**4. General electrophysiological procedures**

Two different sounds duration were synthetically generated with short and long sounds. All of the stimuli were digitally edited to have an equal maximum energy level in dB SPL with the remaining intensity level within each of the stimuli scaled accordingly. The stimuli were digitally edited using the Cool Edit Pro v. 2.0 (Syntrillium Software Cooperation) with 500 ms duration (long sound) and 300 ms duration (short sound). All sounds were identical at their frequencies, thus eliminating any effect due to differences in frequency of occurrence of sound. The sounds were presented binaurally via headphones at a comfortable listening level of ~85 dB. The sound pressure levels of stimulus pairs were then measured at the output of headphones using a Brüel and Kjaer 2230 sound level meter. The standard (S)/deviant (D) pairs for each condition were [Condition 1: long-to-short sounds change] Standard/S-(2), Deviant/D-(1), [Condition 2: short-tolong sounds change] S-(1), D-(2). Thus, in both conditions pairs were designed to contrast short and long sounds. The stimuli were presented in a passive oddball paradigm. Deviant stimuli appeared randomly among the standards at 10% probability. Each condition included 125 deviants. The stimuli were binaurally delivered using SuperLab software (Cedrus Corporation, San Pedro, CA, USA) via headphones (Telephonic TDH-39-P). The inter-stimulus interval (ISI) was 1.25 second (offset-onset). EEG signal recording was time-locked to the onset of the sound. Participants were instructed not to pay attention to the stimuli presented via headphones, but rather to concentrate on a self-selected silent movie. Afterwards, they reported the impression of the movie.

Participants are instructed to sit relaxing in comfortable reclining chair in an electrically and acoustically dampened room. They were told that they would participate in the experiment and that the experimenter would be recording their brain electrical activity. They were given written instructions and provided with a grid for their judgements and a pen. They silently read the instructions and at the end the experimenter verifies that everything was clear. Their histories were taken, including age, educational level, handedness, occupation, current medications, medical history (which included past illness, surgical history, head trauma or accident) and history of alcohol consumption or smoking. If there was any significant history of neurological problems, psychiatric problems or head trauma, that participant was excluded. For the Mismatch Negativity (MMN) study, all participants were

Pre-Attentive Processing of Sound Duration Changes:

minus-standard-tone ERP differences.

Low Resolution Brain Electromagnetic Tomography Study 227

frontal (Fz) electrode site. Peak-picking of the prominent peak (MMN) was accomplished by means of moving an 'enhanced point' cursor through the waveforms displayed on the computer screen, while simultaneously paying attention to the resultant changes in the topographic maps. The mean MMN amplitudes at the frontal (Fz) electrode site were calculated as a mean voltage of the 40 ms intervals (so the peak plus minus 20 ms), centered at the corresponding peak latencies of the left and right frontal electrodes in the grandaveraged waveforms, separately for each stimulus type. The amplitudes were determined by using the 100 ms pre-stimulus baseline. When the participants were watching a silent, subtitled video, MMN to spatial acoustic changes was observed as a significant difference between ERPs to the deviant tones and those to the standard tones. It was at its maximum at the frontal (Fz) electrode site consisting of a negative deflection (note that analyses are based on averaged 40-ms blocks of sample points). MMN amplitudes were measured as the mean amplitude over the 100-300 ms period after the stimulus-onset from the deviant-tone ERP-

