[1] 22w Alpha, beta and gamma electrocorticographic rhythms in somatosensory, motor, premotor and prefrontal cortical areas differ in movement execution and observation in humans

[1] 39w Highlights ECoG activity was recorded in drug-resistant epileptic patients during movement execution and observation.  ECoG desynchronization and synchronization was lower during movement observation than movement execution.  This data support the existence of a mirror system in humans.

[1] 75w Mu rhythm is a human arch-shaped electromagnetic oscillatory activity recordable from central and parietal scalp regions overlying sensorimotor cortex, during muscle relaxation and absence of movement in wakefulness (Pineda, 2005). It is suppressed in amplitude ("desynchronized") during the preparation and execution of voluntary movements, as revealed by magnetoencephalographic (MEG, Salmelin and Hari, 1994a), electroencephalographic (EEG; Babiloni et al, 1999;Pfurtscheller and Lopes da Silva, 1999) and electrocorticographic (ECoG; Crone et al, 1998a;Miller et al, 2007) recordings.

[2] 256w The arch shape of the mu rhythm indicates that it is formed by two main dominant frequencies, namely a higher frequency oscillation (the tip of the arch) superimposed to a lower frequency oscillation (the basis of the arch). A spectral analysis of the mu rhythm unveils two clear peaks of power density in the alpha (i.e. 8-13 Hz) and beta (i.e. 14-24 Hz) bands, respectively (Hari and Salmelin, 1997). In the past, alpha and beta components of the mu rhythm have been investigated individually or together (Tiihonen et al 1989;Hari, 2006, for a review see Hari and Salmelin, 1997). Previous MEG evidence has shown that central alpha rhythms in resting state condition were mainly localized in primary somatosensory cortex, whereas central beta rhythms were localized a bit more anteriorly (Salmelin and Hari, 1994a). Both rhythms were desynchronized during the movement execution and increased substantially in the post-movement period (1-2 s), the beta component being about 300 ms faster than the alpha component (Salmelin and Hari, 1994b). Similar results have been found in scalp EEG recordings (Pfurtscheller and Lopes da Silva, 1999;Stancak and Pfurtscheller, 1995). A more fine spatial analysis of mu rhythm has been performed by means of subdural electrocorticographic (ECoG) recordings in epilepsy patients during their pre-surgical assessment for the localization of epileptic foci. It was shown that alpha and beta desynchronization accompanying motor tasks was mapped on the surface of both primary motor and somatosensory areas, confirming the composite nature of human mu rhythm (Crone et al, 1998a;Pfurtscheller et al, 2003;Miller et al, 2007).

[3] 115w A bulk of MEG and EEG findings have shown that the desynchronization of mu rhythm occurs not only during the execution of passive and active movements, as a reflection of underlying sensorimotor information processing, but also when subjects observe movements performed by other people, possibly as a reflection of the integration between motor and visual information processing underlying our ability of understanding the intentions of others (Pineda et al, 2005). This observation was firstly done by Gastaut and Bert (1954), who reported that central EEG mu rhythm desynchronizes during the observation of films depicting biological motion (e.g. a bike race, boxing, or a funeral) as a function of the identification with the subject on screen.

[4] 165w Later on, a number of MEG/EEG studies have shown that both movement execution and observation decreased the amplitude of beta synchronization following median nerve stimulation in a mathematical source model located in primary motor cortex (Hari et al, 1998;Rossi et al 2002;Muthukumaraswamy and Johnson, 2004a). Moreover, EEG alpha and beta rhythms desynchronized over widespread frontal, central, and parietal areas during the observation of movements performed by other people according to the characteristic of the movements (Cochin et al, 1998(Cochin et al, , 1999;;Babiloni et al, 2002;Martineau and Cochin, 2003;Pfurtscheller et al, 2007;Perry and Bentin, 2009;Neuper et al, 2009). For instance, it has been shown that the observation of a precision grip caused greater EEG alpha desynchronization than the observation of a flat-hand extension (Muthukumaraswamy and Johnson, 2004b) or precision grip gesture without any object (Muthukumaraswamy et al, 2004). Finally, beta rhythm modulations were related to kinematic features of aimed movements both during the execution and the observation of a motor act (Avanzini et al, 2012).

[5] 252w Keeping in mind the above findings, it can be speculated that human mu rhythm reflects the activity of large cortical neural populations oscillating at alpha and beta frequencies to map action observation on action execution motor programs. In monkey brain, single neurons of inferior frontal and parietal areas were termed as "mirror" as they are typically active during both movement execution and observation of the same motor act performed by a subject (di Pellegrino et al, 1992;Gallese et al, 1996;Rizzolatti et al, 1996). In this sense, they are supposed to play a key role in learning and in the ability of understanding the intentions of others from an embodied first-person perspective (di Pellegrino et al, 1992;Gallese et al, 1996;Rizzolatti et al, 1996;Gallese and Sinigaglia, 2011). More recently, "mirror" neurons have been localized even in monkey primary motor cortex (Dushanova and Donoghue, 2010;Vigneswaran et al, 2013). Based on these results, it has been proposed that alpha component of mu rhythm would represent the "negative" counterpart (i.e. suppression of a background EEG oscillatory activity) of the binding processes that synchronize human "mirror" neurons of an observer with those neurons sub-serving the execution of a movement (Pineda, 2005). In other words, alpha component of mu rhythm would reflect the synchronization and desynchronization of the cortical neurons sub-serving the elaboration of perceptual information linked to the movement of a subject as well as the resonance of its own sensorimotor systems from movement execution (Pineda, 2005). For sake of brevity, we call this theory as "Pineda's model".

[6] 44w Pineda's model incorporates a lot of experimental field data and inspired new experiments in this scientific field. However, it does not predict fine brain topography of the above oscillatory mechanism, as it is mainly based on scalp EEG and extra-cranial MEG studies (Pineda, 2005).

