Download as pdf or txt
Download as pdf or txt
You are on page 1of 6

Effect of musical improvisation on inter-brain

wave synchronization
William Javier Betancourt Zapata, Dept. Ingenierı́a Biomédica, Universidad de los Andes, Bogotá, Colombia
wj.betancourt@uniandes.edu.co

Abstract—This article investigates brain synchronization dur- Studies conducted by Lindenberguer et. al. and Müller et.
ing the musical improvisation. Techniques such as phase locking al. [8], [9] employed EEG to examine brain synchronization
value (PLV) and envelope correlation were used to analyze the between two guitarists while they played together following
electroencephalography (EEG) signals of 5 pairs of musicians
playing together. The study found synchronization in frequency a score. The findings revealed increased intra-cerebral phase
bands (δ, θ, α, β1 , and β2 ) during performance and improvisa- locking during the establishment of preparatory tempo
tion above random fluctuations. The results suggest that both and during periods of musical coordination, particularly in
following a score and improvise, involve synchronization between the delta and theta frequency bands at frontal and central
brains and highlight the potential influence of music on brain electrode sites.
activity and interpersonal connections. However, the mechanisms
by which this synchronization occurs are not clear. Therefore,
further research is needed to explore more specific experimental Furthermore, in a study by authors in [11] used
approaches. magnetoencephalography (MEG) when people improvise
music -a process involving the spontaneous generation of
I. I NTRODUCTION novel and aesthetically appealing music. The study found
The brain interacts with the world through sensory that improvisation induced increased theta activity (5-7 Hz)
and motor processes, enabling actions that facilitate these in the left temporal areas associated with rhythm production
interactions. A significant aspect of these interactions and communication, increased alpha activity (8-12 Hz) in
involves coordinated actions towards shared goals among parietal and premotor areas associated with sensorimotor
multiple individuals. Consequently, there has arisen a integration, and decreased beta activity (15-29 Hz) in frontal
necessity to design experiments that assess the behaviour of areas associated with inhibitory control.
electrophysiological signals, providing insights into how the
brain interacts with world. Hyperscanning is a technique that This study aims to assess brain synchronization when mu-
facilitates the simultaneous measurement of neuronal activity sicians engage in musical improvisation after performing a
in two or more individuals. This technique has been applied score, using methods such as PLV and envelope correlation.
in various social contexts, such as during interactive gameplay The results of this research are expected to provide new ideas
[1], non-verbal tasks [2], communication involving speaking into how the brain processes and coordinates with other brains
and listening [3], and coordinated time estimation tasks [4], during musical improvisation.
among others. One commonly studied social environment is
music, and this project aims to investigate how the brain is II. M ETHODOLOGY
involved in situations where people perform music. A. Participants
Four (6) musicians participated (1 woman and 3 men,
Music, an activity practiced in all cultures since ancient mean age = 19 years, range 18-20 years) in this experiment.
times, has long been recognized for its emotional and social With these participants, a total of 5 couples played together.
impact on people. However, its effects on the human brain Previous studies showed that personal closeness [12] can affect
are now being explored. Neuroscience research has shown brain synchronisation, so each couple knew and played with
that music can induce intra-brain synchronization and greater each other before. Each person played instruments such as
functional connectivity between brain areas involved in piano, saxophone, and bass. None of the participants reported
musical perception and production [5]. An interesting line current psychiatric or neurological conditions or use of neu-
of research in this field is the possible brain synchronization ropsychiatric medication. The study protocol was reviewed
between two people when they perform music together. and accepted by the ethics committee of Universidad de los
Several studies have investigated this phenomenon using Andes and all the participants accepted and signed an informed
neuroimaging techniques such as EEG through metrics like consent.
phase coupling or envelope correlation [6]. The former
measures whether two signals are phase-locked in the B. Experimental Setting
observed time window and the later reflects the correlation of We designed an ABAB experiment, also called alternating
energy fluctuations in signals over time [6]. blocks [13]; first (A) performers followed a melody following
a score (control), immediately (B) performers started to play
Fig. 1. A and D) Experimental procedure, and order of experimental tasks. B) Signals obtained from participants. C) Position of the 8 electrodes chosen
according to the international 10-20 system.

