Abstract

The supratemporal sources of the earliest auditory cortical responses (20–80 ms) were identified using simultaneously recorded electroencephalographic (EEG) and magnetoencephalographic (MEG) data. Both hemispheres of six subjects were recorded two or three times in different sessions in response to 8000 right-ear 1 kHz pure tones stimuli. Four components were identified: Pa (28 ms), Nb (40 ms), and two subcomponents of the Pb complex, termed Pb1 (52 ms) and Pb2 (74 ms). Based on MEG data, the corresponding sources were localized on the anatomy using individual realistic head models: Pa in the medial portion of Heschl's gyri (H1/H2); Nb/Pb1 in the lateral aspect of the supratemporal gyrus (STG); and Pb2 in the antero-lateral portion of Heschl's gyri. All sources were oriented antero-superiorly. This pattern was clearest in the contralateral hemisphere, where these three activities could be statistically dissociated. Results agree with previous invasive human intracerebral recordings, with animal studies reporting secondary areas involved in the generation of middle latency auditory-evoked components, and with positron emission tomography and functional magnetic resonance imaging studies often reporting these three active areas although without temporal information. The early STG activity may be attributed to parallel thalamo-cortical connections, or to cortico-cortical connections between the primary auditory cortex and the STG, as recently described in humans.

Introduction

Anatomy of the Human Auditory Cortex

It is well established that the main human cortical auditory areas are located in the superior portion of the temporal lobe, including parts of the supratemporal plane and the supratemporal gyrus (STG). The organization of the human auditory cortex shows similarities with the auditory cortex of the macaque monkey, which has been parcelled into >12 areas organized in three regions: the core line region, containing the primary auditory cortex (PAC) and two more anterior areas, the rostral area (R) and the rostro-lateral area (RL), is surrounded by the belt region (itself subdivided into as many as eight distinct areas), which in turn is boarded laterally by two areas forming the parabelt region in the STG (Pandya, 1995; Rauschecker, 1998aKaas et al., 1999). Cytoarchitectonic studies in humans (Galaburda and Sanides, 1980; Rademacher et al., 1993; Pandya, 1995) have described a highly granular konio-cortex located in the first transverse gyrus of Heschl (H1), occupying about two-thirds of this structure. This area presumably forms the PAC, as also suggested by intracerebral recordings (Liégeois-Chauvel et al., 1991). The first transverse gyrus is delimited anteriorily by the first temporal sulcus (TS1), and posteriorily by Heschl's sulcus (HS1). Although additional transverse gyri (H2, H3) might be present posteriorily to H1 (Rademacher et al., 1993; Penhune et al., 1996; Leonard et al., 1998), it has been suggested that the primary area remains within H1, occasionally extending to the posterior bank of HS1 (Rademacher et al., 1993). The core line is bounded medially by a more primitive area of prokonio-cortex located in the insular circular sulcus, called the ‘root’ area. On its other edges, the PAC is surrounded by several secondary belts of parakonio-cortex (associative areas), anteriorily toward the pole, laterally by the STG, and posteriorily by H2–H3 (if present) and by the planum temporale (PT). Areas H1 and PT have been reported to be highly variable across individuals (Penhune et al., 1996; Leonard et al., 1998; Westbury et al., 1999). Moreover, inter-hemispheric asymmetries have been reported, with both larger PTs, especially in right-handers (Steinmetz et al., 1989; Steinmetz, 1996; Shapleske et al., 1999), and larger H1 volumes of white but not of grey matter (Penhune et al., 1996) in the left hemisphere.

Functional Mapping of the Auditory Cortical Areas

Because of their good spatial resolution, positron emission tomography (PET) and functional magnetic resonance imaging (fMRI) techniques have been contributing greatly to the functional mapping of the cortical structures involved in either simple perceptual tasks or in tasks addressing cognitive processes. In these studies, a prominent activity of the region surrounding Heschl's gyrus was reported (Lauter et al., 1985; Binder et al., 1994; Wessinger et al., 1997a; Scheich et al., 1998; Stippich et al., 1998; Belin et al., 1999, 2000; Lockwood et al., 1999), with some evidence for a tonotopic organization, where high frequencies are represented more postero-medially than low frequencies. Binder et al. (Binder et al., 1994) found that noise bursts and speech stimuli also activate the STG and banks of the supratemporal sulcus (STS), and noted considerable inter-subject variability in the location of the recorded cortical activity. Using pure tones as stimuli in a detection task, Scheich et al. (Scheich et al., 1998) distinguished activity in H1 from activities in more posterior cortical regions probably arising from H2 or H3 and PT.

Chronology of Activation of the Auditory Cortical Areas

An important issue of the investigation of the human auditory cortex is to detect precisely when and in which temporal order the cortical auditory areas are activated after stimulus onset. This important information about sound processing in the central auditory pathway in humans can only be obtained by using high temporal resolution techniques. Electrophysiological studies (Picton et al., 1974; Wood and Wolpaw, 1982) have described that after stimulus onset the early brainstem-evoked responses (occurring before 10 ms) are followed by different evoked components originating from the auditory cortex. The responses elicited between 10 and 70 ms after stimulus onset are termed middle latency components (MLCs), whereas later components, in the range of 80–250 ms, are called long latency components (LLCs). MLCs elicited by clicks are usually described as a succession of four waves labeled as Na (latency: ~19 ms), Pa (~30 ms), Nb (~40 ms), and Pb (50–70 ms). In the literature, Pb is also known as P1 and sometimes P50, especially when stimuli are tone-bursts. In the following, we will use the term Pb when referring to this component. MLCs are followed by the most prominent component of the LLCs, the N1 or N100, having a latency of ~100 ms.

While early electroencephalography (EEG) (Scherg et al., 1989a) and magnetoencephalography (MEG) studies (Elberling et al., 1982; Hari, 1983; Pantev et al., 1988, 1990; Hari, 1990; Reite et al., 1994) explained the generation of N1 with an activation of primary and secondary auditory cortical areas, more recent MEG studies (Pantev et al., 1995; Lütkenhöner and Steinsträter, 1998) or stereotactic EEG (SEEG) intracerebral recordings (Liégeois-Chauvel et al., 1994, 1995) argue for supratemporal sources of N1 most likely located in the PT. Other intracerebral recordings (Celesia, 1976; Howard et al., 2000) suggest that both primary and secondary areas are active in the N1 time range. Additional frontal and parietal generators of the N1 component were also suggested on the basis of EEG and MEG measurements, carried out by Giard et al. and Lävikäinen et al. (Giard et al., 1994; Lävikäinen et al., 1994), respectively.

Less is known about the sources of the MLCs, which, due to their smaller amplitude, are much more difficult to record. The first MEG study carried out by Pelizzone et al. (Pelizzone et al.,1987), for which anatomical reference (MRIs) was not available, showed that Pa had a cortical origin, and that the corresponding sources were located slightly more anterior than N100 sources in the supratemporal plane region. The anatomical locations of Na sources have been studied in MEG using clicks (Scherg et al., 1989b; Hashimoto et al., 1995; Kuriki et al., 1995; Yoshiura et al., 1995), and although they could not be systematically identified in all subjects, it was suggested that they were localized in the postero-medial part of H1. Using either clicks (Scherg et al., 1989b; Mäkelä et al., 1994; Hashimoto et al., 1995; Kuriki et al., 1995; Yoshiura et al., 1995; Gutschalk et al., 1999) or tone-bursts (Pantev et al., 1995), Pa sources have been reported that are located in the same regions as Na sources. These results have been corroborated for Na and Pa by intracerebral SEEG recordings in epileptic patients, with electrode leads within the medial part of H1 (Liégeois-Chauvel et al., 1991). It is thus now commonly suggested that the Na–Pa complex reflects the earliest PAC activity.

By contrast, the anatomical origin of the Pb component still remains controversial. Using MEG, Scherg et al. (Scherg et al., 1989b) studied one subject with clicks and found Pb sources to be more lateral and anterior than Na/Pa sources. This result was confirmed by Yoshiura et al. in several subjects (Yoshiura et al., 1995), and by intracerebral recordings in epileptic patients (Liégeois-Chauvel, 1994, 1995). However, other studies do not agree with such an organization of the Pb sources. Using 1 kHz tone-bursts with a quite long rise time of 15 ms, Reite et al. (Reite et al., 1988) consistently found Pb sources in the PT. EEG mapping and source analysis (Scherg and von Cramon, 1986; Cacace et al., 1990) have further suggested the existence of a radial temporal component at ~40–45 ms, overlapping temporally with other tangential sources, originating probably from the STG, although no clear anatomical correlates have been provided. These studies are in agreement with intracerebral and electrocorticogram recordings (Celesia, 1976; Howard et al., 2000) that also show a component at ~40 ms originating from the STG surface. Reduction of this radial component has been reported by Knight et al. (Knight et al., 1988) in patients with STG lesions. However, this early STG activity has not been described in a series of SEEG studies (Liégeois-Chauvel, 1991, 1994, 1995), in which the authors rather support the existence of multiple MLC subcomponents in the 30–70 ms range, with sources gradually located more antero-laterally along H1.

In general, scalp studies do not provide sufficient anatomical information about the location of MLC sources. Thus, the goal of this investigation was to obtain further anatomical and electrophysiological data on the supratemporal cortical structures first activated by stimulation with pure tone-bursts. In a test–retest paradigm with simultaneously recorded EEG and MEG, non-invasive identification of MLC sources has been performed by use of realistic head models based on individual MRIs.

