Scribd 3
Scribd 3
Scribd 3
David Rosen, Yongtaek Oh, Brian Erickson, Fengqing (Zoe) Zhang, Youngmoo E.
Kim, John Kounios
PII: S1053-8119(20)30119-1
DOI: https://doi.org/10.1016/j.neuroimage.2020.116632
Reference: YNIMG 116632
Please cite this article as: Rosen, D., Oh, Y., Erickson, B., Zhang, F.(Z.), Kim, Y.E., Kounios, J., Dual-
process contributions to creativity in jazz improvisations: An SPM-EEG study, NeuroImage (2020), doi:
https://doi.org/10.1016/j.neuroimage.2020.116632.
This is a PDF file of an article that has undergone enhancements after acceptance, such as the addition
of a cover page and metadata, and formatting for readability, but it is not yet the definitive version of
record. This version will undergo additional copyediting, typesetting and review before it is published
in its final form, but we are providing this version to give early visibility of the article. Please note that,
during the production process, errors may be discovered which could affect the content, and all legal
disclaimers that apply to the journal pertain.
An SPM-EEG Study
Conflicting theories identify creativity either with frontal-lobe mediated (Type-2) executive
control processes or (Type-1) associative processes that are disinhibited when executive control is
relaxed. Musical (jazz) improvisation is an ecologically valid test-case to distinguish between these
views because relatively slow, deliberate, executive-control processes should not dominate during
high-quality, real-time improvisation. In the present study, jazz guitarists (n = 32) improvised to novel
chord sequences while 64-channel EEGs were recorded. Jazz experts rated each improvisation for
during the performances were analyzed in the scalp-frequency domain using SPM12. Significant
clusters of high-frequency (beta-band and gamma-band) activity were observed when higher-quality
with right temporo-parietal and fronto-polar activity. However, after statistically controlling for
experience (defined as the number of public performances previously given), performance quality was a
function of right-hemisphere, largely right-frontal, activity. These results support the notion that
superior creative production is associated with hypofrontality and right-hemisphere activity thereby
supporting a dual-process model of creativity in which experience influences the balance between
executive and associative processes. This study also highlights the idea that the functional
neuroanatomy of creative production depends on whether creativity is defined in terms of the quality
1. Introduction
Much research over the last decade has focused on the contribution of executive, Type-2
processes to creative cognition. The findings paint a complex picture. For example, the engagement of
frontally-mediated, executive control processes can be beneficial in some domains, such as science, but
deleterious in others, such as art and music (Kaufman et al., 2016; Shi et al., 2017). The phase of the
creative process plays an important role, with creative ideation benefiting from less executive control
and idea evaluation benefiting from more (Beaty et al., 2016, Ellamil et al., 2011). Furthermore,
research has shown that greater executive control enhances the creativity of jazz improvisations for
novices but hinders it for those with more expertise (Rosen et al., 2016; 2017). Such findings suggest
that contribution of executive processing to creative cognition depends on differences among both
The present study sought to further delineate the involvement of executive processing in
creative cognition by isolating and comparing real-time neural correlates of creativity and experience
Most studies that have addressed this issue have measured creativity with standardized
psychometric tests or laboratory-based divergent or convergent thinking tasks. While such tasks are
useful for isolating components of creative cognition, they are of limited ecological validity and domain
generality (Zeng et al., 2011). Alternative approaches that employ complex, naturalistic tasks may
better represent creative cognition as it is manifested in real-world situations (Boccia et al., 2015).
Researchers have recently used jazz improvisation for this purpose (Limb & Braun, 2008; Beaty,
2015; Loui, 2018). Improvisation is an excellent test-case for the executive-control debate because the
complexity and temporal demands of expert real-time improvisation should limit reliance on
deliberate, effortful, Type-2 executive processing (Norgaard et al., 2013; Rosen et al., 2017). For
4
Neural Activity During Jazz Improvisation
example, recent research suggests that jazz improvisation benefits from increased Type-1 processes
and default-mode network activity (Beaty, 2015; Rosen et al., 2017). Furthermore, other research
reports that expert jazz improvisation is associated with a distinctive brain-state characterized by
reduced executive control (Limb & Braun, 2008; Pinho et al., 2014).
Self-reports of eminent jazz musicians corroborate this idea. As legendary jazz trumpeter Miles
Davis eloquently put it, “I’ll play, and tell you what it is later” (Szwed, 2012). Jazz saxophonist Charlie
Parker stated, “You've got to learn your instrument. Then, you practice, practice, practice. And then,
when you finally get up there on the bandstand, forget all that and just wail” (Pugatch, 2006, p. 73).
Such statements support the notion that through rigorous training jazz musicians can learn to inhibit
processes. This frees the performer to consciously think about global aspects of his or her performance,
harmonic and melodic priorities, and the development of motifs from previously played material
(Johnson-Laird, 2002; Norgaard, 2011). Professional musicians undoubtedly require some degree of
executive control; however, the function of this control appears to differ between musicians of greater
As for the relevant neural substrates, the executive-control network, including dorsolateral
prefrontal cortex (DLPFC), dorsal premotor cortex (PMD), and the inferior frontal gyrus (IFG) are
thought to underlie Type-2 cognitive-control mechanisms associated with, for example, the evaluation
phase of the creative process (Beaty et al., 2014; 2016; Chrysikou et al. 2014; Ellamil et al., 2012). In
contrast, Type-1 based creativity is thought to result from relative hypofrontality possibly coupled
with posterior disinhibition (Chrysikou et al. 2014; Erickson et al., 2018). In real-time improvisation,
this pattern of brain activity may be transient (Limb & Braun, 2008), with the magnitude of frontal
deactivation predicted by the experience of the musicians (Pinho et al., 2014). Hypofrontality with
posterior disinhibition can also occur as a trait-like correlate of cognitive style present even during the
5
Neural Activity During Jazz Improvisation
resting-state (Erickson et al, 2018).
Recent behavioral and brain-stimulation research suggests that dual-process theory can
integrate these views of creativity. We up-regulated musicians’ Type-2, executive processes via an
explicit instruction to “be creative” (Rosen et al., 2017) or by anodal (excitatory) transcranial direct-
current stimulation (tDCS) targeting right dorsolateral prefrontal cortex (Rosen et al., 2016). Less-
experienced jazz musicians’ improvisation ratings improved and experts were hindered when Type-2
executive processes were ramped up via instructions or stimulation. Based on these results, we
proposed a dual-process model of creativity in which the balance between Type-1 and Type-2
processes is determined by domain expertise. People with less domain expertise rely on deliberate
Type-2 processing; those with greater domain expertise have automatized key elements of their
creative cognition so that they can relax cognitive control and allow Type-1 associative processing to
dominate.
Previous neuroimaging studies that appear to demonstrate that creative cognition results from
transient hypofrontality support this view. However, these studies have important methodological
limitations. For example, Limb and Braun (2008) compared brain activity during improvisation to a
condition in which a melody is played from memory. This may not be a satisfactory control condition
because memory recall recruits the frontal lobe (Wheeler et al., 1997). Thus, apparent frontal
deactivation during improvisation may actually result from increased frontal activation during the
baseline condition of performance from memory. Furthermore, Pinho et al. (2014) examined only
between-subject differences in expertise and not the within-subject variability in creativity and brain
activation that would show that participants’ hypofrontality is transient rather than persistent. To
circumvent these limitations, the present work examined brain activity associated with jazz
given.
1990) as a measure of musicians’ subjective phenomenological experience when improvising. Flow has
been proposed as a positive mental state in which an individual is fully immersed in a task and
experiences energized focus, enjoyment of the process, and total absorption. Although a link between
flow, music improvisation, and peak creative performance has been suggested, there is a lack of
Another variable that we examined was the impact of explicit instructions to be creative. Such
explicit instructions have increased rated creativity in a variety of tasks, possibly by increasing Type-2
Finally, it should be noted that most neuroimaging studies of musical improvisation have used
functional magnetic resonance imaging (fMRI). Although fMRI has excellent spatial resolution, it
requires musicians – almost always pianists – to lie supine in a noisy scanner while using a mirror
above them to view a single hand playing on a small keyboard. These are not trivial constraints.
Pressing (1988) and Goldman (2013) emphasize the importance of embodied cognition and motor
memory in their discussions of jazz improvisation proficiency. Lying down alters spatial and visual
perception, motor responses, and resting-state EEG (Thibault et al., 2014). Because EEG does not
In summary, the present study examined the neural correlates of creative quality and domain
musicians of greater and lesser experience during jazz guitar improvisations. Expert judges later rated
differences between lower-quality and higher-quality performances and between performers of less or
7
Neural Activity During Jazz Improvisation
greater experience. Other analyses examined flow and other behavioral and individual-differences
variables as predictors of improvisation quality. The results of these analyses indicate that
improvisation quality and level of experience may be explained by three axes of neural activity:
anterior versus posterior, dorsal versus ventral, and left versus right hemisphere. The findings also
shed light on the neural substrates of creativity defined in terms of products (i.e., quality of
improvisation) versus processes (i.e., novel and controlled versus familiar and automatic).
