion. Personality and Social Psychology Bulletin, 27, 242253.
Bowers, K. S., Regehr, G., Balthazard, C., & Parker, K. (1990). Intuition
in the context of discovery. Cognitive Psychology, 22, 72110.
Chartrand, T. L., & Bargh, J. A. (1996). Automatic activation of impres-
sion formation and memorization goals: Nonconscious goal priming
reproduces effects of explicit task instructions. Journal of Personality
and Social Psychology, 71, 464 478.
Claxton, G. (1997). Hare brain, tortoise mind: How intelligence increases
when you think less. New York: HarperCollins.
Dijksterhuis, A. (2004). I like myself but I dont know why: Enhancing
implicit self-esteem by subliminal evaluative conditioning. Journal of
Personality and Social Psychology, 86, 345355.
Dijksterhuis, A., & Aarts, H. (2003). On wildebeests and humans: The
preferential detection of negative stimuli. Psychological Science, 14,
14 18.
598 DIJKSTERHUIS

Dijksterhuis, A., & van Knippenberg, A. (1998). The relation between
perception and behavior, or how to win a game of Trivial Pursuit.
Journal of Personality and Social Psychology, 74, 865 877.
Fazio, R. H., Sanbonmatsu, D. M., Powell, M. C., & Kardes, F. R. (1986).
On the automatic activation of attitudes. Journal of Personality and
Social Psychology, 50, 229 238.
Gardner, H. (1993). Creating minds. New York: Basic Books.
Ghiselin, B. (1952). The creative process. New York: New American
Library.
Gilbert, D. T. (1989). Thinking lightly about others: Automatic compo-
nents of the social inference process. In J. S. Uleman & J. A. Bargh
(Eds.), Unintended thought (pp. 189 211). New York: Guilford Press.
Hamilton, D. L., Katz, L. B., & Leirer, V. O. (1980). Organizational
processes in impression formation. In R. Hastie, T. M. Ostrom, E. B.
Ebbesen, R. S. Wyer Jr., D. L. Hamilton, & D. E. Carlston (Eds.),
Person memory: The cognitive basis of social perception (pp. 121153).
Hillsdale, NJ: Erlbaum.
Hassin, R., & Bargh, J. A. (2003). Implicit insight: The case for the on-line
extraction of rules. Manuscript under review.
Jonides, J., Schumacher, E. H., Smith, E. E., Lauber, E. J., Awh, E.,
Minoshima, S., & Koeppe, R. A. (1997). Verbal working memory load
affects regional brain activation as measured by PET. Journal of Cog-
nitive Neuroscience, 9, 462 475.
Kihlstrom, J. F., Barnhardt, T. M., & Tataryn, D. (1992). The psycholog-
ical unconscious: Found, lost, and regained. American Psychologist, 47,
788 791.
Koestler, A. (1964). The act of creation. London: Penguin.
Krosnick, J. A., Betz, A. L., Jussim, L. J., & Lynn, A. R. (1992). Sublim-
inal conditioning of attitudes. Personality and Social Psychology Bulle-
tin, 18, 152162.
Kunst-Wilson, W., & Zajonc, R. (1980, February 1). Affective discrimi-
nation of stimuli that cannot be recognized. Science, 207, 557558.
Levine, G. M., Halberstadt, J. B., & Goldstone, R. L. (1996). Reasoning
and the weighing of attributes in attitude judgments. Journal of Person-
ality and Social Psychology, 70, 230 240.
Lewicki, P., Hill, T., & Czyzewska, M. (1992). Nonconscious acquisition
of information. American Psychologist, 47, 796 801.
Marcel, A. J. (1983). Conscious and unconscious perception: An approach
to the relations between phenomenal experience and perceptual pro-
cesses. Cognitive Psychology, 15, 238 300.
Miller, G. A. (1956). Information theory. Scientific American, 195, 42 46.
Nrretranders, T. (1998). The user illusion: Cutting consciousness down to
size. New York: Viking.
Olton, R. (1979). Experimental studies of incubation: Searching for the
elusive. Journal of Creative Behavior, 13, 9 22.
