Full Terms & Conditions of access and use can be found at
https://www.tandfonline.com/action/journalInformation?journalCode=hcrj20
CREATIVITY RESEARCH JOURNAL
ISSN: 1040-0419 (Print) 1532-6934 (Online) Journal homepage: https://www.tandfonline.com/loi/hcrj20
The Cognitive Neuroscience of Creativity: A Critical
Review
Keith Sawyer
To cite this article: Keith Sawyer (2011) The Cognitive Neuroscience of Creativity:
A Critical Review, CREATIVITY RESEARCH JOURNAL, 23:2, 137-154, DOI:
10.1080/10400419.2011.571191
To link to this article: https://doi.org/10.1080/10400419.2011.571191
Published online: 09 May 2011.
Submit your article to this journal
Article views: 6853
View related articles
Citing articles: 106 View citing articles
The Cognitive Neuroscience of Creativity: A Critical Review
Keith Sawyer
Department of Education, Washington University in St. Louis
Cognitive neuroscience studies of creativity have appeared with increasing frequently in
recent years. Yet to date, no comprehensive and critical review of these studies has yet
been published. The first part of this article presents a quick overview of the 3 primary
methodologies used by cognitive neuroscientists: electroencephalography (EEG), positron
emission tomography (PET), and functional magnetic resonance imaging (fMRI).
The second part provides a comprehensive review of cognitive neuroscience studies of
creativity-related cognitive processes. The third part critically examines these studies;
the goal is to be extremely clear about exactly what interpretations can appropriately
be made of these studies. The conclusion provides recommendations for future research
collaborations between creativity researchers and cognitive neuroscientists.
Cognitive neuroscience studies of creativity have
appeared with increasing frequently in recent years.
These studies have resulted in many intriguing findings,
and a fair amount of media coverage (e.g., Carey, 2006;
Hotz, 2009; Tierney, 2010). Yet to date, no comprehensive
and critical review of these studies has yet been published
(in spite of a few recent treatments: Kaufman,
Kornilov, Bristol, Tan, & Grigorenko, 2010; Skov &
Vartanian, 2009). The purpose of this article is to
provide a critical review of studies that have used the
methodologies of cognitive neuroscience (CN) and that
have implications for creativity researchers.
This article has three goals. The first goal is to provide
creativity researchers with a working understanding
of the three predominant CN methodologies: EEG,
PET, and fMRI. To prepare the methodological summaries
in part 1, three main textbooks (Gazzaniga, Ivry,
& Mangun, 2002; Purves, Brannon, & Cabeza, 2008;
Ward, 2006) as well as two more advanced texts on
fMRI (Huettel, Song, & McCarthy, 2009; Logothesis,
2008) were consulted. Gaps were filled by consulting
with expert cognitive neuroscientists (see the
Acknowledgments).
The second goal is to review CN studies of creativity
that had been published as of October 2010. Part 2
groups these studies into five categories: creative
insight; mind wandering and incubation; differences
between creative and noncreative people; musical
improvisation; and differences with domain-specific
training.
The third goal is to critically interpret these studies.
Part 3 draws on the methodological understanding provided
in part 1 to critically interpret the studies reviewed
in part 2, and to summarize the implications of these
studies for creativity researchers. The article concludes
by providing suggestions for future creativity research
that uses these methodologies.
PART 1: THE METHODOLOGIES OF
COGNITIVE NEUROSCIENCE
The brain is made up of between 100 and 150 billion
neurons. Each neuron connects with between 1,000
and 10,000 other neurons, at connections called
synapses. A neuron receives signals through short tentacles
called dendrites; it sums up those signals to determine
the strength of the signal it sends down its one
single axon. Each axon has as many as a thousand or
more axon terminals that each transfer signals to the
I thank Aaron L. Berkowitz, Edward M. Bowden, Todd Braver,
Mark Jung-Beeman, John Kounios, Bhavin R. Sheth, and the detailed
comments of two anonymous reviewers for help in preparing this
article.
Correspondence should be sent to Keith Sawyer, Washington
University, Department of Education, Campus Box 1183, St. Louis,
MO 63130. E-mail: keith@keithsawyer.com
CREATIVITY RESEARCH JOURNAL, 23(2), 137–154, 2011
Copyright # Taylor & Francis Group, LLC
ISSN: 1040-0419 print=1532-6934 online
DOI: 10.1080/10400419.2011.571191
dendrites of other neurons. Most axons connect to
nearby neighbors, but a small percentage of neurons
have extremely long axons that can send signals across
the brain. All neurons are constantly firing, sending neurotransmitters
from the axon across the synapses to the
dendrites. The strength of the signal is how many times
per second it fires. A relatively calm neuron fires less
than 10 times per second; a highly active neuron fires
between 50 and 100 times per second.
Cognitive neuroscientists focus on the neocortex, the
thin layer of gray matter on the outside of the brain,
because it is responsible for all higher-level mental functions.
The neocortex is about 5 mm thick (Huettel et al.
2009). Inside the brain, below this outer layer, is the
white matter; this large area is filled with the longer
axons that connect distant parts of the brain. It appears
to be white because the axons are covered with myelin, a
fatty substance that increases the efficiency and speed of
the axon’s electrical transmission. The neocortex
appears to be gray in contrast, because it contains the
neuronal cell bodies and the blood vessels that supply
blood to the them.
The cortex is folded, to allow more of the outer cortical
layer to fit into the skull (see Figure 1). The tops of
the folds—the part that is pressed against the skull—are
called the gyri; the crevices are called sulci. The total surface
area of the cortex is about 2300 cm squared—about
the size of a 12-inch pizza—but two-thirds of that is
within the depths of the sulci.
Cognitive neuroscience today uses three methodologies
that allow psychologists to observe changes in brain
activity while people are thinking: EEG, PET, and
fMRI. Each has its own strengths and weaknesses; two
of the methods are sometimes used in combination to
take advantage of their complementary strengths. These
methods use powerful machines—originally developed
for medical diagnoses—to develop temporal and=or spatial
representations that show how brain activity
changes while the mind is engaged in cognitive tasks.
In a CN experiment, the researcher designs a simple task
for the research participant, one that can be done while
the participant’s head is being examined by the machine.
EEG
Neurons transmit signals down the axon and the dendrites
via an electrical impulse. EEG uses sensors placed
on the scalp that measure electromagnetic fields generated
by this neural activity. EEG detects the electrical
activity at the dendrites—the receiving end of the synapse.
If many neurons and their dendrites are lined up
in parallel, and if a sufficiently large number of neurons
are receiving signals at the same time, a tiny electromagnetic
field is created. In the cerebral cortex, neurons and
dendrites are indeed aligned in parallel, and a detectable
change in the electromagnetic field is generated when
neuronal activity increases or decreases. Neurons aren’t
necessarily aligned in the basal regions of the brain; and
even when they are, the electrical signal is weaker
because those neurons are more distant from the scalp.
Thus EEG is most sensitive to cortical activity.
Any number of sensors, up to 256, can be placed on
the head in standard locations. Most studies use
between 32 and 128 sensors. The EEG is often most closely
examined as a person responds to a stimulus event.
The EEG that is recorded right after the stimulus event
is presented, or during the response, is called an event
related potential (ERP).
Because all neurons are constantly firing, the brain
always generates electric waves of amplitudes between
50–200 microvolts. The event related potentials (ERPs)
that psychologists are interested in are much smaller
amplitudes—usually just a few microvolts. As a result,
in an EEG experiment, the participant is given the same
activity at least twenty times, but usually many more, up
FIGURE 1 Gross brain anatomy. (a) lateral view of the left hemisphere;
(b) dorsal view of the cerebral cortex. The major features of
the cortex include the four cortical lobes and various key gyri. Gyri
(singular is gyrus) are separated by sulci (singular is sulcus) and result
from the folding of the cortex. From Gazzaniga, Ivry, and Mangun
(2002, p. 71).
138 SAWYER
to as many as 100 times; then, mathematical algorithms
are used to average over all of the trials (see Figure 2).
The normal brain waves of 50–200 microvolts cancel
each other out, and what remains is the change in brain
activity that is directly related to the cognitive event of
interest—the ERP. EEG signals of interest to cognitive
scientists occur in the frequency range of 1–50 Hz;
ERP signals most often occur in the frequency range
of .5–20 Hz.
Different frequency bands of the brain’s electromagnetic
field indicate different sorts of brain activity:
. delta waves (.5–4 Hz)—during deep sleep;
. theta waves (4–8 Hz)—greater in childhood; implicated
in encoding and retrieval of information;
. alpha waves (8–13 Hz)—occurs while awake, while
relaxed with the eyes closed;
. beta waves (13–30 Hz)—increased alertness and
focused attention;
. gamma waves (>30 Hz)—still not well understood,
but have been implicated in creating the unity of
conscious perception.
Cognitive neuroscientists typically study alpha, beta,
and gamma waves.
The advantage of EEG is that it can detect the brain’s
response to the external stimulus event essentially
immediately—to the microsecond. This is referred to
as a high temporal resolution. The disadvantage is that
EEG cannot support strong claims about where the neurons
are that are causing the change in the electromagnetic
field; this weakness is referred to as a low spatial
resolution. Even though there are as many as 256 electrodes
positioned around the skull, an ERP at any particular
electrode does not necessarily mean that the
ERP was caused by neurons immediately underneath
that electrode, because electromagnetic fields extend
across the brain. To identify the brain regions associated
with neuronal activity, technologies with a higher spatial
resolution are used—PET and fMRI.
PET
When neurons in a particular region of the neocortex
are firing more rapidly, that region is said to have elevated
neuronal activation. As a result of elevated activity,
the neurons require more oxygen, and blood flow is
greater to that region. PET indirectly measures neuronal
activity by detecting local changes in regional cerebral
blood flow (rCBF). PET works by introducing a radioactive
tracer into the bloodstream; where there is more
blood flow, there is more radiation. A radioactive isotope
of oxygen is often used that decays in less than
one minute—a fast decay is important, to reduce the
amount of radiation exposure.