2.1. Participants
This study was approved and followed all necessary requirements set forth by Drexel
University’s Institutional Review Board. Thirty-two jazz guitarists (1 female) participated after
signing a consent form. They received $50 compensation for a session that lasted approximately 90
minutes. Their ages ranged from 18–55 (M = 27.9, SD = 9.38), and their musical training ranged from
Edinburgh Handedness inventory (Oldfield, 1971). They reported no history of neurological disorders
or severe head trauma, substance abuse or dependence, current treatment with mood stabilizing
medications, or any severe hearing impairments. They did not have unusually thick hair or braids that
would have interfered with EEG recordings. Nineteen of the guitarists reported that jazz was their
primary performance genre. Inclusion in the study required guitarists to have performed and
improvised in a live jazz setting at least three times and be able to improvise to novel chord sequences
depicted in jazz notation on a lead sheet. Guitarists ranged in experience from 6 to 1500 live jazz
performances (M = 344.88, SD = 481.49) and included students from local university jazz programs,
professional jazz guitarists, jazz instructors, and jazz novices. Seven additional musicians were tested
but were excluded from data analyses: 2 were eliminated because the guitar was not their primary
8
Neural Activity During Jazz Improvisation
instrument; 2 failed to display sufficient musical competency; 2 were excluded from analyses due to
Our measure of jazz experience, participants’ number of live jazz performances, ranged over
more than 2 orders of magnitude with a skewed distribution (skew = 1.52). We therefore applied the
natural logarithmic transformation to the number of jazz gigs. The power law of practice stipulates
that skill increases logarithmically. Empirical evidence shows that improvement with practice is linear
in a log-log space (Newell & Rosenbloom, 1981). For example, a musician’s second performance gives
him or her twice as much experience over the first, but the 501st performance is only a tiny increase
over the 500th. A secondary motivation for the logarithmic transformation is to improve model fit
optimization for wide ranges of data with substantial skew (Zumel et al., 2014).
Four jazz experts were recruited to judge the improvisations. These judges included a jazz
saxophonist and university jazz instructor, two jazz guitarists who are also university instructors, and
one jazz guitarist who is a private instructor. Judges had a minimum of 25 years of jazz performance
experience, and 2 of the 4 judges had over 40 years of experience. Judges were compensated $300 for
After each participant reviewed and completed his or her consent form, he or she was fitted
with an EEG electrode cap, and impedances were checked and adjusted to below 15 kΩ. Once the setup
was complete, resting-state EEGs (not analyzed here) were collected in 4 two-minute blocks,
alternating between the eyes-open and eyes-closed conditions. A music stand containing a binder of
jazz lead sheets and experiment instructions (see supplementary materials) was positioned to minimize
head movement. Figure 1 depicts the experimental environment, as guitarists improvised while
recording EEG.
9
Neural Activity During Jazz Improvisation
Each guitarist performed an 8-minute jazz improvisation warmup exercise while viewing their
EEGs on a computer monitor in real-time. As guitarists played, the lead researcher provided feedback
to participants, instructing them as to how and when excessive movement and other artifacts are
produced. The purpose of this warmup was threefold: to have guitarists become accustomed to
performance with minimal movement; to understand what types of movement would distort the EEG
data; and to practice improvising with the jazz accompaniment at a comfortable volume.
The improvisation task was programmed using E-Prime 2.0 (Psychology Software Tools,
Sharpsburg, PA). All music and auditory stimuli were recorded and delivered using Logic Pro v.10.3.1
(Apple Inc., Cupertino, CA) digital audio workstation via the M-Audio Fast Track Pro USB Interface
(Cumberland, RI) and studio monitors. The jazz accompaniments included piano, bass, and drums and
were created through iReal Pro for Mac OS X v.7.0.1 (Technimo, New York City, NY), a practice tool
with a full rhythm section for any properly formatted jazz chart (see supplementary materials for
10
Neural Activity During Jazz Improvisation
information about the audio recordings). The chord sequences were written with assistance from a
professional jazz bassist/professor and a professional jazz violinist/professor. The novel lead sheets
were composed with the goal of creating unique, 16-bar sequences that were of approximately equal
difficulty while incorporating some familiar jazz patterns and vocabulary. All songs were set to have a
tempo of 144 beats per minute (i.e., medium swing). Each take consisted of 4 rotations through the
Before the improvisation task, participants were presented with standard instructions to
“Improvise with the music as you normally would as a soloist in a jazz setting.” All instructions were
presented visually and auditorily. Prior to each take, participants had 15 s to examine the chord
sequence. The same chord sequences in the same order were used for all participants. The first song
was considered practice and was not included in subsequent analyses. In an attempt to replicate a
previous study of explicit creativity instructions (Rosen et al., 2017), a staggered baseline design was
used to determine when explicit instructions to “be creative” would be given: “Now, for the following
songs, I want you to try to improvise even more creatively than your past performances. Creativity
should be at the forefront of your mind above and beyond anything else. Based on your experiences
and intuitions as a jazz musician, please try to perform as creatively as possible from this point
forward.”
Each guitarist performed seven improvisations. Upon completion of the improvisation task,
they filled out the Core Flow State Scale (C FSS; Martin & Jackson, 2008) for each of the songs; the
order was counterbalanced. Then, musicians responded to a more detailed demographic survey which
contained questions about their experience during the experiment, perceived complexity and
After all 192 improvisations were recorded, each improvisation was mixed and normalized to
ensure that the guitar and accompaniment had comparable volumes across all subjects and takes.
11
Neural Activity During Jazz Improvisation
Guitarists used their own guitars for the study, so the guitar timbres varied across musicians.
Performances were pseudo-randomized for judging with the constraint that the same musician could
not be heard twice consecutively or more than twice within a single judging block. The judges rated
(Amabile, 1982), judges rated the improvisations on a 7-point Likert scale for creativity, aesthetic
appeal, and technical proficiency. The judges recruited for this study were unaware of the full extent of
the experimental design and research goals. Similar to the participants, judges were asked to utilize
their own expertise in jazz to determine the criterion for their ratings.
Expert ratings have been used in hundreds of creativity studies. The CAT is based on
evaluations of actual creative products, is not dependent on any particular theory of creativity, and
uses the same method for assessing creativity as most domains in the real world (Baer, 2010). The
CAT tasks experts in a domain with rating creative products relative to one another. Interrater
reliability was excellent for all rating scales. The intraclass correlation coefficient (ICC) assessed
reliability for judges’ ratings of creativity (CR; ICC = 0.83, N = 4), technical proficiency (TP; ICC =
0.87, N = 4), and aesthetic appeal (AA; ICC = 0.85, N = 4). Reliability was calculated in SPSS v.24.0.0
(SPSS Inc., Chicago, Il) such that values were computed for consistency where systematic differences
Similar to our previous studies of jazz improvisation, the rating scales were highly correlated.
Strong positive correlations (p < .001) were found for all scale-type comparisons: CR and AA, r(190)=
.97; CR and TP, r(190) = .95, and AA and TP, r(190) = .95. Because these correlations are so high, a
composite quality score was calculated for each improvisation take across all scales and judges. The
12
Neural Activity During Jazz Improvisation
quality rating was our primary outcome measure used for all subsequent analyses. Quality ratings
ranged from 1.16 to 6.33 (M = 4.08, SD = 1.36). No improvisations were rated as ‘1’ or ‘7’; therefore,
2.4. Electroencephalograms
EEGs were recorded with 64 Ag/AgCL active-electrodes embedded in an elastic cap (Brain
Products, Morrisville, NC) with a digitally linked mastoid reference and an electrode montage
Preprocessing was conducted with Matlab 2015b (Mathworks, Inc., Natick, Massachusetts,
USA) using functions from the EEGLAB toolbox version 13.6.5 (Delorme & Makeig, 2004). The
EEGs were epoched into 1-s intervals, and a linear-detrend function was applied using the SIFT
toolbox (Delorme et al., 2011; Mullen et al., 2010 ) to remove linear drift. Bad channels were identified
via visual inspection and replaced by interpolation from surrounding electrodes. Data were passed
through a semi-automated artifact-detection tool, and epochs were classified as clean or artifactual as
follows: threshold (+/- 300 mV); joint-probability (channel/global limit 5SD/3SD); kurtosis (6.5
SD/3.5 SD) and spectral profile (exceeding -100 to 28 db over 20 to 55 Hz); and a final manual review.
These parameters were tuned to detect electromyographic activity and large singular artifacts. The
removal of these artifacts improves independent components analysis (ICA) decomposition quality
while retaining periodic artifacts such as eye blinks and eye movements for correction by ICA.
EEGLAB’s FASTICA algorithm was used to calculate ICA weights. The ADJUST toolbox
automatically detects and removes artifactual ICA components representing blinks, eye movements,
and other spatial discontinuities (Mognon et al., 2011). ADJUST detections included the 33% of the
components with the highest mutual information to ensure that reliable and important components
13
Neural Activity During Jazz Improvisation
were removed (Groppe et al., 2009). The components that survived were then manually reviewed. Data
were then passed through the semi-automated artifact-detection tool again with more conservative
(6.5SD/2SD) and spectral profile (exceeding -100 to 25 dB over 20 to 55 Hz); and final manual review.
