Abstract
The present study aimed at exploring adaptive mechanisms underlying the development of musical competence and, in particular, at qualifying and quantifying differences in cognitive functioning between people with and without musical training, as measured by electroencephalographic (EEG) and behavioral responses during an experimental task tapping into attention and monitoring mechanisms. Eighteen participants took part in the study. EEG responses to an omitted tone task were processed to compute their spatial components and time/frequency dynamics (power spectra, event-related spectral perturbation, and inter-trial coherence). In general, musicians showed greater EEG reactivity than control participants, which might signal adaptive changes linked to trained musical competence. Furthermore, musicians also performed better than controls, suggesting greater cognitive efficiency. Present findings also provide evidence that EEG is a valuable tool to help our understanding of adaptive mechanisms fostered by musical training and that it may complement behavioral methods to test performance.
Introduction
Studying music induces brain adaptations that are poorly understood and have practical implications. In fact, they may inform training programs to enhance performance, as well as rehabilitation and empowerment interventions to preserve cognitive/motor competences in clinical contexts.
Behavioral studies have shown that learning to play a musical instrument enhances cognitive domains related to specific sensorimotor (Forgeard, Winner, Norton, & Schlaug, 2008; Schlaug, 2015; Wan & Schlaug, 2010) and auditory skills (Koelsch, Vuust, & Friston, 2019; Martens, Wierda, Dun, de Vries, & Smid, 2015; Schulze & Koelsch, 2012; Strait & Kraus, 2014; Tierney, Krizman, & Kraus, 2015). Roden et al. (2014) reported that children with musical training performed better in speed tasks than children with science training. Similarly, Bergman Nutley, Darki, and Klingberg (2014) found that musical practice positively influenced processing speed, working memory, and reasoning. Concerning executive functions, there is evidence that learning music improves cognitive flexibility (Zuk, Benjamin, Kenyon, & Gaab, 2014), conflict monitoring (Bialystok & DePape, 2009), and sustained cognitive control (Pallesen et al., 2010).
Neuroimaging studies have shown the effect of musical practice on brain structural and functional organization (see Schlaug, 2015). Concerning white matter research, musicians have unique cross-hemispheric connections, especially at the level of corpus callosum (Elmer & Jäncke, 2018; Schlaug, 2015). Such unique patterns of structural connectivity in musicians have also been shown by diffusion tensor imaging studies, which highlight increased connectivity between the left and right planum temporale of people trained in music (Elmer, Hänggi, & Jäncke, 2016), increased diffusion in professional musicians’ corticospinal tracts (Imfeld, Oechslin, Meyer, Loenneker, & Jancke, 2009), and a unique asymmetry of the superior longitudinal fasciculus in musicians with absolute pitch (Oechslin, Imfeld, Loenneker, Meyer, & Jäncke, 2010). Nevertheless, the insufficiency in longitudinal studies and the heterogeneity in techniques prevent definite conclusions (Moore, Schaefer, Bastin, Roberts, & Overy, 2014). Concerning grey matter, musicians seem to have increased brain volume in auditory and motor areas, as well as the planum temporale, with partly contrasting pieces of evidence in favor of bilateral vs. left-lateralized plastic modification (Amunts, Schmidt-Passos, Schleicher, & Zilles, 1997; Elmer & Jäncke, 2018; Meyer, Elmer, & Jäncke, 2012; Schlaug, 2001). Morpho-volumetric changes have been reported in other cortical areas, such as Heschl’s gyrus (Schneider et al., 2005), Broca’s area, inferior frontal gyrus (Gaser & Schlaug, 2003; Sluming et al., 2002), primary somatosensory cortex, and supramarginal gyrus (Kleber, Veit, Birbaumer, Gruzelier, & Lotze, 2010). The current idea is that individuals who learn music in early childhood develop structural differences compared to the control population (Elmer & Jäncke, 2018; Meyer et al., 2012; Schlaug, 2015; Schulze & Koelsch, 2012).
Even electrophysiological studies have shown specific activation patterns in musicians’ brains. In particular, it has been proposed that musical training leads to improved auditory-evoked potentials and event-related changes of EEG oscillatory activity during listening to music, music-related stimuli (e.g., tones and scales), or even sounds (Baumann, Meyer, & Jäncke, 2008; Elmer & Jäncke, 2018; Ott, Stier, Herrmann, & Jäncke, 2013; Trainor, 2012). Notably, such functional changes associated with music training and expertise are so pervasive that they can be observed even at rest, with increased EEG-based intra- and inter-hemispheric functional connectivity between cortical areas involved in music perception and production in resting expert musicians with respect to non-trained people (Klein, Liem, Hänggi, Elmer, & Jäncke, 2016).