Payne, J. W., Bettman, J. R., & Johnson, E. J. (1988). Adaptive strategy
selection in decision making. Journal of Experimental Psychology:
Learning, Memory, and Cognition, 14, 534 552.
Pelham, B. W., & Neter, E. (1995). The effect of motivation on judgment
depends on the difficulty of the judgment. Journal of Personality and
Social Psychology, 68, 581594.
Pierce, C. S., & Jastrow, J. (1884). On small differences in sensation.
Memoirs of the National Academy of Science, 3, 75 83.
Schooler, J. W., & Melcher, J. (1995). The ineffability of insight. In S. M.
Smith, T. B. Ward, & R. A. Finke (Eds.), The creative cognition
approach (pp. 97134). Cambridge, MA: MIT Press.
Schooler, J. W., Ohlsson, S., & Brooks, K. (1993). Thoughts beyond
words: When language overshadows insight. Journal of Experimental
Psychology: General, 122, 166 183.
Smith, S. M. (1995). Fixation, incubation, and insight in memory and
creative thinking. In S. M. Smith, T. B. Ward, & R. A. Finke (Eds.), The
creative cognition approach (pp. 135156). Cambridge, MA: MIT
Press.
Smith, S. M., & Blankenship, S. E. (1989). Incubation effects. Bulletin of
the Psychonomic Society, 27, 311314.
Wegner, D. M., Fuller, V. A., & Sparrow, B. (2003). Clever hands:
Uncontrolled intelligence in facilitated communication. Journal of Per-
sonality and Social Psychology, 85, 519.
Wegner, D. M., & Smart, L. (1997). Deep cognitive activation: A new
approach to the unconscious. Journal of Consulting and Clinical Psy-
chology, 65, 984 995.
Weisberg, R. W. (1995). Case studies of creative thinking: Reproduction
versus restructuring in the real world. In S. M. Smith, T. B. Ward, &
R. A. Finke (Eds.), The creative cognition approach (pp. 5372). Cam-
bridge, MA: MIT Press.
Whitehead, A. N. (1911). An introduction to mathematics. New York:
Henry Holt.
Wilson, T. D. (2002). Strangers to ourselves: Discovering the adaptive
unconscious. Cambridge, MA: Harvard University Press.
Wilson, T. D., Dunn, D. S., Kraft, D., & Lisle, D. J. (1989). Introspection,
attitude change, and attitude behavior consistency: The disruptive ef-
fects of explaining why we feel the way we do. In L. Berkowitz (Ed.),
Advances in experimental social psychology (Vol. 19, pp. 123205).
Orlando, FL: Academic Press.
Wilson, T. D., Lisle, D., Schooler, J. W., Hodges, S. D., Klaaren, K. J., &
LaFleur, S. J. (1993). Introspecting about reasons can reduce post-choice
satisfaction. Personality and Social Psychology Bulletin, 19, 331339.
Wilson, T. D., & Schooler, J. W. (1991). Thinking too much: Introspection
can reduce the quality of preferences and decisions. Journal of Person-
ality and Social Psychology, 60, 181192.
Yaniv, I., & Meyer, D. E. (1987). Activation and metacognition of inac-
cessible stored information: Potential bases for incubation effects in
problem solving. Journal of Experimental Psychology: Learning, Mem-
ory and Cognition, 13, 187205.
Received September 2, 2003
Revision received May 27, 2004
Accepted June 2, 2004 !

Wanted: Old APA Journals!

APA is continuing its efforts to digitize older journal issues for the PsycARTICLES database.
Thanks to many generous donors, we have made great strides, but we still need many issues,
particularly those published in the 1950s and earlier.

If you have a collection of older journals and are interested in making a donation, please e-mail
journals@apa.org or visit http://www.apa.org/journals/donations.html for an up-to-date list of the
issues we are seeking.

148
Intelligence in Public Literature
T h in k in g, Fast an d S low
Da n iel Ka h n em a n , (New Yor k : Fa r r a r, St r a u s a n d Gir ou x, 2011), 418 pp.