During a PET experiment, a person is given a cognitive
task that can be done within the time it takes for the
oxygen isotope to decay, usually about 40 seconds.
While they engage in this task, the associated brain
regions increase in neuronal activation; regional cerebral
blood flow increases to those regions; and the increased
radioactivity is detected by the PET scanner—a large
donut-shaped device with the head placed at the center.
The result is a three-dimensional representation of the
brain activity associated with the cognitive task.
PET has a fairly high spatial resolution; the technology
is able to measure the neuronal activity associated
with a neocortical region of about 5 millimeters
cubed. This 5 mm3
space is called a voxel for volume
element (the word has an ‘‘x’’ in it because it’s derived
from ‘‘pixel,’’ the term for the two-dimensional ‘‘picture
element’’ that’s used in televisions and computer
screens). On average, neural density in the neocortex is
20,000 to 30,000 neurons per 1 mm cubed, and the number
of synapses in a cortical space of 1 mm cubed is close
to one billion. This means that in the typical CN study,
each voxel effectively contains 5.5 million neurons and
about 50 billion synapses (Logothetis, 2008).
FIGURE 2 In EEG experiments, participants are presented with the
same stimulus or task up to 100 times, and the EEG waves are averaged
across all of the tasks, to reduce the signal-to-noise ratio and
reveal the event-related potential (ERP) associated with the task. These
are waves of the EEG in response to presentation of an audio tone at
time 0. The topmost wave represents a typical EEG from a single trial;
the other waves represents averaging across 10, 50, and 100 trials. This
is done for each electrode, and each will have a slightly different ERP
profile. From Kolb and Whishaw (2002).
THE COGNITIVE NEUROSCIENCE OF CREATIVITY 139
Compared to EEG, PET has a very low temporal resolution;
EEG detects the ERP essentially immediately,
but PET requires 40 seconds or more (the time associated
with the isotope decay) to measure elevated brain
activity.
fMRI
fMRI emerged a few years after PET, but has rapidly
become the most widely used brain imaging technique.
It has been called ‘‘the most important imaging advance
since the introduction of X-rays’’ in 1895 (Logothetis,
2008, p. 869). The fMRI machine uses a magnetic field
to detect the ratio of oxygenated to deoxygenated blood;
each affects the magnetic field differently. The ratio of
oxygenated to deoxygenated blood is referred to as the
blood oxygen level dependent (BOLD) signal. When neuronal
activation increases in a region of the neocortex,
blood flow increases faster than the neurons can use
the oxygen, causing the BOLD signal to increase.
fMRI is used much more than PET for many reasons:
the machines are more readily available; the spatial resolution
is higher; unlike fMRI, PET does not actually
provide a picture of the brain, so it has to be mapped
onto brain space, giving fMRI better spatial identification;
and with fMRI, there is no need to inject radioactive
tracers with each trial, allowing hundreds of trials
which can then be averaged. With a PET experiment, at
most about 30 trials can be done, because the radioactive
isotope must be injected just before each trial. In
addition, the temporal resolution of fMRI is significantly
higher; with PET, the researcher must average
the brain’s activity over 40 seconds of engagement with
a task, because even the fastest isotopes decay over 40
seconds, but an fMRI image can be captured every
two seconds, allowing for an event-related fMRI (similar
to the ERP provided by EEG). As with EEG, neuronal
changes associated with any single trial are
impossible to detect because of the complexity of brain
response; researchers must average the responses to an
identical cognitive task over approximately fifty identical
trials. This allows the use of statistical algorithms
that average out the unrelated brain activity fluctuations
and reveal the signal related to the event being studied.
There are three challenges with fMRI that result from
its dependence on BOLD signal changes. First, BOLD
increases above the resting state only between 1% and
3% at maximum neuronal activation. When cognitive
neuroscientists report increased neuronal activation in
a particular brain region, they are reporting an increase
that is never greater than 3% above the comparison
baseline state of the normal neuronal firing rate within
that region. Second, when neurons increase in activation,
the BOLD signal does not increase immediately;
the initial rise does not occur until several seconds after
the increase in neuronal activity, the peak is 4 to 6 seconds
later, and it does not decline back to baseline for
15 or 20 seconds (Fink, Benedek, Grabner, Staudt, &
Neubauer, 2007; Huettel et al., 2009). This delay in
hemodynamic response varies between individuals, so
each experiment has to correct for that variation. The
hemodynamic response delay varies across different
brain regions even in the same person; there is no known
way to normalize these variations, but they are not
thought to be large. Third, the spatial location of BOLD
does not always correspond exactly to the neurons that
are increasing in activation, because BOLD detects the
anatomical locations of the blood vessels that supply
the neurons, not the location of the neurons themselves
(Huettel et al., 2009). Furthermore, for accurate localization,
fMRI has to detect blood flow in the tiniest
capillaries, the ones immediately next to the neuronal
cell bodies; but there are much larger blood vessels that
feed those capillaries. fMRI technology is largely able to
focus only on the smallest capillaries, but this ability
varies subtly with different cortical regions. Researchers
who use fMRI have developed techniques to account for
these problems, but it will always be an inexact methodology,
because of one final challenge: When neurons
become more active, blood flow increases not only right
next to those neurons, but also over a bigger area that
extends to a few millimeters distant, where there may
be no increase in neuronal activity (Huettel et al., 2009).
Combining Methods
Because these technologies have complementary
strengths, they can be used together to develop fairly
elaborate understandings of how activity in the biological
brain corresponds to human mental functioning.
One of the most common approaches is to use EEG
for its high temporal resolution; then to use fMRI with
the same task for its high spatial resolution; then to
combine the two findings for a more complete picture
of the brain’s activity.
Statistical Averaging
With EEG, PET, and fMRI, it is not possible to study
just one response to a single event, because there are
large changes in the EEG, rCBF, and BOLD signals that
are always occurring as part of the brain’s normal
activity. So each participant does the same task tens or
even hundreds of times, and the brain activity is averaged
across all of these trials. The normal background
variation of the brain’s activity is thus averaged out,
and what remains is the activity of interest.
When research studies report that a specific voxel
shows elevated neuronal activity in a particular task,
the brain is not necessarily engaging these regions every
140 SAWYER
time it engages in the task; what is being reported is an
average over many repetitions of the task.
Standard brains. To account for the ordinary variability
in human brains, cognitive neuroscientists do not
study a single person. Instead, they perform the same
experiment on many people; with PET and fMRI, they
do statistical image averaging across all of the brains;
and they use statistical algorithms to identify the average
location of neuronal activation, across all of the subjects’
brain images, averaged together to generate a
single ‘‘average’’ brain image.
Even though the overall organization of all brains is
quite similar, each human brain is slightly different.
Heads come in different sizes and shapes: some more
narrow, some shorter front to back. Thus, comparing
two brains requires the researcher to mathematically
adjust the size of all of the brains so that they are
roughly the same. Most researchers adjust the brains
to align with a standard brain as published in standard
neuroscience atlases. Otherwise, normal anatomical
variation would make averaging impossible.
Even after doing this, brains differ in the size of the
different gyri and the location of folds in the brain;
the location of sulci can vary by as much as a centimeter.
There is at present no method for manipulating each
brain’s detailed structure to conform to a standard. To
accommodate this natural variation in brain structure,
most studies use a mathematical technique known as
smoothing—which spreads out the observed activation
across nearby voxels, thus increasing the chance of overlap
among different individuals when statistical image
averaging is done.
Movement. Whenever people move their hands, or
bend their knees, or turn their heads, large regions of
the brain are active, including vast areas of the neocortex,
where higher level thought takes place. Even blinking
an eye, or twisting the head a tiny bit, or moving the
eyes to the side even while the head is stationary, or
twitching a leg muscle, causes neuronal activation that
can interfere with the image. With the EEG, eye and
eyelid movements create electric frequencies in the same
range as the EEG signal. For these artifacts, many studies
use a standard ‘‘correction’’ that subtracts out
the artifact. When there are a lot of artifacts at an electrode,
it is often possible to interpolate the signal from
the neighboring electrodes. But sometimes the artifacts
are severe and the trial has to be rejected completely.
As a result, it is important for participants to remain
completely still during these experiments. Typically the
head is physically restrained, and participants are asked
not to move, but even so, muscles can tense or twitch
enough to affect the results. With fMRI, many studies
use algorithms that correct for head movement. And with
fMRI, the participants generally do not talk at the same
time that the image is being taken; the vocal tract causes
electromagnetic activity that can disrupt the BOLD signal.
So for participants to communicate with the
researcher, typically they are given a small handheld
device with a single button, and they are instructed to
push the button depending on what they perceive. This
only requires one finger to move a short distance, reducing
the associated neuronal activation to a minimum.
Paired Image Subtraction
There are three facts about the brain that make CN
challenging.
. First, every neuron is always firing, at least a few
times every second. So researchers always refer to
relative activation levels, rather than neurons being
on or off.
. Second, it is not the case that when people stare off
into space, all of their brain’s neurons are firing
only at a low activation level. Large parts of the
brain are always fairly active. When the mind is
engaged in a cognitively challenging task, the
brain’s energy consumption rarely increases by
more than 5% (Raichle, 2009). One of the most
solid and consistent findings of CN studies is that
the brain’s resting state is quite similar to the problem
solving state; to conceptual processing; and to
memory retrieval (see citations in Smallwood &
Schooler, 2006).
. Third, there are parts of the cortex that always
increase in neuronal activation whenever people
engage in any cognitive task.
Cognitive neuroscientists are interested in all of these
aspects of the brain; but most of the time, they try to
identify specific cortical regions that increase in activation
in one kind of task, but not in others. The methodology
that allows researchers to identify specific brain
regions associated with specific tasks, in spite of these
three challenging facts, is called paired image subtraction.