We analyzed the EEGs both before and after performing a surface Laplacian transformation
(Kayser & Tenke, 2015). The surface Laplacian (i.e., the second spatial derivative) is a reference-free,
high-pass spatial filter that improves localization and spatial resolution by filtering out spatially broad
effects (Fitzgibbons et al., 2013). This reduces the contribution of underlying sources that are distant
from an electrode, leaving activity that is relatively shallow and proximal to the electrodes. Thus, the
surface Laplacian sharpens scalp topography and reflects cortical activity directly beneath each
electrode. The Laplacian transformation also has the benefit of suppressing artifacts that have a broad
Muthukumaraswamy, 2013). The surface-Laplacian transformations were performed using the SSL
SPM. Spectral analysis of Laplacian-transformed spectral power was conducted at the sensor
level using the SPM12 M/EEG software package while controlling for multiple comparisons from
voxel-wise hypothesis testing (Litvak et al., 2011). In SPM-EEG, a General Linear Model (GLM)
approach is used to compare the EEG power at the voxel-level, and clusters of neighboring significant
voxels are compared to a “random field” noise model null criterion to determine significance at the
cluster-level (Erickson et al., 2018). The significance of the effect is determined by thresholding the
size of the cluster of voxels that are larger than would be expected to occur by chance. For each
improvisation from each participant, Fast Fourier Transforms of 1-s regularly epoched data were
14
Neural Activity During Jazz Improvisation
performed from 2-50 Hz in 1-Hz frequency steps (Hanning windowed), robust averaged, and log-
transformed. Then this spectral data were transformed to 3D Scalp x Frequency NIFTI-1 format
images ([x,y], mm; [z], Hz) and were z-score normalized across electrodes within each frequency-step
to equate subjects for global EEG power within each frequency step.
EEG quality contrast analyses were based on the ratings of the 6 improvisations performed by
each participant. Rather than using a median split to compare higher-quality and lower-quality takes,
the EEGs corresponding to the takes rated to be in the middle quintile of quality were omitted to
eliminate takes near the border between higher and lower quality that would reduce the
discriminability between the groups of takes (e.g., Erickson et al., 2018). Thus, the EEG scans
associated with the top 40% and bottom 40% (n = 154 scans) of rated improvisations were analyzed.
A flexible-factorial model was created to determine any confounding factors and to examine the
main effect of quality in high > low and low > high contrasts (Gläscher & Gitelman, 2008). The fact
that both group-level and subject-level factors can be entered into the flexible factorial model is critical
since each participant contributed multiple EEG scans to the analysis. Tests of main effects in each
model were conducted with a conservative cluster-forming threshold of p < .0001 and interpreted at
the cluster-level with family-wise error (FWE) corrected threshold of P_FWE <.05 (Flandin &
Friston, 2019). First, a within-subject model was created with subject and instructions (explicit versus
standard) as factors to determine if instructions significantly affected the EEG spectral characteristic
in the scalp space. No clusters survived the testing for the main effect of instructions; therefore, the
The inclusion of experience as a main effect and as a covariate in the quality models, yielded
significant spectral differences in the EEG between conditions. The four subjects closest to the median
experience value were removed, and high and low experience groups were formed with the remaining
28 subjects (14 in each). In this report, we present three SPM results: the main effect of quality, main
15
Neural Activity During Jazz Improvisation
effect of quality with experience covariate, and the main effect of experience on the EEGs. Each model
was subjected to a group main-effect F-test to test for the presence of the effect and t-tests for
To assess the robustness of our results, K-fold (K = 7) cross-validation was applied. In each of
the 7 “folds,” a new SPM model was created from the Laplacian-transformed EEG data with 22
randomly selected EEG scans (11 high, 11 low) withheld. Significant clusters were identified for the 7
SPM models, and only consistent clusters that appeared in more than half of the folds were retained.
We identified four reliable clusters across the folds. Within each fold, the location of the peak-voxel of
each cluster was noted. For the scans in each fold, the scalp-frequency power at that peak-voxel
location was extracted to represent the cluster. These peak-voxel power values were entered into a
stepwise regression for each fold (Erickson et al., 2018). The 7 stepwise regressions identified which of
the peak-voxel values were predictive of the improvisation quality ratings. Each stepwise regression
produced a model of significant predictors of the quality ratings of the withheld scans in that fold.
Root-mean-squared error (representing the predictive power of the model on untrained data) was
calculated for each fold and then averaged across folds. Additionally, the frequency ranges were
3. Results
3.1.1 Quality
For the high > low quality contrast, SPM t-tests of spectral power revealed several significant
differences (Figure 2), all in the left-hemisphere region including significant clusters around left
16
Neural Activity During Jazz Improvisation
central-parietal electrode CP3 (48 Hz, P_FWE = .005), left-parietal electrode P5 (39 and 47 Hz,
P_FWE < .001), left parieto-occipital electrode PO7 (23 and 25 Hz, P_FWE < .001), and left fronto-
The low > high quality contrast showed a large significant cluster of beta-band and gamma-
band activity (22, 33, and 47 Hz, P_FWE < .001) around right temporo-parietal electrode TP8 and 4
significant clusters (32 and 34 Hz, P_FWE < .001; 43 and 49 Hz, P_FWE < .001; 19 and 24 Hz,
P_FWE = .013; 38 Hz, P_FWE = .032) located around fronto-polar electrodes FP1 and FP2, with
right lateralization.
Figure 2. Topographic SPM significance maps of the main effects of improvisation quality on
Laplacian spectral power for the full factorial model (with improvisations in the middle quintile of
quality omitted). The peak frequencies (in Hz) of significant voxel clusters are shown adjacent to their
location. The image is a flattened topographic scalp map with the anterior-to-posterior head dimension
going from right to left and the left-right head dimension going from top to bottom. Clusters
17
Neural Activity During Jazz Improvisation
associated with higher > lower quality scores are depicted in red, and clusters associated with lower >
higher quality scores are shown in blue.
Figure 3. Topographic SPM significance maps of the main effects of improvisation quality with
experience as a covariate on Laplacian spectral power for the full factorial model (with improvisations
in the middle quintile of quality omitted). The peak frequencies (in Hz) of significant voxel clusters are
shown adjacent to their location. The image is a flattened topographic scalp map with the anterior-to-
posterior head dimension going from right to left and the left-right head dimension going from top to
bottom. Clusters associated with higher > lower quality scores are depicted in red, and clusters
associated with lower > higher quality scores are shown in blue.
Next, experience was statistically controlled for by including it as a covariate in the SPM
18
Neural Activity During Jazz Improvisation
quality models. For the high > low quality contrast, SPM t-tests of spectral power revealed several
significant differences (Figure 3), including three right frontal clusters in discrete frequency bands:
theta (4 Hz, P_FWE = .002), alpha (11 and 13 Hz, P_FWE < .001), and beta (27 and 28 Hz, P_FWE
= .005). Figure 4 shows separate topographic maps for high > low quality contrast for these three
frequency bands. Because the data were Laplacian-transformed, activity recorded at an electrode
originated proximal to that electrode. Activity in each frequency band was focused near a different
electrode (i.e., theta at FC6; alpha at F6; beta at FC4). Additionally, significant gamma-band activity
was found around left fronto-central electrode C3 extending up to left frontal electrode F3 (41 and 43
P_FWE = .031), consistent with the previous quality results when no covariates were included in the
model.
Low > high quality contrasts with experience as a covariate returned two significant clusters
which were nearly identical to the previous model’s output. The overlapping clusters contained
gamma-band activity (32 and 43 Hz, P_FWE = .004) around right temporo-parietal electrode TP8
and three significant clusters of beta (28 Hz, P_FWE = .003) and gamma (33 Hz, P_FWE = .031; 47
Hz, P_FWE = .001) activity located near right fronto-polar electrode FP2. Additionally, this model
yielded a significant cluster of alpha activity at right central electrode FC2 (10 Hz, P_FWE = .003).
19
Neural Activity During Jazz Improvisation
Figure 4. SPM topographic significance maps for peak voxel locations (shown by the black circles) of
the three right frontal clusters of theta (4 Hz), alpha (11-13 Hz), and beta (27-28 Hz) activity in the
high > low quality contrast for the main effects of improvisation quality with experience as a covariate.
The peak frequencies (in Hz) and voxels are depicted beneath each topographic scalp map. The image
is flattened with the anterior-to-posterior head dimension going from top to bottom and the left-right
3.1.2 Experience
For the high > low experience contrast (Figure 5), SPM t-tests of spectral power revealed
significant differences in central and left posterior regions including a significant gamma cluster
around central electrode FCz (32-45 Hz, P_FWE < .001). There were also two left parietal clusters
proximal to electrode CP3 (16-47 Hz, P_FWE < .001; 2-7 Hz, P_FWE = .008). Clusters of low-
frequency activity were detected at posterior midline regions near electrode Cz (2 – 11 Hz, P_FWE =
.001) and a right parietal cluster at electrode CP4 (2, 4, and 9 Hz, P_FWE < .001)
The low > high experience contrast showed a large significant cluster of beta-band activity (14
and 20 Hz, P_FWE < .001) in left frontal regions near electrode AF3 and a beta/gamma-band cluster
(22-47 Hz, P_FWE < .001) around frontal electrode FC6. Additionally, significant clusters of high-
20
Neural Activity During Jazz Improvisation
frequency activity were identified in the beta-band at left temporal electrode FT7 (14 Hz, P_FWE =
.007) and beta/gamma-bands over right parietal cortex around electrode P6 (16-47 Hz, P_FWE <
.001). A marginally significant cluster of alpha activity was revealed at electrode Cz (9 Hz, P_FWE =
.043).