Materials and Methods

Subjects

Six right-handed subjects [according to the Edinburgh handedness questionnaire (Oldfield, 1971)], one male and five females, aged 24–30 years and with neither otological nor neurological disorders, participated in this experiment. Normal audiological status (individual air conduction hearing thresholds between 250 Hz and 4 kHz of <10 dB above normal threshold) was verified by pure-tone audiometry. Informed consent was obtained from each subject after explaining to him/her the nature of the study. The experimental procedures were conducted in accordance with the Ethics Commission of the University of Münster and the Declaration of Helsinki. Subjects were paid for their participation. In order to check the reproducibility of the results, the whole paradigm was repeated three times for three of the test subjects, and duplicated for the other three subjects. These repetitions, or sessions, were carried out on different days.

Stimulation

Stimuli were 1000 Hz short tone-bursts, having a 3 ms rise time, a 20 ms plateau and a 20 ms decay time. Rise and fall ramps were shaped using a Hanning window. The long decay time was chosen to avoid possible off-responses overlapping on-MLCs studied here. Individual thresholds were determined for each subject, and the stimuli were delivered at 60 dB sensation level to the right ear with a randomized inter-stimulus interval (onset-to-onset) ranging from 150 to 350 ms (uniform distribution). Magnetically silent delivery of the stimuli was provided by a special delivery system consisting of speakers (1′′ compression driver, Renkus-Heinz Inc., Foothill Ranch, CA) mounted outside the magnetically shielded room which were connected to a silicon earpiece through 6.3 m of echoless plastic tubing (16 mm inner diameter). The frequency characteristic of this system deviated less than ±10 dB between 200 and 6000 Hz. The transmission delay of ~19 ms was compensated for by an appropriate shift of the trigger signal. Before carrying out the experiments, both the signal spectrum of the stimulus and its correct timing were checked by means of a 2 cm3 artificial ear (Brüel & Kjær model 4157, Naerum, Denmark) equipped with an 0.5′′ condenser microphone (Brüel & Kjær model 4134) and connected to the silicon earpiece at the end of the sound delivery system. For each recorded hemisphere, the total number of stimuli in one session was 8800.

Data Acquisition

Recordings were carried out in a magnetically and acoustically shielded room. Subjects rested in a lateral position on a vacuum cast, with their head lying on a mould to ensure stable fixation throughout the whole experimental session. MEG (37-channel gradiometer; 4D Neuro-imaging, San Diego, CA) and EEG data (32 channels; Neuroscan, Sterling, VA) were recorded simultaneously (bandwidth: 0.16–400 Hz; sampling rate: 1043 Hz). Data were acquired continuously and stored for offline analysis. The detection coils of the neuro-magnetometer were arranged in a circular concave array with a diameter of 144 mm and a spherical radius of 122 mm. The distance between the centers of the coils was 22 mm, and the coil diameter was 20 mm. The sensors were configured as a first-order axial gradiometer with a baseline of 50 mm. The spectral density of the intrinsic noise of each channel was between 5 and 7 fT/√Hz in the frequency range above 1 Hz. The sensor array was placed over the temporal region, centered over a point ~1.5 cm superior to the position T3/T4 of the international 10–20 system for electrode placement, as close to the subject's head as possible. A sensor-position indicator system determined the spatial locations of the sensors relative to the head and indicated whether head movements occurred during the recordings. Because the MEG device covered only one hemisphere at a time, each session consisted of two measurements, one for each hemisphere. The order of recorded hemisphere was balanced between sessions. For EEG, a 32 Ag/AgCl electrode cap was used (Virtanen et al., 1996). The montage was composed of the 19 international 10–20 electrodes, 12 electrodes added at temporal, frontal and mastoid sites, and also one EOG derivation.

During the measurement sessions, subjects watched a self-selected cartoon, whilst lying comfortably on the side opposite to the MEG-recorded hemisphere. Data sets in which the relative position between the head and the MEG instrument changed by >1 cm were discarded from further analysis (this occurred three times over 30 measurements). For the 27 accepted measurements, the head movement was 4.3 mm on average.

Data Processing and Filtering

Trials with EOG artifacts were rejected, using a threshold for MEG and EEG/EOG signals of 3000 fT and 150 μV, respectively. These thresholds were chosen after careful inspection of the continuous data for each subject. On average, this procedure led to 7600 acceptable MEG trials for each subject's hemisphere, and twice as many EEG trials, corresponding to the combination of the individual MEG measurements of the left and the right hemisphere. Artifact-free trials were averaged over a 1300 ms period, including a 600 ms baseline. These averaged epochs were baseline-corrected, using the [–50 ms, 0 ms] time window, digitally band-pass filtered (with a zero-phase-shift filter operating in the frequency domain) between 3 Hz (12dB/octave) and 150 Hz (24 dB/octave), and finally again baseline-corrected using the same 50 ms window.

At this point we would like to stress the importance of the high-pass filter settings. Indeed, for MLC components two types of high-pass cutoff frequencies have been reported in the literature. Some authors who were looking for MLC topographies and sources used values of <5 Hz (Wood and Wolpow, 1982; Deiber et al., 1988; Pelizzone et al., 1987; Reite et al., 1988; Scherg et al., 1989b; Cacace et al., 1990; Woldorff et al., 1993; Mäkelä et al., 1994; Hashimoto et al., 1995; Pantev et al., 1995), while others preferred values of between 10 and 20 Hz (Scherg and Von Cramon, 1986; Kuriki et al., 1995; Yoshiura et al., 1995). Although the latter choice improves the signal-to-noise ratio, it may be critical when the resulting filtered data are used for source analysis. It has been reported that using higher high-pass filter settings deteriorates MLCs (Deiber et al., 1988; McGee et al., 1988). Thus, we found it important for careful source analysis to carry out simulations in order to demonstrate how high high-pass filtering mixes up the topographical contents of time-adjacent components and yields erroneous localization results (cf. Fig. 1).

Three different dipolar sources, mimicking three neighboring active auditory areas, were placed in a spherical model in the left temporal region. Each source was attributed a bell-shaped time course, with maximum amplitudes of 7, 5 and 25 nAm, peaking at 33, 52 and 75 ms, respectively. A simulated overall MEG signal was then computed from the linear contribution of the three sources. The effect of digital band-pass filtering was then studied with the following frequency settings: high-pass of 3 or 20 Hz (12 dB/octave) and low-pass of 150 Hz (24 dB/octave). Filters had zero phase shift. Figure 1A shows the original and both filtered overall signals on two channels of opposite polarity. It can be noted that the 20 Hz high-pass filtering modifies the polarity of certain peaks (arrow on Fig. 1A, channel 35). Figure 1B shows the root-mean-square (RMS) of the individual signals produced by each dipolar source separately, and also those of the overall unfiltered and filtered signals. In addition, it shows original maps corresponding to each source, and maps at the peaks of the 20–150 Hz filtered signal. It is clearly evident that the 3–150 Hz filter hardly affects the data. In contrast, although the 20–150 Hz filter separates different peaks, the corresponding peak maps show considerable distortion compared with the original maps. In particular, the resulting topographies are complex combinations of the individual original maps. Finally, performing a moving dipole fit on both filtered and unfiltered signals yields very different results. While for unfiltered signals this procedure perfectly retrieves each of the three sources in time windows in which their time courses do not overlap (shaded regions, cf. Fig. 1C), it dramatically failed for the 20–150 Hz filtered data for both the overlapping and the non-overlapping time periods (cf. Fig. 1D). For instance, in the 20–150 Hz case, the solution falls closer to source 3 in most of the time ranges when source 2 is actually active (cf. Fig. 1D). Dipole localization errors obtained when using the 3–150 Hz filter were very similar to those obtained when using no filter. The results of this simulation study have justified our choice of setting a value of 3 Hz for the high-pass filter.

Intra-subject MRI–MEG/EEG Registration

For each subject, a full head shape, including eye globe contours, was obtained before each EEG/MEG session (cf. Fig. 2A). The positions of these points, as well as the positions of the EEG electrodes and the MEG coils, were digitized in a coordinate system based on three anatomical landmarks: nasion, and the left and right ear canals. For each subject, sagittal T1-weighted MR images (TE = 4 ms, TR = 9.7 ms, slice thickness = 1.41 mm, pixel size within each slice = 0.98 mm) were acquired using a 1.5 T Siemens Magnetom Vision device. Reconstructed horizontal slices were parallel to the line passing through the anterior and posterior commissures. Prior to acquisition, MRI-visible markers were positioned over the three landmarks. For each session, a simple matching of these three points provided a first, coarse registration of MRI and EEG/MEG sensors. Using a procedure proposed by Schwartz et al. (Schwartz et al., 1996), this registration was improved by matching the full head shape onto the scalp surface segmented from MRIs (cf. Fig. 2B).

MEG/EEG Mapping and Source Localization

For each latency, MEG and EEG data were mapped using the spherical spline algorithm (Perrin et al., 1987), with Tikhonov regularization preventing fictious map rebounds due to noise (Babiloni et al., 1998).

Individual realistic boundary element models (BEMs) were constructed for each subject from a contour stack semi-automatically segmented from MRIs (Yvert et al., 1995). Each BEM consisted of three uniformly meshed surface shells (representing the inner and outer skull and the scalp) with 3.2 triangles/cm2 (with an average element edge length of 8.8 mm). For all subjects, a complete model was made of ~8000 triangles and 4000 points (cf. Fig. 2C). A first-order boundary element method (de Munck, 1992; Ferguson et al., 1994), with an isolated problem approach (Hämäläinen and Sarvas, 1989), was used, and conductivities were set to 0.45, 0.45/80 and 0.45 S/m (Rush and Driscoll, 1968) for the scalp, skull and brain, respectively. Such models have been evaluated recently using simulations (Crouzeix et al., 1999), and it was found that the positions of even radially oriented sources can theoretically be retrieved using realistic BEM models.

