Thirty healthy volunteers (24.4 ± 2.5 years; 11 males) participated in the study. All participants had normal or corrected-to-normal visual acuity, were free of medication and none reported a history of major medical disorders, sustained head injury, psychiatric or neurological disorders, alcohol or drug abuse. They were all right-handed according to a twelve-item version of the Oldfield Edinburgh handedness inventory (Oldfield, 1971). They all provided written informed consent and the study was approved by the Ethics Committees of the Faculty of Psychology and Educational Science (University of Geneva) and of the University of Lausanne.

Participants were comfortably seated in front of a computer-controlled screen at a distance of 60 cm. The motor priming paradigm involved two types of visual stimuli, subtending a visual angle of 4.8° × 1°: a preparatory stimulus (S1) presented for an interval (foreperiod) varying between 1800 and 2200 ms, and an imperative stimulus (S2) lasting until the subject responded or 2000 ms elapsed. The inter-trial interval varied randomly between 1000 and 1500 ms (Figure 1). The tasks consisted of unilateral key presses with the left or right index finger under three different pre-cued conditions: (1) Full, where S1 provided complete advance information about response side; (2) Free, where S1 invited the subject to select the side of response and (3) None, where S1 was uninformative on response side. S2 provided complete response side information in the Full and None conditions and was neutral in the Free condition. Participants performed the Full, Free and None conditions successively (26 trials per condition) and repeated the experiment once (total of 52 trials per condition). The block design ensured that subjects remained concentrated within each condition, reducing errors and omissions. The intrinsic trial-by-trial variance of the choice-reaction-time paradigm helped to reduce the habituation and/or anticipation effects associated with fixed block designs.

Electroencephalogram (EEG) was continuously recorded during the experimental phase using 64 scalp electrodes at a sampling rate of 2048 Hz (BioSemiActiveTwo, Amsterdam, Netherlands). As reference, the system utilizes a feedback loop between two separate electrodes located over the posterior region. EEG channels were then referenced off-line to the mean voltage of all 64 channels (average reference). Electrode impedances were kept below 5 kΩ. The electro-oculogram was recorded using two pairs of bipolar electrodes in both vertical and horizontal directions. Visual stimuli and button presses were automatically documented with markers in the continuous EEG file.

Reaction time was measured as the time interval between S2 onset and response key press. Accuracy was defined as the percentage of correct responses. Incorrect responses included errors (wrong key press) and omissions (absence of response within the 2000 ms-response window).

EEG data analysis was conducted using BrainVision Analyzer 2 software (Brain Products GmbH). Continuous EEG data were downsampled to 256 Hz and corrected for blinks and eye movement artifacts through an independent component analysis (ICA) (Jung et al., 2000). For EEG power analysis of the preparation interval, EEG data were segmented into S1-locked epochs of 5200 ms, starting 1000 ms before S1. Only the 1800 ms after S1 were analyzed (minimum foreperiod common to all trials). To obtain a precise evaluation of the signal preceding S2 for MRAA analysis, data were segmented into S2-locked epochs starting 2750 ms before S2. The epochs were automatically scanned for contamination by muscular or electrode artifacts (criteria for rejection: voltage step >50 μV/ms or peak-to-peak deflection within 300 ms intervals >200 μV) and the remaining trials were inspected visually to control for residual minor artifacts. Only artifact-free EEG trials corresponding to correct responses were analyzed, reaching the average number of 48 per subject in the Full and Free conditions and 47 per subject in the None condition.

A time-frequency (TF) analysis, based on a continuous wavelet transform of the signal (complex Morlet's wavelets), was performed between 8 and 13 Hz in 1 Hz steps (Tallon-Baudry et al., 1998). The mean power of the prestimulus interval was considered as a baseline level and subtracted from prestimulus and poststimulus power, independently for each frequency, providing a measure of relative power. For analysis of the preparation period, prestimulus interval corresponded to −650 to −150 ms before S1 onset. For MRAA analysis, prestimulus interval was selected between –2450 and –2100 ms before S2 onset (thus also preceding S1). TF analysis during the preparation period was performed in each condition for right and left hand responses combined. MRAA were obtained for each condition and response hand from relative power (P) at selected pairs of central electrodes (e.g., C3, C4) according to the following formula: MRAA(C3,C4)=((PC4−PC3)left hand+(PC3−PC4)right hand)/2. Thus, the larger ERD over the region contralateral than ipsilateral to movement was reflected as a negative MRAA. Since the largest alpha-band MRAA was obtained for the CP3/CP4 pair, as previously reported (McFarland et al., 2000), further statistical analysis was restricted to this pair.

