- Split View
-
Views
-
Cite
Cite
Sophie Molholm, Antigona Martinez, Walter Ritter, Daniel C. Javitt, John J. Foxe, The Neural Circuitry of Pre-attentive Auditory Change-detection: An fMRI Study of Pitch and Duration Mismatch Negativity generators, Cerebral Cortex, Volume 15, Issue 5, May 2005, Pages 545–551, https://doi.org/10.1093/cercor/bhh155
- Share Icon Share
Abstract
Electrophysiological studies have revealed a pre-attentive change-detection system in the auditory modality. This system emits a signal termed the mismatch negativity (MMN) when any detectable change in a regular pattern of auditory stimulation occurs. The precise intracranial sources underlying MMN generation, and in particular whether these vary as a function of the acoustic feature that changes, is a matter of some debate. Using functional magnetic resonance imaging, we show that anatomically distinct networks of auditory cortices are activated as a function of the deviating acoustic feature — in this case, tone frequency and tone duration — strongly supporting the hypothesis that MMN generators in auditory cortex are feature dependent. We also detail regions of the frontal and parietal cortices activated by change-detection processes. These regions also show feature dependence and we hypothesize that they reflect recruitment of attention-switching mechanisms.
Introduction
The pre-attentive detection of change in the environment is an essential element of perception and cognition in humans. Such mechanisms mediate the seemingly automatic shift of attention to new and potentially important information, whilst operating without drawing upon limited attentional resources. A well-established dependent measure of pre-attentive auditory change-detection is found in the mismatch negativity.
The mismatch negativity (MMN) is an electrical brain response that is elicited by any discriminable change in the regularity of the acoustic environment. The MMN is typically measured under conditions in which participants are instructed to ignore the auditory stimulation and read a book or watch a movie. Given that the MMN can be elicited in the absence of attention, when attention is directed toward an unrelated and demanding task (Alho et al., 1994a; Woldorff et al., 1998), and even when subjects are sleeping (most reliably during the REM stage of sleep) (e.g. Atienza and Cantero, 2001; Atienza et al., 2002), the MMN system is considered to operate pre-attentively. The system underlying generation of the MMN is considered to maintain a representation of acoustic regularities of the recent past (Cowan, 1984; Cowan et al., 1993), with the MMN elicited when this representation is updated (Winkler et al., 1996; Näätänen, 1992); that is, when there has been a change.
MMN generation is typically studied using an oddball paradigm in which, in its simplest rendition, an infrequent tone (termed the ‘deviant’) is interspersed at random intervals within a train of frequent tones (termed the ‘standard’). The scalp-recorded MMN, best viewed by subtracting the response to the standard from the response to the deviant, is a negative fronto-centrally distributed wave that often inverts at temporal sites around the mastoids, and peaks 100–200 ms post-deviance onset. The principal neural source of the MMN has been determined to be located within the supratemporal plane in or near primary auditory cortex on the basis of dipole analysis of magnetic recordings (Sams and Hari, 1991) and electrical recordings (Giard et al., 1990; Scherg and Berg, 1991) obtained from the scalp of humans, complemented by intracranial recordings in cat (Csépe et al., 1987) and macaques (Javitt et al., 1992, 1994).
An important theoretical question that remains is whether there is a single neural network in auditory cortex that is responsible for generating the MMN regardless of the acoustic dimension that has changed or whether the MMN network differs depending upon the nature of the acoustic change. For example, does a different circuit generate the MMN for a change in frequency versus a change in the duration of a repeating tone? Several models of the MMN have assumed the latter (Näätänen, 1992; Winkler et al., 1996), and there are data from dipole source modeling and scalp topographic mapping consistent with the notion that the neural networks underlying the generation of the MMN vary based upon the characteristics of the eliciting stimuli. For example, Paavilainen et al. (1991) found slightly different ERP topographies, all consistent with generators in auditory cortex, for MMNs elicited by frequency, intensity and duration deviants. This finding is supported by those of Giard et al. (1995), who localized the source dipoles of MMNs elicited by the same three types of deviants to different locations within auditory cortex. Likewise, in MMNm (the magnetic counterpart of the MMN) studies, source modeling has shown different cortical loci on the superior temporal plane for MMNms elicited by inter-stimulus interval deviants (that is, changes in presentation rate) and duration deviants (Levänen et al., 1996), and for MMNms elicited by frequency, duration, and intensity deviants (Rosburg, 2003). In contrast to the above findings, however, other studies using dipole-modeling techniques have failed to find significant differences in the loci of MMN generators for different acoustic features [Sams et al. (1991) and Schairer et al. (2001) both examine frequency, duration and intensity MMNs].