Figure 5. Topographic SPM significance maps of the main effects of experience on Laplacian spectral
power for the full factorial model (with the 4 subjects closest to the median experience value omitted).
The peak frequencies (in Hz) of significant voxel clusters are shown adjacent to their location. The
image is a flattened topographic scalp map with the anterior-to-posterior head dimension going from
right to left and the left-right head dimension going from top to bottom. Clusters associated with high
> low experience scores are depicted in red, and clusters associated with low > high experience scores
are shown in blue.
We used stepwise regression with K-fold (K=7) cross-validation to assess the robustness of the
quality SPM results (with no covariate). K-fold validation is a machine-learning technique which trains
21
Neural Activity During Jazz Improvisation
a statistical model on a subset of the data, in this case withholding 22 scans (11 high-quality, 11 low-
quality) in each fold. The performance of the stepwise multilevel regression model is then assessed on
the observations that were removed, namely, the test set. This procedure is repeated K times to
simulate the model’s predictive ability when applied to new data. The average K-fold training RMSE
was 1.13, and the test (predictive) K-fold RMSE was 1.34. The correlation coefficient between the
RMSE and predicted RMSE for the training and tests sets was 0.401, a medium to large effect size.
Specifically, across the K-fold models, fronto-polar gamma (lower-quality group) was the
strongest predictor, retained by the stepwise regression models in all 7 of the folds. The lower-quality
condition also contributed 2 more components: right temporo-parietal gamma-band clusters in 4 of the
7 folds and beta-band activity in 2 of the folds. The higher-quality group had two predictors retained
in 4 of the 7 folds: theta activity at right fronto-temporal regions and left-posterior gamma. A theta
cluster was observed in some of the folds, but it was not a significant predictor in the SPM quality
contrasts, trending (p = 0.052) when applying a cluster forming threshold of p < .001; however, theta
activity did appear in right frontal regions when experience was included as a covariate in the model. A
The SPM analyses showed the neural predictors of improvisation quality. A parallel analysis
examined non-neural predictors of improvisation quality via multilevel regression (MLR) modeling
implemented with the lme4 (Bates, 2010) software package in R: A Language and Environment for
Statistical Computing v.3.4.4 (R Development Core Team, 2008). MLR models simultaneously assess
group-level and individual-level patterns within a single analysis, taking into consideration fixed-effect
and random-effect parameters. These models are compared using the log-likelihood (LL) goodness-of-
22
Neural Activity During Jazz Improvisation
fit measure. Changes in -2LL are distributed as χ2 with degrees of freedom equal to the number of
parameters added (Mirman, 2016). For all models, the random-effects structures were identical,
determined a priori in accordance with our previous studies. Here, the model accounts for the random
variability of the inter-individual variation of instructions and the inter-item variation of the different
chord changes for each take (Baayen, 2008). Table 1 shows the order that the factors were entered into
the MLR model. ANOVA model comparisons were used to determine the parameters that significantly
predicted improvisation ratings. This method evaluates whether adding or removing a new parameter
Table 1. Chi-square difference tests for jazz improvisation model comparisons. All models took into
account the effects of instructions by subject and improvisation-number as random effects. The best
performing model included primary genre, experience, and flow state as independent fixed effects (*p <
.05, **p < .01, ***p < .001).
Degrees of
The demographic factors of age, sex, race, and ethnicity did not significantly and independently
predict improvisation ratings. Improvisation-order, music, and jazz training failed to significantly and
independently predict improvisation ratings. However, primary genre (yes/no) and experience,
measured by an estimate of the number of live jazz gigs, were strong unique predictors of
improvisation quality scores. The recency of those performances did not improve model fit. Neither the
explicit creativity instruction manipulation nor the interaction of instructions and experience, which
were found to be significant in Rosen et al. (2017), predicted ratings. Only C FSS scores, independent
of experience, significantly improved model fit. Coefficient estimates for primary genre (Estimate =
1.11, SE = 0.33, p < .001) experience (Estimate = 0.30, SE = 0.10, p = .002), and C FSS scores (Estimate
= 0.23, SE = 0.09, p = .008) displayed a positive relationship with the improvisation ratings, such that
4. Discussion
improvisation by examining the interacting effects of performance quality and experience on brain
high-frequency (beta-band and gamma-band) activity recorded over left-hemisphere regions (Figure
2). Low-quality performances were associated with greater beta-band and gamma-band activity over
right-temporo-parietal cortex and right-lateralized fronto-polar cortex (Figure 2). (The right-
lateralized gamma-band activity recorded bilaterally at electrodes FP1 and FP2 may be generated
entirely in right fronto-polar cortex because, anatomically, the right frontal pole protrudes toward the
left hemisphere and the left occipital pole protrudes toward the right hemisphere, an observation
24
Neural Activity During Jazz Improvisation
known as “Yakovlevian torque”; Galaburda et al., 1978.) However, after controlling for differences in
experience, differences in quality were largely associated with right-hemisphere, mostly frontal,
activity (Figure 3). Thus, the association between performance quality and left-hemisphere activity was
a byproduct of the fact that the more experienced musicians generally produced better performances.
Regarding the neural correlates of experience, the guitarists with the most experience were
characterized by relatively greater high-frequency (beta-band and gamma-band) activity recorded over
fronto-central and left posterior regions. Performances by less experienced musicians were associated
with bilateral frontal and prefrontal beta-band and gamma-band activity (right hemisphere > left
hemisphere). Another way to characterize the effect of experience is that high-experience generally
involved more dorsal brain areas and low-experience involved more ventral brain areas (Figure 5).
phenomenological experience of flow, and jazz being the primary musical genre of the performer. We
defer a detailed discussion of flow and primary musical genre to a subsequent report. The present
discussion focuses on the relationship between improvisation quality and brain activity as mediated by
experience.
There were significant effects of experience on the behavioral data. The guitarists with the
highest levels of experience, as determined by number of live jazz performances, received the highest
improvisation ratings, consistent with past jazz improvisation studies (Beaty et al., 2013; Pinho et al.,
2014; Rosen et al., 2016; 2017). Increased levels of flow, reported by the musicians after improvising,
also increased the perceived quality of the improvisations. This suggests there may be a link between
the perceived quality of jazz improvisations and the phenomenological experience of achieving a flow-
state during jazz improvisation. This aligns with theoretical models of jazz improvisation which posit
25
Neural Activity During Jazz Improvisation
that an individual is performing at his or her peak of arousal and ability when entering a flow state
Due to the restrictive nature of our target population (jazz guitarists) and desire to maximize
the sample size and range of jazz expertise, approximately 40% (13 out of 32) of our subjects reported
that jazz was not their primary performance genre. Since our measures for primary genre (PG) and
experience focus on live jazz performance experience, the two factors were highly correlated, as
experience. There are many forms of musical improvisation, and as musicians gain performance
experience in a specific style they develop genre-specific improvisatory approaches and techniques for
In our previous studies of jazz improvisation, PG was not included as a variable; thus, we did
not make any a priori hypotheses about PG. However, the fact that PG correlated with improvisation
ratings is not surprising given that genre-specific demands require different types of training and
practice. Given the assumption that musicians spend more time explicitly training and practicing
(Type-2 processes) in their PG, these musicians may be able to inhibit frontally-mediated, executive
control processes and recruit more implicit, automatic bottom-up (Type 1) processes when improvising
in their PG.
A recent study on musical creativity, improvisation and implicit knowledge suggests that a
musician’s creativity, which is highly-dependent of implicit domain knowledge, will evolve over his or
her lifetime based on the individual’s specific experience and training (Daikoku, 2018). This indicates
not only that PG training is critical to improvisation, but that the specific methods of training in a
given genre will shape creativity on an individual level. Furthermore, musicians’ specific practice
regimens lead to distinct changes within musicians’ auditory perception-action network (Loui, 2018).
For example, jazz musicians develop heightened sensitivity to the tuning of notes within a tonal
26
Neural Activity During Jazz Improvisation
context and are more tolerant of unexpected chords that sound incorrect compared to classical and
non-musicians (Przysinda et al., 2017). Genre-specific music training is a unique example of high-level
motor cognition plasticity such that enduring neural markers are created through years of practice
which best equips a musician with the appropriate cognitive-motor strategies to advance and develop
Within the dual-process model of creativity, in the case of jazz, this mastery requires a decrease
in Type-2 processing. As musicians gain more jazz experience, they must optimize their ability to
efficiently navigate harmonic relationships, allowing for enhanced flexibility when faced with new
chord sequences or harmonic violations (Bianco et al., 2017). Hanson et al. (2016) posit that
recognition of harmonic patterns may be automatized for professional jazz musicians, as domain-
specific, long-term knowledge and expectations allow for the effective use of improvisational strategies
and tools (Huron, 2006; Przysinda et al., 2017). Since our study required guitarists to improvise to
novel, unfamiliar chord sequences in a jazz context, their ability to quickly extract harmonic content
from the chords was paramount and likely contributed to the pronounced effects of genre on the
The most striking feature of the relationship between brain activity and performance quality is
that differences in quality were, after controlling for experience, primarily associated with right-
right frontopolar and temporoparietal regions. This is consistent with the dual-process view linking
superior creative performance with reduced frontal-mediated executive control (i.e., the matched-filter
hypothesis; Chrysikou et al., 2014). (The higher-quality improvisations included a left fronto-central
activation recorded over motor cortex that may indicate more intensive or more elaborate manual
control for expert guitarists’ right hands rather than a more general type of executive control.)