A moving dipole model was applied to the MEG data for each hemisphere separately, at each time sample between 0 and 150 ms. At each time slice, an iterative algorithm starting with an initial guess combined a non-linear part (Marquardt algorithm), searching for position and orientation parameters, and a linear part, estimating the amplitude of the source by a least-square criterion. In order to cope with the problem of local minima, ‘spherical preprocessing’ was used; the initial guess chosen for realistic modeling was the best result found by a spherical model among 50 inverse solutions corresponding to different random initial conditions.

EEG peaks and troughs on fronto-central electrodes (F3/C3 on the left hemisphere, and F4/C4 on the right hemisphere) were used to individually select initial latencies of the different MLC components. These latencies were then adjusted by a few milliseconds, in order to correspond to the most stable EEG topography in time. Finally, the dipole with the best goodness-of-fit value (gof) within a 2 ms time window around the selected latency was considered as the source of the component. Indeed, in cases of multiple simultaneous sources, the best gof occurred at time points when the sources overlap less (cf. Fig. 1). Only sources corresponding to high gof values (>97%) were considered for further analysis. For each subject and component, the source positions obtained in the different sessions were averaged.

Inter-subject Auditory Cortices Registration and Source Analysis

For inter-subject statistical analysis of the source coordinates in each hemisphere, all subjects' source positions were registered in a common frame. This was done to account for gross anatomical differences in location of the auditory cortex across all subjects. All subjects' auditory cortices were aligned by matching the center of three landmarks (A, B, C) identified from MRIs in the auditory cortex (cf. Fig. 3):

Additionally, HS1 and the lateral aspect of the STS were identified from individual sagittal MRIs (cf. Fig. 3). When H1 and H2 were present and had a common stem [as described elsewhere (Leonard et al., 1998)], HS1 was set as the mid-point between TS1 and the second Heschl sulcus (HS2). It thus laid below the sulcus intermedius (SI), separating H1 and H2 (cf. Fig. 3).

For each subject, the center of the head-coordinate system (defined by the x-axis pointing from the left to the right ear canal, the y-axis passing through the nasion, the z-axis being the cross-product of x × y, and the center being the orthogonal projection of the nasion onto the x-axis) was translated to the center of gravity of these three anatomical landmarks. The sources and these anatomical landmarks (points A, B and C and sulci) were thus individually registered in this shifted head coordinate system.

One-way multivariate statistical analysis (MANOVA, Wilk's lambda test) was first performed to account for a global effect of the components on the aligned source positions, and then post-hoc T2 tests were used to account for 2 × 2 significant differences between clusters of sources corresponding to the different components.

Results

Identification of the Components

In the present study, EEG was used to identify the MLC components. This choice was supported experimentally by the fact that EEG data showed more topographical variations over time than did MEG data, thus providing best identifiability of different components. This is illustrated in Figure 4, where the EEG and MEG components from the left hemisphere are shown for two sessions in one subject.

Four components could be consistently identified from EEG recordings, with a high reproducibility across sessions: Pa, peaking at ~28 ms (in 5/6 subjects), Nb (~40 ms, 6/6 subjects) and two subcomponents of the Pb complex, corresponding to two clearly distinct peaks, which we term Pb1 (~52 ms, 6/6 subjects) and Pb2 (~74 ms, 4/6 subjects). All these components showed dipolar pattern topographies consistent with sources pointing antero-superiorly. The Na component could be identified in three subjects from the EEG data, but not from MEG data, and was thus not further evaluated. The four identified waves corresponded to peaks and troughs on fronto-central electrodes: Pa, Pb1 and Pb2 corresponded to successive distinct peaks on the F3/F4 and C3/C4 electrodes, while Nb corresponded to a trough on F3/F4 and often a local peak on C3/C4. The latency of each component was determined according to these peaks, and adjusted by a few milliseconds, in order to correspond to the most stable topography in time.

Latencies and Amplitudes of the Components

For each hemisphere, the latency and signal RMS amplitude were measured for each component. For EEG, the RMS amplitude was computed using only the electrodes located on one hemisphere, excluding those on the midline. Values were then averaged across sessions for each subject, and finally averaged across subjects. These final mean values (± SD) are given in Table 1.

On average, latencies were found to be slightly larger over the ipsilateral right hemisphere than over the contralateral left hemisphere, although these differences were not statistically significant, as assessed by a paired t-test (P > 0.24 for all components).

As shown in Figure 5 and Table 1, a contralateral predominance was indicated by stronger EEG and MEG RMS signals from the left than from the right hemisphere. These inter-hemispheric differences were found to be statistically significant in MEG for the Nb (P = 0.0006), Pb1 (P = 0.0012) and Pb2 (P = 0.027) components, but not for EEG components (P > 0.13).

Topographies of the Components

The activity from Pa to Pb2 components lasted from ~25–30 ms to 80–100 ms after stimulus onset, continuously showing a dipolar pattern compatible with underlying sources always pointing anteriorly and superiorly. However, during this time period, topographies were changing, especially in EEG, as illustrated for one subject in Figure 4, and on the EEG maps grand-averaged across subjects in Figure 5. In particular, Nb showed a much more focal topography on the left hemisphere than did the three other components, thus suggesting a more peripheral source. In the left hemisphere, the Nb component corresponded to a local peak of RMS activity in MEG (Fig. 5). In the ipsilateral right hemisphere, the Nb component corresponded to a trough in the MEG RMS signal, and its EEG topography was not found to be much more focal than for the other components. Finally, positive and negative Pb2 peaks were found to be slightly more anterior than those of Pb1 in both hemispheres (cf. Fig. 5).

Source Analysis

For each component, we selected the dipole source yielding the best gof value no more than 2 ms around the latency determined from most stable EEG topographies around peaks/troughs on fronto-central electrodes (see Materials and Methods). Moreover, sources with gof values <97% were discarded from further analysis. This was mainly the case for the Nb component in the right hemisphere, for which no acceptable sources were obtained in three subjects. The accepted sources were averaged across sessions for each subject, then registered with respect to the three individual landmarks (see Materials and Methods), and finally averaged across subjects. Table 2 gives across-subject mean gof values, mean dipole coordinates and inter-session dispersions for each component, and mean A, B and C landmark positions. Figure 6 shows the final mean locations and orientations across all subjects for each component with respect to individual HS1 and STS sulci. Note that we found inter-session dispersions of the dipole coordinates (1.6–8.9 mm) to be of the order of the head movements (4.3 mm) (see the single case in Fig. 7A). Figure 7B illustrates the mean source positions obtained for one subject, relative to the temporal plane structures (Heschl's gyri, sulci, STG and PT), as reconstructed from MRIs after manual segmentation.

As expected from the topographies, all sources were found to point anteriorly and superiorly (Fig. 6). In the left hemisphere, Pa sources were found on the superio-postero-medial portion of HS1, whereas Nb and Pb1 sources were localized very laterally within the STG or the upper bank of the STS. Pb2 sources were found on HS1 infero-antero-laterally from Pa sources. The multivariate statistical analysis revealed a strong effect of the component factor over the source positions (P < 0.0001) in this hemisphere. Post-hoc T2 tests showed that the cluster of Pa sources was significantly different from those of Nb (P = 0.003), Pb1 (P = 0.001) and Pb2 (P = 0.008) sources, that Nb sources could be dissociated from Pb2 (P = 0.029) but not from Pb1 sources (P > 0.6), and finally that Pb1 sources tended to be different from Pb2 (P = 0.13). In the right hemisphere, the sources of all components were progressively aligned antero-laterally along HS1 when increasing latency, with Nb sources slightly shifted laterally. The inter-subject variability was higher than in the left hemisphere and the MANOVA showed only a global trend (P = 0.098), not allowing reliable post-hoc 2 × 2 comparisons of the source clusters.

Discussion

This study aimed at specifying the anatomical locations of supratemporal MLC sources, using non-invasive simultaneously recorded EEG and MEG data. We deliberately used pure-tone stimuli to allow further extensions of this study regarding the tonotopic organization of the MLC sources. The results obtained in this study offer additional insights into the organization of the earliest activations of the auditory cortex. First, from scalp EEG data, we could reliably identify four distinct MLC components in the time range of the conventional Pa–Pb complex, having characteristic topographies. Results indicate the existence of several middle latency subcomponents. In particular, the Pb, which is often considered as one component peaking at ~40–60 ms, was found to be composed of two subcomponents peaking successively at 52 and 74 ms. Secondly, the localization results confirm the existence of multiple distinct supratemporal areas housing MLC generators. Consideration of the individual anatomy significantly improved the localizability. To our knowledge, this is the first non-invasive study reporting detailed localizations of MLC sources and dissociating multiple supratemporal cortical regions, including associative areas, which underlie their generation.

The following general pattern of successive activities was found:

  • 30 ms (Pa): activation of the postero-medial portion of H1 and H2 extending posteriorily over HS1

  • 40–60 ms (Nb, Pb1): activation of the STG or the upper bank of the STS

  • 70–90 ms (Pb2): activation of more antero-lateral portion of Heschl's gyri and HS1 than for Pa

We tested whether the number of Heschl gyri would influence this pattern. For this purpose, we split the subjects into two groups: two subjects had two Heschl's gyri with posterior duplication (Leonard et al., 1998) and four subjects had only one Heschl's gyrus (with a left duplication anterior to the located sources in two subjects). Source locations were then averaged within each group, which yielded identical spatiotemporal activation patterns in both cases.