The average MMN latency was defined as a moment of the global field power with an epoch of 40-ms time window related stable scalp-potential topography (Pascual-Marqui, 1994). In the next step, low-resolution electromagnetic tomography (LORETA) was applied to estimate the current source density distribution in the brain, which contributes to the electrical scalp field (Pascual-Marqui *et al.,* 1994). Maps were computed with the Low Resolution Electromagnetic Tomography. Two radically oriented point sources (dipoles) in the brain were selected and computed the 21 channels forward solution electric potential map using a 3-shell unit radius spherical head model. The forward solution maps were then used as input for the LORETA computation in order to test the location precision and the ability of the method to separate the two known dipole locations. Scalp potentials rereferenced to the average reference, excluding the EOG electrodes, were interpolated for mapping using the surface spline method. The CSD maps were computed with the spherical spline interpolated data. The maps were computed at a single time point where the component in question was largest in the grand mean waveforms of each stimulus type and condition separately. LORETA computed the smoothest of all possible source configurations throughout the brain volume by minimizing the total squared Laplacian of source strengths. Low-resolution Electromagnetic Tomography (LORETA) is the new implementation of LORETA in the Talairach brain. LORETA makes use of the three-shell spherical head model registered to the Talairach human brain atlas (Talairach and Tournoux, 1988), available as a digitized MRI from the Brain Imaging Center, Montreal Neurologic Institute. Registration between spherical and realistic head geometry use EEG electrode coordinates reported by Towle et al. (1993). The solution space is restricted to cortical gray matter and hippocampus, as determined by the corresponding digitized Probability Atlas also available from Brain Imaging Center, Montreal Neurologic Institute. A voxel is labeled as gray matter if it meet the following three conditions: its probability of being gray matter is higher than that of being white matter, its probability of being gray matter is higher than that of being cerebrospinal fluid, and its probability of being gray matter is higher than 33%. Only gray matter voxels at 7-mm spatial resolution are produced under these neuroanatomical constraints. LORETA computations use the exact head model determined from each individual subject's MRI. The final step in any analysis procedure would be to cross-register

the individual's anatomical and functional image to the standard Talairach atlas.

instructed to ignore the stimuli by watching a silent, subtitled video of their choice (ignore condition). They were asked to avoid body and eye movements and to keep alert. Before the recording session, the task was explained and a practice block of 50 tones (50 deviants) was presented to the participant to ensure a good level performance. In order to avoid alpha rhythm synchronization during the recording session, participants were instructed to remain with their eyes open while watching a silent, subtitled video of their choice and were instructed to avoid eye movement and blinking. The total experimental session was 1-2 h, including approximately 0.20 h. for electrode placement. During the experimental session, participants took a rest breaks (one 15-min break occurring halfway through the recording session and shorter 5-min breaks as needed). Participants were tested in all experimental conditions on the same day.

## **5. Electroencephalographic processing**

EEG data were collected with a Quick-Cap equipped with 64 channels according to the international 10-20 system using Scan system (Scan 4.3, Neurosoft, Inc. Sterling, USA). Reference electrode was at both ear lobes. The signals were bandpass filtered at 0.05-100 Hz and digitized at 1000 Hz. The impedance of the electrode was below 5 kΩ. Eye movements were monitored with two EOG electrodes. Four electrodes monitored horizontal and vertical eye movements for off-line artifact rejection. Vertical and horizontal electrooculogram (EOG) was recorded by electrodes situated above and below the left eye, and on the outer canthi of both eyes, respectively. Epochs with EEG or EOG with a large (>100 μV) amplitude were automatically rejected. The artifact-free epochs were filtered at 0.1-15 Hz, baseline corrected and averaged. EEG was segmented into 1000 ms epochs, including the 100 ms pre-stimulus period. The average waveforms obtained from the standard and deviant were digitally filtered by a 0.1 - 15 Hz band-pass filter and finally baselinecorrected. Grand-averaged difference waveforms were calculated by subtracting the standard from the deviant waveforms. For each condition, presence of a prominent MMN was identified by measuring the integrated power amplitudes over the 40-ms time window centered on the MMN peak in the difference waveform. An MMN component was judged prominent if the amplitude difference between standard and deviant within predefined the window was statistically significant. For each participant, the averaged MMN responses contained 125 accepted deviants.

#### **6. Intracerebral distribution of differences in brain electrical activity**

In order to visualize and to measure the MMN (deviant-tone ERP-minus-standard-tone ERP difference), after the recording, differences were calculated by subtracting the ERP elicited by the standard tones from that elicited by the corresponding deviant tones of the same stimulus class. The MMN was quantified by first determining the MMN peak latency from the frontal (Fz) grand-average difference waves separately for each deviant. The latency windows for picking up the MMN peaks were predefined on the basis of the acrossparticipants peak latency distribution, determined by visual inspection. The MMN component was defined as the most prominent negative peak within the time windows between 100 and 300 ms. Latency and amplitude figures for waveforms were picked at their point of maximal deflection, as seen at their electrode site of maximal voltage distribution of

