Finger tapping to different styles of music and changes in cortical oscillations

Abstract Music has been a therapeutic strategy proposed to improve impaired movement performance, but there remains a lack of understanding of how music impacts motor cortical activity. Thus, the purpose of this study is to use a time–frequency analysis (i.e., wavelet) of electroencephalographic (EEG) data to determine differences in motor and auditory cortical activity when moving to music at two different rates. Twenty healthy young adults tapped their index finger while electroencephalography was collected. There were three conditions (tapping in time with a tone and with two contrasting music styles), and each condition was repeated at two different rates (70 and 140 beats per minute). A time–frequency Morlet wavelet analysis was completed for electrodes of interest over the sensorimotor areas (FC3, FC4, FCz, C3, C4, Cz) and the primary auditory areas (T7, T8). Cluster‐based permutation testing was applied to the electrodes of interest for all conditions. Results showed few differences between cortical oscillations when moving to music versus a tone. However, the two music conditions elicited a variety of distinct responses, particularly at the slower movement rate. These results suggest that music style and movement rate should be considered when designing therapeutic applications that include music to target motor performance.


Statement of Significance
The use of music as a therapeutic is becoming more popular in the treatment of those with Parkinson's disease and other movement disorders. However, there is little understanding of how music impacts sensorimotor activity that contributes to improvements in movement performance in these populations or in healthy populations. This exploratory study is among the first to provide the initial step in understanding how music modulates sensorimotor activity while moving compared to control condition. Results of this study will inform future studies aimed at understanding the use of music to facilitate movement in persons with movement disorders.
previous music experience was collected. See Table 1 for detailed participant demographics and music experience, the latter being obtained through self-report. All procedures were approved by the University Institutional Review Board, and all participants signed informed consent prior to data collection. This study was performed in accordance with the ethical standards as laid down in the 1964 Declaration of Helsinki and its later amendments.

Repetitive movement task
Methods for this study have previously been reported (Stegemoller et al., 2017). In short, participants were instructed to tap their right index finger along with the beat of an acoustic tone and two contrasting forms of music. The dominant forearm and hand were secured in a partial brace in the pronated position. The index finger flexion and extension movements were unconstrained and no tactile feedback was provided. The original pieces of music were composed so no participant had heard it previously. MIDI piano was the only instrumentation and both pieces were composed using part-writing conventions typical of early 19th-century Western classical practices. The musical pieces had distinct forms, one featuring an "activating" arrangement while the other featured a "relaxing" arrangement. While participants did not indicate if they thought the pieces of music elicited an activating or relaxing form, these categories are used to distinguish between the two music conditions in this study. See Table 2 for more specifics of the composition of each music condition. A complete description of the music has been previously published (Stegemoller et al., 2017(Stegemoller et al., , 2018a. Metronome clicks were inserted in the music conditions to ensure that participants were tapping in time to the same beat as the Tone Only condition. Two paces (70 BPM and 140 BPM) were presented for each condition. This resulted in data being obtained for the following six con-