Reviewed by Frank J. Babetski
Few books a r e m u st r ea ds for in t elligen ce
officer s. Fewer st ill a r e m u st r ea ds t h a t m en -
t ion In t elligen ce Com m u n it y fu n ct ion s or t h e
CIA on ly on ce, a n d t h en on ly in pa ssin g. Da n -
iel Ka h n em a n h a s wr it t en on e of t h ese r a r e
books. Th in k in g, Fast an d S low r epr esen t s a n
elega n t su m m a t ion of a lifet im e of r esea r ch in
wh ich Ka h n em a n , P r in cet on Un iver sit y P r ofes-
sor E m er it u s of P sych ology a n d P u blic Affa ir s,
a n d h is la t e colla bor a t or, Am os Tver sky,
ch a n ged t h e wa y psych ologist s t h in k a bou t
t h in kin g. Ka h n em a n , wh o won t h e 2002 Nobel
P r ize in E con om ics for h is wor k wit h Tver sky
on pr ospect t h eor y, a lso h igh ligh t s t h e best
wor k of ot h er r esea r ch er s t h r ou gh ou t t h e book.
T h in k in g, Fast an d S low in t r odu ces n o r evolu -
t ion a r y n ew m a t er ia l, bu t it is a m a st er piece
beca u se of t h e wa y Ka h n em a n wea ves exist in g
r esea r ch t oget h er.
E xper t in t elligen ce officer s a t CIA, a n
a gen cy wit h t h e h u m a n in t elligen ce m ission
a t it s cor e, h a ve com e t h r ou gh exper ien ce a n d
pr a ct ice t o u n der st a n d a n d exploit t h e h u m a n
cogn it ive pr ocesses of wh ich Ka h n em a n wr it es.
Th ese exper t officer s will h a ve m a n y m om en t s
of r ecogn it ion in r ea din g t h is book, wh ich gives
a n em pir ica l u n der pin n in g for m u ch of t h eir
h a r d-won wisdom .
Ka h n em a n a lso m a y ch a llen ge som e st r on gly
h eld beliefs. T h in k in g, Fast an d S low gives
exper t s a n d n ewer officer s, r ega r dless of t h e
in t elligen ce a gen cy in wh ich t h ey ser ve, a n
en or m ou sly u sefu l cogn it ive fr a m ewor k u pon
wh ich t o h a n g t h eir exper ien ces.
A ll statem en ts of fact, opin ion , or an alysis expressed in th is article are th ose of th e au th or. N oth in g in th is article sh ou ld be
con stru ed as assertin g or im plyin g U S govern m en t en d orsem en t of its factu al statem en ts an d in terpretation s.
Studies in Intelligence Vol. 56, No. 2 (Extracts, June 2012)
149
Th e t it le of t h e book r efer s t o wh a t Ka h n e-
m a n , a da pt in g a device t h a t ot h er r esea r ch er s
or igin a lly pr oposed, ca lls t h e t wo syst em s of
t h e h u m a n m in d. Syst em 1, or fa st t h in kin g,
oper a t es a u t om a t ica lly a n d qu ickly wit h lit t le
or n o effor t a n d n o sen se of volu n t a r y con t r ol.
Most Syst em 1 sk illssu ch a s det ect in g t h e
r ela t ive dist a n ces of object s, or ien t in g t o a su d-
den sou n d, or det ect in g h ost ilit y in a voicea r e
in n a t e a n d a r e fou n d in ot h er a n im a ls. Som e
fa st a n d a u t om a t ic Syst em 1 skills ca n be
a cqu ir ed t h r ou gh pr olon ged pr a ct ice, su ch a s
r ea din g a n d u n der st a n din g n u a n ces of socia l
sit u a t ion s. E xper t s in a field ca n even u se Sys-
t em 1 t o qu ickly, effor t lessly, a n d a ccu r a t ely
r et r ieve st or ed exper ien ce t o m a ke com plex
ju dgm en t s. A ch ess m a st er qu ickly fin din g
st r on g m oves a n d a qu a r t er ba ck ch a n gin g a
play sen t t o h im fr om t h e sidelin e wh en h e r ec-
ogn izes a defen sive wea kn ess a r e exa m ples of
a cqu ir ed Syst em 1 t h in kin g.