RT and accuracy were compared across conditions using a one-way repeated measures ANOVA and post-hoc analysis used paired sample t-tests with a statistical threshold of p < 0.01.

EEG power was averaged in the alpha frequency range (8–13 Hz) within eight regions of interest (ROIs) in nine 200 ms time windows starting at S1 onset (0–1800 ms). Regions of interest were defined as follows: Left Motor (LM, C3, C5, CP3, CP5), Right Motor (RM, C4, C6, CP4, CP6), Left Premotor (LPM: FC3, FC5), Right Premotor (RPM: FC4, FC6), Left Parieto-occipital (LPO: PO7, O1), Right Parieto-occipital (RPO: PO8, O2), Midparietal (CUN: Pz, POz) and Midfrontal (SMA: Cz, FCz). Comparison of alpha relative power between conditions was performed using a 3-way repeated measures ANOVA, with Condition, ROI and Time as within-subject factors. Post-hoc analysis used paired sample t-tests with a statistical threshold of p < 0.01. For each condition and 200 ms time window, the CP3/CP4 MRAA difference from 0 was tested using one-sample t-tests with a statistical threshold of p < 0.01. To examine potential relationships between the preparatory alpha activity and the degree of sensorimotor alpha lateralization, the TF data were further analyzed using a tertile split. The MRAA amplitudes in the last 200 ms window preceding S2 in the Full condition were divided into low, medium and high values. Relative alpha power was then compared between the upper (high lateralizers, N = 10) and lower (low lateralizers, N = 10) tertiles, subjects falling into the medium tertile being discarded from analysis. In each Full, Free and None condition, a 3-way repeated measures ANOVA was performed with ROI and Time as within-subject factors and the Lateralization subgroup as a between-subject factor. Post-hoc analysis used independent sample t-tests with a statistical threshold of p < 0.01.

RT for correct responses were similar in the Full (320 ± 63 ms) and Free (318 ± 59 ms) conditions, and longer in the None condition (424 ± 63 ms). A significant effect of condition was evidenced on RT [F_{(2, 58)} = 132.71, p < 0.001]. Post-hoc analysis showed that RT in the None condition was significantly longer than in both Full and Free conditions (p < 0.001). In free-choice trials, subjects made a balanced proportion of right (51.1 ± 4.4%) and left (48.9 ± 4.4%) key presses. A high level of accuracy was observed, with percentages of correct responses of 99.8 ± 0.8% in Full, 99.6 ± 0.8% in Free and 98.6 ± 1.7% in None. Incorrect responses were exclusively omissions in Full and Free (no error by definition in Free) and included 93% of errors and 7% of omissions in None. There was a significant effect of condition on accuracy [F_{(2, 58)} = 8.30, p < 0.005] and post-hoc analysis showed that accuracy was significantly poorer in None as compared to each Full and Free condition (p < 0.01).

Alpha power within the preparation period differed substantially in intensity, topography and temporal course according to the nature of advance information, as expressed by a significant triple interaction between Condition, ROI and Time [F_{(112, 3248)} = 3.15, p < 0.001]. Two main observations were made (Figure 2). (1) A marked alpha ERD occurred bilaterally over the parieto-occipital regions during the whole foreperiod in all conditions. This alpha ERD was larger in the absence of advance information (None), of intermediate size when response had to be freely selected (Free) and smaller in the presence of advance instruction (Full). (2) An alpha ERS developed over the midline parietal region during the Full and Free foreperiods. Results of the post-hoc tests revealed that alpha ERD in Full and Free was significantly smaller than in None in all ROIs, generally at the end of the foreperiod (Figure 3). No significant ERD amplitude difference was observed between Full and Free conditions, although there was a tendency towards a larger alpha ERD in Free than in Full in the right parieto-occipital region between 1200 and 1400 ms (p < 0.05).

No consistent alpha MRAA was obtained in the None condition (non-significant difference from 0). In contrast, alpha MRAA started at the beginning of the foreperiod in both Full and Free conditions and increased in amplitude until their peak at S2 onset, reflecting larger alpha ERD amplitude over the sensorimotor regions contralateral to movement (Figure 4). Full MRAA differed significantly from 0 by −1000 ms to S2 (p < 0.01), whereas Free MRAA differed significantly from 0 by −400 ms to S2 (p < 0.01). In the Full condition, a negative correlation was observed between the amplitude of midparietal alpha ERS and MRAA in the last 200 ms time window preceding S2 (Pearson coefficient: −0.47, p < 0.01). This correlation was not significant in the Free condition.