In every experiment, the first thing that is done is
that a brain image is taken during a carefully selected
control state or baseline state, while the participant lies
still and does nothing, or stares at a target ‘‘X’’ on the
presentation screen, or performs some simple comparison
activity. This baseline is also sometimes called the
rest state. Then, this baseline image is subtracted from
the image that results during the task condition.
The key to designing an effective experiment is to
design two tasks that are identical in every way, except
for one small change that is the cognitive function
of interest. The subtraction cancels out the normal
THE COGNITIVE NEUROSCIENCE OF CREATIVITY 141
activation levels of all of the neurons that do not
undergo any change in activation level (facts 1 and 2)
and it cancels out all of the neurons that change in activation
level the same way in both conditions (fact 3).
The image that results shows the differences in activity
between a task condition, and the baseline of brain
activity (see Figure 3). A specific brain region might be
increasingly active in both the experimental and control
conditions, but if the increase is greater in the experimental
condition (above a threshold of statistical significance),
the visual display will show heightened
activation for the experimental condition.
Paired image subtraction makes a key assumption:
pure insertion, meaning that the additional cognitive process
can be inserted into the baseline process without
indirectly changing any of the activity associated with
the baseline process. But this assumption is probably
rarely the case, because the brain is complex and nonlinear
(Logothetis, 2008). There is no way to detect these
indirect changes, and the implications for interpreting
CN studies remains unclear. In many cases, the violation
of the insertion assumption is not relevant; and other
designs can sometimes be used to ameliorate the problem.
Cognitive Conjunction
Some studies make use of the technique of cognitive conjunction:
participants engage in two slightly different
tasks, such that each of the tasks shares one common
cognitive component. Then, a paired image subtraction
is done for each of the conditions against the resting
state. Finally, the regions of heightened activation that
are shared across the two tasks are identified; these
regions are presumably associated with the common
cognitive component.
PART 2: FINDINGS
The following review groups CN studies relevant to
creativity research into five categories:
. Creative insight: Remote association, hints and
restructuring, hypothesis generation and set shift,
story generation;
. Mind wandering and incubation;
. Creative brains versus noncreative brains;
. Musical improvisation; and
. Differences with training.
Creative Insight
Many psychologists have studied the moment of insight
(e.g., Duncker, 1926; Metcalfe, 1986; Metcalfe & Wiebe,
1987; Perkins, 1981; Schooler, Ohlsson, & Brooks, 1993;
Sternberg & Davidson, 1995; Weisberg & Alba, 1981).
Many of these studies draw from a classic set of insight
problems that are thought to require a moment of insight
to be solved. The exact nature and role of insight in creativity
is widely debated by creativity researchers, with
some researchers emphasizing the critical role played
by insight (Cunningham, MacGregor, Gibb, & Haar,
2009; Ohlsson, 1992), and others arguing that insight
plays no role, and that creativity is essentially identical
to everyday problem solving (Perkins, 1981; Weisberg,
1986, 1993). Brain imaging technologies are ideally
suited to identifying the brain regions associated with
creative insight, and to determining whether or not there
are distinct cognitive processes that are involved when
solving these problems.
The front of the brain is associated with the highest,
most deeply human abilities—what are sometimes called
‘‘controlling’’ and ‘‘executive’’ functions of the brain
(Srinivasan, 2007). Representational systems, such as
symbols and verbal meanings, are processed in the frontal
lobes. Dietrich (2004) hypothesized that conscious
and deliberate creativity is driven by the front of the
brain, but that spontaneous insight emerges from three
cortices behind and under the frontal cortex—the temporal,
occipital, and parietal (TOP). The TOP areas
are devoted primarily to perception and to long-term
memory; they receive many neuronal axon signals from
the lower, sensory brain systems. The frontal lobe does
FIGURE 3 In the upper row of these PET scans, the control condition
(resting while looking at a static fixation point) is subtracted
from the experimental condition of looking at a flickering checkerboard
5.5 degrees from the fixation point. The subtraction produces
a slightly different image for each of five subjects, as shown in the middle
row. Statistical image averaging across these five subjects results in
the image at the bottom. This procedure is always used in both PET
and fMRI studies. From Posner & Raichle (1994, p. 65).
142 SAWYER
not receive direct sensory input; it integrates already
highly processed information from the TOP to enable
even higher-level cognitive behaviors such as abstract
thinking, planning, willed action, working memory,
and attention.
Remote association. One problem with using
insight problems in brain imaging studies is that many
people say they solve them without actually having an
experience of insight. Instead, they say that they worked
systematically and incrementally toward the solution.
For example, the three-word remote associates test
(RAT) triplets (Bowden & Jung-Beeman, 2003b) are
sometimes solved without an accompanying sensation
of insight. Perhaps the sensation of insight is purely a
subjective feeling of emotional intensity or excitement
at having found the problem, but does not actually contribute
to solving the problem. Several CN studies have
examined whether or not insightful solutions result from
different brain activity than solutions with no insight.
Jung-Beeman et al. (2004) conducted a series of experiments
to determine what happens in the brain when
people are solving insight problems. Their studies were
designed to address three questions:
1. Are there different cognitive and neural mechanisms
involved in having an insight solution versus
ordinary problem solving processes?
2. Is there any unconscious processing that immediately
precedes the sudden conscious awareness
of the insight? Previous studies suggest this is
the case (Bowers, Farvolden, & Mermigis, 1995;
Bowers, Regehr, Balthazard, & Parker, 1990;
Kotovsky, 2003).
3. Does the sudden ‘‘Aha!’’ experience reflect a sudden
change in the brain that contributes to the
solution? Or is it simply an affective response
that follows the solution?
From the overview in part 1, we know that all cognitive
tasks result in elevated neuronal activity in many of
the same frontal lobe regions. Likewise, solving problems
with insight and without insight are likely to both
involve many of the same cognitive processes and neural
mechanisms. But insight solutions seem to require distant
or remote associations, and a variety of CN studies
have found that when the brain engages in associative
tasks, there is elevated neuronal activation in the
anterior superior temporal gyrus (aSTG) of the right
hemisphere. For example, language comprehension studies
show that sentences and complex discourse increase
aSTG activation in both hemispheres, and that when
distant semantic relations are used, the right hemisphere
(RH) aSTG is more active. Solving an RAT triplet
requires the simultaneous activation of distant associations,
and there is evidence that the right hemisphere
is more effective at semantic processing of distant associates
(Howard-Jones et al.; Blakemore, Samuel,
Summers, & Claxton, 2005; Bowden & Jung-Beeman,
2003a). And there is some evidence that the prefrontal
part of the RH supports processing of distant associations
(Seger, Desmond, Glover, & Gabrieli, 2000).
Their first experiment used fMRI with 13 subjects
and presented each subject with 124 RAT triplets. Participants
were given 30 seconds to identify each target
word. As soon as they identified the word, they pressed
a button in their hand. The researchers then asked them
to say the word out loud, simply to confirm that they
had the correct word (recall that fMRI is difficult to
use when subjects are talking). Then, they were asked
to press the button again if they had experienced a feeling
of insight accompanying the solution.
Fifty-nine percent of the problems were solved, and
people reported feeling insight for 56% of these solutions.
The researchers examined the brain activity during the 2
seconds before the first button press and the 4 seconds
after. The subjects who solved problems and reported
having a sensation of insight showed elevated neuronal
activation in the RH aSTG during the 6 seconds (both
before and after the button press), compared to those
who solved the problem without an insight sensation.
The temporal resolution of fMRI is low; in this experiment,
one brain image was taken per second. Thus these
findings might simply reflect a subjective experience of
insight, and it is that subjective experience that corresponds
to the elevated activity in the RH aSTG—but that
activity might not contribute to the insight solution; it
might simply be a response to the solution. A more
important question is: Was the problem actually solved
differently when subjects felt they were having an insight?
The researchers wanted confirming evidence with better
timing information, so they did exactly the same experiment
with a new group of subjects, but this time with
EEG, and compared the time-frequency analyses of the
EEGs of insight solutions and noninsight solutions.
When analyzing the EEG signal using the statistical
averaging method described in part 1, they found an
increase in gamma wave activity in the front RH (but
not LH) associated with insight solutions, and this burst
began about .3 seconds before the solution button-press.
No increase was observed for noninsight solutions. The
researchers used the converging evidence from the fMRI
to suggest that this might correspond to the RH aSTG
(the poor spatial resolution of EEG would never allow
such a specific localization). The researchers hypothesized
that the RH aSTG may have contributed to the
solving of a triplet with insight, because the EEG
increase preceded the button press and thus preceded
the conscious realization of a solution.
THE COGNITIVE NEUROSCIENCE OF CREATIVITY 143
However, the elevated EEG activity continued to
increase for a full second after the button was pushed,
suggesting an alternative explanation: it might reflect
the excitement of getting the solution—an effect of the
solution rather than a cause (see Sheth, Sandku¨hler, &
Battacharya, 2008). After all, brain activity that contributed
to identifying a solution would most likely terminate
once that solution were found. Providing evidence
for their causal interpretation, in another study
(Jung-Beeman, Bowden, & Haberman, 2002), when insight
was reported, there was greater neuronal activity
in the RH superior temporal succulus for the final 2 seconds
before participants solve the problems, than when
no insight was reported. Two seconds is long enough
that it most likely precedes the subjective conscious
awareness of knowing the solution.
In a third study (Kounios et al., 2006), different alpha
wave patterns preceded the presentation of the problem
when insight solutions were reported. There was no difference
in the EEG between successfully getting the
answer versus not getting the answer, suggesting that
subjects were doing the same basic sort of mental work
whether or not they got the answer. In Experiment 2,
they did an fMRI to identify the pattern of brain activity
before the problem was presented. Those people who
solved a problem with insight displayed a greater
activity of the anterior cingulate cortex (ACC) just
before the problem was presented. They concluded that
ACC activity is responsible for the alpha-wave midfrontal
activity detected with the EEG.