27
Neural Activity During Jazz Improvisation
Low-quality performances were also associated with right temporo-parietal activity. Berkowitz
and Ansari (2010) reported that musicians deactivated the r-TPJ during melodic improvisation,
hypothesizing that training-induced deactivations in this area involve task focus and goal maintenance,
facilitating creative thought. Consistent with their hypothesis, we found high-frequency beta and
gamma activity in this region which may indicate recruitment, rather than deactivation, of these
band activity proximal to right somatosensory and primary motor cortex. This electrophysiological
activity may suggest inhibition of left-hand motor processes or lack of attention to proprioceptive
feedback, either of which would be antithetical to successful improvisation due to the high
sensorimotor demands of jazz guitar improvisation. For example, Sauseng et al. (2009) used EEG and
transcranial magnetic stimulation (TMS) to show that a motor evoked potential was elicited more
easily when EEG alpha power preceding a magnetic pulse was low, and vice versa. Thus, the amount
of alpha-band activity in sensorimotor cortex may predict motor cortical excitability. This inhibitory
activity may decrease left-hand speed, dexterity, and efficiency. Increased activation and connectivity
of right sensorimotor regions has been linked to higher levels of improvisation experience (Pinho et al.,
2014), together indicating that in a technically demanding, motor-cognitive task like jazz
High-quality performances, after controlling for experience, were associated with right frontal
clusters of theta, alpha, and high-beta activity (Figure 4). The location of peak power differed across
these frequency bands, suggesting that the three clusters of activity originated in non-identical
sources. Speculatively, based on the peak electrode locations, the high-beta cluster focused at electrode
FC4 may have been generated in pre-motor or motor cortex and likely involved in the facilitation of
motor performance. The alpha effect peaking at electrode F6 over right dorsolateral prefrontal cortex
28
Neural Activity During Jazz Improvisation
may reflect deactivation of executive control (Lopata et al., 2017) or could be an example of center-
surround inhibition in which alpha inhibition encapsulates nearby beta activity to isolate it from
interference (Beck & Hallet, 2011). Though we did not have any a priori hypotheses about theta
activity in right frontal regions, low-frequency bands are thought to support long-range oscillatory
communication (Clayton et al., 2015). Several studies indicate that frontal theta EEG activity increases
with increased demands of executive functions, working memory, and control mechanisms (Jensen &
Tesche, 2002; Sauseng et al., 2005; Sauseng et al., 2010. The cluster of right frontal theta activity may
signify activation of the frontoparietal control network and deactivation of the default mode network
(Scheeringa et al., 2008). With regard to our dual-process model of creativity, increased activation of
the fronto-parietal executive control network may not predict high-quality improvisations. Another
interpretation is that long-range theta coupling is associated with task “switching” instead of global
executive functions such as working memory (Sauseng et al., 2005). Long-range oscillatory
communication among regions may underlie creative cognition (Stevens & Kabalina, 2019), especially
in musical performance (Gruzelier et al., 2014). For jazz improvisation, a global increase in executive
functioning may hinder performance, but cognitive switching and increased top-down control may be
necessary to explicitly mediate and adapt to challenges or mistakes during a performance. Jazz
musicians with more experience may be better at reverting to an associative, bottom-up cognitive
Regarding experience, for novices, there was an extensive bilateral-frontal (right > left), high-
frequency cluster of activity with a higher frequency on the right than the left, and a small right
activity functions to maintain the current sensorimotor or cognitive state (Engel & Fries, 2010). In the
present study, novices’ frontal beta oscillations may represent the recruitment and maintenance of
executive functioning, in particular, top-down inhibitory control (Aron et al., 2014; Hwang et al.,
29
Neural Activity During Jazz Improvisation
2014). In support of this hypothesis, a jazz improvisation brain-stimulation study reported anodal
(excitatory) stimulation to right frontal cortex increased the quality ratings for novices but degraded
The cluster of right parieto-occipital beta and gamma activity associated with performances by
low-experience musicians may also involve top-down processing. Previous studies have implicated
network engaged in divergent thinking (Aberg et al., 2016; Beaty et al., 2014). With a right-parietal
neurofeedback training (NFT) protocol, Agnoli et al. (2018) demonstrated that beta NFT increased
both originality and fluency on a divergent thinking task for participants with low levels of creative
achievement.
In contrast, the experts exhibited right parietal and central clusters of broadband activity.
Greater parietal activity may reflect an associative brain-state which relies on multimodal sensory
processing and integration (Csikszentmihalyi, 1996), long-term memory (Wagner et al., 2005), mental
imagery (Kosslyn & Thompson, 2003), spatial coding, sensory-motor transformation, and attention
The high-expertise musicians also exhibited clusters of delta-band activity at left, central, and
right posterior sites. Hlinka et al. (2010) reported that delta power has a strong relationship with
functional connectivity within the default mode network (DMN). The DMN is a set of inferior parietal
and midline brain regions that are activated during low external-task demands (Andrews-Hanna et al.,
2014). The involvement of large-scale networks such as the DMN and the frontoparietal executive
network have been implicated in several recent neuroscientific studies of creativity (Bashwiner et al.,
2016; Mok, 2014; Shi et al, 2018). Increased connectivity across these networks may enable creative
performance in a variety of domains, including music improvisation and creative writing (Beaty et al.,
2014; Wise & Braga, 2014; Loui, 2018). Cooperation between large-scale networks, specifically the
30
Neural Activity During Jazz Improvisation
DMN and the frontal-parietal executive network may support demanding, complex cognitive abilities
such as creative cognition (Kenett et al., 2018). Recent studies of musical improvisation (Pinho et al.,
2014) and creative cognition (Beaty et al., 2016) have highlighted the importance of the DMN in
artistic performance and creativity. In a visual arts task that used an MRI-compatible drawing tablet,
art students’ idea generation was associated with widespread DMN activity while idea evaluation was
associated with both DMN and executive network activity (Ellamil et al, 2012). A main function of the
DMN is episodic memory retrieval (Schacter et al., 2007), and it has been proposed that the DMN’s
relationship with episodic memory plays a key role in creative cognition (Benedek et al., 2014). In a
study of mental imagery during drawing and the viewing paintings, artists displayed enhanced delta
memory (Bhattacharya & Petsche, 2002; 2005). Therefore, one advantage experts may acquire through
training is their ability to efficiently integrate ideas, strategies, and phrases from long-term memory.
In the present study, the novel chord sequences were used as stimuli to place an emphasis on quickly
Experts also exhibited gamma-band activity at right parietal and central locations. Gamma-
band activity in these regions may be associated with various aspects of motor control and motor
planning that would be required for jazz improvisation performance. For example, gamma-band
activity in primary motor cortex occurs during the preparation and execution of movements. It is
stronger for larger movements, weaker for repeated movements, and is thought to contribute to later
stages of motor control (Donner et al., 2009, Muthukumaraswamy, 2010). Thus, this activity may
represent more experienced jazz improvisers’ enhanced motor control and more elaborate action
sequences. Unfortunately, we did not measure the quantity of notes played during the improvisations,
so it is not possible to test this hypothesis. However, because mastery of jazz improvisation is a domain
31
Neural Activity During Jazz Improvisation
that requires a high degree of manual technical proficiency, it is reasonable to assume that experience
is associated with increased motor control and movement (at least in the hands for guitarists).
The neural correlates of high- versus low-experience are consistent with our dual-process
model of creativity in which novices rely more on frontally mediated Type-2 executive processing and
experts rely more on Type-1, associative processing mediated by more posterior brain regions.
The matched-filter hypothesis posits that the frontal lobe exerts cognitive control to direct
posterior associative brain activity in accordance with the current needs of a task (Chrysikou et al.,
2013). For jazz improvisation, domain mastery leads to the automatization of aspects of one’s
performance, involving greater unconscious processing and generation of appropriate musical choices
and actions. This alleviates the need for experts to rely heavily on executive, top-down, controlled
processes, shifting activity to posterior regions that store routines in long-term memory (Krakauer &
Shadmehr, 2006). Our high- versus low-experience results demonstrate such a pattern.
Pressing’s (1988) cognitive model of jazz improvisation emphasizes how experts reduce
cognitive load via mastery of domain-specific technical skills and knowledge to make resources
available for creative ideation, execution, and evaluation of one’s performance. Neural evidence has
shown that domain expertise in musical improvisation is associated with a reduction in frontally
networks and the coupling of the default-mode and executive-control networks (Pinho et al., 2016).