Syst em 2, or slow t h in kin g, a lloca t es a t t en -
t ion t o t h e m en t a l a ct ivit ies t h a t dem a n d
effor t , su ch a s com plex com pu t a t ion s a n d con -
sciou s, r ea son ed ch oices a bou t wh a t t o t h in k
a n d wh a t t o do. Syst em 2 r equ ir es m ost of u s t o
pa y a t t en t ion t o do t h in gs su ch a s dr ive on a n
u n fa m ilia r r oa d du r in g a sn owst or m , ca lcu la t e
t h e pr odu ct of 17x24, sch edu le t r a n spor t a t ion
for a t een a ge da u gh t er s a ct ivit ies, or u n der-
st a n d a com plex logica l a r gu m en t .
Ka h n em a n focu ses m u ch of t h e book on t h e
in t er a ct ion s of Syst em 1 a n d Syst em 2 a n d t h e
pr oblem s in h er en t in t h ose in t er a ct ion s. Bot h
syst em s a r e on wh en we a r e a wa ke. Syst em 1
r u n s a u t om a t ica lly a n d effor t lessly bu t
1
Thinking, Fast and Slow
Syst em 2 idles, beca u se u sin g it r equ ir es effor t
a n d is t ir in g. Syst em 1 gen er a t es im pr ession s
a n d feelin gs, wh ich becom e t h e sou r ce of Sys-
t em 2s explicit beliefs a n d deliber a t e ch oices.
Syst em 1, wh en it en cou n t er s som et h in g it ca n -
n ot qu ickly u n der st a n d a n d did n ot expect (in
ot h er wor ds, a su r pr ise), en list s Syst em 2 t o
m a ke sen se of t h e a n om a ly. Th e a ler t ed Sys-
t em 2 t a kes ch a r ge, over r idin g Syst em 1s a u t o-
m a t ic r ea ct ion s. Syst em 2 a lwa ys h a s t h e la st
wor d wh en it ch ooses t o a sser t it .
Th e syst em s oper a t e t o m in im ize effor t a n d
m a xim ize per for m a n ce a n d a r e t h e r esu lt of
h u n dr eds of t h ou sa n ds of yea r s of h u m a n evo-
lu t ion in ou r en vir on m en t . Th ey wor k
ext r em ely well, u su a lly. Syst em 1 per for m s
well a t m a kin g a ccu r a t e m odels a n d pr edic-
t ion s in fa m ilia r en vir on m en t s. Syst em 1 h a s
t wo sign ifica n t wea kn esses: it is pr on e t o m a ke
syst em ic er r or s in specified sit u a t ion st h ese
a r e bia sesa n d it ca n n ot be t u r n ed off. Sys-
t em 2 ca n , wit h effor t , over r u le t h ese bia ses if
it r ecogn izes t h em . Un for t u n a t ely, Syst em 2 is
dem on st r a bly ver y poor a t r ecogn izin g on es
own bia sed t h in kin g. Tr yin g t o en ga ge Syst em
2 a t a ll t im es t o pr even t Syst em 1 er r or s is
im pr a ct ica l a n d exh a u st in g.
In t er m s of Ka h n em a n s con st r u ct , a sign ifi-
ca n t pa r t of t h e m ission s of in t elligen ce a gen -
cies boils down t o seizin g oppor t u n it ies
pr esen t ed by t h e fla wed in t er a ct ion s of t h e Sys-
t em 1 a n d Syst em 2 t h in kin g of for eign a ct or s
wh ile a t t h e sa m e t im e r ecogn izin g a n d m it i-
ga t in g t h e fla ws of t h eir own Syst em 1 a n d
Syst em 2 in t er a ct ion s. H ost ile ser vices a n d
or ga n iza t ion s t r y t o do t h e sa m e t h in g in
r et u r n . Oper a t ion s officer s r ely on t h e bia ses of
for eign cou n t er in t elligen ce officer s, essen t ia lly
a dvisin g a sset s t o a void excit in g a n y Syst em 2
t h in kin g in people posit ion ed t o do t h em h a r m .