In our results, Pb1 sources could not be dissociated from Nb sources. Separability from Pb2 was not very clear, and only a statistical trend was found. For this reason, we suggest that the Nb–Pb1 complex with a latency of 40–60 ms might in fact reflect the same underlying active region in the STG.

The Nb component is usually described in EEG studies as a trough on a wave recorded at or near the vertex. However, this terminology lets one assume that it corresponds to a source oriented inferiorly. In this study we found that Nb should in fact be characterized by a topography compatible with a source oriented antero-superiorly (cf. Figs 4 and 5 for topographies and Fig. 6 for average source orientations), being more focal than those of Pa, Pb1 and Pb2. We thus believe that this trough near the vertex should rather be considered as the consequence of a lateral shift of activity, leaving less signal on fronto-central electrodes, and that negative frontal Nb EEG topographies are actually artificial consequences of high high-pass filter settings (Fig. 1).

Our localization results in the right hemisphere show a pattern similar to that in the left hemisphere, especially for Pa and Pb2 components. However, Nb and Pb1 sources were not localized as lateral, as they were in the left hemisphere. In the right hemisphere, the Nb sources were identified in only 3/6 subjects, and found to be lateral in two of them. We attribute this finding to our monaural stimulation paradigm: as reported, right ear stimulation resulted in a stronger activation of the contralateral (left) hemisphere (Fig. 5), which, as a consequence, may have improved the separability of the components on that side. From our data, a hemispheric asymmetry irrespective of the stimulated ear cannot be ruled out. However, recent findings using subcortical grids covering the STG in epileptic patients have reported early STG activity around 40 ms in both left and right hemispheres (Howard et al., 2000). Thus we rather attribute this asymetry to our asymmetrical stimulation paradigm.

We did not detect the Na component in MEG recordings, although we could see its EEG counterpart in three subjects. There may be two reasons for this. First, we used 1 kHz pure-tone stimuli, while other MEG studies addressing source localization of Na used broadband click stimuli (Scherg et al., 1989b; Hashimoto et al., 1995; Kuriki et al., 1995; Yoshiura et al., 1995), eliciting more prominent MLCs than tone or noise bursts. For instance, Pellizone et al. (Pellizone et al., 1987), using noise bursts, did not report an Na component in their data. Furthermore, it is known that the Na component is difficult to record in MEG, even with click stimulation (Yoshiura et al., 1995), probably because the underlying sources are weak and located deeply. This interpretation is supported by the results of SEEG investigations in which Na has been reported to stem from deep H1 sources (Celesia, 1976; Liégeois-Chauvel et al., 1991).

In our study, we found Pa sources in the very postero-superior-medial part of the supratemporal plane. This result is consistent with most findings of the literature reporting Pa sources in the postero-medial portion of Heschl's gyri (Scherget al., 1989b; Liégeois-Chauvel et al., 1991; Mäkelä et al., 1994; Hashimoto et al., 1995; Kuriki et al., 1995; Pantev et al., 1995; Yoshiura et al., 1995). It should, however, be noted that all these results did not precisely distinguish between H1, HS1 or H2 regions. In our study, we found Pa sources lying on or posterior to HS1 in all subjects, in a region more likely corresponding to the postero-lateral belt rather than the PAC. Although this result could be attributed to an unknown bias of the inverse procedure, it was found systematically in all subjects. This result can be paralleled with a recent fMRI study reporting activation of the belt region on or posterior to HS1 in the case of speech stimuli (Hashimoto et al., 2000). This finding is also consistent with awake monkey single unit recordings reporting initial spike latencies in the lateral belt adjacent to the PAC at ~25 ms (Pfingst and O'Connor, 1981). In particular, latencies in this area were not reported to be different from those of the PAC (27 ms), suggesting that these two neighboring areas are active simultaneously. The Pa activity recorded here on the human scalp could thus be the overall contribution of these two adjacent areas.

The Pb2 sources were localized antero-infero-laterally to the Pa sources (Fig. 6). This result is in accordance with previous intracerebral findings (Liégeois-Chauvel et al., 1994, 1995), and suggests that this component might reflect the activity of the human counterpart of the anterior areas (R, RL) in the core line and/or of the antero-lateral belt, both described in monkeys (Kaas et al., 1999).

We are aware that using a single moving dipole model might not be optimal in the present case of multiple active regions. However, this simple model allowed for the distinction between three different active areas. To test whether multi-dipole models would have been more efficient, we aimed at identifying the left hemisphere sources of the four components using a model with either three or four simultaneously active stationary dipoles [fixed positions and orientations (Scherg, 1990)] for the first session of the subject of Figure 7. Initial parameter conditions were set to be the results of the moving dipole model at the selected Pa, Nb, Pb1 and Pb2 latencies. For the three-dipole model the Pb1 source was discarded. The dipole parameters were then optimized to best explain the signal over the whole Pa-to-Pb2 time period (30–85 ms). The dipole locations obtained with these models did not differ considerably as compared with the results obtained with the moving dipole model, and variations remained within inter-session variability. The results pointed to the same active anatomical structures, except for the case of four simultaneous dipoles, in which the Pa source was localized far from the auditory cortical region (parietal white matter). However, the time courses of the sources were not convincing enough, each explaining several components, and thus providing no cue to establish a reliable correspondence between the components and their sources. Gutschalk et al. (Gutschalk et al., 1999) have suggested that two sources could be responsible for the MLCs, respectively in the medial HG and in a more lateral part of HG. We further investigated whether our three sources obtained on a moving dipole procedure could be explained by two overlapping dipole sources. For this purpose, a two-dipole model was tested in the same time range (30–85 ms), with initial conditions set at the Pa and Nb locations. Localization results are given in Figure 8, together with the RMS error between experimental and model data (RMS = 1 – gof). The two-dipole model sources fell in the medial PT and the anterolateral Heschl's gyri, at locations quite intermediate with respect to the three-dipole model sources. Although this model nicely explained Pa and Pb1/2 components, it failed to explain the experimental data over the entire time range, and especially at the Nb latency (vertical dashed line), where the RMS error remained important (6%). We are tempted to attribute Gutschalk et al.'s different findings either to the different protocol used (faster stimulation rate) or to their use of higher high-pass filter settings (20 Hz), which might have mixed Pb and Nb components as described in the method section and in Figure 1.

While our localizations of Pa and Pb2 sources in Heschl's gyri are in accordance with the existing electrophysiological literature (although scalp studies seldom provide precise anatomical localizations), a lateral activity at ~40–50 ms has been reported in only a few intracerebral recordings (Celesia, 1976; Howard et al., 2000). This lateral activity has not been reported in other SEEG studies, in which electrode grids covering the STG were not used (Liégeois-Chauvel, 1994, 1995). Rather, these studies suggest an organization of sources gradually aligned along H1 in the lateral direction with increasing latencies between 19 and 70 ms. By contrast, in our study we were able to confirm non-invasively the lateral activation of the STG region at ~40–60 ms. Depending on the subject, we have localized the Nb and Pb1 activity either in the STG or in the upper bank of the STS, with a good reproducibility across sessions. Previous EEG studies (Scherg and Von Cramon, 1986; Cacace, 1990) have suggested the existence of a radial component at ~40–45 ms probably originating from the STG. It is possible that this source is not MEG silent (still having some tangential component) and corresponds to the Nb/Pb1 source we could localize here. However, our EEG topographies had a very dipolar pattern, suggesting instead a mostly tangential underlying source.

This lateral activity might be paralleled to PET and fMRI activations of similar regions. Using PET, Lockwood et al. (Lockwood et al., 1999) showed that a pure-tone sequence presented to the right ear activated PAC in both hemispheres and also the STG/PT region in the left hemisphere. They also reported additional activated regions outside the temporal lobe, namely the posterior cingulate cortex and the lateral cerebellum. Several fMRI studies using either simple stimuli such as pure tones or noise bursts (Binder et al., 1994; Strainer et al., 1997; Scheich et al., 1998; Rauschecker, 1998b; Stippich et al., 1998) or more complex sounds (Binder et al., 1994; Belin et al., 1999, 2000; Hashimoto et al., 2000) also indicated activations in secondary areas including the STG. Interestingly, Rauschecker (Rauschecker, 1998b) has published fMRI results (Wessinger et al., 1997b) showing three neighboring supratemporal activities in medial and antero-lateral HG, and STG areas. This scheme corresponds perfectly with our findings. In the macaque, the STG activity has been shown to be dramatically enhanced when band-pass noise bursts or more complex stimuli than pure tones were applied (Rauschecker et al., 1995; Rauschecker, 1998a).

The possible involvement of secondary areas in the generation of MLCs, as shown in our study, is in agreement with MLC mapping studies in animals. Barth and Di (Barth and Di, 1991; Di and Barth, 1992) have mapped MLC components over the exposed auditory cortex of slightly anesthetized rats with a high spatial resolution. They found two early sharp positive peaks, the earlier one at ~10 ms over the primary auditory cortex and the later one at ~12 ms over a more caudo-dorsal secondary area. All MLC components until 60 ms could be explained by activities in these two areas (Di and Barth, 1992). Juckel et al. (Juckel et al., 1996) have described an early positive MLC component in cats, called P12, peaking at ~12 ms over the primary cortex (AI) and 1 ms later over the secondary cortex (AII). Over AII, this wave was followed by a prominent positive wave, called P25, peaking at ~25 ms. In a series of studies (Buchwald et al., 1981; Erwin and Buchwald, 1986a, 1986b; Chen and Buchwald, 1986), Buchwald and colleagues have shown that these two waves are differently sensitive to arousal state and stimulus rate, in ways similar to the behaviors of Na/Pa and Pb components in humans: while P12 in cats and Na/Pa in humans are weakly affected by these parameters, P25 and Pb are strongly affected. The authors thus paralleled these waves across species. Furthermore, their findings led them to support the existence of parallel thalamo-cortical routes generating separately early and late MLC components: P12 would reflect activation of the primary pathway, and P25 that of secondary pathways modulated by the ascending reticular arousal system. Further studies on guinea pigs (Kraus et al., 1992; McGee et al., 1992; Kraus and McGee, 1995), using pharmacological inactivation of primary (ventral) and secondary (caudo-medial) divisions of the medial geniculate body (MGB) and of the mesencephalic reticular formation, also support two parallel thalamocortical routes underlying MLC components.