Nonetheless, notwithstanding the number of findings indicating that musical training uniquely and extensively shapes functional and structural features of musicians’ brains, the vast majority of available data focuses on differences in neurofunctional correlates of music perception and production between musicians and non-musicians, while neglecting differences in skills not exclusively related to music. In particular, electrophysiological correlates of the effect of musical training on musicians’ cognitive abilities (e.g., attention regulation and working memory) are still understudied. The present work focuses on such a debated issue.
The electroencephalography as a brain-imaging tool
The oscillatory activity of neurons is critical, enabling brain sensorimotor and cognitive functions (Başar, Schürmann, Demiralp, Başar-Eroglu, & Ademoglu, 2001; Buzsáki & Draguhn, 2004; Fries, Nikolić, & Singer, 2007; Klimesch, 1999; Palva & Palva, 2007; Singer, 2009). Electroencephalography (EEG) is a powerful tool that measures electrical brain potentials at the scalp surface with an excellent temporal resolution (Hari & Parkkonen, 2015) and increasing spatial resolution accuracy (Brodbeck et al., 2011; Holmes et al., 2010).
As for the potential for EEG as an investigation tool, while the traditional model of EEG-evoked responses has begun to be considered too simplistic to describe the processes triggered by a sensory/cognitive event, time/frequency models of EEG activity (e.g., event-related spectral perturbations [ERSP] and inter-trial coherence [ITC]), which simultaneously study a brain signal in both time and frequency domains (Bressler & Freeman, 1980; Makeig, 1993; Neuenschwander & Varela, 1993; Pfurtscheller & Aranibar, 1979; Tallon-Baudry, Bertrand, Delpuech, & Pernier, 1996; Weiss & Rappelsberger, 1996), are thought to properly reveal the spatio-temporal variability of such event-related processes.
Concerning the spatial features of EEG signals, it is known that electrical potentials arising from neural sources form time-varying voltage maps that can be recorded by scalp sensors and that – according to the principle of linear superposition – these maps are mixed. Independent component analysis (ICA) allows one to segregate independent signal sources that are linearly mixed at the level of sensors (Hyvärinen, 2013; Hyvärinen & Oja, 2000). Therefore, ICA is able to reveal spatial components corresponding to neural processes associated to different behavioral conditions (Jung et al., 2001a, 2001b; Makeig et al., 1999, 2002).
Therefore, according to recent evidence, cognition results from the mutual exchange of information between brain areas working together as large-scale networks (Bressler & Menon, 2010; Menon, 2015) and, in this vein, time/frequency transforms (i.e., power spectrum, ERSP, ITC) sketch an accurate picture of event-related brain dynamics, whilst ICA reveals their spatial pattern. Both approaches are, thus, powerful and synergistic to study the neurophysiological underpinnings of cognitive skills.
With the present pilot study we aimed at investigating behavioral responses, time/frequency dynamics, as well as spatial components of the oscillatory activity related to orienting of attention in a sample of musicians and non-musicians so as to better explore the effects of musical training on such cognitive skill by integrating electrophysiological and behavioral measures.
In particular, we used an omitted tone paradigm, where subjects had to press a button in response to the expected onset of an omitted tone as accurately as possible. In terms of cognitive domains, the exposure to stimuli with a predictable temporal pattern elicits exogenous temporal orienting of attention (Coull & Nobre, 2008; Doherty, Rao, Mesulam, & Nobre, 2005; O’Reilly, McCarthy, Capizzi, & Nobre, 2008), which is supposed to be more efficient in individuals with musical competence. Therefore, we supposed that musicians would outperform non-musicians in accuracy to predict the onset time of omitted tones, due to the empowerment effect of music training and expertise on auditory-motor-cognitive skills. Concerning time/frequency oscillatory brain responses, we hypothesized that musicians would show: (a) greater theta synchronization in the frontal midline areas, mirroring the efficiency of the attention/working memory network and according to evidence of greater frontal theta power during cognitive tasks (Aftanas & Golocheikine, 2001; Burgess & Gruzelier, 1997; Gevins, Smith, McEvoy, & Yu, 1997; Jensen, Gelfand, Kounios, & Lisman, 2002; Kahana, Seelig, & Madsen, 2001); (b) greater alpha desynchronization, which is positively related to the degree of cognitive performance (Klimesch, 1999); (c) stronger beta synchronization, indicating cortical activation during both cognitive (Fries, 2015) and motor performances (Pfurtscheller, Stančák, & Neuper, 1996; Stančák, 2000). Furthermore, by combining advanced time/frequency analyses and electromagnetic source localization techniques, we also expected that: (a) task-related theta synchronization would be primarily supported by frontal medial structures and anterior cingulate cortex (Ishii et al., 2014; Kahana et al., 2001); (b) task-related spectral modulations in the alpha range would be primarily supported by medial posterior, parietal and occipital structures, according to available functional-anatomical evidence (Hari & Salmelin, 1997; Rosanova et al., 2009); (c) task-related synchronization of beta oscillations mirroring neural activation would be primarily mediated by structures in the frontal and sensorimotor network (Rosanova et al., 2009; Stančák, 2000).