In sum, these experiments provide some evidence for a
distinct pattern of neuronal activation immediately prior
to an insight solution. The pattern of brain activation is
consistent with other cognitive tasks that require distant
semantic associations, including those that are not
thought of as creative tasks per se. But because this pattern
continues even after the solution, it is probably too
simplistic to argue that the associated brain region is
dedicated to solving distant associations like the RAT.
Hints and restructuring. The Gestaltist model of
insight suggests that some sort of fixation leads to an
impasse, and the impasse can only be overcome by a
mental restructuring of the problem representation that
would remove the fixation. Several creativity researchers
have evaluated this model by using an experimental
paradigm in which subjects who are stumped are then
provided with hints designed to prompt a restructuring
(Gick & Lockhart, 1995; Kaplan & Simon, 1990;
Weisberg & Alba, 1981). A similar paradigm has been
used in several CN studies.
Luo and Knoblich (2007) presented insight problems
(Example: ‘‘the thing that can move heavy logs, but
cannot move a small nail’’; answer: ‘‘river’’) followed
by either hints that led to restructuring, or hints that
reinforced the incorrect structuring. When they provided
a restructuring hint, they observed elevated neuronal
activation in the bilateral superior frontal gyrus, the
medial frontal gyrus extending to the cingulate cortex,
and bilateral posterior middle temporal gyrus. They also
found elevated neuronal activation in the ACC with
insight problems compared to noninsight problems.
Interestingly, this ACC activation declined across trials
throughout the course of the experiment, suggesting that
subjects were developing general strategies to deal with
this sort of word problem.
Sheth et al. (2008), using EEG, found brain differences
up to eight seconds before the solution, when
the problem was solved with insight. They observed a
reduction in beta power (15-25 Hz) over the
parieto-occipital and centro-temporal regions, compared
to a 10-second preresponse baseline, with: (a) correct
versus incorrect solutions; (b) solutions without a
hint versus with a hint; (c) success after the hint is provided
versus no success (they compared the ten seconds
before hint and the ten seconds after); (d) self-reported
high insight versus low insight. Gamma band (30–
70 Hz) power was increased in the right fronto-central
and frontal regions for (a) and (c). Lower alpha was
increased for insight versus non-insight solutions in the
central-parieto region. The most intriguing result was
that for those who were stumped and then were presented
with a hint, the brain activation pattern was different
even before the hint was presented for those who
eventually got the answer versus those who did not (cf.
Kounios et al., 2006). This suggests the existence of a
brain state that better prepares a person to receive a hint
and use it successfully.
Fink et al. (2009) found an increase in alpha synchronization
in frontal brain regions, and a diffuse and widespread
pattern over parietal cortical regions, in response
to more free associative tasks (like the alternative uses
test, and a name invention task). More original ideas
were associated with stronger increases in alpha activity
than conventional ideas. A corresponding fMRI study
found strong activation of LH frontal regions in the
same conditions, particularly the left inferior frontal
gyrus (this finding was replicated by Jung-Beeman,
2005).
Using EEG, Sandku¨hler and Bhattacharya (2008)
gave people RAT triplets; they could press a button if
they were stumped, and right away they would be shown
a hint: the first letter of the target word, or half of the
letters of the target word. The first 2 seconds while they
were reading the problem was used as a baseline. They
found strong gamma band responses in parieto-occipital
regions for insight versus noninsight solutions
(38–44 Hz). They also found increased upper alpha band
response (8–12 Hz) in RH temporal regions. Alpha band
144 SAWYER
activity is often associated with cognitive inhibition,
suggesting that for initially unsuccessful trials that
after a hint led to the correct solution, there was active
suppression of weakly activated solution relevant
information.
Hypothesis generation and set shift. Vartanian and
Goel (2007) summarized several studies focused on
hypothesis generation and set shift: a movement from
one state in a problem space to a very different state,
with no obvious incremental step-by-step transition
(also see Goel & Vartanian, 2005). These included
studies on Guilford’s match problems (requiring
re-arranging matches to generate a specified number of
squares) which compared a divergent condition (generate
all of the possible ways this problem could be solved)
to a convergent condition (subjects were presented with
a hypothetical solution and asked to say whether it was
correct). The hypothesis generation required by the
divergent condition activated left dorsal lateral PFC
and right ventral lateral PFC (vs. baseline). When subjects
got a correct solution—which was evidence of a
set shift—only the left dorsal lateral PFC was still
increased in activation versus baseline.
With anagram problems, right ventral lateral PFC
was activated when problems were solved without any
hint (‘‘Can you make a word with CENFAR?’’) versus
given a specific semantic category as a hint (‘‘Can you
make a country with CENFAR?’’). They concluded that
hypothesis generation in open-ended settings activates a
network that includes right ventral lateral PFC, for both
spatial and linguistic stimuli. These are different areas
than the ones implicated in the RAT insight studies—
in those, it is the right temporal lobe. In another study,
they found that activation in right dorsal lateral PFC
covaried with the total number of solutions generated
in response to match problems—which could be the
result of working memory, cognitive monitoring, or conflict
resolution.
Kounios et al. (2008) also studied anagram problems,
and focused on the resting state, the period just before
the anagram was shown. The researchers divided subjects
into two groups based on how they reported having
solved anagram problems: one group with people who
were more likely to report solving a problem with insight
(the high insight group, HI) and another group who
were less likely to report using insight (the low insight
group, LI). High insight people had different resting
state EEGs compared to low insight people. LI people
had more high alpha—which indicates less activity in
the visual cortex—than the HI group. This suppression
of activity was greater in the LH. The LI group had
greater beta-1 EEG as well, suggesting more focused visual
attention. The HI group had more RH activity, in
low alpha, beta-2, beta-3, and gamma frequency ranges.
Kounios et al. (2008) explained these findings by suggesting
that a person’s likelihood of using insight to
solve a problem is influenced by the characteristics of
the prior resting state; they could predict the likelihood
that a person would later use insight to solve an anagram,
by analyzing the EEG during the resting state just
before the anagram was displayed. The tendency to use
insight or not remained stable through the course of the
experiment; people used the same amount of insight in
the second half of the experiment as in the first half. This
confirmed many other studies that have shown that resting
state EEG is relatively stable over time.
In the aforementioned experiments, the brain regions
that display elevated neuronal activity during creative
tasks are also involved in a wide range of non-creative
cognitive tasks. The RH aSTG is implicated in integration
across sentences to extract themes; to form
coherent memories for stories; generating the best ending
for a sentence; and to repair grammatically incorrect
sentences. The ACC is implicated in monitoring for
competition among potential responses or processes; in
voluntary selection; in conflict monitoring; in decision
making; and in unrehearsed movements (see Berkowitz
& Ansari, 2008). Some studies suggest that ACC is
involved in suppressing irrelevant thoughts. Neuronal
activity in the ACC was elevated with insight solutions,
suggesting that shifting the mind away from an answer
that one realizes is incorrect involves cognitive control
mechanisms similar to those involved in suppressing
irrelevant thoughts.
Story generation. Howard-Jones et al. (2005) examined
brain activity while participants generated fictional
short stories, while engaged in a task designed to require
verbal creativity: Participants were presented with three
words and then asked to create a story that contained all
three words. The researchers varied this three-word
short story task to create four conditions in a 2 Â 2
design: (a) instructing subjects to ‘‘be creative’’ or ‘‘be
uncreative’’; (b) providing subjects with three unrelated
words (flea, sing, sword), or three related words
(magician, trick, rabbit). While in an fMRI scanner,
the participants were given 22 seconds to generate each
story. Using paired image subtraction to subtract the
‘‘be uncreative’’ image from the ‘‘be creative’’ image,
an increase in prefrontal activity was observed, including
bilateral medial frontal gyri and left anterior cingulate
cortex (ACC). Using paired image subtraction to
subtract the unrelated words image from the related
words image, elevated neuronal activity was found in
bilateral ACC and right medial frontal gyrus.
As with the insight studies, these brain areas are not
unique to creative tasks; they are involved in a wide
THE COGNITIVE NEUROSCIENCE OF CREATIVITY 145
range of cognitive tasks. Left prefrontal activation is
elevated in word association tasks and sentence completion
tasks. Increased ACC activity has been linked
to a wide range of tasks with increased information processing
demands, including selecting items from episodic
memory. Making divergent associations requires
increased conflict monitoring; the ACC and the prefrontal
cortex are associated with additional conflict
monitoring and with insight solutions.
Mind Wandering and Incubation
One of the oldest observations in the psychology of creativity
is that a creative idea is often preceded by a period
of unconscious incubation (Hadamard, 1945; James,
1880; Poincare´, 1913=1982). Contemporary creativity
researchers have often studied incubation and its role
in creative insight (Beeftink, van Eerde, & Rutte, 2008;
Ellwood, Pallier, Snyder, & Gallate, 2009; Kohn &
Smith, 2009; Patrick, 1986; Sternberg & Davidson,
1995). The majority of studies has confirmed the existence
of an incubation effect, although the exact nature
of the associated unconscious processes remains uncertain.
Hypotheses include mental relaxation, selective
forgetting, random subconscious recombination, and
spreading activation.
Cognitive neuroscientists have studied a closely
related mental phenomenon: mind wandering, when
thoughts drift away from the task at hand to something
completely unrelated. Mind wandering involves a shift
away from a primary task to process some other, personal
goal, but in a way that is not obviously
goal-directed or intentional. Some neuroscientists have
hypothesized that people prone to mind wandering
may score higher on tests of creativity (Hotz, 2009;
Tierney, 2010). Recent studies of the brain’s idle states
can potentially help researchers identify what brain
regions are associated with the mind wandering state,
and potentially have implications for the understanding
of the role of incubation in the creative process.
During waking hours, people’s minds wander
between 15 to 50% of the time, depending on the task.