Aldr ich Am ess Soviet h a n dler s pr efer r ed t h a t
we n ot focu s Syst em 2 t h ou gh t on h ow h e
bou gh t a J a gu a r on a GS-14 pa ych eckSys-
t em 1 fou n d a t a le a bou t h is wifes in h er it a n ce
cogn it ively ea sy t o a ccept . a
a If you t h in k t h a t you cer t a in ly wou ld h a ve k n own Am es wa s a Soviet spy h a d you kn own of h is J a gu a r, t h en you a r e pr oba bly
gu ilt y of h in dsigh t bia s, or t h e t en den cy t o u n der est im a t e t h e ext en t t o wh ich you wer e su r pr ised by pa st even t s. On t h e ot h er h a n d,
you a r e n ot gu ilt y of h in dsigh t bia s if you t h in k t h is (befor e h a vin g r ea d a bou t Am es) a n d h a ve ever r epor t ed a collea gu e t o cou n -
t er in t elligen ce for own in g a J a gu a r.
2 Studies in Intelligence Vol. 56, No. 2 (Extracts, June 2012)
150
A t a r get s bia ses pu t t h e pla u sible in pla u -
sible den ia bilit y du r in g cover t a ct ion s. E ffec-
t ive decept ion s a lso fu n da m en t a lly r ely on a
t a r get s u n ch a llen ged bia ses a n d so m a ke it
ea sy for t h e t a r get t o believe wh a t t h ey a lr ea dy
a r e pr edisposed t o believe. E ffect ive fa br ica -
t or s, especia lly t h ose wit h t a n t a lizin g a ccess,
r ely on ou r bia sed desir e t o believe t h em . On e
or t wo pla u sible r epor t s fr om su ch a per son
m ay be en ou gh t o en ga ge t h e exa gger a t ed em o-
t ion a l coh er en ce or h a lo effect . Rou gh ly pu t ,
on ce la zy Syst em 2 is sa t isfied, it t en ds t o defer
t o Syst em 1, wh ich in t u r n pr oject s posit ive
qu a lit ies in on e a r ea in t o a gen er a lized posi-
t ive a ssessm en t .
Ter r or ist s r ely on t h ese bia ses, bu t t h ey a r e
a lso vu ln er a ble t o t h em . Ter r or ism wor ks
beca u se it pr ovides ext r em ely vivid im a ges of
dea t h a n d dest r u ct ion , wh ich con st a n t m edia
a t t en t ion m a gn ifies. Th ese im a ges a r e im m edi-
a t ely a va ila ble t o a t a r get s Syst em 1.
Syst em 2, even wh en a r m ed wit h r elia ble st a -
t ist ics on t h e r a r it y of a n y t ype of t er r or ist
even t , ca n n ot over com e Syst em 1s a ssocia t ive
r ea ct ion t o specific even t s. If you a r e a CIA offi-
cer wh o wa s wor kin g in La n gley on 25 J a n u a r y
1993, t h en ch a n ces a r e t h a t you ca n n ot m a ke
t h e left t u r n in t o t h e com pou n d fr om Dolley
Ma dison Bou leva r d wit h ou t t h in kin g of Aim a l
Ka si, t h e Pa kist a n i wh o killed t wo CIA officer s
a n d wou n ded t h r ee ot h er s a t t h a t in t er sect ion
t h a t day.
Th e 9/11 h ija cker s on t h e fir st t h r ee pla n es
cou ld cou n t on pa ssen ger s t o st a y sea t ed, r ely-
in g on t h eir a bilit y t o qu ickly r em em ber
a ccou n t s of pr eviou s h ija ckin gs in wh ich t h e
h ija cker s wer e m ot iva t ed t o su r vivet h is is
wh a t Ka h n em a n ca lls t h e a va ila bilit y bia s.