spontaneously without following any music sheet. These two from 0.1 to 50 Hz. Subsequently, each EEG signal were
phases were alternated twice and three times, respectively, the normalized by calculating the Z-score using Equation 1.
improvisation time was set to one minute. Transition between
phase B-A was determined by a chime sound. Each session Channeli − µChannel
Zscorei = (1)
was sound recorded so as to identify the start of recording in σChannel
the EEG signal. • Synchronization measures. We calculated the phase syn-
chrony using PLV, following the methodology made by
C. Music material [14]. This is a measure of phase differences that varies
It was chosen 6 music sheets of jazz songs, so that per- between 0 and 1; where 1 indicates perfect phase locking
formers had the possibility to choose a song from a repertoire and 0 indicates no phase locking [1]. The following
of music sheets such as ’All Blues’, ’All The Things Your equation (Eq. 2) defines the PLV:
Are’, ’Autumn Leaves’, ’Guataca city’, ’Take Five’, and ’My T
Favorite Things’. See supplementary material 1. 1 X i(ϕt,n −ψt,n )
P LVt = | e | (2)
T i=1
D. EEG Data Acquisition
Where T is the number of sample points within the
Micromed LTM 64 was used for record EEG signals. Sam- window, ϕt,n and ψt,n are instantaneous phase values
pling frequency was set to 512 Hz. We chose to have an almost obtained using Hilbert transform, the phases of each time
complete representation of the cortex using the following eight series at time t. The second metric used was envelope
(8) channels from the international 10-20 system: Fp1, Fp2, correlation [15]. Here we obtained the signal amplitude,
T3, T4, C3 C4, O1, and O2 (see figure 1C). The reference defined as the absolute value of the Hilbert transform
electrode (Ref) of each person was placed in the Cz area of of the time series [16], we calculated this for each pair
the brain, and the ground (GND) on the right mastoid bone of channels and correlate these values using the Pearson
of each participant grounds were shared due to the limitation Correlation. The equation (3) [17] is presented below:
of the measure device.
PN
E. EEG Data Analysis (x1,n − x¯1 )(x2,n − x¯2 )
r(x1 ,x2 ) = qP n=1
• Preprocessing. The data were analyzed using custom N 2
PN 2
n=1 (x1,n − x¯1 ) n=1 (x2,n − x¯2 )
programs that were run in MATLAB (version R2020a, (3)
The MathWorks Inc.) Each EEG signal was synchronized
in time with the sound recording signal. Event triggers in Where N is the number of window samples, x1 and x2
EEG data, were identified for each change of block (A-B, are the envelope values of each channel, x1/2
¯ are the
B-A). EEG data was filtered with a band-pass FIR filter mean envelope values within the window.
Fig. 2. A) Representative signals within the data set. The graph above is from the score block and the graph below is from the improvisation block. Each
has pre-processed signals, signals filtered in β2 and the envelopes of the filtered signals. B) Statistical results from synchronization analysis at the different
frequency bands. The matrices contain the FDR-corrected p values from non-parametric permutation tests between each the envelope of each channel of each
participant and each frequency band vs. Surrogate. C) Envelope correlation metrics statistically different are shown.

These measures were calculated between each pair of III. R ESULTS


channels making a one-to-one comparison, that is, the We looked for differences between each block using a
synchronization was estimated between a channel of a t-test. We averaged all windows for each group score and
person (e.g. Fp1) and its corresponding channel of the improvisation to evaluate the differences of each group, that
other participant (Fp1), by using 5 second windows with is, using a single subject approach. However, we did not find
50% overlap. any statistical difference (pF DR > 0.05, see figure 3A in the
• Surrogate datasets. Surrogate data were created with appendix). -
signals taken randomly from different sessions. It was To overcome this, we decided to compare all the windows in
assumed that no significant increase in synchronization each group. Assuming that the synchronization is transitory,
would be found for the surrogate data since these signals we did the permutation test and in the case of the correlation
were not taken simultaneously [14]. These signals were metric by envelope, it was found that with a pF DR < 0.015
also given the same treatment for the analysis of synchro- there were significant differences between some pairs of
nization described for real signals taken by each pair of electrodes in some frequency bands (β2 in channels O2, T4
participants. and T3, and δ in channels O1, O2, T3 and C3). As we can
see in the appendix section in the figure 3B in the matrix on
F. EEG statistical analysis the left. Nevertheless, no statistically significant difference
For the brain correlation measures permutation tests for was found in the metric PLV (see appendix section in the
comparing two populations were adopted to perform statis- figure 3B, in the matrix on the right).