Recent intracerebral stimulation results in humans (Howard et al., 2000) suggest another possible origin of the lateral STG activity. Using intracerebral leads in HG and surface grid electrodes covering the STG, these authors could stimulate a portion of HG, presumably corresponding to PAC, and simultaneously record the activity over the STG. Pronounced STG responses starting as early as 3 ms and peaking at ~15 ms after HG stimulation were reported, suggesting the existence of corticocortical connections between both areas. Such connections could thus mediate auditory information from the PAC region to the STG with a delay of ~10–20 ms, corresponding to the time interval between Pa and Nb/Pb1 components.

Both secondary thalamo-cortical and cortico-cortical connections could thus contribute to the STG activity. Anatomical findings in the primate [recently reviewed by Kaas et al. (Kaas et al., 1999)] confirm that this region receives its major inputs both from secondary divisions of the MGB and from the PAC (Rauschecker et al., 1997). Studies in patients with focal lesions of the PAC could help to raise final conclusions on this issue.

Notes

The authors are very grateful to Bernhard Ross and Andreas Wollbrink for engineering, Karing Berning and Christina Stahl for technical assistance, and Dr P. Fonlupt for helpful discussions concerning the statistical analysis. They also wish to thank Dr T.A. Hackett and Dr J.P. Rauschecker for constructive comments on the manuscript.

This work was supported by the Fyssen Foundation (Paris) through a postdoctoral fellowship awarded to B.Y., by a grant from the Deutsche Forschungsgemeinschaft (Pa 392/6–3) and by an APART-fellowship granted to A.S.-P. by the Austrian Academy of Sciences (APART 524).

Address correspondence to Blaise Yvert, Inserm Unité 280, 151 cours Albert Tomas, F-69424 Lyon cedex 03, France. Email: yvert@lyon151.inserm.fr.

Table 1

Intersubject signal peak latencies and RMS values

 EEG latency (ms) EEG RMS (μV) MEG RMS (fT) ,0,0 No. of subjects 
 LH RH LH RH LH RH LH or RH 
Across subjects mean ( ± 1 SD) peak latencies and signal RMS amplitudes are given for each component and each hemisphere. EEG RMS was computed using only electrodes located strictly on one hemisphere (excluding midline electrodes). The number of subjects in which each component was identified in either the left or the right hemisphere is indicated. LH: left hemisphere; RH: right hemisphere. 
Pa 28.3 ± 2.4 29.2 ± 2.6 0.24 ± 0.12 0.21 ± 0.10 11.6 ± 5.6 8.8 ± 3.6 
Nb 39.3 ± 3.9 41.3 ± 3.5 0.25 ± 0.08 0.23 ± 0.13 13.1 ± 6.0 7.0 ± 4.5 
Pb1 52.2 ± 3.7 53.4 ± 3.1 0.41 ± 0.20 0.37 ± 0.15 23.0 ± 7.2 15.8 ± 5.5 
Pb2 74.1 ± 9.8 75.0 ± 10.5 0.44 ± 0.14 0.42 ± 0.10 26.5 ± 8.2 17.7 ± 4.3 
 EEG latency (ms) EEG RMS (μV) MEG RMS (fT) ,0,0 No. of subjects 
 LH RH LH RH LH RH LH or RH 
Across subjects mean ( ± 1 SD) peak latencies and signal RMS amplitudes are given for each component and each hemisphere. EEG RMS was computed using only electrodes located strictly on one hemisphere (excluding midline electrodes). The number of subjects in which each component was identified in either the left or the right hemisphere is indicated. LH: left hemisphere; RH: right hemisphere. 
Pa 28.3 ± 2.4 29.2 ± 2.6 0.24 ± 0.12 0.21 ± 0.10 11.6 ± 5.6 8.8 ± 3.6 
Nb 39.3 ± 3.9 41.3 ± 3.5 0.25 ± 0.08 0.23 ± 0.13 13.1 ± 6.0 7.0 ± 4.5 
Pb1 52.2 ± 3.7 53.4 ± 3.1 0.41 ± 0.20 0.37 ± 0.15 23.0 ± 7.2 15.8 ± 5.5 
Pb2 74.1 ± 9.8 75.0 ± 10.5 0.44 ± 0.14 0.42 ± 0.10 26.5 ± 8.2 17.7 ± 4.3 
Table 2

Intersubject landmark and source coordinates (mm)

 LH RH 
 x y z gof x y z gof 
Rows 1–4: means ± SEM of individual head system coordinates of points A, B and C, and of their center of gravity. Rows 5–8: mean ± SEM coordinates of all component sources in the head coordinate system, individually shifted to the center of points A, B and C. Mean goodness-of-fit (gof) values (% of explained variance) for each component are also given in columns 4 and 8. Rows 9–12: coordinates given in rows 5–8, shifted back from the average coordinate center given in row 4, providing an estimate of the source coordinates in the head system after accounting for anatomical inter-subject variability. Rows 13–16: dipole head coordinate dispersion across sessions (1 SD), averaged across subjects. LH: left hemisphere; RH: right hemisphere. 
Point A –27.7 ± 1.2  7.6 ± 2.1 62.5 ± 1.5   34.1 ± 2.2  9.8 ± 2.7 60.8 ± 2.6  
Point B –56.7 ± 1.5  8.4 ± 2.5 65.2 ± 0.7   62.8 ± 2.3  10.9 ± 3.5 63.5 ± 1.4  
Point C –44.7 ± 1.0  24.5 ± 2.2 51.9 ± 2.2   52.5 ± 1.6  28.9 ± 2.0 49.0 ± 1.5  
Center –43.0 ± 1.0  13.5 ± 2.3 59.9 ± 1.4   49.8 ± 1.8  16.5 ± 2.7 57.7 ± 1.6  
Sources in the A,B,,C coordinate system 
Pa  8.6 ± 1.7 –10.0 ± 2.5  6.8 ± 2.6 98.5 –20.8 ± 4.7 –11.7 ± 2.7  3.5 ± 1.0 98.4 
Nb  –5.2 ± 3.0 –10.3 ± 1.9 –4.5 ± 2.6 98.8  –2.1 ± 7.9  –9.4 ± 2.1 –1.4 ± 6.9 98.2 
Pb1  –4.5 ± 1.1  –8.1 ± 1.5 –3.3 ± 1.9 99.2  –4.4 ± 4.6  –4.7 ± 2.0 –0.3 ± 3.3 99.0 
Pb2  3.5 ± 3.4  –3.6 ± 1.3 –5.9 ± 3.4 99.2  –2.4 ± 4.6  –1.4 ± 2.2 –2.5 ± 4.0 98.9 
Sources in the mean head coordinate system 
Pa –34.4 ± 1.7  3.5 ± 2.5 66.7 ± 2.6 98.5  29.1 ± 4.7  4.8 ± 2.7 61.2 ± 1.0 98.4 
Nb –48.2 ± 3.0  3.2 ± 1.9 55.4 ± 2.6 98.8  47.7 ± 7.9  7.1 ± 2.1 56.4 ± 6.9 98.2 
Pb1 –47.5 ± 1.1  5.4 ± 1.5 56.6 ± 1.9 99.2  45.5 ± 4.6  11.9 ± 2.0 57.5 ± 3.3 99.0 
Pb2 –39.6 ± 3.4  9.9 ± 1.3 54.0 ± 3.4 99.2  47.4 ± 4.6  15.2 ± 2.2 55.2 ± 4.0 98.9 
Mean inter-session dispersion 
Pa  4.1  6.6  5.4   3.7  4.3  6.2  
Nb  5.3  3.3  4.6   9.0  6.4  5.7  
Pb1  5.9  2.6  5.7   3.0  3.2  2.8  
Pb2  8.9  1.6  3.2   5.4  5.5  5.0  
 LH RH 
 x y z gof x y z gof 
Rows 1–4: means ± SEM of individual head system coordinates of points A, B and C, and of their center of gravity. Rows 5–8: mean ± SEM coordinates of all component sources in the head coordinate system, individually shifted to the center of points A, B and C. Mean goodness-of-fit (gof) values (% of explained variance) for each component are also given in columns 4 and 8. Rows 9–12: coordinates given in rows 5–8, shifted back from the average coordinate center given in row 4, providing an estimate of the source coordinates in the head system after accounting for anatomical inter-subject variability. Rows 13–16: dipole head coordinate dispersion across sessions (1 SD), averaged across subjects. LH: left hemisphere; RH: right hemisphere. 
Point A –27.7 ± 1.2  7.6 ± 2.1 62.5 ± 1.5   34.1 ± 2.2  9.8 ± 2.7 60.8 ± 2.6  
Point B –56.7 ± 1.5  8.4 ± 2.5 65.2 ± 0.7   62.8 ± 2.3  10.9 ± 3.5 63.5 ± 1.4  
Point C –44.7 ± 1.0  24.5 ± 2.2 51.9 ± 2.2   52.5 ± 1.6  28.9 ± 2.0 49.0 ± 1.5  
Center –43.0 ± 1.0  13.5 ± 2.3 59.9 ± 1.4   49.8 ± 1.8  16.5 ± 2.7 57.7 ± 1.6  
Sources in the A,B,,C coordinate system 
Pa  8.6 ± 1.7 –10.0 ± 2.5  6.8 ± 2.6 98.5 –20.8 ± 4.7 –11.7 ± 2.7  3.5 ± 1.0 98.4 
Nb  –5.2 ± 3.0 –10.3 ± 1.9 –4.5 ± 2.6 98.8  –2.1 ± 7.9  –9.4 ± 2.1 –1.4 ± 6.9 98.2 
Pb1  –4.5 ± 1.1  –8.1 ± 1.5 –3.3 ± 1.9 99.2  –4.4 ± 4.6  –4.7 ± 2.0 –0.3 ± 3.3 99.0 
Pb2  3.5 ± 3.4  –3.6 ± 1.3 –5.9 ± 3.4 99.2  –2.4 ± 4.6  –1.4 ± 2.2 –2.5 ± 4.0 98.9 
Sources in the mean head coordinate system 
Pa –34.4 ± 1.7  3.5 ± 2.5 66.7 ± 2.6 98.5  29.1 ± 4.7  4.8 ± 2.7 61.2 ± 1.0 98.4 
Nb –48.2 ± 3.0  3.2 ± 1.9 55.4 ± 2.6 98.8  47.7 ± 7.9  7.1 ± 2.1 56.4 ± 6.9 98.2 
Pb1 –47.5 ± 1.1  5.4 ± 1.5 56.6 ± 1.9 99.2  45.5 ± 4.6  11.9 ± 2.0 57.5 ± 3.3 99.0 
Pb2 –39.6 ± 3.4  9.9 ± 1.3 54.0 ± 3.4 99.2  47.4 ± 4.6  15.2 ± 2.2 55.2 ± 4.0 98.9 
Mean inter-session dispersion 
Pa  4.1  6.6  5.4   3.7  4.3  6.2  
Nb  5.3  3.3  4.6   9.0  6.4  5.7  
Pb1  5.9  2.6  5.7   3.0  3.2  2.8  
Pb2  8.9  1.6  3.2   5.4  5.5  5.0  
Figure 1.