[7] 400w The reason of this limitation is that scalp EEG and extra-cranial MEG techniques have an insufficient spatial resolution to disentangle the fine topographic details of the contribution of primary somatosensory, primary motor, and premotor areas to the generation and modulation of the mu and gamma rhythms during the movement observation and execution. To overcome this limitation, during the pre-surgical monitoring of pharmacoresistant epileptic patients, Tremblay et al. (2004) performed subdural ECoG recordings that are characterized by high spatial and temporal resolution . This study has shown an alpha desynchronization in the primary motor cortex and Broca's area during both execution and observation of an aimless movement. However, this was a case study and the activity of only few brain areas was recorded. More recently, EEG alpha rhythms and functional magnetic resonance imaging (fMRI) were simultaneously measured while participants observed and executed hand actions (Arnstein et al, 2012). The results have shown that fMRI BOLD activity and alpha component of mu rhythm were negatively correlated during both observation and execution of actions in inferior parietal lobe, dorsal premotor, and primary somatosensory cortex (BA1-2) but not in Broca's area (Arnstein et al, 2012). Those findings provided direct support for the hypothesis that suppression of central alpha rhythms reflects the activity of primary somatosensory, premotor, and inferior parietal areas as parts of the" mirror" neuron system in humans. As such, they were not able to assess the role of alpha rhythms in other two putative "mirror" neuron brain regions such as primary motor and ventral premotor areas (Arnstein et al, 2012). This is a key issue in light of the idea that the alpha component of mu rhythm plays a special role in the visuo-motor transformation processes for the understanding of actions performed by other people (Pineda, 2005). Furthermore, several studies indicated that not only alpha component of mu rhythm but also its beta component plays an important role in matching movement execution and observation (Cochin et al, 1998(Cochin et al, , 1999;;Babiloni et al, 2002;Martineau and Cochin, 2003;Pfurtscheller et al, 2007;Perry and Bentin, 2009;Neuper et al, 2009). In addition, it has to be remarked that even though the modulation of gamma rhythms during movement observation was modest in previous scalp EEG findings (Neuper et al, 2009), this could be due to the intrinsic limitations of scalp electrodes in the recording of high-frequency EEG oscillatory activity (Perry et al, 2010;Wriessnegger et al, 2013).

[8] 105w As a further limitation, the Pineda's model does not predict the contribution of alpha and beta sub-bands as well as of gamma (around 40 Hz) band as neural oscillatory substrates of the information processing of movement observation and execution. It is well known that highfrequency alpha and beta components of mu rhythm and gamma rhythm are typically generated by circumscribed cortical sensorimotor regions according to the "motorotopy" representation of primary sensorimotor cortex, while low-frequency alpha and beta components of mu rhythm are generated by larger cortical regions of this sensorimotor cortex reflecting event-related cortical arousal (for a review see Pfurtscheller and Lopes da Silva, 1999).

[9] 110w In the present study, we recorded ECoG activity in epilepsy patients during the presurgical invasive investigation by subdural electrodes, in order to unveil fine details in frequency and spatial domains of the mentioned neurophysiological oscillatory mechanism underlying observation and execution of voluntary aimed movements. In this line, alpha, beta, and gamma sub-bands were taken into account in the ECoG data analysis. The hypothesis was that the oscillatory mechanism sub-serving movement observation and execution is characterized by different frequency and spatial features. It might be characterized not only by desynchronization of alpha rhythms in primary somatosensory, primary motor, premotor and prefrontal areas, but also by parallel beta desynchronization and gamma synchronization.

[1] 121w Electrode localizations Subdural ECoG electrodes were localized using an already described methodology (see Mattia et al., 2012). Briefly, 1 day after electrode grid placement a 3-D high-resolution computed tomography (CT; General Electric Light Speed Multi Slice, Milwaukee, WI, USA) was made. The CT scans of the electrode array were projected onto a brain template using the "Location on Cortex" software package (Miller et al, 2007), which returns the location of each electrode in Talairach coordinates. Finally, Talairach coordinates were converted in the corresponding Montreal Neurological Institute coordinates using a Talairach-Client application. Fig. 2 plots all electrodes of interest in a representative subject, projected onto the MRI brain used by the procedure for the co-registration between these electrodes and the Talairach space.

[2] 51w ECoG trials were used as an input for power spectrum analysis, which was performed with a standard fast Fourier transform (FFT) algorithm using the Welch technique and Hanning windowing function (Matlab; MathWorks, Natick, Massachusetts USA; 1 Hz frequency resolution). We performed the spectrum analysis at the alpha, beta, and gamma bands.

[3] 47w To study the alpha band, the individual alpha frequency (IAF) peak was identified according to the literature guidelines (Klimesch 1996(Klimesch , 1999)). The IAF was defined as the frequency that showed the maximum power within the band 6-13 Hz in the EEG or ECoG spectrum (Table 2).

[4] 55w The most "reactive" alpha frequency was identified subject-by-subject. This "reactive" frequency was defined as the frequency showing the maximum reduction in power density during the event period when compared to the pre-event period. Based on the alpha "reactive" frequency (ARF), the alpha band was subdivided in two alpha sub-bands (Pfurtscheller and Lopes da Silva, 1999).

[5] 28w The low-frequency alpha (low alpha) band ranged between ARF-2 Hz and ARF Hz, while the high-frequency alpha (high alpha) band ranged between ARF Hz and ARF +2 Hz.

[6] 97w To study the beta band, the individual beta frequency (IBF) peak was identified as the frequency that shows the maximum power density within the band 16-24 Hz in the EEG or ECoG spectrum (Table 2). The most "reactive" beta frequency was identified as it was done for the alpha band, i.e. the beta band was subdivided in two beta sub-bands (Pfurtscheller and Lopes da Silva, 1999) and low-frequency beta (low beta) band ranged between beta "reactive" frequency (BRF)-2 Hz and BRF Hz, while the high-frequency beta (high beta) band ranged between BRF Hz and BRF +2 Hz.