H owever, beca u se of t h eir su ccess a t t h e Wor ld
Tr a de Cen t er a n d t h e Pen t a gon , t h e t er r or ist s
u n wit t in gly a n d im m edia t ely r en der ed h ija ck-
in g a less effect ive t a ct ic. Th e pa ssen ger s on
F ligh t 93, qu ickly a r m ed wit h kn owledge of t h e
ot h er t h r ee fligh t s, wer e a ble t o en ga ge Sys-
t em 2 t o over com e Syst em 1s exist in g a va il-

a bilit y bia s a n d m a ke t h e decision t o ph ysica lly
over power t h e t er r or ist s.
Ka h n em a n s in sigh t s per t a in t o t h e en t ir e
spect r u m of in t elligen ce oper a t ion s. We a ccept
in for m a t ion secu r it y pr a ct ices t h a t dem on st r a -
bly im pede pr odu ct ivit y in or der t o r edu ce t h e
da n ger of wor se losses posed by cyber a t t a ck or
pen et r a t ion . At t h e sa m e t im e, we wou ld
a lm ost cer t a in ly con sider t h e sa m e a m ou n t of
lost pr odu ct ivit y a m a jor defea t if a h a cker h a d
in flict ed it on u s. Th is is wh a t Ka h n em a n ca lls
t h e loss a ver sion bia s. Syst em 2 does n ot a sser t
con t r ol over Syst em 1s cogn it ive ea se a t im a g-
in in g a disa st er beca u se in cr ea sed pr odu ct ivit y
is m u ch m or e difficu lt for Syst em 2 t o im a gin e.
An y in t elligen ce officer m a kin g bu dget deci-
sion s sh ou ld r ea d Ka h n em a n s t h ou gh t s on t h e
bia ses u n der lyin g t h e su n k-cost fa lla cy, or t h e
decision t o in vest a ddit ion a l r esou r ces in los-
in g en dea vor s wh en bet t er in vest m en t s a r e
a va ila ble. People fin d it difficu lt t o en ga ge Sys-
t em 2 t o cu t t h eir losses in su ch sit u a t ion s,
especia lly wh en Syst em 1 ca n ea sily con vin ce
t h em of t h e loss of pr est ige t h a t wou ld su r ely
follow. H ow oft en does t h e sa m e officer wh o
st a r t ed a n expen sive m a jor pr oject a lso decide
t o kill it ? You likely did n ot h a ve t o en ga ge Sys-
t em 2 t o a n swer t h e qu est ion .
Likewise, n on e of u s a r e im m u n e t o wh a t
Ka h n em a n ca lls t h e pla n n in g fa lla cy, wh ich
descr ibes pla n s a n d for eca st s t h a t a r e u n r ea lis-
t ica lly close t o best -ca se scen a r ios a n d cou ld be
im pr oved by con su lt in g st a t ist ics in sim ila r
ca ses. Th is r eview, for exa m ple, t ook t wice a s
lon g t o wr it e a s I t h ou gh t it wou ld, ju st like
a lm ost ever y ot h er pa per I h a ve ever wr it t en .
In t elligen ce a n a lyst s sh ou ld pay pa r t icu -
la r ly close a t t en t ion t o Ka h n em a n s ch a pt er s
on t h e n est ed pr oblem s of pr edict ion , in t u it ion ,
a n d exper t ise. a For eca st in g a n d pr edict ion a r e
cor e m ission elem en t s for a n a lyst s. Ka h n em a n
br ea ks t h em down in t o t wo m a in va r iet ies. Th e
fir st , su ch a s t h ose en gin eer s m a ke, r ely on
a Ma n y in t elligen ce a n a lyst s a r e fa m ilia r wit h som e of t h ese t h eor ies fr om Rich a r ds J. H eu er, J r.s Psych ology of In telligen ce A n alysis
(Wa sh in gt on , DC: Cen t er for t h e St u dy of In t elligen ce, 1999), wh ich is ba sed in pa r t on ea r lier ver sion s of Ka h n em a n s a n d Tversk ys
wor k. Th is pu blica t ion is a va ila ble on lin e a t h t t ps://www.cia .gov/libr a r y/cen t er-for-t h e-st u dy-of-in t elligen ce/csi-pu blica t ion s/book s-
a n d-m on ogr a ph s/psych ology-of-in t elligen ce-a n a lysis/in dex.h t m l.