Another possibility is that creative jazz improvisation benefits from a moderate degree of
executive control and evaluative processes but that Type-1 processes are predominant for the highest
levels of performance (Lopata et al., 2017; Rosen et al, 2017). Due to the fact that idea generation and
evaluation likely occur in parallel during improvisation, this makes sense because expending cognitive
32
Neural Activity During Jazz Improvisation
resources on evaluation and self-judgment could detract from those required for creative ideation.
Behaviorally, these neurocognitive differences between novices and experts may explain why jazz
musicians with the most experience are able to produce more creative improvisations. Over time,
expert improvisers are able to seamlessly integrate and draw from a hierarchical structure of learned
and novel ideas, form associative links between choices, and select, retrieve and execute ideas activated
in associative memory (Clarke, 1988). Learned routines and technical prowess provide expert
musicians with a baseline level of improvisation abilities that result in a product perceived to be of
higher quality by an expert rater separate from the mental processes that generated the performance.
Lopata et al. (2017) recorded high-alpha (10-12 Hz) EEG activity over frontal cortex during
musical improvisation and interpreted this electrophysiological activity as cortical inhibition. It was
positively correlated with performance quality ratings and formal improvisation training. In contrast,
our study found that lower-quality improvisations were associated with higher-frequency beta-band
and gamma-band oscillations, indicative of cortical activation (Lachaux, et al., 2007; Scheeringa et al.,
2011) rather than inhibition-related, alpha-band activity. When we controlled for experience, however,
the higher-quality performances did contain increased synchronization of right frontal alpha, thereby
representing a neural substrate of creative jazz improvisation that is not primarily driven by
experience.
One key difference between our study and Lopata’s was in the musical lead sheets on which the
performers based their improvisations, specifically, in the complexity and novelty of the 16-bar chord
changes. Although our chord sequences included some common jazz vocabulary, they were designed to
be unique and unfamiliar to participants; therefore, greater levels of cognitive control may have been
required for our musicians relative to those participating in the aforementioned study (Lopata et al.,
33
Neural Activity During Jazz Improvisation
2017). In fact, our findings align theoretically with those of Lopata et al. We used the surface-
Laplacian transformation to filter out spatially diffuse signals, and we were able to identify a focal
cluster of alpha activity in the high-quality condition near r-DLPFC. The results of both studies are
consistent with the view that higher-quality musical improvisation in expert performers is enabled by
neurocognitive model in which the right hemisphere is critical for dealing with novel tasks or
situations and the left hemisphere is recruited for routinized processing in familiar situations. The
present results (Figure 5) provide some support for the idea that, for less experienced musicians,
improvisation is a less familiar task which recruits greater right-hemisphere activity relative to highly
automatize key processes involved in improvisation resulting in a shift from right-hemisphere to left-
hemisphere activity. According to this view, there are two forms of creative cognition: one for
comparatively novel situations and one that applies well-learned routines to familiar situations.
4.4 Limitations
Although the high-density EEG recordings, surface-Laplacians, SPM analyses, and refined
experimental procedures used in this study offer substantive advances over previous efforts, there are
First, this study did not find an effect of explicit creativity instructions on brain activity nor did
it replicate the interaction we found between experience and explicit instructions on quality ratings in
a previous behavioral study of jazz piano improvisation (Rosen et al., 2017). These different results
may be due to the fact that the musicians in the present study improvised to different chord
progressions during each of the 6 takes. In the previous study, one lead sheet was used, and musicians
34
Neural Activity During Jazz Improvisation
improvised to that lead sheet four times. It is possible that repeated exposure to the chords assisted
novices more than experts since experts are likely better able to adapt to novel chord sequences. Since
the standard-instructions condition always preceded the explicit creativity instructions condition,
there may have been experience-dependent ordering effects even though we attempted to minimize
these effects with a staggered baseline design. The use of 6 unique chord progressions may have
restricted variability across subjects’ improvisation ratings compared to the previous study, thereby
suppressing the expected interaction effect. Nevertheless, when comparing within-subject quality-
rating variability across the two studies, we found that they were equivalent. Therefore, even though
Second, although the use of a jazz improvisation task was motivated by a concern for ecological
validity, the ecological validity of this study was still somewhat limited by the stimuli and setting. The
chord sequences were novel to performers and did not include a written melody or “head.” Before
improvising, jazz musicians typically familiarize themselves with a series of chord changes and often
use the melody of a piece as a foundation from which to improvise. We chose novel chord sequences
and the absence of melodies to eliminate potentially confounding factors such as sight reading, melody
interactions, familiarity, and cohesion among musicians, which are central to jazz improvisation
(Berliner, 1994).
Third, the available statistical power, though substantial, did not afford the opportunity to
examine improvisation quality with even higher resolution. Although we were able to demonstrate
differences in brain activity between higher and lower levels of improvisation quality (with the middle
quintile removed), brain activity may change in complex ways as quality increases. For example,
specific brain activations associated with changes in quality may have different transition points
35
Neural Activity During Jazz Improvisation
between low and high quality. Another possibility is that there may be additional transitional brain
states that are not simple mixtures of the lower and higher quality states. Future research should
Finally, although the surface Laplacian transformation sharpened the detail and enhanced the
interpretation of our EEG results, it does not afford precise localization of effects, although it does
preserve frequency information that is invisible to hemodynamic imaging. Future fMRI research or
EEG or MEG research employing a denser sensor array and source localization algorithms should
reveal more specific details about the structures underlying the observed effects.
The findings from this study provide neural evidence for a dual-process model of creativity in
jazz improvisation. Here, the EEGs of low-quality improvisations and low-expertise are interpreted as
representing increased activity in the frontoparietal network and increased recruitment of executive
Type-2 processes (Marek & Dosenbach, 2018), as we had hypothesized. An increase in cognitive
control is likely necessary for novice jazz musicians to perform successfully, especially when
improvising to novel sequences of 16-bar chord changes at a medium tempo (144 beats per minute).
The combination of the challenging improvisation task and the technical demands of jazz generate an
intensive cognitive load for less experienced jazz musicians may contribute to this highly salient effect.
Thus, they are not able to adapt quickly enough to generate solos with the qualities needed to receive a
high-quality rating. When asked to define “What makes a jazz improvisation creative?,” the judges in
this study reported that “exploration over presets,” the “originality of ideas within the framework of a
relatively common jazz vocabulary,” and the “establishment of themes, variations, rhythmic phrases
musicians yielded two theoretically interesting results. First, high-frequency brain activity was
identified at central and posterior brain regions. No clusters of frontal activity were found. We
interpret this finding as further evidence for our dual-process model of creativity whereby the highest
levels of jazz improvisation benefit from unconscious, associative, Type-1 processes. Second, the
clusters of activity which characterized high-quality improvisations were left-lateralized. These results
align with neurocognitive models which posit that the left hemisphere is utilized more for familiar
tasks and habituated behaviors and processing (Goldberg et al., 1994). In this context, one would
expect that familiarity with the jazz improvisation task and habituated behaviors (e.g., licks, jazz
vocabulary, melodic patterns for specific chord sequences) would be demonstrated by jazz musicians
However, after controlling for experience, performance quality was associated with activation
differences in the right hemisphere. For example, with experience as a covariate, high-quality
improvisations were associated with alpha-band activity over r-DLFPC. This finding may suggest that
improvisations are perceived as having higher quality when musicians inhibits executive control
processes, aligning with previous jazz improvisation and creativity studies (Limb & Braun, 2008;
Lopata et al., 2017; Pinho et al, 2014). The analysis of the neural correlates of high- versus low-quality
improvisations provide multiple neural markers that are in accord with our dual-process model of
creativity.
Additionally, this study raises an important theoretical issue regarding the neural substrates
and definition of creativity. Ignoring experience, performances that were judged to be more creative
were associated with left-hemisphere processing; right-hemisphere processing was associated with
less-creative performances. The consensual assessment technique has expert judges evaluate the
quality of creative products. According to this product-based conceptualization, the left hemisphere
37
Neural Activity During Jazz Improvisation
would be considered to play a special role in creativity. However, if the effectiveness of cognitive
processes recruited to deal with novel or less-familiar situations is considered central to a definition of
creativity, after taking into account the role of experience, the right hemisphere would be considered to
play a special role in creative cognition. Thus, the functional neuroanatomy of creativity depends on
whether creativity is defined in terms of creative products or the mental processes that generate those
products.
38
Neural Activity During Jazz Improvisation
References
Aberg, K. C., Doell, K. C., & Schwartz, S. (2016). The “Creative right brain” revisited:
individual creativity and associative priming in the right hemisphere relate to hemispheric
asymmetries in reward brain function. Cerebral Cortex, 27(10), 4946-4959.
Agnoli, S., Zanon, M., Mastria, S., Avenanti, A., & Corazza, G. E. (2018). Enhancing
creative cognition with a rapid right-parietal procedure. Neuropsychologia, 11, 99-106.
Amo, C., De Santiago, L., Luciáñez, D. Z., Alonso-Cortés, J. M. L., Alonso-Alonso, M.,
Barea, R., & Boquete, L. (2017). Induced gamma band activity from EEG as a
possible index of training-related brain plasticity in motor tasks. PloS one, 12(10),
e0186008.