[7] 21w For each selected band, we evaluated the ERD/ERS values at each BA of interest (i.e. BA1-2, BA4, BA6, BA44, and BA45).

[8] 122w For each subject and each experimental condition, we performed the following computations. Thirdly, we tested whether the same electrode in a given BA showed maximum reactivity in all frequency bands and in the two conditions (EXECUTION, OBSERVATION). In most cases, different electrodes of a given BA showed maximum reactivity at different frequency bands, in line with the assumption that alpha, beta and gamma sub-bands represent different aspects of the same neurophysiological process. Furthermore, different electrodes of a given BA showed maximum reactivity for the EXECUTION and the OBSERVATION condition, in line with the idea that the neural substrates underlying movement execution and observation do not perfectly overlap. Fig. 3 plots the localization of all "reactive" electrodes of interest in a representative subject.

[9] 101w As shown in Figure 4, the grand average of ECoG power density spectra in the alpha band showed similar features during movement execution and observation. In particular, the ECoG power density spectra of pre-event period were characterized by an IAF peak at all BAs, particularly evident at BA1-2, BA4 and BA6. In both execution and observation conditions, and at all BAs, the event period (i.e. the period of movement execution or observation) was characterized by a clear reduction of the power density in the frequency range around IAF peak with respect to the pre-event period, especially at BA1-2, BA4 and BA6.

[10] 91w Insert here Figure 4 Also the grand average of the ECoG power density spectra in the beta band showed similar features in the two conditions, with the exception of BA1-2 (Figure 5). The pre-event period was characterized by a clear IBF peak at BA1-2, BA4, and BA6. During the event period, a reduction of the ECoG power density in the frequency range around IBF peak occurred at BA4, BA6 and BA44, but not at BA1-2. It is noteworthy that the IBF peak had a slightly lower amplitude than the IAF peak.

[11] 64w In the high gamma band the increase of the ECoG power density during movement execution with respect to the pre-event period occurred only at BA4, BA6, and BA45 (Figure 6). Furthermore, it was smaller than that observed at the low gamma band (Figure 7). During the movement observation, there was difference of gamma power density at any BAs with respect to the pre-event period.

[12] 17w Insert here Figure 7 Changes of ECoG power density during the event period compared to pre-event period

[13] 44w Using ANOVA designs described in the "Materials and Methods", we assessed whether ECoG power density at "reactive" electrodes differed between the event and the pre-event period. Table 3 plots full details of the statistically significant effects (p<0.05). Main facts are reported in the following.

[14] 103w Compared to pre-event period, alpha power density was lower (p<0.05) during the event period for both sub-bands and for the two conditions (EXECUTION, OBSERVATION). This was true at BA1-2, BA4, and BA6. In the same line, beta power density was lower at BA1-2, BA4, BA6 (in this instance there was a marginally significance) and BA44 in the events of the two conditions. Finally, gamma power density was higher (p<0.05) during the event period of the two conditions at BA6 and BA45 (both sub-bands) as well as at BA44 (low frequencies; p<0.005). At BA4, gamma power density was higher only during movement execution (p<0.005).

[15] 46w Using ANOVA designs described in the "Materials and Methods", we also assessed whether ERD/ERS at "reactive" electrodes differed in amplitude between the two conditions (EXECUTION, OBSERVATION). Table 4 plots all frequency bands and BAs showing statistically significant differences (p<0.05). Main facts are reported in the following.

[16] 65w Alpha ERD (both frequency bands) was higher during the movement execution than the movement observation at BA4 and BA6 (p<0.05). At BA44, this holds true only for the highfrequency alpha (p<0.005). The beta ERD was also higher during the movement execution than the movement observation at BA6 (p<0.05). Finally, gamma ERS was higher during movement execution than during movement observation at BA4 and BA6 (p<0.05).

[1] 163w Here subdural ECoG activity was recorded in drug-resistant epileptic patients during the execution of actions to reach and grasp common objects according to their affordances, as well as during the observation of the same actions performed by an experimenter. Extending the Pineda's model ( 2005), we tested the hypothesis that movement observation and execution emerge from different oscillatory neurophysiologic mechanisms and induce not only a desynchronization of alpha rhythms, but also a parallel beta desynchronization and gamma synchronization in primary somatosensory (BA1-2), primary motor (BA4), ventral premotor (BA6), and prefrontal (BA44, BA45) all areas encompassing human "mirror" neuron systems. These mechanisms are expected to sub-serve visuo-guided movements and understanding of movements performed by others (see fMRI studies by Buccino et al, 2004;Calvo-Merino et al, 2005, 2006;Cross et al, 2006). For the first time, this study used a methodological approach including (1) fine frequency and spatial analysis of ECoG activity and (2) an ecological social setting for the investigation of human "mirror" neuron system.

[1] 112w Subdural ECoG activity was recorded in humans to test the hypothesis that movement observation and execution show different neurophysiologic oscillatory activity as revealed by the modulation of alpha, beta, and gamma rhythms in a large portion of the core cortical network typically engaged in understanding the movements performed by others (i.e. primary somatosensory, primary motor, ventral premotor, dorsolateral and inferior frontal areas). The main results suggest that across frontal and primary somatosensory areas, alpha and beta desynchronization, together with more circumscribed gamma synchronization, were related to the execution of reaching and grasping common objects. Compared to movement execution, the observation of the same actions performed by an experimenter induced the following correlates:

[2] 233w (i) a smaller desynchronization of alpha in primary motor (BA4), in ventral premotor (BA6) and in inferior frontal (BA44) areas, (ii) a smaller desynchronization of beta in primary motor (BA4) area; (iii) a smaller synchronization of gamma in primary motor (BA4) and ventral premotor (BA6) areas. Overall our data support the existence of a mirror neuron system in humans, and they add important details about the way it works. Based on our findings, we put forward the following tentative explanation. On one hand, alpha and beta desynchronization (mu rhythm) would reflect the release of a background inhibitory mechanism gating visual, somatosensory, and motor inputs to widespread fronto-parietal and temporal areas in the control of movement execution and, to a lesser degree, during the understanding of movements performed by other people. On the other hand, gamma synchronization across premotor area and the inferior frontal gyrus would reflect the recruitment and binding of specific neural assemblies in the integration of motor and visual signals more specifically related to the control of movement execution based on somatosensory and visual stimuli. The role of gamma synchronization during the movement observation should be further investigated by intracerebral ECoG recordings ensuring a higher spatial resolution to capture more sparse and small-scale neurophysiological mechanisms and possible "inhibitory" neurophysiologic mechanisms. These results can be a gold reference of clinical applications including a neurophysiological diagnostic imaging of social cognition in epileptic patients.