Studies in Intelligence Vol. 56, No. 2 (Extracts, June 2012)
151
Thinking, Fast and Slow
look-u p t a bles, pr ecise ca lcu la t ion s, a n d explicit
a n a lyses of ou t com es obser ved on sim ila r occa -
sion s. Th is is t h e a ppr oa ch a n a n a lyst u ses t o
pr edict t h e a m ou n t of explosive for ce n eeded t o
pen et r a t e a cer t a in t h ickn ess of con cr et e, or
ca lcu la t e h ow m u ch fu el a cer t a in t ype of a ir-
pla n e n eeds t o com plet e a cer t a in t ype of m is-
sion .
Ot h er for eca st s a n d pr edict ion s in volve in t u -
it ion a n d Syst em 1 t h in kin g. Ka h n em a n fu r-
t h er br ea ks down t h is va r iet y of pr edict ion in t o
t wo su bva r iet ies. Th e fir st dr a ws on t h e skills
a n d exper t ise a cqu ir ed by r epea t ed exper ien ce,
in wh ich a solu t ion t o t h e cu r r en t pr oblem
com es qu ickly t o m in d beca u se Syst em 1 a ccu -
r a t ely r ecogn izes fa m ilia r cu es. Th e secon d
su bva r iet y of in t u it ive pr edict ion , wh ich is
oft en in dist in gu ish a ble fr om t h e fir st , is ba sed
on bia sed ju dgm en t s. Th is t ype of in t u it ive pr e-
dict ion , t ypica lly for wa r ded wit h con sider a ble
con fiden ce, ver y oft en lea ds t o t r ou ble. Th e
expa n ded u se in in t elligen ce a n a lysis of st r u c-
t u r ed a n a lyt ic t ech n iqu es a n d a ppr oa ch es
a dopt ed in t h e wa ke of t h e 9/11 a t t a cks a n d t h e
Na t ion a l In t elligen ce E st im a t e on Ir a qi wea p-
on s of m a ss dest r u ct ion r epr esen t s in pa r t a n
effor t t o elim in a t e t h is la t t er t ype of pr edict ion .
Th e t r ick is in u sin g st r u ct u r ed t ech n iqu es
a n d a ppr oa ch esor a pplied Syst em 2 t h in k-
in gin a wa y t h a t elim in a t es bia sed in t u it ive
for eca st s a n d pr edict ion s wit h ou t a lso discou r-
a gin g, dela yin g, or even elim in a t in g t h e in t u i-
t ive in sigh t s t h a t t r u e exper t ise pr ovides. Th is
dilem m a pr oba bly expla in s in pa r t wh y som e
exper t s in t h e CIAs Sen ior An a lyt ic Ser vice
r em a in a m biva len t a bou t st r u ct u r ed a n a lyt ic
t ech n iqu es a n d a ppr oa ch es.
Ka h n em a n , despit e h is st a t ed pr efer en ce for
st a t ist ics a n d a lgor it h m s, ca n n ot dism iss ou t of
h a n d t h e va lu e of in t u it ive pr edict ion bor n e of
t r u e exper t ise. H is E xper t In t u it ion : Wh en
Ca n We Tr u st It ? ch a pt er cen t er s on wh a t h e
ca lls h is a dver sa r ia l colla bor a t ion wit h Ga r y
Klein , a lea din g pr opon en t of N a t u r a list ic Deci-
3
Thinking, Fast and Slow
sion Ma kin g, wh o r eject s Ka h n em a n s em ph a -
sis on bia ses a n d focu ses in st ea d on t h e va lu e
of exper t in t u it ion a n d on h ow in t u it ive skills
develop. It is n ot difficu lt t o im a gin e t h a t t h eir
colla bor a t ion wa s m or e difficu lt t h a n Ka h n e-
m a n gen er ou sly por t r a ys it t o h a ve been , wh ich
m a kes t h e a r ea s on wh ich t h ey wer e a ble t o
a gr ee even m or e n ot ewor t h y.