Materials and methods
Participants
Nine non-musicians and nine musicians gave their informed written consent to participate in the experiment. The pilot study was approved by a local ethics committee and conformed to the Declaration of Helsinki. Musicians and non-musicians did not differ in age or level of education. Table 1 summarizes demographic data. Musicians had received at least eight years of formal music lessons, could sight-read music, and practiced at least one musical instrument for more than 2 h per day. Non-musicians had never received any formal music lessons and could neither play nor read music. To verify normal hearing, audiograms were obtained from all participants. Musicians and non-musicians did not show significantly different intelligence or cognitive profiles. Exclusion criteria were: (a) hearing deficits; (b) prior episodes of head injury, cerebrovascular events or other clinical conditions that could have led to brain damage; (c) prior episodes of depression or other psychiatric disturbances; (d) concurrent therapies with psychoactive drugs.
Group Characteristics.
n.s.: not significant, M: male, F: female, L: left, R: right, * = unpaired t-test, °Chi-square test, °°Mann-Whitney test.
Experimental protocol and procedure
Spontaneous resting eyes-closed EEG was firstly recorded for 5 minutes, in order to determine the qualitative characteristics and the arousal state at the beginning of the experiment. The Epworth sleepiness scale was administered to rule out a strong tendency asleep during the day (Johns, 1991). The actual recording session was characterized by 10-minute resting EEG recording and 2.6-minute ERPs recording. Subjects performed two EEG/ERPs recording sessions with a 5-minute pause in between.
Omitted tone detection task
At the beginning of the task, subjects were instructed to anticipate forthcoming omitted stimuli by pressing a button in response to their virtual onset. Tones were presented at 3 s regular intervals; every fourth tone was omitted. ERPs were elicited with binaural 1000 Hz tones sent via headphones (75 dB SPL, 10 ms rise/fall, 60 ms plateau). The session was considered valid if the subject correctly detected at least 90% of the omitted tones. Two recording sessions were performed. Each session included 120 trials. Tone and no-tone trials were separated offline.
EEG recordings
EEG was recorded with a 32-channel DC amplifier (Mizar BElight device, EBN, Florence, Italy; GalileoNT1.5 acquisition software). The signal was recorded at Fpz, Fz, Cz, Pz, Oz, Fp1, F3, C3, P3, O1, F7, T3, T5, Fp2, F4, C4, P4, O2, F8, T4, and T6 electrode sites (International 10–20 system) with non-polarizable Ag/AgCl electrodes. The reference electrode was at linked mastoids with a forehead ground. Electrodes wires were braided together to reduce the effect of electromagnetic fields on the recording. An electrooculogram (EOG) was recorded bipolarly with electrodes placed above and at the outer corner of the left eye. Electrode impedance was less than 5KΩ and it was checked occasionally during the recording conditions.
EEG signals were recorded by using a 1–40 Hz bandpass filter and were subjected to an anti-aliasing filterbank (cut-off frequency: 30 Hz, 110 dB/octave). The amplified EEG analogical signals were digitized with a sampling frequency of 1024 Hz (16 bit). The presence of adequate EEG traces was determined by visual inspection on the computer screen.
Artifact rejection and data analysis
ERP epochs were time locked to tone and omitted tone onsets (starting at -2000 ms, ending at 2000 ms) and baseline corrected. We used the approach used in EEGLAB (Delorme & Makeig, 2004) for artifact rejection, which uses statistical thresholding to suggest to-be-rejected epochs. Such a semi-automated rejection approach was coupled with visual inspection. Therefore, after the automated routine, the epochs marked for rejection were confirmed by visual inspection. This procedure was integrated with an ICA approach, rejecting artifactual components by map profile. Trials were also rejected if the subject failed to press the button to omitted tones. Again, trials with long reaction time (more than 500 ms), as well as those in which the omitted stimulus was excessively anticipated (>1000 ms from its virtual onset), were rejected. A minimum of 70 artifact-free trials was required for analysis (average rejection rates < 10%; there were no statistically significant between-group differences).
Independent component analysis
The runica algorithm (Makeig, Jung, Bell, Ghahremani, & Sejnowski, 1997) was applied to the artifact-free EEG data from each subject to separate the EEG recordings, a combination of individual brain and non-brain sources mixed by volume conduction, into spatially static independent components (ICs) of maximal temporal independence. Runica was performed on 307 sec of EEG data for each subject. Thus, the number of time points used to estimate the weighting matrix was 35.6 times the minimum recommended amount for satisfactory decomposition (Delorme & Makeig, 2004). The total of 82 components for the 18 subjects was separated into clusters. For each dataset, the following component activity measures were calculated: spectra in the 3–25 Hz band (the dimension was reduced to 10 using PCA), ERPs, equivalent dipole location in spherical coordinates (3 dimension), scalp maps (the dimension was reduced from the number of channels to 10 using PCA), ERSP, ITC. The neural network algorithm (from the Matlab Neural Network Toolbox) was used to cluster ICs based on the 3-D locations of their equivalent dipoles.