Data processing
All EEG data were processed in Matlab using custom code with standard analysis practices (Krigolson, 2018). The code can be found at https://github.com/Neuro-Tools. Initially, data was inspected and all excessively noisy and/or faulty channels were removed from analysis.
The EEG data were then down sampled to 256 Hz and re-referenced using an average reference. Data were filtered using a dual-pass Butterworth filter with a passband of 0.1 to 30 Hz and a notch filter at 60 Hz. Following this, a restricted Infomax independent components analysis (ICA) was conducted to identify ocular artifacts (Delorme & Makeig, 2004;Luck, 2014  Following ICA, data were then reconstructed using the remaining ICA components. Specifically, following the removal of any ICA components that had the characteristics of a blink, inverse ICA was conducted whereby the matrix of ICA components is multiplied by the mixing matrix. The mixing matrix is simply the inverse of the matrix which was computed by the ICA algorithm to separate the original data into components which were maximally independent (see Makeig & Onton, 2011 for more details). Following the inverse ICA step, all removed channels were interpolated using the method of spherical splines. Any channel that was removed previously due to excessive noise was topographically interpolated through an interpolation algorithm which estimated the removed electrode's activity as per a weighted average of activity of the surrounding electrodes. The method of spherical splines is a form of spline interpolation which works by weighting the electrodes in a manner that best accounts for the dipole fields of the scalp when computing the removed electrode (Ferree, 2006). A time-frequency wavelet analysis using custom scripts (https: //github.com/Neuro-Tools), adapted from Cohen (2014), was implemented. All time-frequency analyses were conducted on single trials prior to averaging. The time-frequency wavelets were conducted on the pre-processed, segmented data by multiplying fast Fourier transformed EEG data with complex Morlet wavelets. As per the recommendation of Cohen (2014), we convolved the observed EEG signal with the product of a complex sine wave tapered by a Gaussian window.
The convolution window used a 4 ms step size. The number of cycles was varied across each frequency from 3 to 8 cycles. Specifically, the cycle parameter was 3 at 1 Hz, and the cycle parameter increased in a logarithmic manner until reaching 8 cycles at 30 Hz. We chose to vary the number of cycles to appropriately balance the time-frequency precision trade-off (Cohen, 2014;2019). In addition, the frequencies of the wavelet were between 1 and 30 Hz, with a step size of 30 linear steps. The window size for the Morlet wavelets was between −500 to 1000 ms, centered around EMG onset for both the Fast (140 BPM) and the Slow (70 BPM) tapping conditions. For the permutation test, we choose a reduced window size of −300 to 800 ms to avoid edge artifacts in the time domain.
The time-frequency wavelets were normalized within each condition through the use of a baseline. The Slow tapping condition (70 BPM) was baseline corrected between −500 and −300 ms pre-movement onset. In contrast, the Fast tapping condition (140 BPM) was baseline corrected using a window of −300 to −100 ms pre-movement onset.
We choose these separate baselines for the Slow and Fast conditions due to the possible overlap of previous finger taps in the Fast tapping condition if −500 to −300 ms had been used. The baseline procedure was divisive.

Data analysis
A cluster-based permutation testing was applied to all conditions (Cohen, 2014; see also https://github.com/Neuro-Tools). In order to compute the permutation test, at each time point and frequency, the average EEG activity across each condition for each participant were computed. We compared the following conditions: (1)  conditions. In order to avoid edge artifacts due to smearing, a permutation window that was between −300 and 800 ms was chosen. After extracting the data, the outputs were z-score corrected for each indi-  Figure 1 summarizes the results of all comparisons. Tables 3 and 4 show individual participant data.

Activating minus Tone Only
The analyses revealed no significant differences between the Tone Only condition and the Activating condition ( Figure 2 (Figure 3).

Activating minus Relaxing
Finally, the Activating and Relaxing conditions were compared and a number of differences were observed, with the Activating