tical analysis [18]. All musical improvisation windows were
analyzed along with the same number of musical performance We wanted to discard that the differences were due to
windows. Furthermore, the aforementioned groups were tested random fluctuations. So we created a surrogate distribution
together with the same number of randomly selected surrogate with which we compare each group. These comparisons were
windows. Subsequently, the results were corrected using the made with the permutation test as before. The results are
false rate discovery (FDR) method [19] to correct possible shown in the appendix section (see figure 3C) and the figure
type one (1) errors. 2B. -
Regarding the significant results, we found differences in the analysis was performed while considering each value from
level of synchronization for almost all the bands and pairs of each window throughout the session. As shown in figure 3B,
channels, both in improvisation and in score with respect to the significant results were found in some frequency bands (β2
surrogate group. The delta band in the improvisation group did and δ) and in certain channels (O2, T4, T3, and T3, C3, O1,
not, for the most part, have significantly different results than O2, respectively), but only in the metric of synchronization by
the surrogate data (figure 2B). On the other hand, in figure envelope correlation. Besides, we found that synchronization
2C, we have the channels whose differences are significant values, in frequency band β2 , were statistically different in
between improvisation and score groups (δ and β2 frequency both blocks regarding surrogate data. The activity of the β2
bands). frequency band has been linked to mirror neuron activity
Finally, in the figure 2A two representative signals can be [21]. In our study, When a person plays an instrument, it
observed within the data set. In this part, the signal with is possible that the fellow exhibits similar brain activity. In
the preprocessing, the filtered signal in the β2 band and these cases, the second person would be brain-simulating the
the envelopes for each of the subjects in each block are same action as the first person. This phenomenon can occur
shown. In each of the figures, both dark and a light signals in response to different sensory modalities [21].
can be observed from the two individuals. Furthermore, it
is evident that the correlation is higher in the score block It has been recognized that the way in which the brain
compared to the improvisation block. Notably, in the score regions communicate is through brain synchronization [22].
block, synchronization is potentially occurring in the form of This allows the precise temporal coupling between different
transient spikes, which are marked with arrows. structures to give rise to cognitive processes. It has been
proposed that social interaction would be mediated by similar
IV. D ISCUSSION mechanisms, inter-brain synchronization would be needed for
As it was discussed in the introduction, the inter-brain a successful interaction.
synchronization has been correlated with a wide range of
social behaviors and characteristics. Specifically, there is It was found that there was brain synchronization when
evidence of joint activation (brain-to-brain) of brain regions compared with the surrogate data, the synchronization values
when people are playing music together [10], [11], [13]. were higher in the score block than in the improvisation
In the current study we validate and found that there is an block. This is possibly due to the types of synchronization that
inter-brain synchronization not only when people perform occur when following a score. According to literature, four
music together but also, when people improvise music in types of synchronization might occur: reciprocal, induced,
frequency bands such as δ, θ, α, β1 and β2 . driven, and coincidental [1]. When two people are interacting,
reciprocal or driven synchronization exist [22]. This means
First, all signals filtered were passed through the Hilbert that the brain signals of one person modulate or adapt to
transform [20], which allows obtaining an analytic the signals of the other person. In our case these types of
representation of a signal. It provides a way to extract communication are needed to achieve coupling. However, in
the instantaneous amplitude and phase information. the case of score performance, there is an additional element
This information was used for synchronization analysis, that possibly contributes to higher synchronization values.
either through PLV or envelope correlation, respectively. This element could be the score itself, they were following
Synchronization analyses were conducted based on the the same musical piece. This could provide another type of
comparison between groups. The first comparison involved synchronization - induced -. When two people are engaged
averaging the envelope correlation or PLV values for each in the same activity, their brain signals exhibit similar
block (A and B). The results of this comparison are shown characteristics in regions associated with that activity. In
in Figure 3A in the appendix section, where no significant hyper-scanning investigations, when two people are watching
findings were observed. This could possibly be attributed a movie together [24] their signals would synchronize even
to the behavior of synchronization in this type of task. In without any interaction between them.