Source localization
After performing ICA decomposition, equivalent current dipole models (ECD) were calculated for each component by applying the boundary element model in the DIPFIT toolbox (Oostenveld & Oostendorp, 2002). Electrode coordinates according to the standard 10–20 configuration were warped to the head model. Automated coarse-fitting to the boundary element model returned a single dipole model for each of the 82 ICs. Dipole localization implicates a back projection of the signal to a potential source that could have generated the signal, followed by calculating the best forward model from that source which includes the highest proportion of the signal recorded from the scalp (Delorme, Palmer, Onton, Oostenveld, & Makeig, 2012). The analysis was conducted to dipoles originating within the head, and the residual variance threshold was set to 15%. The RV is the mismatch between the original scalp-recorded signal and this forward projection of the ECD model and indicates the accuracy of fit measure for the ECD model. Group-level analysis was performed via EEGLAB module. It allows for the comparison of ICA data across individuals and conditions.
EEG data analysis
ERSP measures quantify average changes in power of the EEG frequency spectrum as a function of time relative to an experimental event. ITC assesses the consistency of local phase of a brain waveform across time-locked single trials compared with the baseline (Makeig, Debener, Onton, & Delorme, 2004a; Makeig et al., 2004b). Each trial contained samples from 2000 ms before to 2000 ms after the timelocking event. Both ERSP and ITC in the 3–30 Hz frequency range was computed in EEGLAB using Morlet wavelet decomposition, applied over 200 overlapping windows, starting with a 3 cycle wavelet at the lowest frequency (Delorme & Makeig, 2004), generating 200 time points (-1442.9 to 1441.9). The window size used was 1141 samples wide. ERSP and ITC values were transformed into log-units and converted to decibel units (dB), by multiplying the log ratio with the factor 10 (Grandchamp & Delorme, 2011). Concerning the ERSP, a reduction of the power relative to the baseline (i.e., event-related desynchronization) is displayed in blue, an increase in power relative to the baseline (i.e., event-related synchronization) is displayed in red. Concerning the ITC, strong phase synchronization is displayed in dark red. The power spectrum of the EEG was calculated for statistical assessment.
Statistical analysis
Group differences in age, gender, years of education, handedness, reaction time, and error rate data were assessed using parametric or non-parametric tests, where appropriate (SPSS 21.0, IBM Corporation, Armonk, NY, USA). Chi-square test was applied for the analysis of differences in gender, level of education, and handedness. Unpaired t-test with Welch correction was applied for values sampled from Gaussian distributions, whilst a Mann-Whitney test was performed when data did not follow a Gaussian distribution.
As for EEG data, to reveal changes across the power spectrum, the mean baseline log power spectrum from each spectral estimate was subtracted, obtaining the baseline-normalized ERSP. The bootstrap method was used to test the significance of deviations from baseline (EEGLAB Matlab toolbox, Delorme & Makeig, 2004).
Differently from the traditional ANOVA statistic, the bootstrap approach does not make the following assumptions: (a) normal distribution of the value of interest; (b) error terms are normally distributed; (c) known behavior of the parameters. Moreover, it is appropriate to solve the multiple comparisons problem (Delorme & Makeig, 2004; Efron, 1982). Practically speaking, a surrogate data distribution is established by selecting spectral values for each trial from randomly determined latency windows according to the epoch baseline (e.g., immediately preceding the stimulus onset), and then averaging these. Repeating this process several hundred times produces a surrogate baseline amplitude distribution whose specified percentiles are then taken as significance thresholds (Delorme & Makeig, 2004).
Results
Behavioral responses
Two non-musician participants gave, respectively, one and two out-of-time responses, whereas all musicians always gave their responses within the target time window. Musicians performed better than control participants at estimating the onset of omitted tones, t(16) = 8.20, p < 0.0001, Cohen’s d = 0.61, pow = 0.24 (Mcontr = 317.4, SDcontr = 234.0; Mmusic = 191.3, SDmusic = 174.2). Again, musicians showed lower intra-individual variability of reaction time over the trials than control participants, t(16) = 3.38, p = 0.004, Cohen’s d = 1.68, pow = 0.91 (Mcontr = 222.1, SDcontr = 38.9; Mmusic = 149.2, SDmusic = 47.0). All participants, regardless of the group, tended to signal the omitted tone before its actual omission, thus anticipating their responses (73% of trials), χ2(1) = 93.30, p < 0.0001, w = 2.28, pow = 0.99. More specifically, control participants anticipated the omitted tone onset in 76.9% of cases, whereas the musicians anticipated omitted tones in 71.1% of the cases, χ2(1) = 61.48, p < 0.0001, w = 1.85, pow = 0.98.