[3] 28w Furthermore, it could be used as a reference for the rehabilitation or mapping of brain functions of scalp EEG markers in neurological patients with motor and social disorders.

[4] 93w Table Legends Table 1. Clinical data of patients affected by pharmaco-resistant epilepsy participating in the experiment. For each patient, age at the time of the surgery, gender, years since diagnosis, years of education, total intelligence quotient (IQ), positioning of the electrode grid, location of the seizure focus, and specimens pathology are reported. Table 2. Mean values (±standard error, SE) of individual alpha frequency (IAF) peak and of individual beta frequency (IBF) peak at 5 Brodmann areas (BAs) of interest (i.e. BA2, BA4, BA6, BA44, and BA45) for the EXECUTION and the OBSERVATION condition.

[5] 79w Table 3. Full details (F and p values, degrees of freedom) of the statistically significant effects of the ANOVA designs comparing ECoG power density at "reactive" electrodes between the event and the pre-event period of the two conditions (EXECUTION; OBSERVATION). The frequency bands of interest were alpha, beta, and gamma, while the BAs of interest were BA1-2, BA4, BA6, BA44, and BA45. The ANOVA designs are described in detail in the section on the statistical analysis ("Materials and Methods").

[6] 238w Table 4. Full details of the statistically significant effects of the ANOVA designs comparing ERD/ERS power density at "reactive" electrodes between the two conditions (EXECUTION; OBSERVATION). The frequency bands of interest were alpha, beta, and gamma, while the BAs of interest were BA1-2, BA4, BA6, BA44, and BA45. The ANOVA designs are described in detail in the section on the statistical analysis ("Materials and Methods"). Figure 2. The localization of all electrodes of interest in a representative subject (subj. #2). In the figure, these electrodes are projected onto the MRI brain used by the software package for the coregistration between the electrodes and the Talairach space (i.e. "Location on Cortex", Miller et al, 2007). The colors of the electrodes code Brodmann Areas (BAs) of interest. #2). In the figure, these "reactive" electrodes are projected onto the MRI brain used by the software package for the co-registration between the electrodes and the Talairach space. The colors of the electrodes code BAs of interest for all frequency bands of interest such as low-and high-frequency alpha, low-and high-frequency beta, and low-and high-frequency gamma. The equivalent figures for the remaining subjects are available as Supplementary Material, Figures S1-S6. Grand-average of electrocorticographic (ECoG) power density spectra in low-frequency gamma frequency range (from 36 Hz to 44 Hz) for 5 BA (i.e. BA1-2, BA4, BA6, BA44, and BA45), 2 time periods (pre-event , event), and 2 conditions (EXECUTION, OBSERVATION). Legend: "low" stands for "low-frequency".

[7] 44w Grand-average of electrocorticographic (ECoG) power density spectra in highfrequency gamma frequency range (from 55 Hz to 100 Hz) for 5 BA (i.e. BA1-2, BA4, BA6, BA44, and BA45), 2 time periods (pre-event , event), and 2 conditions (EXECUTION, OBSERVATION). Legend: "high" stands for "high-frequency".

[8] 111w Table 1. Patients Age at surgery (yrs) Sex Years since diagnosis Education (yrs) Total IQ Grid position Epileptogenic zone Pathology 1 41 F 28 8 98 Left frontotemporal Left mesial temporal lobe Hippocampal sclerosis 2 36 M 9 8 86 Right frontotemporal undetermined ---3 30 M 11 13 89 Left frontotemporal Left mesial temporal lobe Hippocampal sclerosis 4 36 M 6 13 93 Right temporooccipital Right posterior dorsolateral temporal Cortical microdisgenesis 5 47 M 17 13 81 Right lateralmesial frontal Right dorsolateral frontal Gliosis 6 28 M 8 8 92 Left frontotemporal undetermined ---7 22 F 1 13 80 Left frontotemporal Left frontal Focal cortical dysplasia Type II B (Taylor Type)

[9] 60w Table 2. Individual alpha and beta frequency peak (IAF, IBF) BA IAF IBF Execution Observation Execution Observation BA1-2 8.8 (±0.7) 8.7 (±1.4) 17.7(±0.9) 17.5 (±0.6) BA4 8.2 (±0.6) 7.5 (±0.9) 16.7(±0.9) 16.5 (±0.8) BA6 8.4 (±0.9) 7.9 (±1) 16 (±0.6) 17 (±0.8) BA44 8 (±1) 7.8 (±1) 16.8 (±0.9) 17 (±0.8) BA45 8.8 (±1.2) 7.7 (±0.9) 16.5 (±1.3) 16.7 (±0.8)

[10] 44w Table 4. Brodmann Area alpha band beta band gamma band BA1-2 not significant not significant not significant BA4 Condition factor: F(1,5)=12.9 p=0.015 not significant Condition factor: F(1,5)=13.5 p=0.01 BA6 Condition factor: F(1,6)=44.6 p=0.0006 Condition factor: F(1,6)=7.6 p=0.03 Condition factor: F(1,6)=6.36 p=0.05 BA44 interaction Condition

[11] 15w x Sub-band F(1,5)=23.6 p=0.004 not significant not significant BA45 not significant not significant not significant

[1] 75w Seven right-handed, adult volunteers (3 females, 4 males; age of 47.5 years ±5.9 standard error, SE) with intractable epilepsy were included in the experimental group. All subjects gave their informed consent and were free to withdraw from the study at any time. The general procedures were approved by the local Institutional Ethics Committee (IRCCS Neuromed, Pozzilli, Italy) and were performed in accordance with the ethical standards laid down in the Declaration of Helsinki of 1964.