Andrews-Hanna, J. R., Smallwood, J., & Spreng, R. N. (2014). The default network and self-
generated thought: component processes, dynamic control, and clinical relevance. Annals
of the New York Academy of Sciences, 1316(1), 29.
Aoki, F., Fetz, E. E., Shupe, L., Lettich, E., & Ojemann, G. A. (1999). Increased gamma-
range activity in human sensorimotor cortex during performance of visuomotor
tasks. Clinical Neurophysiology, 110(3), 524-537.
Arkin, C., Przysinda, E., Pfeifer, C., Zeng, H., & Loui, P. (2019). Grey Matter Correlates of
Creativity in Musical Improvisation. Frontiers in Human Neuroscience, 13, 169.
Aron, A. R., Robbins, T. W., & Poldrack, R. A. (2014). Inhibition and the right inferior
frontal cortex: one decade on. Trends in cognitive sciences, 18(4), 177-185.
Baer, John. (2010). Is creativity domain specific. The Cambridge handbook of creativity.
J.C. Kaufman & R.J. Sternberg (Eds.). New York, NY: Cambridge University Press, 321-341.
Bashwiner, D. M., Wertz, C. J., Flores, R. A., & Jung, R. E. (2016). Musical creativity “revealed” in
brain structure: interplay between motor, default mode, and limbic networks. Scientific
reports, 6, 20482.
Beaty, R. E., Smeekens, B. A., Silvia, P. J., Hodges, D. A., & Kane, M. J. (2013). A first look
at the role of domain-general cognitive and creative abilities in jazz
improvisation. Psychomusicology: Music, Mind, and Brain, 23(4), 262.
Beaty, R. E., Benedek, M., Wilkins, R. W., Jauk, E., Fink, A., Silvia, P. J., ... & Neubauer, A.
39
Neural Activity During Jazz Improvisation
C. (2014). Creativity and the default network: a functional connectivity analysis of
the creative brain at rest. Neuropsychologia, 64, 92-98.
Beaty, R. E., Benedek, M., Silvia, P. J., & Schacter, D. L. (2016). Creative cognition and
brain network dynamics. Trends in cognitive sciences, 20(2), 87-95.
Beck, S., & Hallett, M. (2011). Surround inhibition in the motor system. Experimental brain
research, 210(2), 165-172.
Benedek, M., Jauk, E., Sommer, M., Arendasy, M., & Neubauer, A. C. (2014). Intelligence,
creativity, and cognitive control: The common and differential involvement of
executive functions in intelligence and creativity. Intelligence, 46, 73-83.
Berkowitz, A. L., & Ansari, D. (2008). Generation of novel motor sequences: the neural
correlates of musical improvisation. Neuroimage, 41(2), 535-543
Berliner, P. F. (1994). Thinking in jazz: Composing in the moment. Jazz Educators Journal,
26, 28-34.
Bhattacharya, J., & Petsche, H. (2002). Shadows of artistry: cortical synchrony during
perception and imagery of visual art. Cognitive Brain Research, 13(2), 179-186.
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),
1-14.
Bianco, R., Novembre, G., Keller, P. E., Villringer, A., & Sammler, D. (2017). Musical
genre-dependent behavioural and EEG signatures of action planning. A comparison
between classical and jazz pianists. NeuroImage.
Boasen, J., Takeshita, Y., Kuriki, S., & Yokosawa, K. (2018). Spectral-spatial differentiation
of brain activity during mental imagery of improvisational music performance using
MEG. Frontiers in human neuroscience, 12, 156.
Boccia, M., Piccardi, L., Palermo, L., Nori, R., & Palmiero, M. (2015). Where do bright
40
Neural Activity During Jazz Improvisation
ideas occur in our brain? Meta-analytic evidence from neuroimaging studies of
domain-specific creativity. Frontiers in psychology, 6, 1195.
Börgers, C., Epstein, S., & Kopell, N. J. (2008). Gamma oscillations mediate stimulus
competition and attentional selection in a cortical network model. Proceedings of
the National Academy of Sciences, 105(46), 18023-18028.
Clayton, M. S., Yeung, N., & Kadosh, R. C. (2015). The roles of cortical oscillations in sustained
attention. Trends in cognitive sciences, 19(4), 188-195.
Cooper, P. S., Wong, A. S., McKewen, M., Michie, P. T., & Karayanidis, F. (2017).
Frontoparietal theta oscillations during proactive control are associated with goal-
updating and reduced behavioral variability. Biological psychology, 129, 253-264.
Csikszentmihalyi, M. (1996). Flow and the psychology of discovery and invention. New
York: Harper Collins.
Daikoku, T. (2018). Entropy, uncertainty, and the depth of implicit knowledge on musical
creativity: Computational study of improvisation in melody and rhythm. Frontiers in
computational neuroscience, 12, 97.
Delorme, A., & Makeig, S. (2004). EEGLAB: an open source toolbox for analysis of single-
trial EEG dynamics including independent component analysis. Journal of neuroscience
methods, 134(1), 9-21.
Delorme, A., Mullen, T., Kothe, C., Acar, Z. A., Bigdely-Shamlo, N., Vankov, A., & Makeig, S. (2011).
EEGLAB, SIFT, NFT, BCILAB, and ERICA: new tools for advanced EEG processing.
Computational intelligence and neuroscience, 10.
Deng, S., Winter, W., Thorpe, S., Srinivasan, R., 2011. EEG surface Laplacian using
realistic head geometry. Int. J. Bioelectromagn. 13, 173–177.
Donner, T. H., Siegel, M., Fries, P., & Engel, A. K. (2009). Buildup of choice-predictive
activity in human motor cortex during perceptual decision making. Current
Biology, 19(18), 1581-1585.
41
Neural Activity During Jazz Improvisation
Ellamil, M., Dobson, C., Beeman, M., & Christoff, K. (2012). Evaluative and generative
modes of thought during the creative process. Neuroimage, 59(2), 1783-1794.
Engel, A. K., & Fries, P. (2010). Beta-band oscillations—signalling the status quo? Current
opinions in neurobiology, 20(2), 156-165.
Erickson, B., Truelove-Hill, M., Oh, Y., Anderson, J., Zhang, F. Z., & Kounios, J. (2018).
Resting-state brain oscillations predict trait-like cognitive styles. Neuropsychologia,
120, 1-8.
Fink, A., & Benedek, M. (2014). EEG alpha power and creative ideation. Neuroscience & Biobehavioral
Reviews, 44, 111-123.
Fink, A., Graif, B., & Neubauer, A. C. (2009). Brain correlates underlying creative
thinking: EEG alpha activity in professional vs. novice dancers. NeuroImage, 46(3),
854-862.
Fink, A., Slamar-Halbedl, M., Unterrainer, H. F., & Weiss, E. M. (2012). Creativity:
Genius, madness, or a combination of both? Psychology of Aesthetics, Creativity,
and the Arts, 6(1), 11.
Fitzgibbon, S. P., Lewis, T. W., Powers, D. M., Whitham, E. W., Willoughby, J. O., &
Pope, K. J. (2013). Surface laplacian of central scalp electrical signals is insensitive
to muscle contamination. IEEE Transactions on Biomedical Engineering, 60(1), 4-9.
Flandin G, Friston KJ. (2016). Analysis of family-wise error rates in statistical parametric
mapping using random field theory. https://arxiv.org/abs/1606.08199.
Galaburda, A. M., LeMay, M., Kemper, T. L., & Geschwind, N. (1978). Right-left
asymmetries in the brain. Science, 199(4331), 852-856.
Gläscher, J., Gitelman D., 2008. Contrast weights in flexible factorial design with multiple
groups of subjects. Retrieved: August 20, 2018 from〈http://www.sbirc.ed.ac.uk/
Cyril/download/Contrast_Weighting_Glascher_Gitelman_2008.pdf〉.
Goldberg, E., Podell, K., & Lovell, M. (1994) Lateralization of frontal lobe functions and
cognitive novelty. The Journal of Neuropsychiatry and clinical neurosciences, 6(4),
371-376.
Groppe, D. M., Makeig, S., & Kutas, M. (2009). Identifying reliable independent
42
Neural Activity During Jazz Improvisation
components via split-half comparisons. NeuroImage, 45(4), 1199-1211.
Hansen, N. C., Vuust, P., & Pearce, M. (2016). " If You Have to Ask, You'll Never Know":
Effects of Specialised Stylistic Expertise on Predictive Processing of Music. PloS
one, 11(10), e0163584.
Hwang, K., Ghuman, A. S., Manoach, D. S., Jones, S. R., & Luna, B. (2014). Cortical
neurodynamics of inhibitory control. Journal of Neuroscience, 34(29), 9551-9561.
Huron, D. B. (2006). Sweet anticipation: Music and the psychology of expectation. MIT
press.
Jensen, O., & Mazaheri, A. (2010). Shaping functional architecture by oscillatory alpha
activity: gating by inhibition. Frontiers in human neuroscience, 4, 186.
Jensen, O., & Tesche, C. D. (2002). Frontal theta activity in humans increases with memory load
in a working memory task. European journal of Neuroscience, 15(8), 1395-1399.