In the Full condition, the two lateralization subgroups displayed distinct preparatory alpha patterns (Figure 5A), as confirmed by a significant Lateralization subgroup × ROI × Time interaction [F_{(56, 1008)} = 1.72, p < 0.001]. Compared with subjects with small MRAA amplitude (low lateralizers), subjects with large MRAA amplitude (high lateralizers) showed larger ERS over the midparietal region, in parallel with smaller ERD over the right parieto-occipital and left premotor regions (post-hoc independent sample t-tests, Figure 6).

In the Free condition, a similar distinction between alpha patterns was observed according to the degree of sensorimotor alpha lateralization (Figure 5B), confirmed by a significant Lateralization subgroup × ROI × Time interaction [F_{(56, 1008)} = 1.48, p < 0.01]. As in Full, high lateralizers displayed larger ERS in the midparietal region and smaller ERD in the right parieto-occipital and left premotor regions compared with low lateralizers. However, the differences between high and low lateralizers were not as large as in the Full condition (p < 0.05, Figure 6).

In the None condition, no significant difference in alpha power was observed between the lateralization subgroups.

Although no specific preparation could be achieved, alpha ERD amplitude was larger in the None than in both the Full and Free conditions. This suggests that instead of a passive process, a general activation state occurred in widespread regions to prepare for reaction at S2. The posterior maximum of this activation is concordant with regional cerebral blood flow data showing larger posterior parietal activation in conditions with absent or limited advance information as compared to full information about movement (Deiber et al., 1996). A difference of this kind was interpreted as a need for enhanced visuospatial attention to S2 for correct movement selection in conditions with restricted preparation. Few electro-cortical data are available on spectral power modulation during unspecified motor preparation, as most studies explore preparation intervals with variable degrees of response uncertainty, including the absence of response requirement. These studies have generally described larger alpha and beta ERD over sensorimotor regions in the S1–S2 interval when a response was required at S2 and when no uncertainty remained on the response (Kaiser et al., 2001; Alegre et al., 2004; Deiber et al., 2005; Klostermann et al., 2007; Tzagarakis et al., 2010; Sabate et al., 2012). However, no difference in the posterior regions but a reduced alpha ERD in the SMA was observed in the “None” condition compared with completely or partially pre-specified conditions for five possible spatiotemporal configurations of bimanual finger responses (Deiber et al., 2005). Hence, the alpha oscillatory pattern related to unspecific motor preparation may vary according to the characteristics of the upcoming motor response, and in particular to the mobilization of visuospatial attentional resources. Further investigation should determine whether large posterior alpha activation denoting high attentional level is more generally associated with a limited number of motor choices, and potentially high anticipatory behavior, as the present data tend to suggest.

The Full and Free conditions displayed comparable amplitude of alpha ERD in all cerebral regions, indicating that both modes of selection were activating the neuronal networks corresponding to motor preparation in similar ways. There was a tendency toward a larger alpha ERD in Free than in Full in the right parieto-occipital cortex, suggesting additional attentional resources associated with free choice. Moreover, in contrast with None, both Full and Free conditions were associated with a medio-posterior alpha ERS occurring in the second half of the foreperiod. During selective attention and working memory tasks, alpha ERS has been shown to develop over areas not engaged in the task, reflecting inhibition of task-irrelevant regions (Klimesch et al., 2007; Jensen and Mazaheri, 2010; Foxe and Snyder, 2011). A recent analysis of alpha oscillatory activity during working memory localized a midposterior alpha ERS in the cuneus (Brodmann areas 18/19), interpreted as the functional inhibition of visual areas no longer relevant to the task (Michels et al., 2008). In our paradigm, while S1 lasted for the whole delay, it was only in the two conditions where movement could be prepared that the midparietal alpha ERS developed. Moreover, its amplitude was proportional to the lateralization of sensorimotor alpha activity at the end of the foreperiod in the Full condition, suggesting that the inhibition of visual processing was linked to a state of motor readiness. In other words, once selection of movement is achieved, motor preparation processes would occur in parallel to inhibition of S1 visual guidance, which would no longer be necessary. This finding supports the view that a balance between engagement of motor areas and suppression of ongoing distracting visual information is taking place in the top-down control of motor preparation reflected in alpha-band oscillations (Jensen and Mazaheri, 2010; Foxe and Snyder, 2011).

