Behavioral Data

Correct response rates did not differ between attention and control tasks (mean value, 93.1% and 92.2%, respectively, p > .05), or within the attention task, between spatial and pitch variation processing (93.2% and 92.9%, respectively, p > .05). All subjects reported that they were mentally following spatial or pitch variations of the auditory streams in the AS and AP conditions while they were actively ignoring the auditory stream trajectories in the control task. Moreover, the EOG signal was drifting during stimulus presentation and no attention or acoustic variation effect was noticed in the EOG time courses (p > .05) until 700 msec after variation onset.

Topographical Analysis

The main component elicited by the auditory stimulus onset, N1, showed a classical ERP topography with a fronto-central maximum and a polarity reversal across the sylvian fissure, in both A and C conditions (Figure 1). Computing scalp current densities (SCDs) by surface laplacian has proved to be efficient in separating temporal, frontal (Giard et al., 1994), and parietal (Kaiser & Bertrand, 2003) components of auditory responses. Applied to the present data, SCD maps showed, in both tasks, bilateral current sink/source patterns over temporal regions and a current sink over fronto-central areas (Figure 1). These components were maintained throughout the stimulus duration.

Open in a separate window

Figure 1

Stream-onset N1 topographies. Left hemisphere ERP and SCD maps (mean value between 80 and 130 msec, 14 subjects) in attention and control conditions. SCD allows to better disclose temporal and frontal components. Black ovals indicate electrode groups chosen for the analysis: fronto-central current sink (F), left superior temporal current sink (LST), and left inferior temporal current source (LIT). Very similar topographies were obtained over the right hemisphere and symmetrical electrode groups were chosen for the right temporal components.

After this N1 component, we found a parieto-occipital component largely prominent in the attention condition (Figure 2). SCD analysis permitted to dissociate two subcomponents: a current source over superior parietal cortices and bilateral sources over temporo-parieto-occipital (TPO) junctions, clearly visible on the attention-minus-control maps (Figure 2).

Open in a separate window

Figure 2

Topographies during the varying period. ERP and SCD maps (left, back, and right views, mean value between 385 and 595 msec after variation onset, 14 subjects) in attention and control conditions, and difference maps between conditions. In ERP maps, a positive component is present in the attention condition and is topographically distributed over parietal and occipital areas. SCD maps allow to disclose distinct and focal current sources over the superior parietal cortex for both tasks and over TPO junctions in the attention condition only, clearly visible on the attention-minus-control maps. Ovals indicate the electrode groups chosen for the analysis: superior parietal current source (P), right and left TPO current sources (RTPO and LTPO, respectively), in addition to the frontal and temporal components already defined in the N1 topography (Figure 1).

In order to estimate the time courses of those components, SCD maps were averaged across subjects and subsets of electrodes were selected to cover eight regions. Fronto-central (F: Fz/Cz), right/left superior temporal (RST: C4 and LST: C3), and right/left inferior temporal (RIT: T8/P8 and LIT: T7/P7) electrode groups were defined from the N1 topography (Figure 1), whereas superior parietal (P: CP1/Pz/CP2) and right/left temporo-parieto-occipital (RTPO: PO4 and LTPO: PO3) electrode groups were from the topographies of the varying period (Figure 2). SCD curves were averaged within each group for subsequent statistical analysis (Figures 3 and ​and44).

Open in a separate window

Figure 3

Temporal and frontal SCD time courses. Mean SCD curves (14 subjects) are drawn over the three analysis periods (defined in Figure 6C) for the frontal (F), left superior and inferior temporal (LST and LIT, respectively), and right superior and inferior temporal (RST and RIT, respectively) electrode groups. The time structure of the stimulus is shown at the top of the figure (gray ovals represent additional noise-bursts used in the control task). During the stationary part, only mean attention (A) and control (C) curves are drawn, whereas in the varying part all conditions are depicted with different line symbols. The evoked N1/P2 complex to auditory stream onset is clearly visible at all electrode groups. N1/P2 to acoustic variations onset can also be noticed, although less pronounced. Note the polarity reversal throughout the stimulus duration between superior and inferior temporal components bilaterally. Significant differences between the attention and the control conditions are indicated by gray bars (ANOVA: *p < .05, **p < .01, ***p < .001). During the varying part, the left temporal components show a significantly more pronounced sustained response for pitch (AP and CP) than for spatial variations (AS and CS) with stars symbolizing the level of significance (ANOVA: *p < .05, ***p < .001). The black horizontal bars at the LST site indicate interaction effects between attention and acoustic variation (p < .05).