Kaas, J. H., & Stepniewska, I. (2016). Evolution of posterior parietal cortex and parietal‐
frontal networks for specific actions in primates. Journal of Comparative
Neurology, 524(3), 595-608.
Kaufman, S. B., Quilty, L. C., Grazioplene, R. G., Hirsh, J. B., Gray, J. R., Peterson, J. B., &
DeYoung, C. G. (2016). Openness to experience and intellect differentially predict creative
achievement in the arts and sciences. Journal of Personality, 84(2), 248-258.
Kayser, J., Tenke, C.E., 2015. Issues and considerations for using the scalp surface
Laplacian in EEG/ERP research: a tutorial review. Int. J. Psychophysiol. 97(3), 189–
209.
Kenett, Y. N., Medaglia, J. D., Beaty, R. E., Chen, Q., Betzel, R. F., Thompson-Schill, S. L., & Qiu,
J. (2018). Driving the brain towards creativity and intelligence: A network control theory
analysis. Neuropsychologia, 118, 79-90.
43
Neural Activity During Jazz Improvisation
Kosslyn, S. M., & Thompson, W. L. (2003). When is early visual cortex activated during
visual mental imagery?. Psychological bulletin, 129(5), 723.
Kristeva, R., Patino, L., & Omlor, W. (2007). Beta-range cortical motor spectral power and
corticomuscular coherence as a mechanism for effective corticospinal interaction
during steady-state motor output. Neuroimage, 36(3), 785-792.
Lachaux, J. P., Fonlupt, P., Kahane, P., Minotti, L., Hoffmann, D., Bertrand, O., & Baciu,
M. (2007). Relationship between task‐related gamma oscillations and BOLD signal:
New insights from combined fMRI and intracranial EEG. Human brain
mapping, 28(12), 1368-1375
Limb, C. J., & Braun, A. R. (2008). Neural substrates of spontaneous musical performance:
An fMRI study of jazz improvisation. PLoS one, 3(2), e1679.
Litvak, V., Mattout, J., Kiebel, S., Phillips, C., Henson, R., Kilner, J., ... & Penny, W. (2011).
EEG and MEG data analysis in SPM8. Computational intelligence and
neuroscience.
Lopata, J. A., Nowicki, E. A., & Joanisse, M. F. (2017). Creativity as a distinct trainable
mental state: An EEG study of musical improvisation. Neuropsychologia, 99, 246-
258.
Loui, P. (2018). Rapid and flexible creativity in musical improvisation: review and a
model. Annals of the New York Academy of Sciences.
Marek, S., & Dosenbach, N. U. (2018). The frontoparietal network: function, electrophysiology,
and importance of individual precision mapping. Dialogues in clinical neuroscience, 20(2),
133.
Martin, A. J., & Jackson, S. A. (2008). Brief approaches to assessing task absorption and
enhanced subjective expertise: Examining ‘short’and ‘core’ flow in diverse
performance domains. Motivation and Emotion, 32(3), 141-157.
McGraw, K. O., & Wong, S. P. (1996). Forming inferences about some intraclass
correlation coefficients. Psychological Methods,1(1), 30.
Mirman D. (2016). Growth curve analysis and visualization using R : The R series. Boca
Raton, FL: CRC Press.
Mognon, A., Jovicich, J., Bruzzone, L., & Buiatti, M. (2011). ADJUST: An automatic EEG
artifact detector based on the joint use of spatial and temporal features.
Psychophysiology, 48(2), 229-240.
44
Neural Activity During Jazz Improvisation
Mok, L. W. (2014). The interplay between spontaneous and controlled processing in creative
cognition. Frontiers in human neuroscience, 8, 663.
Mullen, T., Delorme, A., Kothe, C., & Makeig, S. (2010). An electrophysiological information flow
toolbox for EEGLAB. Biol. Cybern, 83, 35-45.
Newell, A., & Rosenbloom, P. S. (1981). Mechanisms of skill acquisition and the law of
practice. Cognitive Skills and their Acquisition, 1-51.
Norgaard, M., Spencer, J., & Montiel, M. (2013). Testing cognitive theories by creating a
pattern-based probabilistic algorithm for melody and rhythm in jazz improvisation.
Psychomusicology: Music, Mind, and Brain, 23(4), 243.
Pinho, A. L., de Manzano, Ö., Fransson, P., Eriksson, H., & Ullén, F. (2014). Connecting
to create: Expertise in musical improvisation is associated with increased functional
connectivity between premotor and prefrontal areas. The Journal of Neuroscience,
34 (18), 6156-6163.
Pinho, A. L., Ullén, F., Castelo-Branco, M., Fransson, P., & de Manzano, Ö. (2016).
Addressing a paradox: dual strategies for creative performance in introspective and
extrospective networks. Cerebral Cortex, 26(7), 3052-3063.
Pressing, J. (1988). Improvisation: Methods and models. In: J.A. Sloboda (Ed.), Generative
processes in music: The psychology of performance, improvisation, and
composition. Oxford: Clarendon Press, 129-178.
Pugatch, J. (2006). Acting is a job: Real life lessons about the acting business (p. 73). New
York: Allworth Press.
Rosen, D., Youngmoo E. Kim, Daniel Mirman, and John Kounios. (2017). All you need to
do is ask? The exhortation to be creative improves creative performance more for
non-expert than expert jazz musicians. Psychology of Aesthetics, Creativity, and
the Arts. 11 (4), 420.
45
Neural Activity During Jazz Improvisation
Rosen, D. S., Erickson, B., Kim, Y. E., Mirman, D., Hamilton, R. H., & Kounios, J. (2016).
Anodal tDCS to right dorsolateral prefrontal cortex facilitates performance for
novice jazz improvisers but hinders experts. Frontiers in Human Neuroscience, 10,
579.
Sauseng, P., Griesmayr, B., Freunberger, R., & Klimesch, W. (2010). Control mechanisms in
working memory: a possible function of EEG theta oscillations. Neuroscience & Biobehavioral
Reviews, 34(7), 1015-1022.
Sauseng, P., Klimesch, W., Freunberger, R., Pecherstorfer, T., Hanslmayr, S., & Doppelmayr, M.
(2006). Relevance of EEG alpha and theta oscillations during task switching. Experimental
Brain Research, 170(3), 295-301.
Sauseng, P., Klimesch, W., Gerloff, C., & Hummel, F. C. (2009). Spontaneous locally
restricted EEG alpha activity determines cortical excitability in the motor
cortex. Neuropsychologia, 47(1), 284-288.
Sauseng, P., Klimesch, W., Schabus, M., & Doppelmayr, M. (2005). Fronto-parietal EEG
coherence in theta and upper alpha reflect central executive functions of working
memory. International Journal of Psychophysiology, 57(2), 97-103.
Schacter, D. L., Addis, D. R., & Buckner, R. L. (2007). Remembering the past to imagine
the future: the prospective brain. Nature Reviews Neuroscience, 8(9), 657.
Scheeringa, R., Bastiaansen, M. C., Petersson, K. M., Oostenveld, R., Norris, D. G., & Hagoort, P.
(2008). Frontal theta EEG activity correlates negatively with the default mode network in
resting state. International journal of psychophysiology, 67(3), 242-251.
Scheeringa, R., Fries, P., Petersson, K. M., Oostenveld, R., Grothe, I., Norris, D. G., ... &
Bastiaansen, M. C. (2011). Neuronal dynamics underlying high-and low-frequency
EEG oscillations contribute independently to the human BOLD signal. Neuron, 69(3), 572-583.
Shi, B., Cao, X., Chen, Q., Zhuang, K., & Qiu, J. (2017). Different brain structures associated with
artistic and scientific creativity: a voxel-based morphometry study. Scientific Reports, 7, 42911.
Shi, L., Sun, J., Xia, Y., Ren, Z., Chen, Q., Wei, D., ... & Qiu, J. (2018). Large-scale brain
network connectivity underlying creativity in resting-state and task fMRI: cooperation
between default network and frontal-parietal network. Biological psychology, 135, 102-
111.
Stevens Jr, C. E., & Zabelina, D. L. (2019). Creativity comes in waves: an EEG-focused
exploration of the creative brain. Current Opinion in Behavioral Sciences, 27, 154-
162.
Szwed, J. (2012). The man who recorded the world: a biography of Alan Lomax. New
York, NY: Random House.
Wagner, A. D., Shannon, B. J., Kahn, I., & Buckner, R. L. (2005). Parietal lobe
46
Neural Activity During Jazz Improvisation
contributions to episodic memory retrieval. Trends in Cognitive Sciences, 9(9), 445-
453.
Wheeler, M. A., Stuss, D. T., & Tulving, E. (1997). Toward a theory of episodic memory:
the frontal lobes and autonoetic consciousness. Psychological Bulletin, 121(3), 331-
354.
Wise, R. J., & Braga, R. M. (2014). Default mode network: the seat of literary creativity?. Trends in
cognitive sciences, 18(3), 116-117.
Zeng, L., Proctor, R. W., & Salvendy, G. (2011). Can traditional divergent thinking tests be
trusted in measuring and predicting real-world creativity? Creativity Research
Journal, 23(1), 24-37.
Zumel, N., Mount J., & Porzak, J. (2014). Practical data science with R. Greenwich, CT:
Manning Publications Co.