 Simulations evaluating the effect of digital high-pass filtering. Three dipolar sources positioned in the left hemisphere of a spherical model with bell-shaped activities, peaking successively at 33, 52 and 75 ms. (A) Overall signal on two channels when no filter, a 3–150 Hz band-pass filter or a 20–150 Hz band-pass filter is applied. Signals corresponding to no filter and to the 3–150 Hz band-pass filter are overlapping. Note that polarity reversal might occur in the case of 20–150 Hz filtering (arrow). (B) RMS values across all channels for the unfiltered (thick solid line) and both filtered (dashed lines) overall signals, and for signals created by each source alone (solid thin lines). At latencies when each source is maximally active, the original map is plotted above the map obtained from the 20–150 Hz filtered signal. Note the strong distortion of the filtered maps. (C) Distance between the moving dipole solution and the closest of the three sources (either 1, 2 or 3), when no filter is applied. Each source is retrieved perfectly when source activities do not overlap in time (shaded areas), and within some ms in adjacent latencies. Right: top view, representing the actual locations of the three sources (circles) and the moving solutions (filled squares). (D) The same as in (C), but with 20–150 Hz filtering. Note that source 3 is mostly retrieved instead of source 2 at the second RMS peak (the 52 ms peak on the thick dashed curve in B).

Figure 1.

 Simulations evaluating the effect of digital high-pass filtering. Three dipolar sources positioned in the left hemisphere of a spherical model with bell-shaped activities, peaking successively at 33, 52 and 75 ms. (A) Overall signal on two channels when no filter, a 3–150 Hz band-pass filter or a 20–150 Hz band-pass filter is applied. Signals corresponding to no filter and to the 3–150 Hz band-pass filter are overlapping. Note that polarity reversal might occur in the case of 20–150 Hz filtering (arrow). (B) RMS values across all channels for the unfiltered (thick solid line) and both filtered (dashed lines) overall signals, and for signals created by each source alone (solid thin lines). At latencies when each source is maximally active, the original map is plotted above the map obtained from the 20–150 Hz filtered signal. Note the strong distortion of the filtered maps. (C) Distance between the moving dipole solution and the closest of the three sources (either 1, 2 or 3), when no filter is applied. Each source is retrieved perfectly when source activities do not overlap in time (shaded areas), and within some ms in adjacent latencies. Right: top view, representing the actual locations of the three sources (circles) and the moving solutions (filled squares). (D) The same as in (C), but with 20–150 Hz filtering. Note that source 3 is mostly retrieved instead of source 2 at the second RMS peak (the 52 ms peak on the thick dashed curve in B).

Figure 2.

 (A) The digitized head shape obtained from one subject. (B) White dots represent the head shape points registered onto the MRI by using either the coarse registration method based only on the three anatomical landmarks nasion, and the left and right tragi (left picture), or the surface registration method fitting the whole head shape onto the MR scalp surface (right picture). (C) A realistic BEM model (top to bottom: scalp, outer and inner skull surfaces) used for this subject, represented with electrode positions (white dots) and left and right coil arrays. The Sylvian fissure is displayed on the side view of the inner skull surface (bottom right).

Figure 2.

 (A) The digitized head shape obtained from one subject. (B) White dots represent the head shape points registered onto the MRI by using either the coarse registration method based only on the three anatomical landmarks nasion, and the left and right tragi (left picture), or the surface registration method fitting the whole head shape onto the MR scalp surface (right picture). (C) A realistic BEM model (top to bottom: scalp, outer and inner skull surfaces) used for this subject, represented with electrode positions (white dots) and left and right coil arrays. The Sylvian fissure is displayed on the side view of the inner skull surface (bottom right).

Figure 3.

 Identification of landmarks A, B and C, and HS1 and STS sulci on individual MRIs. When H1 and H2 exist with a common stem, they are separated by SI, and thus HS1 does not exist. In this case, HS1 was logically set in the middle of TS1 and HS2 on sagittal slices.

Figure 3.

 Identification of landmarks A, B and C, and HS1 and STS sulci on individual MRIs. When H1 and H2 exist with a common stem, they are separated by SI, and thus HS1 does not exist. In this case, HS1 was logically set in the middle of TS1 and HS2 on sagittal slices.

Figure 4.

 Left: EEG (top) and MEG (bottom) evoked responses over the left (contralateral) hemisphere showing the reproducibility of the data across two sessions in one subject. Evoked responses are shown in each modality on all channels for the first session and also on two channels of opposite polarity: F3 and IM1 electrodes for EEG, and channels 18 and 25 for MEG. The stimulus pure tone (stim) is represented below the curves. Right: Maps of each component over the left hemisphere for both sessions in EEG and MEG. Note the focal EEG topography of Nb. The tilt in MEG topographies should be attributed to different positioning of the sensors between both sessions.

Figure 4.

 Left: EEG (top) and MEG (bottom) evoked responses over the left (contralateral) hemisphere showing the reproducibility of the data across two sessions in one subject. Evoked responses are shown in each modality on all channels for the first session and also on two channels of opposite polarity: F3 and IM1 electrodes for EEG, and channels 18 and 25 for MEG. The stimulus pure tone (stim) is represented below the curves. Right: Maps of each component over the left hemisphere for both sessions in EEG and MEG. Note the focal EEG topography of Nb. The tilt in MEG topographies should be attributed to different positioning of the sensors between both sessions.

Figure 5.

 Top: The grand average of RMS signals across subjects and sessions in MEG (left and right hemispheres) and EEG. Below: EEG topographies over left and right hemispheres for the four components. Note that on the left hemisphere Nb is more focal and that in both hemispheres Pb2 is more anterior than Pb1.

Figure 5.

 Top: The grand average of RMS signals across subjects and sessions in MEG (left and right hemispheres) and EEG. Below: EEG topographies over left and right hemispheres for the four components. Note that on the left hemisphere Nb is more focal and that in both hemispheres Pb2 is more anterior than Pb1.

Figure 6.

 The across-subject average source locations and orientations for all components, with respect to individual HS1 and STS sulci. Coordinates (mm) are given in the head coordinate system individually shifted to the center of landmarks A, B and C, and aligned between subjects. Ellipsoids represent 3-D SEM volumes across subjects. Arrows indicate source orientations and amplitudes. Post = posterior, ant = anterior, lat = lateral, med = medial, inf = inferior, sup = superior.

Figure 6.

 The across-subject average source locations and orientations for all components, with respect to individual HS1 and STS sulci. Coordinates (mm) are given in the head coordinate system individually shifted to the center of landmarks A, B and C, and aligned between subjects. Ellipsoids represent 3-D SEM volumes across subjects. Arrows indicate source orientations and amplitudes. Post = posterior, ant = anterior, lat = lateral, med = medial, inf = inferior, sup = superior.

Figure 7.

 Localization results for one subject. (A) Dipole locations found in the different sessions are shown on individual MRIs. Note that the localization reproducibility was best in the left (contralateral) hemisphere, especially for the Pa component. (B) Across-session average locations projected onto the supratemporal plane (manually segmented from MRIs). Note the larger PT in the left hemisphere. Post = posterior, ant = anterior, lat = lateral, med = medial.