Open in a separate window

Figure 4

Parietal SCD time courses. Mean SCD curves (14 subjects) are presented for the three analysis periods (defined in Figure 6C) for the LTPO and RTPO, respectively, and the superior parietal (P) electrode groups. The time structure of the stimulus is shown at the top of the figure (gray ovals represent additional noise-bursts used in the control task). During the stationary part, only mean attention (A) and control (C) curves are drawn, whereas in the varying part all conditions are depicted with different line symbols. Note the large TPO activities developing in the attention condition only and a sustained parietal activity enhanced by attention. Significant differences between the attention and the control conditions are indicated by gray bars (ANOVA: *p < .05, **p < .01, ***p < .001).

Fronto-Central Component

The frontal component (F in Figure 3) started with a transient N1/P2 complex followed by a sustained negative deflection throughout the stimulus duration (emerging at 210 msec for all conditions, p < .001), on which a small N1/P2 complex was superimposed at variation onset. During the stationary period, both stream-onset N1 mean amplitude (80–130 msec) and the sustained negative deflection (210–420 msec, p < .05, and 525–735 msec, p < .05) were more negative for the attention condition (p < .001). No effect was detected for the P2 component. During the varying part, no attention, acoustic variation, or interaction effect was found significant for the sustained component.

Temporal Components

The temporal N1/P2 complexes were followed by a sustained bilateral deflection with polarity reversal, maintained during the entire stream duration (emerging at 210 msec for all conditions, p < .01 at RST, p < .05 at LST, and p < .01 at LIT).

During the stationary period, N1 mean amplitude (80–130 msec) to auditory stream onset was significantly more negative for the attention than the control condition at LST (p < .05), RST (p < .05), and LIT (p < .001) sites. The sustained waves were partially enhanced in the A condition during the stationary period as indicated by gray bars in Figure 3. No effect was detected for the P2 component.

At stream variation onset, a small N1/P2 complex was superimposed on the sustained waves. During the varying period, pitch variations induced a larger sustained activity than spatial variations over the left auditory cortex, irrespective of attention (LST: all 210-msec windows from 385 to 910 msec, p < .01 and LIT: all 210-msec windows from 595 to 910 msec, p < .05). Additionally, an interaction effect (Attention × Acoustic variations) was found at the LST site (all 210-msec windows from 490 to 910 msec, p < .05). Wilcoxon tests showed that the negativity was significantly larger for the AP condition than for AS and CP (p < .01), and for CP than CS (p < .05), whereas no significant difference was found between the AS and CS conditions. The same effects were found during the final period as indicated in Figure 3.

The apparent “noise” on these temporal waveforms actually corresponded to the periodic steady-state response (35 msec) elicited by the amplitude modulation of the auditory stream.

Temporo-Parieto-Occipital Components

Bilateral positive sustained components (LTPO and RTPO in Figure 4) were significantly emerging, over the TPO junctions, in the A condition only, from 250 msec to the end of the stationary period (p < .05) and from 350 msec after the variation onset until the end of the stream (p < .05). In the A condition, irrespective of the type of acoustic variations, these components were significantly more positive during the stationary (all 210-msec windows from 210 to 735 msec, p < .05), varying (all 210-msec windows from 280 to 910 msec, p < .01), and final (all 210-msec windows from −840 to −420 msec, p < .01) periods. No difference between pitch and spatial variations, no interaction effect, and no interhemispheric difference were found significant for these components.

Attention Network

As parietal and frontal activities were maintained during the whole duration of the stream and were only modulated by the orientation of attention and not by the acoustic feature to be integrated, they are likely to be related to high-level task-related processes.