[2] 108w To localize epileptogenic brain regions, all subjects underwent a comprehensive presurgical protocol described in details previously (Quarato et al, 2005), with the addition of an invasive ECoG investigation by subdural electrodes. Aside from the chronic epilepsy, clinical neurological examination was unremarkable (i.e. no sign of abnormality). Specifically, none of the subjects showed major overt cognitive deficits during the examination, as revealed by an interview about the general cognitive functions made by expert neurologists and psychologists of the IRCCS Hospital Neuromed. All subjects fully understood experimenters' instructions and easily performed the experimental task. Table 1 reports relevant demographic, clinical, and neuroradiological data in all patients included in the study.

[3] 307w During the ECoG recordings, the patients seated comfortably in their beds. Two experimenters were in the room. One was sitting in front of the patient, while the other was standing on the patient's right side. Two experimental conditions were administrated in pseudorandom order, namely OBSERVATION and EXECUTION. In both conditions, each trial started when experimenter I positioned a common object that differed in size, shape and function (cup, glass, and phone) on the plate of a mechanical device, which was placed on a small table between the patient and the experimenter II. After placing the object, the experimenter I gave a verbal instruction, which was either "Observe" or "Execute". In the OBSERVATION condition, the experimenter II reached, grasped and lifted the object following its affordances, while the patient stayed still. In the EXECUTION condition, the same action was performed by the patient, while the experimenter II stayed still. In both conditions, the mechanical device detected the first lifting of the object with respect to the plate sensor, and released a TTL pulse (reduced in amplitude by a voltage divider and decoupled through an optoisolator chip) to the ECoG data acquisition system signalling the zero-time of each trial. Afterward, the object was put back to the table by the agent (i.e. patient or experimenter II) and took away by the experimenter I, who changed the objects trial-by-trial in a pseudorandom order. Temporal gap between the onset of the verbal command of the experimenter I and the zero-time was on average of 2 seconds. Furthermore, the intertrial interval could randomly vary but it was never less than 15-20 seconds. The OBSERVATION and the EXECUTION condition included 60 actions each. Figure 1 illustrates the objects used for the experiment, the position of the patient and of the experimenter II, and the ECoG periods of interest for the subsequent data analysis.

[4] 75w The recorded ECoG, EOG and EMG data were analyzed and segmented in single trials of 10 s, each spanning -5 s to + 5 s. The ECoG epochs with ocular, muscular and other types of artifact were preliminarily identified by a computerized automatic procedure (Moretti et al 2003). The software package included procedures for (i) EOG artifact detection; (ii) EMG analysis for the detection of involuntary muscle activity of the patients during the OBSERVATION condition;

[5] 65w (iii) ECoG artifact analysis; and (iv) optimization of the ratio between artifact-free ECoG channels and ECoG single trials to be rejected. The ECoG epochs contaminated by ocular artifacts were then corrected by an autoregressive method (Moretti et al 2003). Finally, two expert electroencephalographists manually confirmed this automatic selection and only the ECoG epochs totally free from artifact residuals were accepted for the subsequent data analyses.

[6] 60w To quantify the event-related changes of ECoG power, we calculated the so-called eventrelated desynchronization/ synchronization (ERD/ERS; Pfurtscheller and Aranibar, 1979;Pfurtscheller and Neuper, 1994;Pfurtscheller et al, 1997). The ERD/ERS was defined as the percentage decrement/increment in the instant power density during a given event period as compared to the instant power density during a pre-event period. The formula was the following:

[7] 69w where E indicates the power density during the event period and R indicates the power density during the pre-event period. In the present study, the event period was the movement period (from 1 s before to 1 s after the zero-time), whereas the pre-event period was defined as the time interval between -4 s and -3 s before the zero-time. The procedure was repeated for all the selected bands.

[8] 50w Percent negative values (i.e. weaker ECoG power density during the event than pre-event period) represented ERD (Pfurtscheller and Lopes da Silva, 1999;Pfurtscheller et al 1997). On the contrary, percent positive values (i.e. higher ECoG power density during the event than pre-event period) represented ERS (Pfurtscheller and Lopes da Silva, 1999).

[9] 212w It should be remarked that the time periods for the analysis of EEG rhythms related to the movement execution or observation (e.g. pre-event from -4 s and -3 s before the zero-time; event from -1 s to +1 s) were chosen to avoid confounding effects of the EEG changes related to verbal command of the experimenter I, and the relative go (EXECUTION trials) or no/go (OBSERVATION trials) processes. Indeed, these verbal command and go or no/go processes are supposed to occur between -2.5 s and -1.5 s before the zero-time (e.g. first lifting of the object with respect to the plate sensor). For example, let us consider the extreme case that onset of the verbal command is given at -1.5 s before the zero-time, which means that a period of only 1.5 s passes from perception of the verbal command to first lifting of the object. According to previous evidence, verbal command and relative go and no/go processes are expected to induce the modulation of EEG rhythms in the following time windows: 1) theta from -1.4 s to -1.2 s (Huster et al., 2013); 2) alpha from -1.2 s to -1.0 s (Oniz and Başar, 2009); and 3) beta from -1.3 s to -1.0 s (Ritter et al., 2009;Swann et al., 2011).

[10] 9w The following two sessions of statistical analysis were performed.