For example, people’s minds are wandering 20% of the
time they are reading, and half of those times people
are not even aware that their minds are wandering
(Smallwood & Schooler, 2006, p. 956). The content of
mind wandering is dominated by typical life events
and is rarely focused on fantasy (Andrews-Hanna,
Reidler, Huang, & Buckner, 2010; Singer & Antrobus,
1963). Mind wandering varies with fatigue, with alcohol,
and with difficulty of the task. Kane et al. (2007) found
that it averaged 30% in an experiential sampling of
everyday life, and that it varied depending on working
memory capacity (WMC). In people with high WMC,
their minds wandered less when the task required
focused concentration, but their minds wandered more
when task demands were low.
Klinger (2009) found that, on average, people have
about 4,000 thoughts of all kinds during a typical day,
each averaging 14 seconds in length, and that half of
these qualify as ‘‘daydreaming’’ defined as undirected
mind wandering or thought that is at least partially
fanciful (p. 228). He hypothesized that even in cognitively
demanding tasks, there would be a minimum rate
of daydreaming of about 10% of the time.
Raichle (2009) identified a default network—parts of
the brain that are active in the resting state, but that
become less active when engaged in various cognitive
tasks. Somewhat paradoxically, these same regions
reduce in activity during unconscious states
(Andrews-Hanna et al., 2010). Brain imaging has found
that the default network continues to be active in tasks
that involve passive sensory processing, but tends to
reduce its activity with tasks with high central executive
demand—a pattern similar to mind wandering, thus
suggesting that the default network may be involved in
mind wandering. Mason et al. (2007) found that when
people were engaged in a task they had practiced on,
allowing them to daydream, the default network was
more active than when they were engaged in a novel
task. Christoff, Gordon, Smallwood, Smith, and
Schooler (2009) found that the default network was
most active when people’s minds were wandering and
they were not aware of it. There is evidence that task
performance is more severely disrupted by mind wandering
when one is not aware that it is occurring.
The previously mentioned studies of mind wandering
suggest that people spend more of their daily lives
engaged in an incubation-like state than they probably
realize: People typically are only consciously aware of
one-half of their mind wandering episodes. This suggests
an interesting possibility that creativity researchers
might study further: these brief episodes of mind wandering
may provide the mind with moments of ‘‘mini
incubation’’ that contribute to creative thought, by
temporarily taking conscious attention away from the
problem at hand and providing a brief opportunity for
insight to occur.
Creative Brains Versus Noncreative Brains
The methods of CN are not able to reliably analyze the
activity within a single person’s brain. But they can be
used to identify differences between groups of people,
so long as there are enough people in each group to
do statistical image averaging. A few studies have examined
differences in neuronal activity between people who
get high scores on creativity tests, compared with people
who get low scores. In one of the earliest studies, using
EEG, Martindale and Hines (1975) found that creative
146 SAWYER
people show higher levels of alpha wave activity when
engaged in creative tasks like the Alternate Uses Test
and the RAT, whereas medium- and low-creative
groups had lower alpha wave activity.
Carlsson, Wendt, and Risberg (2000) used the Creative
Functioning Test to select a high creative and a
low creative group (each with 12 male right-handed students).
They then presented three tasks which were
expected to activate the frontal lobes increasingly: the
lowest expected activation was for an automatic speech
task (count aloud, starting with 1), the next higher activation
was a word fluency task (FAS: say all words you
can think of that start with the letter F or A or S), and
the final was a divergent thinking task (say as many uses
as you can think of for a brick). They measured rCBF
by using a special helmet with 254 sensors to detect radiation
emitted by the isotype. They found that low creatives
had more elevated LH during the word fluency
task; high creatives had more elevated RH during the
brick test. (Strangely, the automatic counting task
resulted in higher blood flow than either of the other
two tasks; the researchers could not think of a plausible
explanation for this, so their analysis focused on the
FAS-brick contrast.) The biggest differences, when comparing
brain activity on FAS and on brick, were elevated
anterior prefrontal activation in creatives (both
hemispheres), and decreased fronto-temporal and
anterior prefrontal activity for low creatives (particularly
in the RH). They concluded that high creatives
use bilateral prefrontal regions on the brick task, while
low creatives used mostly LH. High creatives had more
increased activity in these regions, compared to FAS,
than low creatives—whose brains looked about the
same in FAS and in brick.
Cha´vez-Eakle (2007) and her team used PET to compare
six individuals with Torrance Test of Creative
Thinking (TTCT) scores in the 99th percentile with six
individuals at the 50th percentile, by giving them the
Unusual Uses test. The high scorers on verbal TTCT
had greater rCBF (regional cerebral blood flow) in the
right precentral gyrus; the high scorers on the figural
TTCT had greater rCBF in right postcentral gyrus, left
middle frontal gyrus, right rectal gyrus, right inferior
parietal lobe, and right parahippocampal gyrus—
indicating that ‘‘a bilaterally distributed brain
system is involved in creative performance’’ (p. 217)
although most of the elevated activity is in the right
hemisphere.
These studies provide some evidence that in less creative
people, the right hemisphere is slightly less active.
But it is misleading to say that creativity is ‘‘in’’ the right
hemisphere (also see Feist, 2010; Kaufman et al., 2010);
the aforementioned studies found that high creatives
show patterns of bilateral hemispheric activation. And
as always, one must keep in mind that these differences,
resulting from paired image subtraction, are never greater
than 3% in any single voxel.
Musical Improvisation
A series of intriguing experiments have recently been
conducted with trained musicians engaged in a variety
of musical tasks. In the first such study, 11 professional
pianists were imaged with fMRI as they improvised a
simple melody, based on an 8-note melody that was
displayed to them, and then were imaged as they
reproduced the improvised melody (Bengtsson,
Csikszentmihalyi, & Ullen, 2007). This resulted in
‘‘improvise minus reproduce’’ subtraction images. Then,
the pianists freely improvised but without memorizing
and reproducing the improvisation, and this resulting
in ‘‘freeimp minus rest’’ subtraction images. Their keyboard
had one octave of 12 keys (white and black),
and the musicians could hear what they were playing
through scanner-safe headphones.
When the averaged image of a brain reproducing an
improvised melody was subtracted from the averaged
image of the brain improvising that melody for the first
time (the ‘‘improvise minus reproduce’’ condition), there
were significant brain differences in 14 regions, including:
right DLPFC, pre-SMA, bilaterally in the rostral
portion of the dorsal premotor cortex (PMD); temporal
lobe activations in the left posterior superior temporal
gyrus (STG), and the fusiform gyrus; bilateral occipital
activity in the middle occipital gyrus. Essentially, all
of these areas were also activated in the conjunction
between improv-minus-reproduce and freeimpminus-rest.
The right DLPFC is activated in many other
free choice tasks, including word generation, number
generation, word-stem completion, and sentence completion.
A range of studies show that the DLPFC is centrally
involved in planning and performing novel and
complex behavioral sequences, including language and
thought. Several of the other active areas are also activated
in movement sequence production.
Berkowitz and Ansari (2008) studied 12 classically
trained pianists engaged in four different tasks. The
researchers designed a special five-note keyboard that
the subjects could play with the fingers of one hand, only
moving the fingers and not the hand. The keyboard had
middle C through G, the white keys only. The subjects
listened through scanner-safe headphones.
As with all CN studies, the goal was to identify the
subtle differences between tasks with paired image
subtraction; each task required a slightly different
degree of improvisation. Before the experiment, the subjects
were taught seven different five-note patterns that
were extremely simple: either five presses of the same
key (CCCCC, DDDDD, EEEEE, FFFFF, GGGGG),
an ascending scale (CDEFG), or a descending scale
THE COGNITIVE NEUROSCIENCE OF CREATIVITY 147
(GFEDC). In their first task, the pianists played any of
the 5-note patterns, in any order they chose. Thus they
had to make a decision every five notes, resulting in a
rather small degree of melodic improvisation. In the
second melodic improvisation task, the pianists continuously
invented five-note melodies—thus making a
decision every note.
Both of these tasks were performed with or without a
metronome that clicked two beats each second. With the
metronome, subjects were told to play only one note per
click. With no metronome, subjects were told to improvise
rhythmically as well as melodically. This design
allowed the researchers to isolate the brain regions associated
with three different activities: rhythmic improvisation
alone, melodic improvisation alone, and both
types of improvisation combined.
Using the cognitive conjunction technique, the conjunction
of the brain images during the two melodic
improvisation tasks was associated with increased
neural activity in the dorsal premotor cortex (dPMC),
anterior cingulate cortex (ACC), and inferior frontal
gyrus=ventral premotor cortex (IFG=vPMC), all in the
left hemisphere (which was expected, since the task
was performed with the right hand).
As with all studies discussed, these brain regions are
the same ones that are used in a wide variety of everyday
cognitive tasks. The dPMC is involved in a wide variety
of motor tasks, including selection and performance of
movements. The ACC, which is implicated in remote
associations, is involved in many cognitive tasks, including
unrehearsed movements, decision making, voluntary
selection, and willed action. The third region includes
part of Broca’s area, typically associated with language
production and understanding, or more generally
with producing and processing sequential auditory
information.
In sum, improvisation involves brain regions that are
involved in the generation and comprehension of
sequences, making decisions among competing alternatives,
and the creation of a plan for the motor execution
of that sequence. These are domain-general brain
regions, suggesting a role for domain-general mental processes
in creativity.
Limb and Braun (2008) used a more realistic improvisational
musical task. They compared two conditions,
using six trained jazz musicians: (a) Subjects played
a previously memorized jazz composition, while
accompanied by a jazz quartet they could hear through
headphones; (b) subjects improvised over the same
chord sequence, while hearing the same accompaniment
through their headphones. Their keyboard had 35
full-sized keys.