Temporo-Parieto-Occipital Components

By means of SCD mapping, it has also been possible to detect large and very focal bilateral TPO activities in the attention condition. To our knowledge, no similar activity has ever been found in neuroimaging studies on sound motion, which were generally based on passive listening or simple stimuli. In contrast, we used more complex varying stimuli in a demanding task. In the attention condition, to be able to follow both types of acoustic variations, most of the subjects (11 among 14) asserted that they had mentally generated a visual representation of the sound. It should be noted that TPO electrode sites are close to V5 visual areas (Figure 5) known to be activated during visual motion perception and mental imagery (Goebel, Khorram-Sefat, Muckli, Hacker, & Singer, 1998). The bilateral TPO activation could thus correspond to visual mental imagery induced by both types of acoustic variation processing. The existence of these same activities with small amplitude before the acoustic variation onset could be associated with a preparatory attentional process selectively involving the brain areas which will be later engaged in visual mental imagery.

Open in a separate window

Figure 5

Surface electrode positions on a normalized brain with neuroimaging findings. The Talairach coordinates of activations induced by visual motion in area V5 (diamonds) were collected in several neuroimaging studies (Dumoulin et al., 2000; Lewis et al., 2000; Sunaert, Van Hecke, Marchal, & Orban, 1999, 2000; Chawla et al., 1999; Goebel et al., 1998). These activations were superimposed on a 3-D rendered magnetic resonance image template from the Montreal Neurological Institute (MNI), back view. Dashed ovals cover V5 activities. The EEG electrodes of the present experiment (10–20 and 10–10 electrode systems and additional intermediate electrodes) are positioned on a realistic scalp surface (Oostenveld & Praamstra, 2001), and head and brain contours are coregistered (rotation and scaling) to the MNI template. Electrodes corresponding to RTPO and LTPO activities are indicated in black.

Temporal Components

At last, sustained bilateral temporal components were characterized by SCD polarity reversals throughout the stimulus duration. These activities most likely originate from superior temporal auditory areas (Gutschalk, Patterson, Rupp, Uppenkamp, & Scherg, 2002) and reflect auditory integration. We found these bilateral temporal activities for spatial and pitch processing in both tasks. Previous studies have suggested bilateral (Hart et al., 2004; Warren & Griffiths, 2003; Pavani et al., 2002; Warren et al., 2002; Zatorre, Bouffard, et al., 2002; Lewis et al., 2000) or right-lateralized (Baumgart et al., 1999) activation of the planum temporale when contrasting moving or spatially distributed sounds to stationary sounds. For pitch processing, right-lateralized (Patterson, Uppenkamp, Johnsrude, & Griffiths, 2002; Zatorre et al., 1994) or bilateral (Hart et al., 2004; Warren & Griffiths, 2003; Thivard, Belin, Zilbovicius, Poline, & Samson, 2000) activation in superior temporal areas has been found when contrasting sequences with changing pitch to sequences with fixed pitch.

During the varying part, left temporal activities were particularly more important for pitch than for spatial variations. This result suggests a differential processing of spatial and pitch variations in the left temporal auditory cortex. This dissociation could correspond either to an increased neuronal activity for pitch processing or to different configurations of neural sources for spatial and pitch processing. To our knowledge, few studies have been interested in the difference between pitch and spatial processing within the temporal cortex. By contrasting sound sequences with changing pitch to sequences with changing location, Hart et al. (2004) and Warren and Griffiths (2003) have shown a dissociation between pitch and spatial processing in the right and left superior temporal planes. Nevertheless, our left-lateralized effect is consistent with Hart et al.’s additional activity for pitch processing in the left planum polare. According to these studies, a posteromedial (spatial)–anterolateral (pitch) dissociation would exist within the superior temporal cortex of both hemispheres. Close configurations of neural sources for spatial and pitch processing that could not be disentangled with the spatial resolution of SCD computation could explain why we could not find any difference in the right hemisphere.

As an interaction effect between attention and acoustic variations has been found during the same period for the left superior component, pitch and spatial feature integration within the auditory cortex could correspond to automatic processes that are differentially modulated by attention. In addition, there was a small attention effect on the left temporal activity during the stationary part. This effect prior to stream variation could be due to an attention-dependent preactivation of the auditory areas which will be subsequently involved in differential feature processing.