[11] 61w Firstly, we evaluated the statistical significance of the low-frequency alpha, highfrequency alpha, low-frequency beta, high-frequency beta, low-frequency gamma and highfrequency gamma reactivity (p<0.05). To this end, 15 3-way analyses of variance (ANOVAs) were performed, namely 5 BAs (BA1-2, BA4, BA6, BA44, BA45) X 3 bands (alpha, beta, gamma). The dependent variable of these ANOVAs was ECoG power density. The factors (level)

[12] 23w were Time (Pre-event, Event), Condition (EXECUTION, OBSERVATION), and Sub-band (low frequency, high frequency). Post hoc tests were performed by Duncan test (p<0.05, one-way).

[13] 71w Secondly, we evaluated the statistically significant difference of the alpha, beta, gamma reactivity between the EXECUTION and the OBSERVATION condition (p<0.05). To address this issue, 15 2-way ANOVAs (5 BAs X 3 bands as the first session) were performed. The dependent variable of these ANOVAs was the ERD/ERS. The factors (level) were Condition (EXECUTION, OBSERVATION) and Sub-band (low frequency, high frequency). Post hoc tests were performed by Duncan test (p<0.05, one-way).

[14] 48w Unfortunately, we could not select the mentioned 5 BAs for all patients as the subdural electrodes were not positioned exactly in the same way in all participants. Only BA6 electrodes were represented in all 7 patients, while the BA1-2, BA4, BA44, and BA45 were represented in 6 patients.

[1] 139w The implanted ECoG electrode grids (Ad-Tech Medical Corp., Racine, WI, USA) consisted of circular (1 mm height, 4 mm diameter) platinum-iridium contacts embedded in a thin (0.5 mm) flexible transparent silastic plate and evenly spaced at 10 mm centers. All patients, but two had 64-contacts grids; two had a 48-contacts grid. The number and placement of electrodes were determined solely by clinical considerations. Electrodes were placed for pre-surgical monitoring in primary somatosensory (Brodmann area 1 and 2, BA1-2), primary motor (BA4), and prefrontal (BA44, and BA45). The electrode grid covered also lateral premotor cortex (BA6), mainly in its ventral part. Such localization covered most of core network for the control of visuoguided movements and for the understanding of movements performed by others, as revealed by previous fMRI studies (Buccino et al, 2004;Calvo-Merino et al, 2005, 2006;Cross et al, 2006).

[2] 48w ECoG data acquisition relied on a 128-AC-channel Beehive Millennium monitoring system (Grass Telefactor, West Warwick, RI, USA), with sampling rate of 400 Hz associated to an automatic anti-aliasing filter. All recording contacts were referenced to an electrode placed outside the skull (at the mastoid bones). Data were unfiltered.

[3] 42w Exploiting the same apparatus we also recorded the electromyogram (EMG) of the flexor digitorum superficialis of the dominant hand, the electrocardiogram, and both eye movements and blinking via an electro-oculogram (EOG) derived from electrodes placed over the lateral canthus of the eye.

[4] 137w Finally, given that there is not a unique definition of what frequencies characterize the gamma band, as other works have already done (e.g. see Crone et al 1998b;Szurhaj et al 2006), we considered two gamma sub-bands. Namely, we defined a low-frequency gamma (low gamma) band ranging between 36 Hz and 44 Hz and a high-frequency gamma (high gamma) band ranging between 55 Hz and 100 Hz (e.g. Miller et al. 2007;Miller et al. 2010). Our choice was grounded upon the evidence that temporal differences on low and high gamma might reflect relatively different neurophysiological mechanisms. Noteworthy, both gamma frequency bands of interest have the highest frequency limits well lower than the Nyquist frequency of 200 Hz, which is half of the sampling rate we used for the digitalization of the EEG data with an automatic anti-aliasing filter.

[5] 76w As far as the ECoG power density spectra during the pre-event period in the low and high gamma range is concerned, they were remarkably lower in amplitude than those of the alpha and the beta range. In the low gamma band, the ECoG power density during movement execution had a net increase at all BAs, but at BA1-2. In the OBSERVATION condition, the increase of gamma power density was limited to BA6 and BA44 (Figure 6).

[6] 205w In the above ECoG data analyses, we considered the single most "reactive electrodes" at each BA. One might argue that this "reactive electrode" could not reflect the overall BA reactivity. To address this issue, a control analysis assessed that overall reactivity by computing the mean values of ERD/ERS across all electrodes placed in a given BA. This analysis was done in those areas showing statistically significant (p<0.05) differences of the reactivity between the EXECUTION and the OBSERVATION (BA4, BA6, and BA44). In general, the results of this control analysis confirmed those obtained considering the single "reactive electrodes". However, overall alpha and beta ERD/ERS was generally reduced by more than 30% with respect to the corresponding values obtained using single "reactive electrode". This finding was due to the fact that some electrodes of a given BA exhibited ERD, while others showed ERS. This picture is in line with the well-known mechanism of focused alpha/beta ERD surrounded by alpha/beta ERS (Pfurtscheller and Lopes da Silva, 1999). On the whole, we concluded that peculiar oscillatory processes underlying movement execution and observation are better probed by the main analysis based on single "reactive electrodes" than by the control analysis averaging of ERD/ERS across all electrodes of a given BA.

[7] 120w Main analysis showed some statistically significant differences (p<0.05) in ERD/ERS between the two conditions (EXECUTION, OBSERVATION). The validity of this result depends on the assumption that the pre-event ECoG power density did not differ in the two conditions. To test this assumption, a control analysis was performed. T paired test (p<0.05) was used to compare the pre-event ECoG power density between the two conditions for the ERD/ERS values showing statistically significant differences in the relative BAs (see the section entitled "Changes of ERD/ERS at the different frequency bands between the movement execution and the movement observation"). Results of this control analysis showed no statistically significant difference of these pre-event ECoG power density values between the two conditions (EXECUTION, OBSERVATION) (p<0.05).