Cortical IC source clusters
According to cluster properties (i.e., scalp maps, dipole location, power spectrum profiles, ERSP and ITC indices) four clusters were associated to relevant brain processes. Table 2 summarizes the numbers of participants and sources constituting each cluster, their Tailarach coordinates, and the estimated association to Brodmann areas and cortical structures (Talairach & Szikla, 1980).
Synopsis of Main Features of Selected Clusters.
Cluster 6
Cluster 6 was a posterior cluster characterized by a peak in the upper-alpha band and a scalp map showing a left parietal lobe projection (Figure 1a). The dipole centroid was located in the left Brodmann area 31 (Talairach coordinates: x = 0, y = -43, z = 36; see Figure 1b). During the tone condition, musicians showed greater alpha power than controls participants across the 10.75 to 13.25 Hz range, p < 0.05. Similarly, during the omitted tone condition, musicians presented greater alpha power than control participants across the 10.5 to 12.25 Hz range, p < 0.05 (see Figure 1c).
Left Parieto-Occipital Cluster (Cluster 6). A: cluster scalp map. B: component cluster power spectrum (1–30 Hz). C: dipole locations for each individual contributing to the cluster displayed in blue, with the centroid location displayed in red. D: ERSP time/frequency plots for controls and musicians groups, respectively; significant differences between groups are based on bootstrapped statistics.
About ERSP indices, both groups showed an early alpha ERD and a mostly beta late ERS during the omitted tone condition. Nevertheless, alpha reactivity was significantly greater in musicians than control participants, p < 0.001 (frequency range: 7 Hz to 13 Hz; time window: 100 ms to 400 ms, see Figure 1d), suggesting a stronger cortical activation in correspondence to the expected onset of the omitted tone.
Cluster 2
Cluster 2 was a frontal cluster characterized by a peak in the upper-alpha band and a scalp map showing a frontal projection (Figure 2a). The dipole centroid was located in the right Brodmann area 24 (Talairach coordinates: x = 1, y = -23, z = 18, see Figure 2b). The analysis of ERSP indices comparing omitted tone versus actual tone conditions highlighted that only musicians showed a clear twofold ERS process. In particular, during the tone condition they showed greater early theta ERS (frequency range: 5 Hz to 8 Hz; time window: 130 to 400 ms), whilst during the omitted tone condition they showed a broad greater beta ERS (frequency range: 17 Hz to 28 Hz; time window: 150 ms to 1400 ms; see Figure 2c). These findings suggest that, comparing the tone and the omitted tone condition, Cluster 2 ERSP reactivity was relevant only in the musicians group.
Frontal Cluster (Cluster 2). A: cluster scalp map. B: dipole locations for each individual contributing to the cluster displayed in blue, with the centroid location displayed in red. C: ERSP time/frequency plots for control and musician groups, respectively; significant differences between groups are based on bootstrapped statistics. D: ITC time/frequency plots for control and musician groups, respectively; significant differences between groups are based on bootstrapped statistics.
Moving to ITC indices, musicians presented greater phase coherence than controls participants in the theta frequency band around the estimated onset of omitted tones, p < 0.001 (frequency range: 3 to 5 Hz; time window: -50 ms to 300 ms; see Figure 2d), suggesting a stronger neuronal phase synchrony of theta oscillatory responses.
Cluster 3
Cluster 3 was a posterior cluster characterized by a peak in the upper-alpha band and a scalp map showing a left dominant occipital projection (Figure 3a). The dipole centroid was located in the left Brodmann area 17 (Talairach coordinates: x = 0, y = -85, z = 7; see Figure 3b). During the tone condition, musicians showed greater alpha power than control participants, especially within the 12.75 to 14.5 Hz range, p < 0.01. Again, during the omitted tone condition, musicians showed greater alpha power than control participants, especially within the 12.75 to 15.5 Hz range, p < 0.01 (see Figure 3c). These findings suggest that musicians were characterized by greater upper-alpha reactivity than non-musicians during both the tasks.
Posterior Cluster (Cluster 3). A: cluster scalp map. B: dipole locations for each individual contributing to the cluster displayed in blue, with the centroid location displayed in red. C: component cluster power spectrum (1–30 Hz). D: ERSP time/frequency plots for control and musician groups, respectively; significant differences between groups are based on bootstrapped statistics.