All these modulations of the temporal activities are consistent with a current hypothesis (Griffiths & Warren, 2002) suggesting a role of the planum temporale in differentially processing auditory spectro-temporal information.

Task

Subjects had to perform two different tasks in separate blocks. In each block, the stimuli consisted of pitch-varying streams randomly alternating with spatially varying streams with equal probabilities. In the stream-oriented attention task (called attention task), subjects were asked to follow sound variations and report its final direction (high/low pitch or left/right) with a four-direction joystick (up/down or left/right, respectively). In the diverted attention task (called control task), subjects were asked to compare two successive noise-bursts superimposed to the end of the stream, thus orienting their attention away from the varying stream. These two bursts randomly differed in location or pitch, independently of the underlying stream variations. Subjects had to report with the same joystick whether the second burst was to the right or to the left, or lower- or higher-pitched than the first one. Performances were balanced between tasks and also between spatial and pitch feature processing with difficulty levels adapted to each subject (see Procedure section). Each new stimulus started between 3.3 and 4.1 sec after the subject’s response and no specific speed instructions were given.

Stimuli

Auditory streams consisted of 35-msec band-passed noise-bursts [5-semitone (st) wide, starting from a frequency F, and cosine tapered] and were presented through headphones (Sennheiser PX30, Wedemark, Germany). In both tasks, they were composed of two successive parts, a stationary part of random duration (735, 805, 875 msec with equal probabilities) followed by an acoustically varying part (duration ranging from 980 to 2205 msec).

Stream trajectories were piecewise-linear functions, identical for pitch and spatial variations (Figure 6A). Trajectories of pitch-varying streams were following piecewise-linear functions ΔF(t) defined in semitones (st). These functions were equal to zero during the stationary part and were increasing or decreasing between −ΔFm and +ΔFm during the varying part. Trajectories of spatially varying streams were following piecewise-linear functions ΔA(t) defined in degrees. These functions were equal to 0° (ahead) during the stationary part and were increasing or decreasing between −ΔAm and +ΔAm during the varying part. Depending on the difficulty level in the attention task, ΔAm was ranging between 30° and 60°, and ΔFm between 6 and 14 st. Nevertheless, the speed of variation remained about the same across difficulty levels (around 30 st/sec and 160°/sec). To minimize predictability, there were, for each difficulty level, at least six different trajectories varying by the number of segments of the functions ΔF(t) and ΔA(t) (randomized between 4 and 7). Moreover, the trajectories were interrupted at variable times, either when the variations were reaching an extremum or four bursts before. This led to about 12 different time courses, identical for spatial and pitch variations, consisting of 3 to 6 direction changes for each difficulty level.

Open in a separate window

Figure 6

Stimulus structure. (A) Stream trajectories. Auditory streams were composed of two successive parts, a stationary and a varying part. Streams varied in location or pitch, following zigzag trajectories. Each trajectory of pitch-varying stream was following a piecewise-linear function ΔF(t) (gray line) defined in semitones. In the case of spatial variations, the functions ΔA(t) were varying between −ΔAm and +ΔAm. These functions were identical for pitch and spatial variations. They consisted of 4 to 7 segments and could be interrupted at variable times, either when the variations were reaching an extremum (black diamonds) or four bursts before (black circles). (B) Stream frequency content. Gray bars represent the 35-msec, band-passed and 5-st wide noise-bursts which composed of auditory streams. In the case of pitch-varying streams, the lower frequency F of each burst was defined in semitones by the function: F(t) = ΔF(t) + F0 ± 1 with ΔF(t) as described in B. The black dashed line represents the trajectory ΔF(t)+F0, and the black dashes represent the actual lower frequency F(t) of each burst. (C) Analysis periods. Three periods were chosen for signals analysis: stationary ([−150, 735 msec], time 0 being the onset of the stream), varying ([−150, 910 msec], time 0 being the onset of the varying part), and final ([−840, −420 msec], time 0 being the offset of the stream). They correspond to the periods having the shortest duration for each part and common to both tasks, that is, before noise-burst (gray ovals) occurrence in the control condition. Freq = frequency; loc = location; st = semitones.