[8] 52w Alpha rhythms showed a clear power peak in all primary and secondary sensorimotor cortical regions of interest during the pre-event periods, and a remarkable desynchronization (i.e. cortical activation) during both movement execution and observation. Of note, the most prominent alpha power peaks and desynchronization were found in BA1-2, BA4, and BA6 areas.

[9] 208w Furthermore, the alpha desynchronization during movement execution and observation was found in both low-and high-frequency sub-bands. These results extend in terms of frequency and spatial details previous low-resolution EEG findings showing an alpha desynchronization over widespread scalp regions overlying prefrontal, sensorimotor, and parietal areas during the observation of movements performed by other people (Cochin et al, 1998(Cochin et al, , 1999;;Babiloni et al, 2002;Martineau and Cochin, 2003), as well as strict relationships between scalp central alpha desynchronization and relevant features of the observed movements (Muthukumaraswamy et al, 2004;Muthukumaraswamy and Johnson, 2004b;Avanzini et al, 2012). Furthermore, the present results on a more consistent number of subjects extended to BA4 and prefrontal cortex the findings of Tremblay et al. (2004) obtained from single subject. Finally, the present results complement fMRI-EEG data of Arnstein et al. (2011) who showed that the alpha component of the mu rhythm during movement observation and execution covaried in inferior frontal lobule and in dorsal premotor cortex and BA2. As a novelty, the present findings indicate that high-and lowfrequency alpha desynchronization occurs in BA1-2, BA4, and BA6 during movement observation and execution, indicating that this might be a feature of the "mirror" neuron system mechanism (unfortunately, we could not record activity in the posterior parietal cortex).

[10] 163w Differently from alpha rhythms, beta rhythms during the pre-event periods showed clear power peaks only in BA1-2, BA4, and BA6 areas. Furthermore, low-and high-frequency beta desynchronization (i.e. cortical activation) during movement execution and observation was maximum in BA1-2, BA4, BA6, and BA44. These results extend to premotor cortex previous (low-resolution) MEG and scalp EEG findings showing prominent source estimation in primary motor cortex of beta synchronization following a median nerve stimulation associated to both movement execution and observation (Hari et al, 1998;Rossi et al 2002;Muthukumaraswamy and Johnson, 2004a). The present results also extend previous MEG and scalp EEG findings showing beta desynchronization over widespread scalp regions overlying prefrontal, sensorimotor, and parietal areas during the observation of movements performed by other people (Cochin et al, 1998(Cochin et al, , 1999;;Babiloni et al, 2002;Martineau and Cochin, 2003;Avanzini et al, 2012). Finally, these results enlarge previous evidence showing beta desynchronization during action observation in premotor area F5 in monkeys (Kilner et al., 2014;Caggiano et al., 2015).

[11] 82w In the present study, alpha and beta desynchronization during the two conditions were not always localized at the same electrode within a given cortical region. This finding confirms previous MEG and scalp EEG findings showing that source estimation of central alpha and beta rhythms did not coincide in primary somatosensory and motor cortex (Salmelin and Hari, 1994a;Pfurtscheller and Lopes da Silva, 1999;Stancak and Pfurtscheller, 1995), as such it confirms the composite nature of human mu rhythm as frequency and functional source topography.

[12] 309w What is the role of alpha and beta desynchronization during movement execution and observation? There is consensus that alpha and beta rhythms (mu rhythm) reflect a tonic inhibition of somato-motor systems in the resting state condition and a release of this inhibition during cortical information processing associated to visual, somatosensory, and motor inputs in primary somatosensory, primary motor, and ventral premotor areas (Pfurtscheller and Lopes da Silva, 1999). In this theoretical framework, human alpha rhythms have a dominant role and affect our ability to accurately select, consciously perceive, judge, and memorize visual stimuli (Klimesch, 1999;Babiloni et al, 2004Babiloni et al, , 2005Babiloni et al, , 2006Babiloni et al, , 2007Babiloni et al, , 2008Babiloni et al, , 2009Babiloni et al, , 2010;;Mathewson et al, 2011), while beta rhythms might have a major role in motor functions (Ronnqvist et al, 2013). There is consensus that low-frequency alpha (8-10 Hz) and beta (16-20 Hz) rhythms might reflect global cortical arousal, tonic attention, and general activation of motor systems with a widespread cortical distribution. In contrast, high-frequency (10-12 Hz) alpha and beta (20-24 Hz) rhythms might reflect the activation of task-specific neural circuits with a more circumscribed cortical representation (Klimesch, 1999;Pfurtscheller and Lopes da Silva, 1999). The higher the alpha and beta desynchronization, the higher the local cortical activation. Furthermore, previous EEG-fMRI evidence showed the distributed cortical generation of alpha and beta rhythms (Mantini et al, 2007) and emphasized the important role of thalamus in the generation of alpha rhythms (Gonçalves et al, 2006). The higher the alpha synchronization (i.e. cortical inhibition), the higher the thalamic activation (Gonçalves et al, 2006). Finally, single neuron recordings in animal models showed that alpha rhythms are mainly generated by oscillatory activity of pyramidal neurons in response to salient visual inputs delivered by thalamo-cortical neurons (Lorincz et al, 2008(Lorincz et al, , 2009;;Hughes and Crunelli, 2005).

[13] 79w Keeping in mind the above findings, it can be speculated that widespread low-frequency alpha and beta desynchronization reflects a release of cortical inhibition associated to increased vigilance (alpha) and arousal of motor systems (beta) during movement execution and observation. In parallel, high-frequency alpha and beta desynchronization might reflect a release of cortical inhibition in neural circuits associated to motor commands and processing of somatosensory and/or visual feedback during movement execution and observation (Klimesch, 1999;Pfurtscheller and Lopes da Silva, 1999).

