The set of structures identified in each condition over the first 900 ms of recording is displayed in Table 1. In all subjects, some areas were found active in both conditions and others were more particularly involved in the “illusion” condition only. Dipole locations were found in S1, in the superior parietal gyrus, in the supramarginal gyrus and in the right angular gyrus in both conditions. The left angular gyrus, the left SMA and the precentral gyrus were activated in the “illusion” condition only.#

Fig. 1 illustrates the superimposed waveforms recorded from the 151 channels in one representative subject, in the two conditions. It reveals that MEG signals recorded in both conditions clearly differed. In the “illusion” condition, after minor peaks observed in the first 200 ms, two major deflections appeared, the first beginning at 300 ms after the vibration onset and the largest one at 400 ms after the vibration onset. From 600 ms, the MEG signals were kept stable. In the “no illusion” condition, the major deflection began 500 ms after the vibration onset.#

In order to more precisely determine structures associated to the different peaks of activity, we identified the set of active structures in each of the four 200-ms windows: [0–200 ms], [200–400 ms], [400–600 ms] and [600–800 ms] in both conditions.#

Before 200 ms, S1 and the superior parietal gyrus were found active in both conditions whereas sources of activity were found in the supramarginal gyrus, in the angular gyrus and in M1 only in the “illusion condition.” Between 200 and 400 ms, S1, the supramarginal gyrus and the angular gyrus were found active in both conditions while the superior parietal gyrus was found active in the “no illusion” condition only and M1 and the supplementary motor area were found active in the “illusion” condition only. After 400 ms, sources of activity were identified in S1 and in the supramarginal gyrus in both conditions whereas the angular gyrus and M1 were found active in the “illusion” condition only.#

In summary, in each condition, the sources of activity could change from the first to the second time window, but stabilized after 400 ms. Therefore, Fig. 2 illustrates these data by disõplaying the locations of estimated dipoles on the MR images for each condition in one representative subject, before and after 400 ms. In the “no illusion” condition, before 400 ms, activity was mainly found in the superior parietal gyrus and in S1, and after 400 ms, activity was observed in the supramarginal gyrus and in S1. In the “illusion” condition, before 400 ms, activity was mainly located in M1 and in S1, and after 400 ms, in addition to the sources still present in M1 and S1, two additional locations were found in the supramarginal gyrus and in the angular gyrus.#

Fig. 3 represents dipole moments for each structure in each time window in both conditions, which indicates the strength of each ECD over time. A general remark is that the evolution of the activity level with time was different from one structure to the other within and between conditions. This was confirmed by statistics. An ANOVA including factors Condition, Time-window and Structure was carried out by using all the sources amplitudes for each subject, each time-window and each condition. The results revealed a significant Condition×Time-window×Structure interaction (F(15,135)=2.17, p<0.01), a significant Condition×Structure interaction (F(5,45)=4.96, p<0.001) and a significant Time-window×Structure interaction (F(15,135)=3.64, p<0.001). These interactions justified to carry out two separate ANOVAs including the Time-window and Structure factors, in each condition. In the “no illusion” condition, ANOVA revealed a significant effect of Time-window (F(3,27)=4.70, p<0.001), and of Structure (F(5,45)=23.12, p<0.0001) and a significant Time-window×Structure interaction (F(15,135)= 4.59, p<0.0001). The activity level of S1 (F(3,27)=7.32, p<0.01) and of the superior parietal gyrus (F(3,27)=4.50, p<0.01) increased with time, while the activity level of the supramarginal gyrus remained stable from 200 to 800 ms (F(3,27)=1.77, no significant). The angular gyrus was activated only between 200 and 400 ms and its activity level was the lowest. In the “illusion” condition, ANOVA revealed a significant effect of Time-window (F(3,27)=4.39, p<0.01), and of Structure (F(5,45)=3.67, p<0.01) and a significant Time-window×Structure interaction (F(15, 135)=1.8, p<0.05). T-tests were performed to test the differences of activity level observed between two consecutive time-windows. The activity level of M1 and of S1 increased after 400 ms (M1: t(1,9)=6.43, p<0.01; S1: t(1,9)=3.2, p<0.05) whereas the activity level of the supramarginal gyrus increased after 200 ms (t(1,9)=2.25, p<0.05) then remained stable over time. Concerning the angular gyrus, its activity level increased from the first time-window to the second one (t(1,9)=1.89, p<0.05) and from the third one to the last one (t(1,9)=1.67, p=0.09).#