They observed activation in the same three brain
regions as Berkowitz and Ansari (2008). But because
the tasks were so much more complex, they found
changes in activity in over forty brain regions. Many
of these are likely to be not specific to music or to improvisation,
but related to general cognitive activity such as
attention, working memory, and task complexity. One
particularly interesting result was a decrease in activity
in almost all of the lateral prefrontal cortices, particularly
in the lateral orbital prefrontal cortex (LOFC)
and the dorsolateral prefrontal cortex (DLPFC), suggesting
inhibition of regions involved in monitoring
and correction. They observed increased activation in
superior and middle temporal gyri (STG and MTG)
and ACC, as well as many other areas. They observed
increased activity in the medial prefrontal cortex
(MPFC), which has been associated with autobiographical
narrative. The decreased activity is in the regions
which are associated with consciously monitoring
goal-directed behaviors.
As Berkowitz (2010) pointed out, ‘‘the brain imaging
results from these two studies correlate quite well with
artists’ experiences of improvisation’’ (p. 144). Musical
improvisation involves brain regions associated with
the skills that underlie everyday improvisational action:
selection and performance of movement, decision making,
language processing and sequential auditory information,
and inhibition of monitoring. However, note
that no brain areas are uniquely associated with improvisation;
all of these brain areas are involved in a wide
variety of cognitive tasks, many of them not considered
to be creative.
Differences with Training
Brain imaging studies have found that people with musical
training process music differently, people with artistic
training process art differently, and people with
dance training process dance differently.
Music training. When listening to music and when
generating music, the brains of trained musicians show
different patterns of activation than nonmusicians.
Berkowitz and Ansari (2010) compared trained musicians
with nonmusicians in the simple five-note improvisational
task described earlier. The key difference was
that the musicians deactivated the right temporoparietal
junction (rTPJ) while the nonmusicians did not. The
authors cite evidence that this region is engaged in
bottom-up stimulus driven processing (although it is
involved in many other tasks as well); deactivation of
this region seems to inhibit attentional shifts toward
task-irrelevant stimuli during top-down, goal-driven
behavior. Thus, musical training seems to result in a
shift toward inhibition of stimulus-driven attention,
allowing for a more goal-directed performance state.
Schlaug (2006) demonstrated that trained musicians
148 SAWYER
process a pitch memory task using different brain
regions than non musicians.
Art training. Bhattacharya and Petsche (2005) used
EEG to compare artists (MFA graduates from the
Academy of Fine Arts in Vienna) and nonartists mentally
composing a drawing (while staring at a white
wall), and found significantly different patterns of functional
cooperation between cortical regions. Comparing
the tasks to rest, the artists showed stronger short- and
long-range delta band synchronization, whereas the
non-artists showed enhanced short-range beta and
gamma band synchronization primarily in frontal
regions; comparing the two groups during the task, the
artists showed stronger delta band synchronization
and alpha band desychronization, and strong RH dominance
in synchronization. These findings correspond to
the more general finding that well-mastered tasks typically
show greater coherence or synchronization across
cortical regions. For example, expert chess players show
stronger delta band coherence than novices when anticipating
chess movements. The researchers interpreted
these differences as due to more advance long-term visual
memory, and extensive top-down processing.
Dance training. Fink et al. (2009) compared expert
professional dancers with beginning dancers who had
just completed a first class in basic dance. They asked
them to wear EEG electrodes while they mentally performed
either an improvised dance, or a classic waltz.
They also did an Alternative Uses test (tin, brick, sock,
ballpoint pen), and during these tests, the dancers
showed stronger alpha synchronization in post-parietal
brain regions. During improvisation imagery, dancers
showed more RH alpha synchronization than the
novices, while there were no differences with the waltz.
The researchers interpreted the increased alpha synchronization
as inhibition of processes not directly relevant,
or of top-down control.
These studies that compare experts with nonexperts
provide support for an intriguing hypothesis: Neuronal
activation patterns may change in response to experience
and learning (e.g., Posner & Raichle, 1994;
Reiterer, Pereda, & Bhattacharya, 2009). However,
causation has not yet been proven; it might be the case
that people who become experts in a domain have different
brains to start with. Longitudinal studies will be
required to determine causation.
PART 3: CRITICAL REVIEW
Although CN is still in its infancy, it has already contributed
to the understanding of creativity.
. The entire brain is active when people are engaged
in creative tasks. The studies reported herein tend
to obscure the diffuse nature of most brain activity,
due to the paired image subtraction technique. At
issue in all of these studies is not ‘‘which areas of
the brain are active’’ (the answer is almost always
‘‘most of them’’) but rather a more specific technical
question: which brain areas display statistically
significant differential activation across two tasks?
. These studies generally do not reveal differential
activation across the two hemispheres; left and right
hemispheres are equally activated in most creative
tasks (although with some minor variations). Thus
CN studies confirm what is already known from
previous methodologies (see Runco, 2007; Sawyer,
2006): there is no evidence for the popular belief
that creativity is located in the right hemisphere of
the brain. Many regions of the brain, in both
hemispheres, are active during creative tasks.
. When people are engaged in creative tasks, the
same brain areas are active that are active in many
everyday tasks—even in ordinary tasks that people
do not associate with creativity. Every normal,
healthy human being engages in these brain processes;
they are required for everyday functioning.
These findings are consistent with research in the
creative cognitive approach (Smith, Ward, &
Finke, 1995) that has found that creativity involves
a wide variety of everyday cognitive abilities.
. Mind wandering is likely to be the brain state that
corresponds to incubation. But it seems to occur in
very brief periods that could be thought of as micro
incubation, and it occurs even when a person is
unaware it is happening.
. The importance of association, and in particular,
distant association, is confirmed by these studies.
. The role of unconscious progress toward insight is
partially confirmed, consistent with the existence of
an incubation effect, and with studies of intuition
(e.g., Bowers et al., 1995).
. The importance of domain specific expertise is confirmed.
Extensive training in a domain is associated
with different patterns of brain activation.
These findings suggest that CN has much promise as
a methodology for creativity research. However, the
methodology has several limitations:
1. For the most part, what CN has discovered are
facts that are largely already known from the
classic experimental methods of cognitive psychology
(Carey, 2006). For example, in the
1970s, these experiments had discovered that
verbal and visual information were represented
differently in the brain; so when brain imaging
THE COGNITIVE NEUROSCIENCE OF CREATIVITY 149
shows that these two types of information result
in different patterns of neuronal activation, no
one is surprised. To take another example,
researchers also discovered that implicit memory
and explicit memory were distinct, long before
CN identified the associated brain regions.
2. Neuroscientists agree that all cognitive function
involves many regions of the brain. The colorful
images that are displayed in journal articles, with
small colored dots showing ‘‘the locations’’ associated
with a task, result from averaging and subtraction;
what is not shown in that image are the
many brain regions that are active in both conditions,
before the subtraction (see Figure 3).
An overly strong emphasis on localization has
the potential to distract from the reality that
what goes on in the brain is diffuse and distributed.
These studies cannot be used to make claims
that creativity is ‘‘in the right brain,’’ for
example; even if heightened activation had been
observed in the RH—which it generally is not—
the heightened activation is never more than 3%
above the activation level of that region during
the comparison task or the baseline state.
3. It is hard to use brain imaging studies to make
claims about causation, because an area may be
activated during a task but not play a critical role
in performing the task; rather, it might be listening
to other brain areas that provide the critical
computations.
4. These findings result from averaging across many
trials, typically about 50 per experimental condition.
The voxels that show statistically significant
elevation after averaging and subtraction
do not necessarily display elevated activation in
every trial. So it is always incorrect to interpret
these studies as showing that ‘‘creativity is located
in the ACC’’ or even ‘‘the right hemisphere’’
although a quick read through a particular study
might misleadingly give that impression. A more
guarded conclusion is always in order for all CN
studies, to wit: Over large numbers of trials and
over many subjects, on average, we observe a statistically
significant difference (generally 3% or
less) in brain activation across two tasks.
5. The scanners used for PET and fMRI are large
and expensive. People have to lie down and
remain completely still, while they listen to the
extremely loud whirring of the scanner’s motor.
Bodily movements, even quite small ones, activate
large regions of the brain and overwhelm
the signal associated with the mental processing
of interest. So there is no hope of being able to
study people engaged in normal activities in their
everyday contexts.
This makes the study of collaborative groups,
and of people engaged in creative work with
external representations—for example, the problem
finding process used by many painters
(Getzels & Csikszentmihalyi, 1976)—impossible
to study.
6. The standard definition of creativity as ‘‘the general
of something novel and useful’’ is not used in
any of these studies. It is probably not possible to
operationalize ‘‘usefulness’’ in CN terms; and
‘‘novelty’’ is not sufficient to define creativity,
because the generation of novel behaviors and
thoughts is a component of almost all cognitive
activity (e.g., almost every sentence people speak
during a day is one that they have never spoken
before). So at present, there is a fairly large gap
between creativity as operationalized in these
experiments, and creativity as operationalized
by most creativity researchers.
7. Finally, the approach is, by nature, reductionist;
the mental processes studied are quite small, compared
to the mental complexity and long-term
cognitive processes that are associated with
real-world creativity. The reductionist approach
of CN is appropriate for many brain functions—
for example, perceptual systems like vision are
processed in a fairly regular way through a specific
set of neurons. But with higher cognitive functions
such as problem solving, language, decision making,
memory, and creativity, one’s thoughts and
behaviors are emergent complex phenomena: they
involve many distinct neural groups, scattered
throughout the brain (Bechtel & Richardson,
1993). Any meaningful creative product is likely
to have behind it tens or even hundreds of these
brief mental events. Imagine a writer composing
a poem; each selection of a single word is likely
to result from multiple events of association and
insight. And after a first draft is completed, the
process of editing and revising will involve hundreds
more such mental events. Thus in a creator’s
life, these mental moments occur over a long period
of time, where the mind’s processing is interspersed
with solitary interactions with external
representations, and social interactions with
others working in the area. These two latter processes
can’t be localized in the brain—the role of
external representations requires attention to
externalization processes, and the role of social
interaction requires study of the interactional
dynamics of groups.
Cognitive neuroscience has the potential to contribute
a valuable perspective to creativity researchers.
These studies show that creativity is not localized in
150 SAWYER
one brain region; rather, creativity emerges from a complex
network of neurons firing throughout the brain.