As for ERSP indices, both control participants and musicians showed: (a) an early theta ERS occurring immediately after the tone omission; (b) an upper-theta/alpha ERD in the period immediately preceding and following the expected onset of omitted tones; (c) a later alpha/beta ERS response after the tone omission. Conversely, only musicians showed a late theta ERD, p < 0.001 (frequency range: 3 Hz to 4.5 Hz; time window: 670 to 870 ms; see Figure 3d), marking a late cortical response in the theta band after tone omission.
Cluster 5
Cluster 5 was a posterior cluster characterized by a peak in the upper-alpha band and a scalp map showing an occipital scalp projection (Figure 4a). The dipole centroid was located in the left Brodmann area 18 (Talairach coordinates: x = 0, y = -72, z = 2; see Figure 4b). The power spectra accounted for a dominant left occipital alpha 10.5 Hz activity. During the tone condition, musicians showed greater alpha power than control participants, especially across the 11.5 to 13.5 Hz range, p < 0.01. Similarly, during the omitted tone condition, musicians showed greater alpha power than control participants, especially across the 10.75 to 13.25 Hz range, p < 0.01 (see Figure 4c). These findings once again suggest that musicians show unique upper-alpha reactivity during both the tasks. About ERSP indices, both groups showed alpha ERD and weak theta ERS during, respectively, the omitted tone and the tone condition. Furthermore, only musicians presented a clear large alpha ERD during the omitted tone condition with respect to the actual tone condition, p < 0.001 (frequency range: 7 Hz to 13 Hz; time window: 100 to 900 ms; see Figure 4d), which suggests a specific cortical activation when the tone is expected but not presented.
Posterior Cluster (Cluster 5). A: cluster scalp map. B: dipole locations for each individual contributing to the cluster displayed in blue, with the centroid location displayed in red. C: component cluster power spectrum (1–30 Hz). D: ERSP time/frequency plots for control and musician groups, respectively; significant differences between groups are based on bootstrapped statistics.
Discussion
The present paper aimed at qualifying and quantifying specific differences in cognitive functioning between musicians and non-musicians, as measured by electrophysiological and behavioral responses in an experimental task tapping into attention and monitoring mechanisms, and at exploring potential EEG markers of such differences. With respect to the vast majority of available literature on neurofunctional and behavioral differences between people with or without formal musical training, the reported study did not focus on differences in perception, processing or execution of music or musical stimuli. It was instead devised to properly explore objective markers of potential cognitive empowerment effects that musical training might have provided with regard to participants’ cognitive profile and, in particular, to attention regulation and working memory skills.
More specifically, we analyzed both reaction times and time/frequency EEG oscillatory responses of a group of musicians and non-musicians during an omitted tone experimental paradigm. We found that musicians were more accurate to predict the onset of omitted tones, confirming that their auditory-motor skills were more efficient. Furthermore, the analysis of time/frequency oscillatory responses showed that expertise in music is associated to a greater theta, alpha, and beta reactivity, which may reflect an optimized response of brain oscillatory activities.
Behavioral responses
Musicians were more accurate than non-musicians in predicting the onset time of the omitted tone, indicating a greater efficiency of auditory-motor skills related to music expertise. Moreover, they had a lower degree of intra-individual variability in reaction time over the trials, which would further confirm their greater auditory-motor capacity. In fact, intra-individual variability measures usually have a negative correlation with intelligence scores (Barrett, Eysenck, & Lucking, 1986) and are known as a valuable biomarker of cognitive performance (Berkson & Baumeister, 1967; Jensen, 1992)
The results also showed that in both groups there was a tendency to anticipate the behavioral response (i.e., pressing the button before the instant of tone omission); this phenomenon is known as mean negative asynchrony (Aschersleben, Gehrke, & Prinz, 2001; Aschersleben & Prinz, 1995; Repp, 2005). As reported in previous research (Aschersleben, 2002; Franěk, Mates, Radil, Beck, & Pöppel, 1991), our data showed that the tendency for taps to precede the pacing stimulus is smaller in musicians than non-musicians, confirming a greater efficiency of the auditory-motor responses.
Power spectra: Alpha power is enhanced in musicians
Both groups had an alpha peak at the level of parieto-occipital region (i.e., clusters 3, 5, 6), according to the knowledge that the alpha rhythm is mostly generated at the level of parieto-occipital cortex (Hari & Salmelin, 1997).
In both tone and omitted tone conditions, musicians showed a greater alpha power than controls. It is known that EEG power reflects the number of neurons that discharge synchronously and, within the alpha frequency range, it is related to a greater cognitive performance (Klimesch, 1999). From a neurophysiological point of view, synchronized neuronal activity might regulate both local and inter-areal processing (Cardin et al., 2009; Lakatos, Karmos, Mehta, Ulbert, & Schroeder, 2008; Singer, 2009; Womelsdorf & Fries, 2006, 2007). It seems that alpha synchrony is positively related the individual working-memory capacity, particularly during memory maintenance, as well as increasing task difficulty (Palva, Monto, Kulashekhar, & Palva, 2010). Accordingly, our finding would indicate enhanced working-memory reactivity in individuals with musical competence.