Th ey a gr eed t h a t t h e con fiden ce t h a t exper t s
expr ess in t h eir in t u it ive ju dgm en t s is n ot a
r elia ble gu ide t o t h eir va lidit y. Th ey fu r t h er
a gr eed t h a t t wo ba sic con dit ion s m u st be pr es-
en t befor e in t u it ive ju dgm en t s r eflect t r u e
exper t ise: a n en vir on m en t t h a t is su fficien t ly
r egu la r t o be pr edict a ble a n d a n oppor t u n it y t o
lea r n t h ese r egu la r it ies t h r ou gh pr olon ged
pr a ct ice. An exper t fir efigh t er s sen sin g t h e
n eed t o or der h is m en t o eva cu a t e a bu r n in g
bu ildin g ju st befor e it colla pses or a r a ce
dr iver s kn owin g t o slow down well befor e t h e
m a ssive a cciden t com es in t o view a r e du e t o
h igh ly va lid clu es t h a t ea ch exper t s Syst em 1
h a s lea r n ed t o u se, even if Syst em 2 h a s n ot
lea r n ed t o n a m e t h em .
Lea r n in g, in t u r n , r elies on r eceivin g t im ely
a n d u n a m bigu ou s feedba ck . Ma n y if n ot m ost
of t h e issu es wit h wh ich in t elligen ce a n a lyst s
a r e seized a r e wh a t Ka h n em a n a n d Klein
wou ld pr oba bly ca ll low-va lidit y en vir on -
m en t s, in wh ich t h e in t u it ive pr edict ion s of
exper t s sh ou ld n ot be t r u st ed a t fa ce va lu e,
4 Studies in Intelligence Vol. 56, No. 2 (Extracts, June 2012)
ir r espect ive of t h e con fiden ce wit h wh ich t h ey
a r e st a t ed. Mor eover, t h ey wou ld pr oba bly con -
sider t h e feedba ck a va ila ble t o a n a lyst sfr om
policym a ker s a n d even t sin a dequ a t e for effi-
cien t lea r n in g a n d exper t ise developm en t . Ka h -
n em a n wa s n ot r efer r in g specifica lly t o
in t elligen ce a n a lyst s wh en h e wr ot e, it is
wr on g t o bla m e a n yon e for fa ilin g t o for eca st
a ccu r a t ely in a n u n pr edict a ble wor ld, bu t h e
h a s given in t er views in wh ich h e discu sses
in t elligen ce a n a lyst s in t h is con t ext . At t h e
sa m e t im e, h e a lso wr ot e, h owever, it seem s
fa ir t o bla m e pr ofession a ls for believin g t h ey
ca n su cceed in a n im possible t a sk. In sh or t ,
Ka h n em a n con cedes t h a t in t u it ion h a s t o be
va lu ed, bu t it ca n n ot n ecessa r ily be t r u st ed.
T h in k in g, Fast an d S low pr ovides in t elli-
gen ce officer s wit h a n a ccessible voca bu la r y t o
discu ss t h e pr ocesses of h u m a n cogn it ion t h e
in t er a ct ion s bet ween Syst em 1 a n d Syst em 2
t h in kin gwh ich a r e a t t h e cen t er of t h eir
wor k. It does n ot , h owever, pr ovide solu t ion s or
r elia ble a ppr oa ch es t o bia s m it iga t ion . Accor d-
in g t o Ka h n em a n , t h e best we ca n h ope t o do is
lea r n t o r ecogn ize sit u a t ion s in wh ich m is-
t a kes a r e likely, a n d t r y h a r der t o avoid spe-
cific er r or s wh en t h e st a kes a r e h igh .
Ka h n em a n a lso spen ds ver y lit t le t im e discu ss-
in g h ow bia ses wor k in colla bor a t ive en vir on -
m en t s, despit e h is own ver y in sigh t fu l a ccou n t s
of h is colla bor a t ion wit h Tver sky. We ca n h ope
h e will explor e t h a t in h is n ext wor k.
152