Figure 7.

 Localization results for one subject. (A) Dipole locations found in the different sessions are shown on individual MRIs. Note that the localization reproducibility was best in the left (contralateral) hemisphere, especially for the Pa component. (B) Across-session average locations projected onto the supratemporal plane (manually segmented from MRIs). Note the larger PT in the left hemisphere. Post = posterior, ant = anterior, lat = lateral, med = medial.

Figure 8.

 Moving dipole solutions corresponding to Pa, Nb and Pb2 components (white circles) served as initial conditions in a three-stationary-dipole model explaining the signal over the 30–85 ms time window. This latter model yielded solutions (filled circles) very close to the moving dipole solutions. By contrast, when only two stationary dipoles were considered, solutions (crosses) fell at intermediate locations. Furthermore, the corresponding RMS error functions characterizing the part of the signal unexplained by each model show that the two-dipole model does not explain satisfactorily the signal over the entire time range, expecially around the Nb latency. The error function obtained with the moving dipole model is also given for comparison, and vertical dashed lines indicate the selected moving dipole latencies for Pa, Nb and Pb2.

Figure 8.

 Moving dipole solutions corresponding to Pa, Nb and Pb2 components (white circles) served as initial conditions in a three-stationary-dipole model explaining the signal over the 30–85 ms time window. This latter model yielded solutions (filled circles) very close to the moving dipole solutions. By contrast, when only two stationary dipoles were considered, solutions (crosses) fell at intermediate locations. Furthermore, the corresponding RMS error functions characterizing the part of the signal unexplained by each model show that the two-dipole model does not explain satisfactorily the signal over the entire time range, expecially around the Nb latency. The error function obtained with the moving dipole model is also given for comparison, and vertical dashed lines indicate the selected moving dipole latencies for Pa, Nb and Pb2.

References

Babiloni F, Carducci F, Babiloni C, Urbano A (
1998
) Improved realistic Laplacian estimate of highly-sampled EEG potentials by regularization techniques.
Electroencephalogr Clin Neurophysiol
 
106
:
336
–343.
Barth D, Di S (
1991
) The functional anatomy of middle latency auditory evoked potentials.
Brain Res
 
565
:
109
–115.
Belin P, Zatorre RJ, Hoge R, Evans AC, Pike B (
1999
) Event-related fMRI of the auditory cortex.
NeuroImage
 
10
:
417
–429.
Belin P, Zatorre RJ, Lafaille P, Ahad P, Pike B (
2000
) Voice-selective areas in human auditory cortex.
Nature
 
403
:
309
–312.
Binder JR, Rao SM, Hammeke TA, Yetkin FZ, Jesmanowicz A, Bandettini PA, Wong EC, Estkowski LD, Goldstein MD, Haughton VM, Hyde JS (
1994
) Functional magnetic resonance imaging of the human auditory cortex.
Ann Neurol
 
35
:
662
–672.
Buchwald JS, Hinman C, Norman RJ, Huang CM, Brown KA (
1981
) Middle- and long-latency auditory evoked responses recorded from the vertex of normal and chronically lesioned cats.
Brain Res
 
205
:
91
–109.
Cacace AT, Satya-Murti S, Wolpaw JR (
1990
) Human middle-latency auditory evoked potentials: vertex and temporal components.
Electroencephalogr Clin Neurophysiol
 
77
:
6
–18.
Celesia GG (
1976
) Organization of auditory cortical areas in man.
Brain
 
99
:
403
–414.
Chen BM, Buchwald JS (
1986
) Midlatency auditory evoked responses: differential effects of sleep in the cat.
Electroencephalogr Clin Neurophysiol
 
65
:
373
–382.
Crouzeix A, Yvert B, Bertrand O, Pernier J (
1999
) An evaluation of dipole reconstruction accuracy with spherical and realistic head models in MEG.
Clin Neurophysiol
 
110
:
2176
–2188.
Deiber MP, Ibanez V, Fischer C, Perrin F, Mauguiere F (
1988
) Sequential mapping favours the hypothesis of distinct generators for Na and Pa middle latency auditory evoked potentials.
Electroencephalogr Clin Neurophysiol
 
71
:
187
–197.
de Munck JC (
1992
) A linear discretization of the volume conductor boundary integral equation using analytically integrated elements.
IEEE Trans Biomed Engng
 
39
:
986
–990.
Di S, Barth DS (
1992
) The functional anatomy of middle-latency auditory evoked potentials: thalamocortical connections.
J Neurophysiol
 
68
:
425
–431.
Elberling C, Bak C, Kofoed B, Lebech J, Saermark K (
1982
) Auditory magnetic fields from the human cerebral cortex: location and strength of an equivalent current dipole.
Acta Neurol Scand
 
65
:
553
–569.
Erwin RJ, Buchwald JS (
1986
) Midlatency auditory evoked responses: differential recovery cycle characteristics.
Electroencephalogr Clin Neurophysiol
 
64
:
417
–423.
Erwin R, Buchwald JS (
1986
) Midlatency auditory evoked responses: differential effects of sleep in the human.
Electroencephalogr Clin Neurophysiol
 
65
:
383
–392.
Ferguson AS, Zhang X, Stroink G (
1994
) A complete linear discretization for calculating the magnetic field using the boundary element method.
IEEE Trans Biomed Engng
 
41
:
455
–460.
Galaburda A, Sanides F (
1980
) Cytoarchitectonic organization of the human auditory cortex
J Comp Neurol
 
190
:
597
–610.
Giard M, Perrin F, Echallier J, Thévenet M, Pernier J (
1994
) Dissociation of temporal and frontal components in the human auditory N1 wave: a scalp current density and dipole model analysis
Electroencephalogr Clin Neurophysiol
 
92
:
238
–252.
Gutschalk A, Mase R, Roth R, Ille N, Rupp A, Hähnel S, Picton T, Scherg M (
1999
) Deconvolution of 40 Hz steady-state fields reveals two overlapping source activities of the human auditory cortex.
Clin Neurophysiol
 
110
:
856
–868.
Hari R (
1983
) Auditory evoked magnetic fields of the human brain.
Rev Laryngol Otol Rhinol (Bord)
 
104
:
143
–151.
Hari, R (1990) The neuromagnetic method in the study of the human auditory cortex. In: Auditory evoked magnetic fields and electric potentials (Grandori F, Hoke M, Romani GL, eds), pp. 222–282. Basel: Karger Verlag.
Hashimoto I, Mashiko T, Yoshikawa K, Mizuta T, Imada T, Hayashi M (
1995
) Neuromagnetic measurements of the human primary auditory response.
Electroencephalogr Clin Neurophysiol
 
96
:
348
–356.
Hashimoto R, Homae F, Nakajima K, Miyashita Y, Sakai KL (
2000
) Functional differentiation in the human auditory and language areas revealed by a dichotic listening task.
NeuroImage
 
12
:
147
–58.
Howard MA, Volkov IO, Mirsky R, Garell PC, Noh MD, Granner M, Damasio H, Steinschneider M, Reale RA, Hind JE, Brugge JF (
2000
) Auditory cortex on the human posterior superior temporal gyrus.
J Comp Neurol
 
416
:
79
–92.
Hämäläinen MS, Sarvas J (
1989
) Realistic conductivity geometry model of the human head for interpretation of neuromagnetic data.
IEEE Trans Biomed Engng
 
36
:
165
–171.
Juckel G, Csepe V, Molnar M, Hegerl U, Karmos G (
1996
) Intensity dependence of auditory evoked potentials in behaving cats.
Electroencephalogr Clin Neurophysiol
 
100
:
527
–537.
Kaas JH, Hackett TA, Tramo MJ (
1999
) Auditory processing in primate cerebral cortex.
Curr Opin Neurobiol
 
9
:
164
–170.
Knight RT, Scabini D, Woods DL, Clayworth C (
1988
) The effects of lesions of superior temporal gyrus and inferior parietal lobe on temporal and vertex components of the human AEP.
Electroencephalogr Clin Neurophysiol
 
70
:
499
–509.
Kraus N, McGee T (
1995
) The middle latency response generating system.
Electroencephalogr Clin Neurophysiol Suppl
 
44
:
93
–101.
Kraus N, McGee T, Littman T, Nicol T (
1992
) Reticular formation influences on primary and non-primary auditory pathways as reflected by the middle latency response.
Brain Res
 
587
:
186
–194.
Kuriki S, Nogai T, Hirata Y (
1995
) Cortical sources of middle latency responses of auditory evoked magnetic field.
Hear Res
 
92
:
47
–51.
Lauter JL, Herscovitch P, Formby C, Raichle ME (
1985
) Tonotopic organization in human auditory cortex revealed by positron emission tomography.
Hear Res
 
20
:
199
–205.
Lävikäinen J, Huotilainen M, Pekkonen E, Ilmoniemi R, Näätänen R (
1994
) Auditory stimuli activate parietal brain regions: a whole-head MEG study.
NeuroReport
 
6
:
182
–184.
Leonard CM, Puranik C, Kuldau JM, Lombardino LJ (
1998
) Normal variation in the frequency and location of human auditory cortex landmarks. Heschl's gyrus: where is it?
Cereb Cortex
 
8
:
397
–406.
Liégeois-Chauvel C, Musolino A, Chauvel P (
1991
) Localization of the primary auditory area in man.
Brain
 
114
:
139
–153.
Liégeois-Chauvel C, Musolino A, Badier JM, Marquis P, Chauvel P (
1994
) Evoked potentials recorded from the auditory cortex in man: evaluation and topography of middle latency components.
Electroencephalogr Clin Neurophysiol
 