the literature, we found articles that based the analysis on
small moments of a certain activity, in this case [9], the Although various studies have established a relationship
authors focused their analysis on the time period related to between synchronization and success in social interaction, the
the adjustment of preparatory tempo and the periods around specific underlying mechanisms for this phenomenon are still
the coordinated onset of musical performance. However, unknown. As mentioned earlier, two signals can synchronize
synchronization was not evaluated throughout the whole even without direct interaction. This highlights the importance
activity. Our findings are possibly indicating a transient of considering the stimuli experienced during the interaction
synchronization over time. and how individuals process activities and motor actions at the
level of the nervous system. In other words, the causal mech-
As mentioned in the previous section, assuming that anism responsible for harmonic social interaction has yet to
synchronization in this activity is transient and does not be identified. Furthermore, while effective synchronization has
persist throughout the session, the same synchronization been observed between signals using different measures, these
measures demonstrate symmetrical effects [24], indicating that [12] A. Czeszumski, S. Eustergerling, A. Lang, D. Menrath, M. Gersten-
both brains exhibit similar patterns of change. However, it’s berger, S. Schuberth, F. Schreiber, Z. Z. Rendon, and P. König, ”Hyper-
scanning: A Valid Method to Study Neural Inter-brain Underpinnings
important to note that during interactions, our signals may of Social Interaction,” Front. Hum. Neurosci., vol. 14, p. 39, 2020. doi:
not always be symmetrical, suggesting that interactions can 10.3389/fnhum.2020.00039.
be unidirectional rather than bidirectional. [13] M. Sasaki, J. Iversen and DE Callan (2019). Music Improvisation
Is Characterized by Increase EEG Spectral Power in Prefrontal and
Perceptual Motor Cortical Sources and Can be Reliably Classified From
A. Limitations Non-improvisatory Performance. Front. Hum. Neurosci. 13:435. doi:
10.3389/fnhum.2019.00435
The present study presents limitations and raises questions [14] A. Pérez, M. Carreiras, and J. A. Duñabeitia, ”Brain-to-brain entrain-
that could be addressed in future research. Firstly, the sample ment: EEG interbrain synchronization while speaking and listening,”
Sci. Rep., vol. 7, no. 1, p. 4190, Jun. 2017., doi: 10.1038/s41598-017-
size used was small. Additional studies with larger samples 04464-4
could provide additional and more reliable information about [15] A.G. Guggisberg, S. Rizk, R. Ptak, et al., ”Two Intrinsic Coupling Types
the mechanisms of synchronization during the coordination for Resting-State Integration in the Human Brain,” Brain Topography,
vol. 28, no. 2, pp. 318-329, 2015.
of social interactions. Besides, the synchronization analysis [16] A. Zamm, C. Palmer, A-KR Bauer, MG. Bleichner, AP. Demos and
performed in this article could benefit from focusing on shorter S. Debener. (2021) Behavioral and Neural Dynamics of Interpersonal
time periods in order to specifically understand which actions Synchrony Between Performing Musicians: A Wireless EEG Hyper-
scanning Study. Front. Hum. Neurosci. 15:717810. doi: 10.3389/fn-
allow for the coupling of signals, as reported in the literature. It hum.2021.717810
is also important to consider that the participating musicians [17] Z. Šverko, M. Vrankić, S. Vlahinić, and P. Rogelj. (2022).
in our study played different instruments, which raises the Complex Pearson Correlation Coefficient for EEG Connectivity
Analysis. Sensors (Basel, Switzerland), 22(4), 1477. https://doi-
question of whether this could affect brain synchronization. org.ezproxy.uniandes.edu.co/10.3390/s22041477
The majority of existing studies have focused on duets of [18] F. Butar, and J. Park, (2008). Permutation Tests for Comparing Two
musicians playing the same instrument. Populations. Journal of Mathematical Sciences amp, Mathematics Edu-
cation, vol. 3, 2: 19-30.
[19] Y. Benjamini, and Y. Hochberg. Controlling the false discovery rate: A
V. S UPPLEMENTARY M ATERIAL practical and powerful approach to multiple testing. Journal of the Royal
Statistical Society: Series B (Methodological). 1995;57(1):289–300.
The Supplementary Material for this article can be found [20] P. Clochon, J.M. Fontbonne, N. Lebrun, and P. Etévenon 1996. A
online at: Suplementary new method for quantifying 12. EEG event-related desynchronization:
amplitude envelope analysis. Electroencephalogr. Clin. Neurophysiol.
98: 126– 129.