instructed to ignore the stimuli by watching a silent, subtitled video of their choice (ignore condition). They were asked to avoid body and eye movements and to keep alert. Before the recording session, the task was explained and a practice block of 50 tones (50 deviants) was presented to the participant to ensure a good level performance. In order to avoid alpha rhythm synchronization during the recording session, participants were instructed to remain with their eyes open while watching a silent, subtitled video of their choice and were instructed to avoid eye movement and blinking. The total experimental session was 1-2 h, including approximately 0.20 h. for electrode placement. During the experimental session, participants took a rest breaks (one 15-min break occurring halfway through the recording session and shorter 5-min breaks as needed). Participants were tested in all experimental

EEG data were collected with a Quick-Cap equipped with 64 channels according to the international 10-20 system using Scan system (Scan 4.3, Neurosoft, Inc. Sterling, USA). Reference electrode was at both ear lobes. The signals were bandpass filtered at 0.05-100 Hz and digitized at 1000 Hz. The impedance of the electrode was below 5 kΩ. Eye movements were monitored with two EOG electrodes. Four electrodes monitored horizontal and vertical eye movements for off-line artifact rejection. Vertical and horizontal electrooculogram (EOG) was recorded by electrodes situated above and below the left eye, and on the outer canthi of both eyes, respectively. Epochs with EEG or EOG with a large (>100 μV) amplitude were automatically rejected. The artifact-free epochs were filtered at 0.1-15 Hz, baseline corrected and averaged. EEG was segmented into 1000 ms epochs, including the 100 ms pre-stimulus period. The average waveforms obtained from the standard and deviant were digitally filtered by a 0.1 - 15 Hz band-pass filter and finally baselinecorrected. Grand-averaged difference waveforms were calculated by subtracting the standard from the deviant waveforms. For each condition, presence of a prominent MMN was identified by measuring the integrated power amplitudes over the 40-ms time window centered on the MMN peak in the difference waveform. An MMN component was judged prominent if the amplitude difference between standard and deviant within predefined the window was statistically significant. For each participant, the averaged MMN responses

**6. Intracerebral distribution of differences in brain electrical activity** 

In order to visualize and to measure the MMN (deviant-tone ERP-minus-standard-tone ERP difference), after the recording, differences were calculated by subtracting the ERP elicited by the standard tones from that elicited by the corresponding deviant tones of the same stimulus class. The MMN was quantified by first determining the MMN peak latency from the frontal (Fz) grand-average difference waves separately for each deviant. The latency windows for picking up the MMN peaks were predefined on the basis of the acrossparticipants peak latency distribution, determined by visual inspection. The MMN component was defined as the most prominent negative peak within the time windows between 100 and 300 ms. Latency and amplitude figures for waveforms were picked at their point of maximal deflection, as seen at their electrode site of maximal voltage distribution of

conditions on the same day.

contained 125 accepted deviants.

**5. Electroencephalographic processing** 

frontal (Fz) electrode site. Peak-picking of the prominent peak (MMN) was accomplished by means of moving an 'enhanced point' cursor through the waveforms displayed on the computer screen, while simultaneously paying attention to the resultant changes in the topographic maps. The mean MMN amplitudes at the frontal (Fz) electrode site were calculated as a mean voltage of the 40 ms intervals (so the peak plus minus 20 ms), centered at the corresponding peak latencies of the left and right frontal electrodes in the grandaveraged waveforms, separately for each stimulus type. The amplitudes were determined by using the 100 ms pre-stimulus baseline. When the participants were watching a silent, subtitled video, MMN to spatial acoustic changes was observed as a significant difference between ERPs to the deviant tones and those to the standard tones. It was at its maximum at the frontal (Fz) electrode site consisting of a negative deflection (note that analyses are based on averaged 40-ms blocks of sample points). MMN amplitudes were measured as the mean amplitude over the 100-300 ms period after the stimulus-onset from the deviant-tone ERPminus-standard-tone ERP differences.