These findings paint a complex picture of the relationship
between brain science and creativity. Future technology
can be expected to give us better and better
images; however, there are three limitations of these
methods that cannot be overcome.
1. The spatial resolution of EEG will never get
much higher due to the diffuse nature of the electric
fields in the skull.
2. The temporal resolution of PET and fMRI can
never be increased dramatically because the
hemodynamic response is so sluggish. It takes
15 seconds for the brain’s blood flow to return
to its resting state, so it will never be possible
to do trials more often than every 15 seconds.
3. The spatial resolution of fMRI will get much
higher with increasing magnetic field strength.
Logothetis (2008) predicted that slice thicknesses
will decline to one-half of a millimeter, with voxel
sizes two or three orders of magnitude smaller
than at present (p. 871). But an increase in spatial
resolution will not advance CN very much, due
to the normal variations among brains that
require averaging and smoothing, and also due
to the fact that BOLD is itself not localized to
one narrow location.
RECOMMENDATIONS FOR FUTURE STUDY
The overall message emerging from these studies is that
creativity is not dependent on any particular mental process
or brain region. Over twenty different brain regions
are implicated in the studies reviewed here. Partly, these
findings reflect a lack of consistency in how creativity is
operationalized. The studies use a wide range of tasks,
and each of these tasks is likely to differentially activate
the brain. Thus, these studies suggest that what people
think of as creativity involves a wide variety of cognitive
processes, each of which results in distinct patterns of
brain activation.
Some of the definitions of creativity used by creativity
researchers—such as novelty plus appropriateness—are
not suitable for CN methodologies. Another widely used
concept—divergent thinking—does not seem to be localizable
in the brain. Association (in the convergent
thinking task of the RAT) is partially localizable, but
not completely. For creativity research to effectively utilize
CN methodologies, it is likely to be necessary to
decompose the concept of creativity into more specific,
more experimentally tractable constructs. Existing
CN studies of cognition, memory, attention, etc. can
potentially provide creativity researchers with new ways
to think about how to subdivide the concept of creativity
into a set of brain-based constructs. For example, it
might be the case that there are many different neural
pathways that result in behavior that is characterized
as creative—just as there are many different neural pathways
for what is called memory, depending on whether it
is implicit or explicit, imagery or verbal, etc. As such,
collaboration with cognitive neuroscientists could
contribute to theory development in creativity research.
However, there is a risk that could limit the potential
for links between CN and creativity research: once creative
behavior is decomposed into experimentally tractable
cognitive functions, these component functions
may turn out to be the same functions involved in behavior
that is generally thought to be noncreative, routine,
or everyday. It may be the case that the construct of
creativity simply cannot be defined in terms of cognitive
events that occur in one minute or less.
The previous review suggests several promising lines
of future study:
1. More studies should use EEG to identify the temporal
process of creative thought. EEG should
more frequently be used in conjunction with
fMRI, as done with the RAT triplet studies of
(Bowden et al., 2005).
2. Studies that focus on spatial location can provide
intriguing insight into the brain’s cognitive structure;
but these studies should not be interpreted
as identifying ‘‘the location’’ for a particular cognitive
activity. Rather, the promise of high spatial
resolution technologies such as fMRI is that
researchers can identify similar patterns of elevated
neuronal activity in a variety of tasks, and
potentially identify cognitive processes and
mechanisms not formerly thought to be associated
with creativity. Identifying such links will
require working with experienced cognitive neuroscientists
who are aware of the full range of
studies, not just studies focused on creative
thought.
3. Mind wandering studies could be applied more
specifically to incubation and creativity.
4. Most studies to date have focused on the neocortical
regions associated with higher level cognitive
processing. Future studies might extend the
focus to brain regions associated with emotion,
motivation, and the subjective experience of flow,
because other methodologies have shown that
these are associated with creativity.
5. There is much potential in studies of creativity in
specific domains, both with experts and nonexperts.
Creativity research has concluded that
creative performance is due to both domain
THE COGNITIVE NEUROSCIENCE OF CREATIVITY 151
general and domain specific components, with
the majority due to domain specific components
(Kaufman & Baer, 2005). If so, the majority of
CN studies should likewise focus on creativity
in the domains. This has not been the case to
date.
6. A broad range of creativity research shows the
important role of domain-specific knowledge,
acquired over time and represented in long term
memory (LTM), thought to be primarily in the
temporal, occipital, and parietal lobes. Thus,
future studies should explore the role of LTM
in creative thought, and the complementary relationships
between LTM and working memory
(cf. Dietrich, 2004).
7. Domain-specific expertise also involves automatized
routines, developed as a result of training
and expertise. Future studies should examine
the role of automatized routines in creativity;
these are thought to be represented in the cerebellum
(see Vandervert, Schimpf, & Liu, 2007).
Cognitive neuroscience has made great strides in a
short period of time, and it has potential to enhance
the understanding of creative cognition. These experiments
demonstrate the benefits of collaboration between
creativity researchers and cognitive neuroscientists.
Continued interdisciplinary collaboration has the potential
to further advance understanding of the mental processes
and structures associated with creative thought
and behavior.
REFERENCES
Andrews-Hanna, J. R., Reidler, J. S., Huang, C., & Buckner, R. L.
(2010). Evidence for the default network’s role in spontaneous cognition.
Journal of Neurophysiology, 104, 322–335.
Bechtel, W., & Richardson, R. C. (1993). Discovering complexity:
Decomposition and localization as strategies in scientific research.
Princeton, NJ: Princeton University Press.
Beeftink, F., van Eerde, W., & Rutte, C. G. (2008). The effect of interruptions
and breaks on insight and impasses: Do you need a break
right now? Creativity Research Journal, 20, 358–364.
Bengtsson, S. L., Csikszentmihalyi, M., & Ullen, F. (2007). Cortical
regions involved in the generation of musical structures during
improvisation in pianists. Journal of Cognitive Neuroscience, 19, 830.
Berkowitz, A. L. (2010). The improvising mind: Cognition and creativity
in the musical moment. Oxford, UK: Oxford University Press.
Berkowitz, A. L., & Ansari, D. (2008). Generation of novel motor
sequences: The neural correlates of musical improvisation. Neuroimage,
41, 535–543.
Bhattacharya, J., & Petsche, H. (2005). Drawing on mind’s canvas:
Differences in cortical integration patterns between artists and
non-artists. Human Brain Mapping, 26, 1–14.
Bowden, E. M., & Jung-Beeman, M. (2003a). Aha! insight experience
correlates with solution activation in the right hemisphere. Psychonomic
Bulletin and Review, 10, 730–737.
Bowden, E. M., & Jung-Beeman, M. (2003b). One hundred forty-four
Compound Remote Associate Problems: Short insight-like problems
with one-word solutions. Behavioral Research, Methods,
Instruments, and Computers, 35, 634–639.
Bowden, E. M., Jung-Beeman, M., Fleck, J., & Kounios, J. (2005).
New approaches to demystifying insight. Trends in Cognitive
Science, 9(7), 321–328.
Bowers, K. S., Farvolden, P., & Mermigis, L. (1995). Intuitive antecedents
of insight. In S. M. Smith, T. B. Ward, & R. A. Finke (Eds.),
The creative cognition approach (pp. 27–51). Cambridge, MA: MIT
Press.
Bowers, K. S., Regehr, G., Balthazard, C., & Parker, K. (1990). Intuition
in the context of discovery. Cognitive Psychology, 22, 72–110.
Carey, B. (2006, February 5). Searching for the person in the brain.
New York Times, pp. WK 1, 4.
Carlsson, I., Wendt, P. E., & Risberg, J. (2000). On the neurobiology
of creativity: Differences in frontal activity between high and low
creative subjects. Neuropsychologia, 38, 873–885.
Cha´vez-Eakle, R. A. (2007). Creativity, DNA, and cerebral blood
flow. In C. Martindale, P. Locher, & V. M. Petrov (Eds.), Evolutionary
and neurocognitive approaches to aesthetics, creativity, and the
arts (pp. 209–224). Amityville, NJ: Baywood.
Christoff, K., Gordon, A. M., Smallwood, J., Smith, R., & Schooler, J.
W. (2009). Experience sampling during fMRI reveals default network
and executive systems contribution to mind wandering. Proceedings
of the National Academy of Sciences, 106, 8719–8724.
Cunningham, J. B., MacGregor, J. N., Gibb, J., & Haar, J. (2009).
Categories of insight and their correlates: An exploration of relationships
among classic-type insight problems, rebus puzzles,
remote associates, and esoteric analogies. Journal of Creative
Behavior, 43, 262–280.
Dietrich, A. (2004). The cognitive neuroscience of creativity. Psychonomic
Bulletin and Review, 11, 1011–1026.
Duncker, K. (1926). A qualitative (experimental and theoretical) study
of productive thinking (solving of comprehensive problems). Pedagogical
Seminary and Journal of Genetic Psychology, 33, 642–708.
Ellwood, S., Pallier, G., Snyder, A., & Gallate, J. (2009). The incubation
effect: Hatching a solution? Creativity Research Journal,
21, 6–14.
Feist, G. J. (2010). The function of personality in creativity: The nature
and nurture of the creative personality. In J. C. Kaufman & R. J.
Sternberg (Eds.), The Cambridge handbook of creativity (pp.
113–130). Cambridge, UK: Cambridge University Press.
Fink, A., Benedek, M., Grabner, R. H., Staudt, B., & Neubauer, A. C.
(2007). Creativity meets neuroscience: Experimental tasks for the
neuroscientific study of creative thinking. Methods, 42, 68–76.
Fink, A., Grabner, R. H., Benedek, M., Reishofer, G., Hauswirth, V.,
Fally, M., et al. (2009). The creative brain: Investigation of brain
activity during creative problem solving by means of EEG and
fMRI. Human Brain Mapping, 30, 734–748.
Gazzaniga, M., Ivry, R. B., & Mangun, G. R. (2002). Cognitive neuroscience:
The biology of the mind (2nd ed.). New York: Norton.