Different locations of dipoles were found depending on whether tendon vibration elicited or not illusory movements. SI and some posterior parietal areas (supramarginal gyrus, superior parietal gyrus and right angular gyrus) were activated in both conditions while the left angular gyrus, the SMA and MI were activated only when vibration yielded illusory movements. As the vibrations always produced the same types of sensory afferents (skin and muscle spindle primary afferents) with the same energy, the major difference between conditions consisted in different perceptual effects due to differences in extensor/flexor balance of proprioceptive stimulation in each condition. Therefore, the activation of motor areas and of the left angular gyrus seems specifically associated with the experience of kinesthetic sensations. This suggests that motor areas as well as posterior parietal areas are involved in the perception of illusory movements. These data are consistent with the fMRI foci obtained in our previous study (Romaiguère et al., 2003). In the fMRI study, cortical activities associated to illusory movements were found in SMA, MI, SI and the left inferior parietal areas as in the present study, but also in cingulate motor areas and in premotor areas. These two last regions were not found active here for at least two reasons linked to the technical disadvantages of the MEG. The first is that it is difficult for MEG to detect brain dipoles radial to the skull, which are mainly generated in the gyrus, which could explain that premotor areas (located in the middle frontal gyrus) were not found here. The second is that dipoles generated in deep structures are almost not detected by MEG since magnetic fields recorded from outside of the scalp rapidly decrease with increasing depth. This probably explains why no dipoles were found in cingulate motor areas in the present study. Nonetheless, our data suggest, contrary to the proposition made by Naito et al. (1999), that motor areas alone do not underlie movement illusion perception but rather that these areas participate among a complete set of neural structures including inferior parietal areas and motor areas. Collectively, all these structures convey the illusion of movement.#

The involvement of posterior parietal structures in perception of illusory movements has also been found by Radovanovic et al. (2002) in a PET study where they compared brain structures involved in the perception of passive limb movement and illusory movement induced by muscle tendon vibration. Moreover, the activation of posterior parietal areas has been reported in several motor imagery tasks when imagination condition was compared to rest condition (Decety et al. 1994; Stephan et al. 1995; Grafton et al. 1996). In 2000, using a fMRI study, Gerardin et al. (2000) have demonstrated additional activation in parietal areas during imagination with respect to execution of movement. They have proposed that parietal areas could constitute the neural substrate for the storage of visual and kinesthetic limb postures. All these data suggest that the parietal cortex could play an important role in the elaboration of motor images and kinesthetic sensations (Roll et al., 1994).#

The major interesting result brought up by the present study concerns the dynamics of cortical activities. When looking at the different structures activated in each time-window, it appears that the structures activated during each condition changed from the first interval to the second one as if sources jumped from one location to the other while they remained stable during the following two intervals. The first 400 ms seems to correspond to a transient phase during which different perceptual and decisional processing should occur whereas these processes would be stabilized after 400 ms. More specifically, from 400 ms, the supramarginal gyrus and S1 were activated during both conditions while the angular gyrus and M1 were activated only during the “illusion” condition. These data suggest that S1 and the supramarginal gyrus could support the processes related to the processing of somatosensory afferents that are shared by the two experimental conditions while M1 and the angular gyrus would rather support processes more specifically associated to the perception of illusory movements.#

Twelve healthy right-handed subjects (7 women and 5 men, 24–38 years of age) participated in this study. They were chosen for their ability to feel clear vibration-induced movement illusions. They were paid for their participation.#

All the participants gave informed consent to the experimental procedure, as required by the Helsinki Declaration (1964).#

Participants were seated in a magnetically shielded room with their heads placed inside the neuromagnetometer. Vibration was delivered via two specifically designed pneumatic devices independently driven by a software program that permits synchronization between the vibration trains and the MEG acquisition. Vibration was applied perpendicularly to the distal tendons of the right wrist extensor and flexor muscles.#