The average MMN latency was defined as a moment of the global field power with an epoch of 40-ms time window related stable scalp-potential topography (Pascual-Marqui, 1994). In the next step, low-resolution electromagnetic tomography (LORETA) was applied to estimate the current source density distribution in the brain, which contributes to the electrical scalp field (Pascual-Marqui *et al.,* 1994). Maps were computed with the Low Resolution Electromagnetic Tomography. Two radically oriented point sources (dipoles) in the brain were selected and computed the 21 channels forward solution electric potential map using a 3-shell unit radius spherical head model. The forward solution maps were then used as input for the LORETA computation in order to test the location precision and the ability of the method to separate the two known dipole locations. Scalp potentials rereferenced to the average reference, excluding the EOG electrodes, were interpolated for mapping using the surface spline method. The CSD maps were computed with the spherical spline interpolated data. The maps were computed at a single time point where the component in question was largest in the grand mean waveforms of each stimulus type and condition separately. LORETA computed the smoothest of all possible source configurations throughout the brain volume by minimizing the total squared Laplacian of source strengths.

Low-resolution Electromagnetic Tomography (LORETA) is the new implementation of LORETA in the Talairach brain. LORETA makes use of the three-shell spherical head model registered to the Talairach human brain atlas (Talairach and Tournoux, 1988), available as a digitized MRI from the Brain Imaging Center, Montreal Neurologic Institute. Registration between spherical and realistic head geometry use EEG electrode coordinates reported by Towle et al. (1993). The solution space is restricted to cortical gray matter and hippocampus, as determined by the corresponding digitized Probability Atlas also available from Brain Imaging Center, Montreal Neurologic Institute. A voxel is labeled as gray matter if it meet the following three conditions: its probability of being gray matter is higher than that of being white matter, its probability of being gray matter is higher than that of being cerebrospinal fluid, and its probability of being gray matter is higher than 33%. Only gray matter voxels at 7-mm spatial resolution are produced under these neuroanatomical constraints. LORETA computations use the exact head model determined from each individual subject's MRI. The final step in any analysis procedure would be to cross-register the individual's anatomical and functional image to the standard Talairach atlas.

Pre-Attentive Processing of Sound Duration Changes:

expressed as mean ± S.D and all significant

possibly represents and the processes the sounds.

sound and a more medial distribution for the short one.

values were reported.

Low Resolution Brain Electromagnetic Tomography Study 229

elicited. One-sample *t*- tests were used to verify the presence of the MMN component, by comparing the mean amplitude of the 100 - 300 ms interval against a hypothetical zero, separately in each condition. The MMN latency values was also compared. Repeated measure ANOVA was carried out on the topographic descriptors of the MMN. In order to gather information on cortical sources specifically involved in the MMN generation, LORETA images for deviant sounds were compared with those for standard sounds using paired *t*-test statistics, after logarithmic transformation of the data. All results were

For the LORETA analyses, the average LORETA images were constructed across participants: the brain electric activity during the ERPs amplitude waveforms for each condition and the voxel-by-voxel *t* test differences between conditions. The voxel-by-voxel paired *t* tests were run to assess in which cortical regions the conditions differed. The *t* maps were threshold at *p* < 0.0001*.* As pointed out above, reliable differences in the scalp ERP field configuration can unambiguously be interpreted as suggesting that at least partially different neuronal populations are active during the conditions. LORETA assesses in which brain regions the conditions differed. The Structure-Probability Maps Atlas (Lancaster *et al.,* 1997a; 1997b) was also used to determine which brain regions were involved in differences between conditions. Brodmann area(s) (BA) and brain regions closet to the observed locations identified by the Tarairach coordinates were reported. Overall, one sample *P*-

**8. Pre-attentive processing and lateralization of sound duration changes** 

The finding indicated that the prominent response to both sounds elicited MMN peaking at 128 to 212 ms from stimulus onset. The grand-averaged ERPs showed that the MMN mean amplitude of both sounds was statistically significant (*t*-test). The paired sample *t*-test revealed a significant difference between conditions (*t* (10) = 73.00; *p* < 0.0001) showing that both sounds equally elicited a MMN. The magnitude of the acoustic difference between the stimulus pairs was reflected by the MMN amplitude, showing larger MMN amplitudes in long sound compared to the short one. The difference in MMN latencies to both sounds might reflect differential processing of the human auditory cortex. The delay in the MMN to the long sound might reflect additional time required to process sound perception. This processing apparently involves activation of a memory trace, or cell assembly, which