Getzels, J. W., & Csikszentmihalyi, M. (1976). The creative vision. New
York: Wiley.
Gick, M. L., & Lockhart, R. S. (1995). Cognitive and affective components
of insight. In R. J. Sternberg & J. E. Davidson (Eds.), The nature
of insight (pp. 197–228). Cambridge, MA: MIT Press.
Goel, V., & Vartanian, O. (2005). Disassociating the roles of right ventral
lateral and dorsal lateral prefrontal cortex in generation and
maintenance of hypotheses in set-shift problems. Cerebral Cortex,
15, 1170–1177.
Hadamard, J. (1945). The psychology of invention in the mathematical
field. Princeton, NJ: Princeton University Press.
Hotz, R. L. (2009, Friday, June 19). A wandering mind heads straight
toward insight. The Wall Street Journal, p. A11.
152 SAWYER
Howard-Jones, P. A., Blakemore, S.-J., Samuel, E. A., Summers, I. R.,
& Claxton, G. (2005). Semantic divergence and creative story
generation: An fMRI investigation. Cognitive Brain Research, 25,
240–250.
Huettel, S. A., Song, A. W., & McCarthy, G. (2009). Functional magnetic
resonance imaging (2nd ed.). Sunderland, MA: Sinauer Associates.
James, W. (1880). Great men, great thoughts, and the environment.
Atlantic Monthly, 46(276), 441–459.
Jung-Beeman, M. (2005). Bilateral brain processes for comprehending
natural language. Trends in Cognitive Science, 9, 512–518.
Jung-Beeman, M., Bowden, E. M., & Haberman, J. (2002, April). The
Aha! experience and semantic activation in the cerebral hemispheres.
Paper presented at the 9th Annual Meeting of the Cognitive
Neuroscience Society, San Francisco, CA.
Jung-Beeman, M., Bowden, E. M., Haberman, J., Frymaire, J. L.,
Arambel-Lui, S., Greenblatt, R., et al. (2004). Neural activity when
people solve verbal problems with insight. PLoS Biology, 2(4),
0500–0510.
Kane, M. J., Brown, L. H., McVay, J. C., Silvia, P. J., Myin-Germeys,
I., & Kwapil, T. R. (2007). For whom the mind wanders, and when.
Psychological Science, 18(7), 614–621.
Kaplan, C. A., & Simon, H. A. (1990). In search of insight. Cognitive
Psychology, 22, 374–419.
Kaufman, A. B., Kornilov, S. A., Bristol, A. S., Tan, M., &
Grigorenko, E. L. (2010). The neurobiological foundation of
creative cognition. In J. C. Kaufman & R. J. Sternberg (Eds.),
The Cambridge handbook of creativity (pp. 216–232). Cambridge,
UK: Cambridge University Press.
Kaufman, J. C., & Baer, J. (Eds.). (2005). Creativity across domains:
Faces of the muse. Mahwah, NJ: Lawrence Erlbaum Associates.
Klinger, E. (2009). Daydreaming and fantasizing: Thought flow and
motivation. In K. D. Markman, W. M. P. Klein, & J. A. Suhr
(Eds.), Handbook of imagination and mental simulation (pp.
225–239). New York: Psychology Press.
Kohn, N., & Smith, S. M. (2009). Partly versus completely out of your
mind: Effects of incubation and distraction on resolving fixation.
Journal of Creative Behavior, 43, 102–118.
Kolb, B., & Whishaw, I. Q. (2003). Fundamentals of human neuropsychology
(5th edition). New York: Worth Publishers.
Kotovsky, K. (2003). Problem solving: Large=small, hard=easy,
conscious=nonconscious, problem-space=problem-solver. In J. E.
Davidson & R. J. Sternberg (Eds.), The psychology of problem solving
(pp. 373–383). New York: Cambridge.
Kounios, J., Fleck, J., Green, D. L., Payne, L., Stevenson, J. L.,
Bowden, E. M., et al. (2008). The origins of insight in resting-state
brain activity. Neuropsychologia, 46, 281–291.
Kounios, J., Frymaire, J. L., Bowden, E. M., Fleck, J., Subramaniam,
K., Parrish, T. B., et al. (2006). The prepared mind: Neural activity
prior to problem presentation predicts subsequent solution by
sudden insight. Psychological Science, 17, 882–290.
Limb, C. J., & Braun, A. R. (2008). Neural substrates of spontaneous
musical performance: An fMRI study of jazz improvisation. PLos
One, 3(2), e1679.
Logothetis, N. K. (2008). What we can do and what we cannot do with
fMRI. Nature, 453, 869–878.
Luo, J., & Knoblich, G. (2007). Studying insight problem solving with
neuroscientific methods. Methods, 42, 77–86.
Martindale, C., & Hines, D. (1975). Creativity and cortical activation
during creative, intellectual, and EEG feedback tasks. Biological
Psychology, 3, 71–80.
Mason, M. F., Norton, M. I., Van Horn, J. D., Wegner, D. M.,
Grafton, S. T., & Macrae, N. (2007). Wandering minds: The
default network and stimulus-independent thought. Science, 315,
393–395.
Metcalfe, J. (1986). Feeling of knowing in memory and problem solving.
Journal of Experimental Psychology: Learning, Memory, and
Cognition, 12, 288–294.
Metcalfe, J., & Wiebe, D. (1987). Intuition in insight and noninsight
problem solving. Memory and Cognition, 15, 238–246.
Ohlsson, S. (1992). Information-processing explanations of insight and
related phenomena. In M. T. Keane & K. J. Gilhooly (Eds.),
Advances in the psychology of thinking (Vol. 1, pp. 1–44). London:
Harvester Wheatsheaf.
Patrick, A. S. (1986). The role of ability in creative ‘‘incubation.’’
Personality and Individual Differences, 7(2), 169–174.
Perkins, D. N. (1981). The mind’s best work. Cambridge, MA: Harvard
University Press.
Poincare´, H. (1913=1982). The foundations of science. Washington,
DC: University Press of America.
Posner, M. I., & Raichle, M. E. (1994). Images of mind. New York:
Scientific American Library.
Purves, D., Brannon, E. M., & Cabeza, R. (2008). Principles of cognitive
neuroscience. Sunderland, MA: Sinauer Associates.
Raichle, M. E. (2009). A paradigm shift in functional brain imaging.
Journal of Neuroscience, 29, 12729–12734.
Reiterer, S., Pereda, E., & Bhattacharya, J. (2009). Measuring second
language proficiency with EEG synchronization: How functional
cortical networks and hemispheric involvement differ as a function
of proficiency in second language speakers. Second Language
Researcher, 25(1), 77–106.
Runco, M. A. (2007). Creativity: Theories and themes: Research, development,
and practice. Burlington, MA: Elsevier Academic Press.
Sandku¨hler, S., & Bhattacharya, J. (2008). Deconstructing insight:
EEG correlates of insightful problem solving. PLoS One, 3(1),
e1459.
Sawyer, R. K. (2006). Explaining creativity: The science of human innovation.
New York: Oxford.
Schlaug, G. (2006). The brain of musicians: A model for functional
and structural adaptation. Annals of the New York Academy of
Sciences, 930, 281–299.
Schooler, J. W., Ohlsson, S., & Brooks, K. (1993). Thoughts beyond
words: When language overshadows insight. Journal of Experimental
Psychology: General, 122(2), 166–183.
Seger, C. A., Desmond, G. H., Glover, J. D. E., & Gabrieli. (2000).
Functional magnetic resonance imaging evidence for righthemisphere
involvement in unusual semantic relationships.
Neuropsychology, 14, 361–369.
Sheth, B. R., Sandku¨hler, S., & Bhattacharya, J. (2008). Posterior beta
and anterior gamma oscillations predict cognitive insight. Journal of
Cognitive Neuroscience, 21, 1269–1279.
Singer, J. L., & Antrobus, J. S. (1963). A factor-analytic study of
daydreaming and conceptually related cognitive and personality
variables. Perceptual and Motor Skills, 17, 187–209.
Skov, M., & Vartanian, O. (Eds.). (2009). Neuroaesthetics. Amityville,
NY: Baywood.
Smallwood, J., & Schooler, J. W. (2006). The restless mind. Psychological
Bulletin, 132(6), 946–958.
Smith, S. M., Ward, T. B., & Finke, R. A. (Eds.). (1995). The creative
cognition approach. Cambridge, MA: MIT Press.
Srinivasan, N. (2007). Cognitive neuroscience of creativity: EEG based
approaches. Methods, 42, 109–116.
Sternberg, R. J., & Davidson, J. E. (Eds.). (1995). The nature of insight.
Cambridge, MA: MIT Press.
Tierney, J. (2010, June 29). Discovering the virtues of a wandering
mind. The New York Times, p. D1.
Vandervert, L. R., Schimpf, P. H., & Liu, H. (2007). How working
memory and the cerebellum collaborate to produce creativity and
innovation. Creativity Research Journal, 19, 1–18.
THE COGNITIVE NEUROSCIENCE OF CREATIVITY 153
Vartanian, O., & Goel, V. (2007). Neural correlates of creative
cognition. In C. Martindale, P. Locher, & V. M. Petrov
(Eds.), Evolutionary and neurocognitive approaches to aesthetics,
creativity, and the arts (pp. 195–207). Amityville, NJ:
Baywood.
Ward, J. (2006). The student’s guide to cognitive neuroscience. Hove,
UK: Psychology Press.
Weisberg, R. W. (1986). Creativity: Genius and other myths. New York:
W. H. Freeman.
Weisberg, R. W. (1993). Creativity: Beyond the myth of genius. New
York: W. H. Freeman.
Weisberg, R. W., & Alba, J. W. (1981). An examination of the alleged
role of ‘‘fixation’’ in the solution of several ‘‘insight’’ problems.
Journal of Experimental Psychology: General, 110, 169–192.
154 SAWYER