The vibrators were held on the wrist of the subjects by rubber straps. The hand was maintained in a constant position by a light plastic splint attached to the hand and forearm of the subjects by adjustable straps. The soft transient noise caused by the pneumatic device activation was masked by fitting earplugs into the subject's ear canals and by delivering continuous white noise. Participants were told not to blink during stimulus presentation and were required to keep immobile and to fixate to a point marked by a cross on the screen placed in front of them.#

According to the protocol used in Romaiguère et al. (2003), simultaneous tendon vibration of extensor and flexor muscle groups was applied following two patterns:

• –In the “Illusion” condition, covibration was applied at 30 Hz on wrist extensor tendons and at 110 Hz on wrist flexor tendons (frequency difference: 80 Hz).
• –In the “No illusion” condition, covibration was applied at 70 Hz on both extensor and flexor tendons (frequency difference: 0 Hz).

The perceptual effects of the two conditions of stimulation were carefully controlled just before beginning the MEG recordings. All of the subjects clearly reported movement illusion in the condition where vibration with different frequencies was applied over each tendon and no movement illusion when the vibration frequency was the same on both tendons.#

Each trial consisted of a 5-s vibration train, followed by 4 s of inter-stimulus interval. There were 102 repetitions of each condition, presented in a pseudo-random order that differed for each subject. A total of 204 trials were delivered in four different blocks of 51 trials each.#

The magnetic fields were measured over the entire head using a 151-DC-SQUID whole-head type MEG system (Omega 151, CTF Systems, Port Coquitlam, B.C., Canada). The signals were recorded continuously, with a sampling frequency of 625 Hz. Three small coils were attached to reference landmarks on the participant's head: at the left and right pre-auricular points and at the nasion. At the beginning of each block, the head position relative to the coordinate system of the MEG helmet was calculated from the position of those coils in order to register possible head movements during the experimental session.#

The averaging epochs lasted from 1 s before stimulus onset up to 7 s thereafter. The pre-stimulus epoch served as a baseline. Epochs contaminated by eye blinks were rejected and trials with MEG changes exceeding 0.5 pT were discarded. For further analysis, data were then low-pass filtered at 20 Hz.#

High-resolution 3-D T1-weighted structural images were acquired for only ten subjects (two subjects were unable to participate in a MRI examination) using a 3-T whole-body imager Medspec 30/80 AVANCE (Brucker, Ettlingen, Germany) for anatomical localization (inversion-recovery 3D sequence, 1×0.75×1.22 mm). The anatomical slices covered the whole brain and were acquired parallel to the sagittal plane. MRI visible markers were placed at the locations of the head-positioning coils used during MEG registration (nasion and two periauricular points). The MEG sensor coordinate system was later aligned with the MRI images using these three head markers.#

Neuronal sources were modeled as equivalent current dipoles (ECD). The ECD approach is based on the assumption that the activated cortical area is relatively small and can therefore be approximated by a pointlike current source with a location, orientation, and strength. The ECD parameters can be determined directly from the local distribution of the magnetic field by a least-squares search. CTF's DipoleFit software package (dipolar method) was used for the computation of source generators. Spatiotemporal modeling of simultaneous sources was performed using a spherical head model whose parameters were fitted to the MRI of each subject. The residual variance was below 20% for all dipoles. We did not carry out a “group modeling” of MEG sources because this would have needed a “grand average” over all MEG signals obtained for each subject, and such a “grand average” would have disturbed the relative accurate localization of sources due to the variability in head positions of subjects with respect to the MEG sensors. The MEG source modeling was thus performed separately for each subject.#

ECDs identification was made in three stages. First, for each participant and each condition, between 0 and 900 ms, which corresponds to our time period of interest, source modeling of magnetic fields was performed at each main MEG peak. One to three sources were identified per subjects for each MEG peak. Secondly, CTF's MRIViewer software package was used to localize each dipole in the own brain of each subject by using the anatomical images obtained by MRI. Brain structures corresponding to each dipole location were then identified with an anatomical atlas (Duvernoy, 1991). Finally, for each subject and each condition, a set of the previously identified structures was constituted over each of four successive 200-ms windows: [0–200 ms], [200–400 ms], [400–600 ms] and [600–800 ms]. The 200-ms windows corresponded to the period during which dipole localizations remained quite stable.#