[14] 55w The present ECoG recordings provide an ideal methodology for detecting human gammaband activity (>40 Hz, see), when compared to previous scalp EEG recordings (e.g. Sheer et al 1966, Pfurtscheller et al., 1993, Shibata et al., 1999;Jerbi et al 2009). Indeed, small amplitude voltage fluctuations are not easily and reliably recordable with scalp electrodes (Singer, 1993).

[15] 178w We found that low-frequency (36-44 Hz) gamma synchronization (i.e. cortical activation) occurred in several frontal areas (e.g. BA4, BA6, BA44, and BA45) during movement execution, while it was generally negligible during movement observation. Furthermore, high-frequency gamma synchronization (i.e. cortical activation) showed higher amplitude in BA4 and BA6 during movement execution than during movement observation. In contrast, it showed similar amplitude in BA45, considered the homologous of the Broca's area in humans (Petrides et al 2005). These results extend previous evidence showing gamma synchronization during action observation in premotor area F5 in monkeys (Caggiano et al., 2015). It can be speculated that low-frequency gamma synchronization reflects active local motor processing along a distributed frontal circuit of small neural populations during movement execution (Crone et al, 1998a,b;Pfurtscheller et al, 2003;Engel and Singer, 2001;Fries et al, 2007;Srinivasan et al, 2013). On the contrary, highfrequency gamma synchronization might reflect the processing of both events in even smaller neural populations of ventral frontal lobe, where mirror neurons were first discovered in monkeys (di Pellegrino et al, 1992;Gallese et al, 1996;Rizzolatti et al, 1996).

[16] 303w An interesting finding of the present study is that ECoG desynchronization or synchronization in the frontal regions was generally lower during movement observation than during movement execution. At the present early stage of research, we cannot provide a conclusive explanation of this finding. Based on previous studies of single neuron recordings in BA4 (Vigneswaran et al. 2013) and in BA6 (Kraskov et al. 2009), it can be speculated that this finding is related to the features of neurons with "mirror" properties. . In monkey brain, one subgroup of these neurons showed an increase of the discharge during movement execution and observation (facilitation-type), while a second (less common) another population exhibited an increase of the discharge during the movement execution and to a decrease of the discharge during the movement observation (suppression type). It has been hypothesized that this second class of neurons prevents movement execution during the observation of actions performed by others (Kraskov et al. 2009;Vigneswaran et al. 2013). An alternative explanation posits a summation of cortical information processing (reflected by higher ECoG desynchronization/synchronization) related to movement observation and execution during the condition of movement execution. In the present study, experimental subject could see his/her own hand during the movement execution. This is a natural condition of human brain functioning, as most of human voluntary transitive hand movements of reaching and grasping are performed under visual control. However, this condition should be taken into account in the interpretation of the present results, as mirror neurons are view dependent as shown in humans by fMRI (Oosterhof et al., 2012) and in nonhuman primates by single neuron recordings (Caggiano et al. 2011). In monkey brain, mirror neurons sensitive to the observation of own grasping action were observed in BA4 and BA6 (Fadiga et al, 2013) and in posterior parietal cortex (Maeda et al. 2014).

[17] 173w Keeping in mind the above considerations, changes of alpha, beta, and gamma rhythms during the present condition of movement execution did not reflect only motor commands and movement-related somatosensory feedback, but also a visual brain response to own operating hand from a subjective perspective. With respect to the ECoG rhythms recorded during movement observation, those occurring during movement execution showed peculiar changes at all frequency bands both in terms of amplitude and functional topography, unveiling the modulation of a neurophysiological oscillatory mechanism in the most complex integrative processes related to the movement execution under visual control. The present findings call for future studies investigating the extent to which the present differences of ECoG activity between movement execution and observation are due to a peculiar synchronization/desynchronization of mirror neurons in relation to the mere observer's perspective view (perspective view centered on subject's head during the movement execution) vs. the relation to the correspondence between the observed moving hand and the somatosensory feedback (re-afferent somatosensory signals processed together with visual signals during the movement execution).

[18] 26w We found that during movement execution, there was a much more widespread alpha and beta desynchronization and gamma synchronization than that shown in previous ECoG studies.

[19] 106w Most of the previous studies explored just the sensorimotor cortex (Crone et al, 1998a;Crone et al, 1998b;Miller et al 2007;Pfurtscheller et al., 2003;Cheyne et al., 2008). There is just one study, but conducted on a single patient, showing alpha desynchronization during movement execution in an area involved in speech production (the putative Broca's area, Tremblay et al 2004). Thus, as far as we know, this is the first time that an alpha and beta desynchronization is reported in BA1-2, BA4, and BA6 simultaneously explored. The same is true for the observation of low-frequency and/or high-frequency gamma synchronization in BA4, BA 6, BA44 and BA45 simultaneously explored.

[20] 217w Differently from Tremblay et al (2004), we did not find a remarkable alpha desynchronization in the Broca's area (BA44 and BA 45). Possibly, the discrepancy with previous evidence could be explained with the type of movement performed in the present study, namely a goal-directed reaching movement. All the others studies employed non-goal directed actions such as finger movements (Szurhai et al 2006;Miller et al., 2007;Cheyne et al., 2008), elbow flexions (Cheyne et al., 2008), muscle contractions (Crone et al., 1998a, Crone et al., 1998b), and tongue protraction/retraction (Miller et al., 2007, Pfurtscheller et al., 2003). All these movements are likely to have at least partially different neural control systems with respect to reaching movements. In the real world (i.e. outside laboratory settings), only goal-related arm movements as those of the present study are crucial to sub-serve physical interactions with the environment by grasping and manipulating food. In addition, they are crucial to sub-serve direct physical contacts between individuals. Thus, it is reasonable that goal-directed reaching arm movements have at least partially different neural control systems with respect to the other mentioned movements (for experimental evidence, see Mirabella et al., 2009;Mirabella et al., 2011;Federico and Mirabella, 2014). Future studies should investigate the fine topographic organization of neurophysiologic oscillatory mechanisms underlying the neural control of different kinds of movements.