Estimated source localization of the average MMN responses evoked by both sounds was clearly identified. The current source density values in the time frame 128-212 ms poststimulus were calculated with LORETA. Stronger activation for long sound was found at 212 ms in the left middle temporal gyrus (MTG) (-59, -32, 1; *t*-value, 1.81), while the short sound most strongly activated at 128 ms in the left superior temporal gyrus (STG) (-59, -39, 8; *t*-value, 1.03) (see Figure 1). Analysis of the MMN responses indicated left-hemispheric laterality in both sound durations (*F* (3,30) = 47.02; *p* < 0.0001). The source analysis indicated strongest MMN response tentatively originating in the left hemisphere and possibly involving the perisylvian area in both sounds, with a more superior distribution for the long

The individual momentary potential measures from 21 electrodes at the MMN latency were analyzed with LORETA to determine the MMN source loci (Pascual-Marqui, 1994). These latencies were between 100-140 ms for long- and short-sound duration changes. LORETA calculated the current source density distribution in the brain, which contributed to the electrical scalp field, at each of 2395 voxels in the gray matter and the hippocampus of a reference brain (MNI 305, Brain Imaging Centre, Montreal Neurological Institute) based on the linear weighted sum of the scalp electric potentials (Pascual-Marqui, 1994). LORETA chooses the smoothest of all possible current density configurations throughout the brain volume by minimizing the total squared Laplacian of source strengths. This procedure only implicates that neighboring voxels should have a maximally similar electrical activity, no other assumptions were made. The applied version of LORETA used a three-shell spherical head model registered to the Talairach space and calculated the three-dimensional localization of the electrical sources contributing to the electrical scalp filed for all subjects and conditions, defining the regions of interest on the basis of local maxima of the LORETA distribution. Stereotaxic coordinates of the voxels of the local maxima were determined within areas of significant relative change associated with the tasks. The anatomical localization of these local maxima was assessed with reference to the standard Stereotaxic atlas, and validation of this method of localization was obtained by superimposition of the SPM maps on a standard MRI brain provided by the SPM99. Peaks located within superior temporal gyrus was also identified by using published probability maps following a correction for the differences in the coordinate systems between the Talairach and Tournoux atlas and the Stereotaxic space employed by SPM99.

Regarding to the Brodmann areas(s) and brain regions localization, the Talairach Daemon (TD) will be taken into consideration. The Talairach Daemon (TD) is a high-speed database server for querying and retrieving data about human brain structure over the Internet (http://ric.uthscsa.edu/td\_applet/). The TD server data is searched using x-y-z coordinates resolved to 1x1x1 mm volume elements within a standardized stereotaxic space. An array, indexed by x-y-z coordinates, that spans 170 mm (x), 210 mm (y) and 200 mm (z), provides high-speed access to data. Array dimensions are selected to be approximately 25% larger than those of the Co-planar Stereotaxic Atlas of the Human Brain (Talairach and Tournoux, 1988). Coordinates tracked by the TD server are spatially consistent with the Talairach Atlas. Each array location stores a pointer to a relation record that holds data describing what is present at the corresponding coordinate. Presently, the data in relation records are either Structure Probability Maps (SP Maps) or Talairach Atlas Labels, though others can be easily added. The relation records are implemented as linked lists to names and values for brain structures. The TD server is run on a Sun SPARCstation 20 with 200 Mbytes of memory. Intention is to provide 24-hour access to the data using a variety of client applications, as well as continue to add more brain structure information to the database.

#### **7. Statistical evaluation**

The statistical significance of MMN (deviant-minus-standard difference) was tested with one-sample *t*-tests by comparing the mean MMN amplitude at the frontal (Fz) electrode site, where the MMN was most prominent. The MMN was measured using the mean frontal (Fz) amplitude in the 100 - 300 ms interval of the deviant-minus-standard difference curves. This interval included the grand mean MMN peak latencies in those conditions where MMN was