The absence of alpha power difference between instructed and free motor preparation raised the question of the sensitivity of the method and/or the parameter under study. To our knowledge, alpha activity has not been formally compared between internal and external motor preparation states in delayed response tasks. In simple RT tasks performed by only six subjects, Tremblay et al. (2008) did not observe any clear-cut difference in alpha ERD preceding internal and external responses, although there was a wider choice of finger movements than in the present study and thus potentially more confounding factors linked to increased effort. Instead, they showed that the timing of beta activity was sensitive to the mode of action selection. In the same vein, ERP mapping showed identical electrical brain activity configurations within the S1–S2 interval for internal and external motor selection, the only differences being in their duration (Thut et al., 2000). Neuroimaging studies have also described a similar network of brain areas activated in externally- and internally-driven motor responses, with more activation for the latter in restricted cerebral regions such as the pre-SMA or anterior cingulate cortex (Deiber et al., 1996; Hoffstaedter et al., 2012), which cannot be easily resolved using EEG techniques. The present EEG data are consistent with the current literature in showing a global similarity of alpha activity in external and internal motor preparation, but fail to demonstrate a regional difference in alpha power between the two selection modes. Further studies should determine whether a distinction of this kind in selection control is reflected in other frequency bands and/or other parameters of oscillatory activity, such as phase synchronization (Schyns et al., 2011).

The present data confirmed previous observations on the sensitivity of the alpha lateralization index to advance movement preparation. Early lateralization of alpha amplitude was observed over the motor regions following reliable advance information on response laterality, independently of response execution probability at S2 (de Jong et al., 2006; Gladwin et al., 2008). While replicating these findings, we also showed that the free condition displayed a significant alpha amplitude lateralization pattern, indicating that participants had indeed made an early selection of their movement when required to do so. Alpha lateralization reached statistical significance later in the Free than in the Full condition, suggesting that selecting movement side on internal decision took more time than it did on external guidance. This interpretation supports recent ERP results showing that the update of motor plans impacted differently on free and instructed choices and suggesting that, in free selection, multiple choices remained available until a late stage in motor preparation (Fleming et al., 2009). While RTs were the same in the two selection modes, as previously observed (Fleming et al., 2009; Hoffstaedter et al., 2012), our data showed that alpha lateralization reached a similar amplitude for internal and external motor selection at S2 occurrence, indicating comparable motor readiness. In addition, the significant alpha lateralization index in the Free condition where S1 was uninformative further supports the general link between alpha lateralization and motor processes rather than visuospatial attention shifts, which could potentially be triggered by S1 in the Full condition (Doyle et al., 2005). Lateralization of alpha ERD over motor cortical areas has also been repeatedly described during motor imagery, demonstrating the sensitivity of alpha oscillations to covert motor processes (Pfurtscheller and Neuper, 1997; McFarland et al., 2000; Nikulin et al., 2008). The present findings further emphasize the relevance of alpha MRAA in assessing the covert motor preparation processes involved in free choice.

In an attempt to explore further the relations between alpha oscillatory patterns, the degree of sensorimotor alpha lateralization and the mode of motor selection, we used the amplitude of the alpha lateralization index preceding S2 in the Full condition to split the subjects into high- and low- lateralizers. The Full condition served as reference, assuming that motor preparation would be achieved at the end of the foreperiod. The oscillatory alpha patterns were found to differ between the two subgroups, providing indications about underlying cerebral functioning. Furthermore, they replicated in the Full and Free conditions, indicating similarities in external and internal selection strategies. Two types of cerebral activation patterns could be distinguished. First, participants with large lateralization of sensorimotor alpha activity (high lateralizers) exhibited dominating midparietal alpha ERS, suggesting that they primarily relied on a strategy of motor activation/task-irrelevant visual activity suppression; they also showed less activation in the premotor regions, probably subsequent to activation of the motor regions. Second, participants with weak lateralization of sensorimotor alpha activity (low lateralizers) displayed predominant right parieto-occipital alpha ERD, indicating the primary recruitment of right dominant visuospatial attentional resources (Mesulam, 1981). These observations were consistent with our initial finding that, for a simple between-hand choice, no fundamental difference could be observed in alpha activity pattern when the selection was internal as opposed to external. Instead, they revealed that, independently of the selection mode, distinct cerebral alpha patterns can be observed in relation to the lateralization of sensorimotor alpha activity preceding the response. We distinguished a motor alpha pattern, characterized by elevated lateralization of sensorimotor activity and concomitant suppression of visual activity, from an attentional alpha pattern featuring reduced sensorimotor lateralization paralleled by increased activation of posterior attentional networks. In the young population presently studied, both alpha patterns led to similar RTs in each external and internal selection mode, indicating that they were equally effective on motor execution. Alternatively, the two alpha patterns are likely to reflect distinct preparatory cerebral strategies, the functional relevance of which remains to be clarified.

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.