Event-related spectral perturbations
Alpha induced responses are greater in musicians
During the omitted tone condition, musicians showed a twofold alpha ERD that was stronger than that observed in controls. The first process (i.e., cluster 6) had a dipole centroid located in the left BA 31, the second process (i.e., cluster 5) had a dipole centroid located in the left BA 18.
The role of alpha oscillations in the processing of cognitive information is well known (Başar, 2012; Fink, Schwab, & Papousek, 2011; Klimesch, 1999, 2012). Several studies showed that it reflects the focusing of attention during visual (Jensen et al., 2002; Jokisch & Jensen, 2007; Medendorp et al., 2007; Tuladhar et al., 2007), auditory, and somatosensory working-memory tasks (Haegens, Osipova, Oostenveld, & Jensen, 2010). From a functional point of view, the most accepted hypothesis is that an increased alpha power indicates a suppression of irrelevant information, whilst an attenuation of alpha power facilitates processing of relevant information in task-relevant brain regions (Klimesch, 2012; Klimesch, Sauseng, & Hanslmayr, 2007). More specifically, the alpha activity provides pulsed inhibition reducing the processing capabilities of a given area (Jensen & Mazaheri, 2010). An alternative hypothesis states that the alpha increase reflects active processing related to memory maintenance (Palva & Palva, 2007, 2011). In this debated theoretical context, it is not possible to attribute a definite functional significance of the observed alpha power changes. Therefore, our data only confirm the important role of alpha activity in cognitive paradigms. Furthermore, the greater ERD observed in musicians would indicate stronger activation of alpha oscillators along the cortico-cortical and cortico-thalamo-cortical circuits (Klimesch et al., 2007). Concerning the spatial analysis of alpha responses, it revealed two components. The first (i.e., cluster 6) was located in the left BA 31, corresponding to posterior cingulate cortex, which is an epicenter of the default mode network (Buckner, Andrews-Hanna, & Schacter, 2008; Greicius, 2008; Leech, Kamourieh, Beckmann, & Sharp, 2011; Margulies et al., 2009). To date, there is no full agreement about its functions (Leech, Braga, & Sharp, 2012). One authoritative hypothesis states that it plays a crucial role during tasks of internally directed attention (Buckner et al., 2008; Raichle et al., 2001). Accordingly, the greater alpha ERD observed in musicians would reflect a strong activation of this network during a task requiring endogenous orienting of temporal attention. The second component (i.e., cluster 5) was located in the left BA 18, corresponding to the visual associative occipital cortex. It had a frequency range in the lower alpha band and its temporal pattern was characterized by a sustained alpha ERD from 50 ms before the tone omission through the whole post-stimulus epoch. We argue that this sustained alpha activity corresponds well to the arrest reaction, which is a known EEG phenomenon elicited by a variety of tasks (Pfurtscheller & Aranibar, 1977) and indicating cortical excitation (Klimesch, 2012; Klimesch et al., 2007). From a functional point of view, it is assumed to correspond to general task demands and attentional processes (Klimesch, 1999, 2012). In this vein, the greater alpha ERD observed in musicians would indicate a more robust attentional response.
Musicians showed a greater theta induced activity
Comparing the omitted tone versus the tone condition, only musicians showed a stronger theta ERS during the latter. This process (i.e., cluster 2) had a dipole centroid located at the level of the anterior cingulate cortex. The greater theta synchronization during the tone condition could reflect the attention/working memory network in maintaining the information to accomplish the task. The finding of frontal theta synchronization is in agreement with previous research demonstrating that, during working memory tasks, theta power enhances over the frontal midline (Aftanas & Golocheikine, 2001; Burgess & Gruzelier, 1997; Gevins et al., 1997; Ishii et al., 2014; Jensen & Tesche, 2002; Kahana et al., 2001). Accordingly, the observation that musicians had a stronger theta ERS indicates a greater cortical activation and this reactivity might be related to their better behavioral performance.