92
:
204
–214.
Liégeois-Chauvel C, Laguitton V, Badier JM, Schwartz D, Chauvel P (
1995
) Cortical mechanisms of auditive perception in man: contribution of cerebral potentials and evoked magnetic fields by auditive stimulations.
Rev Neurol (Paris)
 
151
:
495
–504.
Lockwood AH, Salvi RJ, Coad ML, Arnold SA, Wack DS, Murphy BW, Burkard RF (
1999
) The functional anatomy of the normal human auditory system: responses to 0.5 and 4.0 kHz tones at varied intensities.
Cereb Cortex
 
9
:
65
–76.
Lütkenhöner B, Steinsträter O (
1998
) High-precision neuromagnetic study of the functional organization of the human auditory cortex.
Audiol Neurootol
 
3
:
191
–213.
McGee T, Kraus N, Manfredi C (
1988
) Toward a strategy for analyzing the auditory middle-latency response waveform.
Audiology
 
27
:
119
–130.
McGee T, Kraus N, Littman T, Nicol T (
1992
) Contributions of medial geniculate body subdivisions to the middle latency response.
Hear Res
 
61
:
147
–154.
Mäkelä J, Hämäläinen M, Hari R, McEvoy L (
1994
) Whole-head mapping of middle-latency auditory evoked magnetic fields.
Electroencephalogr Clin Neurophysiol
 
92
:
414
–421.
Oldfield R (
1971
) The assessment and analysis of handedness: The Edinburgh Inventory.
Neuropsychologia
 
9
:
97
–113.
Pandya DN (
1995
) Anatomy of the auditory cortex.
Rev Neurol (Paris)
 
151
:
486
–494.
Pantev C, Hoke M, Lehnertz K, Lütkenhöner B, Anogianakis G, Wittkowski W (
1988
) Tonotopic organization of the human auditory cortex revealed by transient auditory evoked magnetic fields.
Electroencephalogr Clin Neurophysiol
 
69
:
160
–170.
Pantev C, Hoke M, Lehnertz K, Lütkenhöner B, Fahrendorf G, Stöber U (
1990
) Identification of sources of brain neuronal activity with high spatiotemporal resolution through combination of neuromagnetic source localization (NMSL) and magnetic resonance imaging (MRI).
Electroencephalogr Clin Neurophysiol
 
75
:
173
–184.
Pantev C, Bertrand O, Eulitz C, Verkindt C, Hampson S, Schuierer G, Elbert T (
1995
) Specific tonotopic organizations of different areas of the human auditory cortex revealed by simultaneous magnetic and electric recordings.
Electroencephalogr Clin Neurophysiol
 
94
:
26
–40.
Pelizzone M, Hari R, Makela JP, Huttunen J, Ahlfors S, Hamalainen M (
1987
) Cortical origin of middle-latency auditory evoked responses in man.
Neurosci Lett
 
82
:
303
–307.
Penhune VB, Zatorre RJ, MacDonald JD, Evans AC (
1996
) Interhemispheric anatomical differences in human primary auditory cortex: probabilistic mapping and volume measurement from magnetic resonance scans.
Cereb Cortex
 
6
:
661
–672.
Perrin F, Pernier J, Bertrand O, Giard MH, Echallier JF (
1987
) Mapping of scalp potentials by surface spline interpolation.
Electroencephalogr Clin Neurophysiol
 
66
:
75
–81.
Pfingst BE, O'Connor TA (
1981
) Characteristics of neurons in auditory cortex of monkeys performing a simple auditory task.
J Neurophysiol
 
45
:
16
–34.
Picton TW, Hillyard SA, Krausz HI, Galambos R (
1974
) Human auditory evoked potentials. I. Evaluation of components.
Electroencephalogr Clin Neurophysiol
 
36
:
179
–190.
Rademacher J, Caviness VS Jr, Steinmetz H, Galaburda AM (
1993
) Topographical variation of the human primary cortices: implications for neuroimaging, brain mapping, and neurobiology.
Cereb Cortex
 
3
:
313
–329.
Rauschecker JP (
1998
) Cortical processing of complex sounds.
Curr Opin Neurobiol
 
8
:
516
–521.
Rauschecker JP (
1998
) Parallel processing in the auditory cortex of primates.
Audiol Neurootol
 
3
:
86
–103.
Rauschecker JP, Tian B, Hauser M (
1995
) Processing of complex sounds in the macaque nonprimary auditory cortex.
Science
 
268
:
111
–114.
Rauschecker JP, Tian B, Pons T, Mishkin M (
1997
) Serial and parallel processing in rhesus monkey auditory cortex.
J Comp Neurol
 
382
:
89
–103.
Reite M, Teale P, Zimmerman J, Davis K, Whalen J (
1988
) Source location of a 50 msec latency auditory evoked field component.
Electroencephalogr Clin Neurophysiol
 
70
:
490
–498.
Reite M, Adams M, Simon J, Teale P, Sheeder J, Richardson D, Grabbe R (
1994
) Auditory M100 component 1: relationship to Heschl's gyri.
Cogn Brain Res
 
2
:
13
–20.
Rush S, Driscoll DA (
1968
) Current distribution in the brain from surface electrodes.
Anesth Analg
 
47
:
717
–723.
Scheich H, Baumgart F, Gaschler-Markefski B, Tegeler C, Tempelmann C, Heinze HJ, Schindler F, Stiller D (
1998
) Functional magnetic resonance imaging of a human auditory cortex area involved in foreground-background decomposition.
Eur J Neurosci
 
10
:
803
–809.
Scherg M (1990) Fundamentals of dipole source potential analysis. In: Advances in Audiology Vol. 6. Auditory evoked magnetic fields and electric potentials (Grandori F et al., eds), pp. 40–69. Basel: Karger.
Scherg M, Von Cramon D (
1986
) Evoked dipole source potentials of the human auditory cortex.
Electroencephalogr Clin Neurophysiol
 
65
:
344
–360.
Scherg M, Vajsar J, Picton TW (
1989
) A source analysis of the late human auditory evoked potentials.
J Cogn Neurosci
 
1
:
336
–355.
Scherg M, Hari R, Hämäläinen M (1989b) Frequency-specific sources of the auditory N19–P30–P50 response detected by a multiple source analysis of evoked magnetic fields and potentials. In: Advances in biomagnetism (Williamson S, ed.), pp. 97–100. New York: Plenum Press.
Schwartz D, Lemoine D, Poiseau E, Barillot C (
1996
) Registration of MEG/EEG data with 3D MRI: methodology and precision issues.
Brain Topogr
 
9
:
101
–116.
Shapleske J, Rossell SL, Woodruff PW, David AS (
1999
) The planum temporale: a systematic, quantitative review of its structural, functional and clinical significance.
Brain Res Brain Res Rev
 
29
:
26
–49.
Steinmetz H (
1996
) Structure, functional and cerebral asymmetry: in vivo morphometry of the planum temporale.
Neurosci Biobehav Rev
 
20
:
587
–591.
Steinmetz H, Rademacher J, Huang YX, Hefter H, Zilles K, Thron A, Freund HJ (
1989
) Cerebral asymmetry: MR planimetry of the human planum temporale.
J Comput Assist Tomogr
 
13
:
996
–1005.
Stippich C, Freitag P, Kassubek J, Soros P, Kamada K, Kober H, Scheffler K, Hopfengartner R, Bilecen D, Radu EW, Vieth JB (
1998
) Motor, somatosensory and auditory cortex localization by fMRI and MEG.
NeuroReport
 
9
:
1953
–1957.
Strainer JC, Ulmer JL, Yetkin FZ, Haughton VM, Daniels DL, Millen SJ (
1997
) Functional MR of the primary auditory cortex: an analysis of pure tone activation and tone discrimination.
Am J Neuroradiol
 
18
:
601
–610.
Virtanen J, Rinne T, Ilmoniemi RJ, Naatanen R (
1996
) MEG-compatible multichannel EEG electrode array.
Electroencephalogr Clin Neurophysiol
 
99
:
568
–570.
Wessinger CM, Buonocore MH, Kussmaul CL, Mangun GR (
1997
) Tonotopy in human auditory cortex examined with functional magnetic resonance imaging.
Hum Brain Mapp
 
5
:
18
–25.
Wessinger CM, Tian B, VanMeter JW, Platenberg RC, Pekar J, Rauschecker JP (
1997
) Hierarchical processing within human auditory cortex examined with fMRI.
Soc Neurosci Abstr
 
23
:
2073
.
Westbury CF, Zatorre RJ, Evans AC (
1999
) Quantifying variability in the planum temporale: a probability map.
Cereb Cortex
 
9
:
392
–405.
Woldorff M, Gallen C, Hampson S, Hillyard S, Pantev C, Sobel D, Bloom F (
1993
) Modulation of early sensory processing in human auditory cortex during selective attention.
Proc Natl Acad Sci USA
 
90
:
8722
–8726.
Wood CC, Wolpaw JR (
1982
) Scalp distribution of human auditory evoked potentials. II. Evidence for overlapping sources and involvement of auditory cortex.
Electroencephalogr Clin Neurophysiol
 
54
:
25
–38.
Yoshiura T, Ueno S, Iramina K, Masuda K (
1995
) Source localization of middle latency auditory evoked magnetic fields.
Brain Res
 
703
:
139
–144.
Yvert B, Bertrand O, Echallier JF, Pernier J (
1995
) Improved forward EEG calculations using local mesh refinement of realistic head geometries.
Electroencephalogr Clin Neurophysiol
 
95
:
381
–392.