To minimize habituation effects for both types of variation, the lower frequency F of each burst was randomly varying by ±1 st around F0, with F0 equiprobably chosen between 1318, 1480, and 1661 Hz for each stimulus. In the case of pitch-varying streams, the lower frequency of each burst was thus defined in semitones by the function: F(t) = ΔF(t) + F0 ± 1 with ΔF(t) as described above (see Figure 6B). In the case of spatially varying streams, F(t) is only varying around F0, F(t) = F0 ± 1, throughout stimulus duration. Pitch-varying streams were localized ahead at 0° azimuth.

To give the impression of a continuously moving sound source, each burst was convoluted with an average-listener, head-related transfer function (see http://sound.media.mit.edu/KEMAR.html) with a 5°-step in azimuth. Compared to other studies using only interaural time differences to position sounds in space with earphones, this procedure creates a better percept of sound objects located in the external space by modulating phase, amplitude, and frequency content of the sounds.

In the control condition, two additional noise-bursts (5-st wide, starting from a frequency F, and lasting 100 msec) appeared successively 420 and 140 msec before the end of the stream (Figure 6C). They differed either in location [azimuths: ±ΔAb; frequency: F] or in pitch [azimuth: 0°; frequency: F ± ΔFb]. Depending on the difficulty level, ΔAb was ranging from 5° to 35° and ΔFb from 1 to 8 st. Because the trajectories were randomized and interrupted at variable times, subjects could not predict the end of the stream, and thus, the occurrence of the noise-bursts.

More generally, to minimize habituation and predictability, many acoustic parameters were randomized: the central pitch of the stationary part (F0), the pitch of each burst (F), and the stream trajectory. Furthermore, by randomizing stream duration, the subject’s attention was maintained focused throughout the stimulus in both tasks, and sustained processing of the stream trajectories was ensured in the attention task. With such a randomization strategy, it seems unlikely that subjects could memorize and recognize all these trajectory patterns during a recording session.

Procedure

In the first session, subjects were selected on the basis of their ability to perceive spatial variations, to perform the different tasks, and to minimize blinks and eye movements assessed by electrooculogram. First, they were familiarized with sound sequences and tasks in a stepwise learning procedure using visual feedback. Then, they performed blocks of each task with different difficulty levels, until we could determine the difficulty level corresponding to an 80% correct response rate for each kind of stimuli in each task (spatially or pitch-varying streams in the attention task and location- or pitch-differing noise-bursts in the control task). Only 1 subject out of 17 had to be excluded because of excessive eye blinks.

In the second session, subjects’ EEG were recorded during 12 blocks of 30 trials in both attention and control tasks, yielding a total of 360 trials per task (180 with spatial and 180 with pitch variations). Attention (A) and control (C) blocks were presented in an AA–CC–AA–CC–. . . design. For each subject, the difficulty level was determined during the selection procedure as mentioned above. In some cases, because of subjects’ performance improvement or tiredness, the difficulty level was adapted after each block to maintain a constant performance (around 80% correct response rate) throughout the session. Subjects were instructed to keep their eyes open while fixating a point, and to refrain from blinking during stimulus presentation.

EEG Recordings

The subjects were seated in an armchair in a sound-attenuated and electrically shielded chamber. EEG was recorded from 36 scalp electrodes (Easy Cap, FMS, Herrsching, Germany) referenced to the nose (ground electrode at forehead and impedances below 5 kΩ) and EOG from two electrodes were placed centrally above the right eye and at its outer canthus. The electrode positions are shown schematically in Figure 7. Signals were amplified, filtered (0.05–200 Hz bandwidth), and sampled at 1000 Hz (Synamps, Neuroscan Labs, Sterling, VA).

Open in a separate window

Figure 7

EEG recording sites. Electrodes were placed according to the following scheme: 19 locations of the 10–20 electrode system, 13 locations of the 10–10 electrode system, and 4 additional intermediate locations. One should note the high density of electrodes over parietal areas required for precise SCD estimates.

We thank Jochen Kaiser for helpful discussions, and Pierre-Emmanuel Aguera and Jean-François Echallier for technical support.