The individual momentary potential measures from 21 electrodes at the MMN latency were analyzed with LORETA to determine the MMN source loci (Pascual-Marqui, 1994). These latencies were between 100-140 ms for long- and short-sound duration changes. LORETA calculated the current source density distribution in the brain, which contributed to the electrical scalp field, at each of 2395 voxels in the gray matter and the hippocampus of a reference brain (MNI 305, Brain Imaging Centre, Montreal Neurological Institute) based on the linear weighted sum of the scalp electric potentials (Pascual-Marqui, 1994). LORETA chooses the smoothest of all possible current density configurations throughout the brain volume by minimizing the total squared Laplacian of source strengths. This procedure only implicates that neighboring voxels should have a maximally similar electrical activity, no other assumptions were made. The applied version of LORETA used a three-shell spherical head model registered to the Talairach space and calculated the three-dimensional localization of the electrical sources contributing to the electrical scalp filed for all subjects and conditions, defining the regions of interest on the basis of local maxima of the LORETA distribution. Stereotaxic coordinates of the voxels of the local maxima were determined within areas of significant relative change associated with the tasks. The anatomical localization of these local maxima was assessed with reference to the standard Stereotaxic atlas, and validation of this method of localization was obtained by superimposition of the SPM maps on a standard MRI brain provided by the SPM99. Peaks located within superior temporal gyrus was also identified by using published probability maps following a correction for the differences in the coordinate systems between the Talairach and Tournoux

Regarding to the Brodmann areas(s) and brain regions localization, the Talairach Daemon (TD) will be taken into consideration. The Talairach Daemon (TD) is a high-speed database server for querying and retrieving data about human brain structure over the Internet (http://ric.uthscsa.edu/td\_applet/). The TD server data is searched using x-y-z coordinates resolved to 1x1x1 mm volume elements within a standardized stereotaxic space. An array, indexed by x-y-z coordinates, that spans 170 mm (x), 210 mm (y) and 200 mm (z), provides high-speed access to data. Array dimensions are selected to be approximately 25% larger than those of the Co-planar Stereotaxic Atlas of the Human Brain (Talairach and Tournoux, 1988). Coordinates tracked by the TD server are spatially consistent with the Talairach Atlas. Each array location stores a pointer to a relation record that holds data describing what is present at the corresponding coordinate. Presently, the data in relation records are either Structure Probability Maps (SP Maps) or Talairach Atlas Labels, though others can be easily added. The relation records are implemented as linked lists to names and values for brain structures. The TD server is run on a Sun SPARCstation 20 with 200 Mbytes of memory. Intention is to provide 24-hour access to the data using a variety of client applications, as

The statistical significance of MMN (deviant-minus-standard difference) was tested with one-sample *t*-tests by comparing the mean MMN amplitude at the frontal (Fz) electrode site, where the MMN was most prominent. The MMN was measured using the mean frontal (Fz) amplitude in the 100 - 300 ms interval of the deviant-minus-standard difference curves. This interval included the grand mean MMN peak latencies in those conditions where MMN was

well as continue to add more brain structure information to the database.

atlas and the Stereotaxic space employed by SPM99.

**7. Statistical evaluation** 

elicited. One-sample *t*- tests were used to verify the presence of the MMN component, by comparing the mean amplitude of the 100 - 300 ms interval against a hypothetical zero, separately in each condition. The MMN latency values was also compared. Repeated measure ANOVA was carried out on the topographic descriptors of the MMN. In order to gather information on cortical sources specifically involved in the MMN generation, LORETA images for deviant sounds were compared with those for standard sounds using paired *t*-test statistics, after logarithmic transformation of the data. All results were expressed as mean ± S.D and all significant

For the LORETA analyses, the average LORETA images were constructed across participants: the brain electric activity during the ERPs amplitude waveforms for each condition and the voxel-by-voxel *t* test differences between conditions. The voxel-by-voxel paired *t* tests were run to assess in which cortical regions the conditions differed. The *t* maps were threshold at *p* < 0.0001*.* As pointed out above, reliable differences in the scalp ERP field configuration can unambiguously be interpreted as suggesting that at least partially different neuronal populations are active during the conditions. LORETA assesses in which brain regions the conditions differed. The Structure-Probability Maps Atlas (Lancaster *et al.,* 1997a; 1997b) was also used to determine which brain regions were involved in differences between conditions. Brodmann area(s) (BA) and brain regions closet to the observed locations identified by the Tarairach coordinates were reported. Overall, one sample *P*values were reported.