During tone omission, musicians showed a late theta ERD (cluster 3) indicating a cortical inhibitory response. According to Staudigl, Vollmar, Noachtar, and Hanslmayr (2015), low-frequency power decreases would enable a neural assembly to express a stimulus-specific code by enabling a more information-rich temporal phase trajectory. Differently, when the same neural assembly is entrained to the same rhythm in situations of high synchrony, such a coding mechanism would work less efficiently (Hanslmayr, Staudigl, & Fellner, 2012). Therefore, theta ERD could represent a mechanism to enhance the neural coding capacity through the decorrelation of neural activity (Hanslmayr, Staresina, & Bowman, 2016). In support of this idea, other studies demonstrated an inverse relationship between theta power magnitude and memory capacity (Greenberg, Burke, Haque, Kahana, & Zaghloul, 2015; Long, Burke, & Kahana, 2014). In this vein, the late theta ERD observed in musicians during the tone omission could indicate a higher neuronal computational power. Concerning the spatial analysis of this theta responses, it showed a dipole centroid located at the level of the left BA 17 in the occipital cortex whose “natural” frequency of oscillation is in the alpha band (Rosanova et al., 2009). Recently, single pulse transcranial magnetic stimulation has demonstrated that, during specific attentional tasks, the occipital cortex may show periodic theta bursts reflecting periodic inter-areal communication for attentional exploration and selection (Dugué, Roberts, & Carrasco, 2016; Dugué & VanRullen, 2017; VanRullen, 2016).
Beta-induced responses are greater in musicians
During the omitted tone condition, musicians had a long-lasting beta ERS. It is known that a discrete enhancement of beta activity is observed across the sensorimotor network after termination of voluntary movement: the so-called post-movement beta synchronization (PMBS) (Pfurtscheller et al., 1996; Stančák, 2000). The finding that musicians showed a stronger beta ERS indicates a greater cortical activation during the task and this reactivity might be related to their better behavioral performance. Moreover, the finding that the process was located at the level of the right Brodmann area 24 (within the prefrontal lobe) would indicate the activation of the motor network.
Inter-trial coherence: Musicians showed an early theta response during the omitted tone condition
In the grand mean ITC plot, during the omitted tone condition, musicians presented early theta phase coherence (i.e., neural synchrony). From a functional point of view, theta synchronized activity has been related to working-memory and long-term memory processes (see Fell & Axmacher, 2011). In clinical research, it is well known that theta activity increases during attentional and working memory tasks (Gevins et al., 1997; Grimault et al., 2009; Jensen & Tesche, 2002; Klimesch, 1999; Raghavachari et al., 2001). In a visual evoked potential study, Kim, Grabowecky, Paller, Muthu, and Suzuki (2007) demonstrated that voluntary sustained attention enhanced stimulus-driven neuronal electrophysiological activity. In this vein, musicians’ early theta synchronization during the omitted tone condition might reflect increased sensitivity of attentional/working-memory system and better accuracy to predict omitted tone occurrence. Furthermore, even taking into account the spatial resolution limits of an EEG recording with only 21 electrodes, it is noteworthy that this process had a dipole localized in the anterior cingulate area, which is strongly involved during attentional/working memory tasks (Bissonette, Powell, & Roesch, 2013; Carter et al., 1998; Cole & Schneider, 2007; Constantinidis & Procyk, 2004; Wager & Smith, 2003), as well as in temporal orienting of attention (Nobre, 2001).
Conclusions
Findings show that musicians are more accurate to predict the onset of an omitted tone, highlighting their superior auditory-motor skills. Results also show that the analysis of time/frequency oscillatory responses is a valuable quantitative approach to study differences in cognitive performances. The focus on such advanced measures of EEG event-related modifications is actually among the main strengths of this work. The integrated set of time/frequency metrics, indeed, allowed us to sketch a clearer picture of the investigated cognitive processes and to overcome some of the limitations of more traditional time-domain quantifications. Namely, ITC measures may be used to investigate inter-areas communications, timing of cortical excitability changes, and cortical evoked activity, while ERSP measures may be used to investigate information complementary to those obtained by traditional averaging-based approach. In our case, expertise in music was related to greater theta, alpha, and beta reactivity. The network of cortical structures involved in the task – and, in particular, its medial frontal and parietal components – is known to modulate attention orientation processes and attention resources. Present findings hint at the potential of musical training in modulating and strengthening cognitive and attention skills and add to limited evidence on electrophysiological markers of cognitive improvements induced by systematic musical training.
Nonetheless, we acknowledge that the sample size of the present pilot study is limited. While reported findings are promising in the possibility to outline the electrophysiological profile of an expertise, larger-scale studies are required to confirm such first observations, to improve the statistical power of critical comparisons, and to strengthen related interpretation and implications. Again, strict definition of enrolled musicians and a comparison with different specialized musicians would allow sketching a better picture of the empowerment effects of systematic music training and of potential differences due to music specialization. Furthermore, future research should include dense-array EEG recordings, which are the conditio sine qua non to address the issue of a reliable spatial resolution of neural sources underlying an EEG signal. Again, future investigations would also benefit from including additional experimental tasks tapping on different cognitive abilities. This would allow to better clarify the extent of the effects of training on musicians’ cognitive profile, both at behavioral and at neurofunctional levels. As stated in the introduction, the investigation of the modulation of cognitive and electrophysiological profiles due to expertise (i.e., in this case, musical competence) has important practical implications, providing a quantitative measure to test markers of task performance that might be applied even in other domains.