R EFERENCES [21] G. Caetano, V. Jousmäki, and R. Hari. Actor’s and observer’s pri-
mary motor cortices stabilize similarly after seen or heard mo-
[1] A. Burgess. (Dec. 24 2013). On the interpretation of synchronization tor actions. Proc Natl Acad Sci U S A. 2007;104(21):9058-9062.
in EEG hyperscanning studies: a cautionary note. Frontiers in human doi:10.1073/pnas.0702453104
neuroscience vol. 7 881. doi:10.3389/fnhum.2013.00881 [22] F. Varela, JP. Lachaux, E. Rodriguez, et al.The brainweb: Phase syn-
[2] M. Ménoret, L. Varnet, R. Fargier, et al. (2013). Neural correlates of chronization and large-scale integration.Nat Rev Neurosci 2, 229–239
non-verbal social interactions: a dual-EEG study. Neuropsychologia. (2001). https://doi.org/10.1038/35067550
2014;55:85-97. doi:10.1016/j.neuropsychologia.2013.10.001 [23] J.G. Klinzing, N. Niethard and J. Born. Mechanisms of systems mem-
[3] A. Pérez, M. Carreiras, and J. A. Duñabeitia, ”Brain-to-brain entrain- ory consolidation during sleep. Nat Neurosci 22, 1598–1610 (2019).
ment: EEG interbrain synchronization while speaking and listening,” Sci. https://doi.org/10.1038/s41593-019-0467-3
Rep., vol. 7, p. 4190, 2017. 7:4190. Doi: 10.1038/s41598-017-04464-4 [24] A. F. de C. Hamilton, ”Hyperscanning: Beyond the Hype,” Neuron, vol.
[4] Y. Mu, C. Guo, and S. Han, ”Oxytocin enhances inter-brain synchrony 109, no. 3, pp. 404-407, 2021. DOI: 10.1016/j.neuron.2020.11.008.
during social coordination in male adults,” Soc. Cogn. Affect. Neurosci.,
vol. 11, pp. 1882-1893, 2016. Doi: 10.1093/scan/nsw106 VI. A PPENDIX
[5] S, Koelsch. (2014). Brain correlates of music-evoked emotions. Nature
Reviews Neuroscience, 15(3), 170-180. doi: 10.1038/nrn3666
[6] A. Zamm, S. Debener, A. R. Bauer, M. Bleichner, A. Demostraciones,
and C. Palmer. (my. 14, 2018). Amplitude envelope correlations syn-
chronous cortical oscillations in performing musicians. The New York
Academy of Sciences, 1423(1). Doi:
[7] I. Konvalinka, P. Vuust, A. Roepstorff, and C. D. Frith. (2010). Follow
you, follow me: continuous mutual prediction and adaptation in joint
tapping. The Quarterly Journal of Experimental Psychology, 63(11),
2220-2230. doi: 10.1080/17470218.2010.497843
[8] U. Lindenberger, S. C. Li, W. Gruber, and V. Müller. (2009). Brains
swinging in concert: cortical phase synchronization while playing guitar.
BMC Neuroscience, 10(1), 22. doi: 10.1186/1471-2202-10-22
[9] Sänger, J., V. Müller and U. Lindenberger. 2012. Intra- and interbrain
synchronization and network properties when playing guitar in duets.
Front. Hum. Neurosci. 6: 312.
[10] D. Dolan, J. Sloboda, H. Jensen, B. Cruts, and E. Feygelson. (2013). The
improvisatory approach to classical music performance: an empirical
investigation into its characteristics and impact. Music Perform. Res. 6,
1–38.
[11] J. Boasen, Y. Takeshita, S. Kuriki, and K. Yokosawa. (2018). Spectral-
spatial differentiation of brain activity during mental imagery of improvi-
sational music performance using MEG. Front. Hum. Neurosci. 12:156.
doi: 10.3389/fnhum.2018.00156
Fig. 3. A-B) Statistical results, averaging the window values in each group and comparing all the windows in each group, respectively, from the brain-to-brain
synchronization (Envelope Correlation and PLV) analysis at the different frequency bands. The matrices contain the FDR-corrected p values from nonparametric
permutation tests between each channel in environment of improvisation and following a score. None of the channel pairs showed any significant difference
pF DR < 0.05 for part A and in part B no statistically significant difference was found for the PLV metric and for the envelope correlation metric in some
channels (O1, O2, T3, C3, T4) in some bands (δ and β2 ) statistical differences were found. Rows represent frequency bands and columns each channel pair.
C) Statistical results from synchronization analysis at the different frequency bands. The matrices contain the FDR-corrected p values from nonparametric
permutation tests between each channel of each participant. Besides, each matrix contains the comparison between each group (A-B) with surrogate data for
PLV synchronization metric)

You might also like