DISCUSSION
The purpose of this study was to compare the timing of cortical oscillations when moving to music versus moving to a Tone Only at two F I G U R E 4 Activating minus Relaxing for (a) frontal electrodes, (b) central electrodes, and (c) temporal electrodes for 70 and 140 beats per minute (BPM). Black contour lines indicate differences that survived the cluster-based permutation test. Black lines indicate statistical differences in the power spectrum between conditions different rates. Results revealed that timing was mostly unaffected by the experimental condition. For movement to activating music versus a tone, there were no significant differences. For movement to relaxing music versus a tone, differences occurred in the beta band over electrodes FCz (70 and 140 BPM) and the alpha and beta band over electrode Cz (140 BPM). This is in keeping with our previous study that showed a significant increase in beta band power at 70 BPM and a significant increase in alpha band at 140 BPM when comparing both music conditions to the Tone Only condition (Stegemöller et al., 2018). Our previous study examined evoked activity from the motor response by comparing differences across the power spectrum without accounting for changes in power over time. Analyses in this study was intended to capture the single trial activity (including both induced and evoked activity). Thus, the differences revealed may indicate similarities in induced and evoked activity over the sensorimotor cortex when moving with music. However, these results are far from conclusive, and there is still a need for future studies to parse out the effect of music on sensorimotor activity.
Cortical activity differs between low and high rate repetitive movements. During low rate repetitive movements cortical activity is characterized by a desynchronization of oscillations in the alpha and beta followed by synchronization between movements (Erbil & Ungan, 2007;Stegemöller et al., 2016).
At higher rate movements, sensorimotor cortical activity is characterized by near continuous desynchronization (Muthukumaraswamy, 2010;Stegemöller et al., 2016;Toma et al., 2002). Interestingly, in this study most differences in alpha and beta band oscillations recorded over the sensorimotor areas (electrodes C3, C4, Cz, FC3, FC4, FCz) occurred throughout the movement cycle, from roughly 200 ms before to 600 ms after movement onset for both movement rates. Given that the alternating sequence of desynchronization and synchronization of alpha and beta band oscillations is thought to reflect the sensorimotor activity associated with suppression and release of movement (Pfurtsheller et al., 1999), the results of this study may suggest that moving to music may impact both the suppression and release of movement. Yet, there were no differences revealed in cortical oscillations recorded over sensorimotor areas for the activating versus Tone Only condition suggesting that other factors, such as stylistic components of the music were not accounted for. Alternatively, the observed differences in cortical activity could be related to differences due to the participants' preference in the music used in the present study.
Interestingly, a number of differences were revealed when comparing the two music conditions (i.e., activating versus relaxing music).
For those comparisons, significant differences occurred in electrodes FC3 (140 BPM), Cz (140 BPM), C4 (70 BPM),and T7 (70 BPM). Previous research has suggested that music style may impact sensorimotor activity (Janata et al., 2012;Witek et al., 2014;Stegemöller et al., 2018b;Izbicki & Stegemöller, 2020). Faster tempo, moderate syncopation, and repetitive rhythm elicit a greater urge to move while slower tempo, excessive syncopation, and non-repetitive rhythm elicit little to no urge to move (Janata et al., 2012;Witek et al., 2014). Thus, the two contrasting styles of music used in this study were designed with these details in mind. The intention was that the activating music would elicit a greater urge to move than the relaxing music, which may be reflected in differences in sensorimotor cortical oscillations.
Our results revealed that there was a significant decrease in both alpha band and beta band power for multiple electrodes during the Activating condition compared to the Relaxing condition. Given that a decrease or desynchronization in alpha and beta band power may indicate release of movement, the results of this study may indeed support the notion that music style that is designed to increase the urge to move may be reflected in sensorimotor oscillations. However, participants did not indicate if the activating style elicited an urge to move over the relaxing style. Thus, an alternative explanation for differences between the two styles of music may be related to music preference, as we posited above.
The only difference in cortical oscillations recorded over auditory regions emerged when comparing the two music conditions and was in the alpha band. Previous research has suggested that changes in alpha band power over the auditory regions represent a change in listening effort (Marsella et al., 2017;Wisniewski et al., 2017;Wöstmann et al., 2015). A decrease in power may indicate a decrease in listening effort and may suggest that participants in this study displayed more listening effort during the Relaxing condition. Given that the order of the conditions were randomized, the differences may be driven by mechanisms other than fatigue. Given that the activating style was designed with the intent to elicit a greater desire to move than the Relaxing condition, perhaps participants did not need as much listening effort to determine when to synchronize movement. Conversely, the differences in alpha power may be reflective of music preferences in which participants devoted less listening effort to the less preferred style of music. Indeed, participant cohort in this study tended to prefer the relaxing music over the activating music. However, continued research is needed to parse out if the responses revealed in this study are due to differences in the music stimuli or differences in participant factors, such as preference.
Nonetheless, results of this study indicate that cortical oscillations over the auditory and sensorimotor areas are influenced by differing styles of music.