While the methods of source modeling and scalp topographic mapping are highly useful tools to estimate the location of the neural sources underlying a given electro-magnetic scalp topography, they only provide general estimations, and have shown mixed results concerning the question of interest. Therefore, in the present study, we used the substantially better anatomical localization afforded by functional magnetic resonance imaging (fMRI) to test the hypothesis that the neural network that generates the MMN response depends upon the characteristics of the eliciting stimuli.
To our knowledge, only two previous imaging studies have investigated the anatomy of the neural networks underlying MMNs to changes in different acoustic features. Unfortunately, in one of these an alternative explanation may account for findings of different neural networks underlying different MMNs (Mathiak et al., 2002), and in the other the question could not be addressed because no MMN-related activations were detected for one of the two deviants that were tested (Doeller et al., 2003). Mathiak et al. (2002) examined fMRI MMN activations to intensity and duration deviants, compared with activity elicited by standard stimuli. Whereas an infrequent decrease in stimulus intensity activated bilateral primary and secondary auditory cortices, only right hemisphere secondary auditory areas were activated by an infrequent decrease in stimulus duration. While this finding generally supports the notion that the neural generator of the MMN is dependent on the acoustic feature that is varied, differences in intensity maps within auditory cortex (Clarke and Rivier, 1998; Schreiner et al., 1992) could well account for the greater MMN-related activation in the intensity compared with the duration conditions. In a separate study, Doeller et al. (2003) used fMRI to examine MMN activations as a function of changes in both tone frequency and tone location. While frequency deviants resulted in significant changes in the BOLD signal within the right STG, the authors failed to find any MMN activations for the location condition. This was the case despite reliable location-MMNs being elicited in the same set of subjects during a separate ERP recording session.
In the present study we used a MMN paradigm designed specifically to eliminate a problem inherent in many of the previous imaging studies of the MMN (Celsis et al., 1999; Opitz et al., 1999; Wible et al., 2001; Müller et al., 2002; Sevostianov et al., 2002; Liebenthal et al., 2003). In these studies, all of which employed frequency deviants, activations elicited by blocks of stimulation comprising both standard and deviant tones were compared with activations elicited by blocks of stimulation in which only the standard tone was presented. Any differences in observed patterns of activations resulting from these two conditions were interpreted as reflecting MMN-related activity. However, due to the tonotopic organization of primary and secondary auditory cortices (Merzinich and Brugge, 1973; Kaas and Hackett, 1998; Schonwiesner et al., 2002), the reported differences in activations might well be explained by differences in the sensory stimulation given during oddball and control blocks. That is, in the oddball (standard and deviant) blocks of these studies, two tones of different frequencies are presented and thus two distinct regions of tonotopic auditory cortex will necessarily be activated, resulting in a larger overall area of activation. On the other hand, in the control condition, only the standard tone is presented and as such only one tonotopic region is activated. Thus, in the control condition, one class of generators would be activated (i.e. those associated with the specific frequency of the standard), whereas in the oddball condition, three classes of generators are putatively activated; those associated with the frequency of the standard, those associated with the frequency of the deviant and those associated with the MMN. Problematically, there is no way to precisely determine the extent to which the additional activations of the oddball condition are due to frequency specific activation associated with the deviant versus activation associated with the MMN. Similar considerations apply to event-related fMRI designs, in that the standard will elicit activity associated with its frequency and the deviant will elicit activity associated with its frequency as well as activity associated with the MMN (Opitz et al., 2002; Doeller et al., 2003).
We have surmounted this problem in the present design by using a sequencing technique that equates for stimulus energies across both MMN and control blocks. Using this modified design, we show that separable neural circuits within auditory cortices underlie the MMN to frequency change and duration change. Further, our data strongly suggest that the neural circuitry activated by the duration-MMN and subsequent processing, recorded while subjects ignore auditory stimulation and watch a silent video, engages much of the neural circuitry activated by active processing of temporal information, as seen in unrelated studies on temporal processing (Schubotz and von Cramon, 2001; Macar et al., 2002; Lewis and Miall, 2003; Coull et al., 2004).
Materials and Methods
Subjects
Twenty subjects participated (age 27 ± 8, 14 female). All were right-handed and neurologically normal. Subjects provided written, informed consent according to institutional guidelines, and were paid for their participation. The Institutional Review Board of the Nathan Kline Institute approved all procedures.
Stimuli
Tones differing in frequency or duration served as stimuli. The stimuli were presented at a comfortable listening level that was clearly audible above the MRI scanner noise; intensity level was adjusted on an individual subject basis (∼90 dB). Tone 1 was a 500 Hz tone of 100 ms duration, tone 2 was a 400 Hz tone of 100 ms duration and tone 3 was a 500 Hz tone of 50 ms duration.
Procedure
Subjects were fitted with pneumatic ear inserts, and their ears were then covered by custom-built foam and plastic earmuffs that served to further attenuate magnet noise while allowing for delivery of auditory stimulation. Once subjects were placed in the bore of the magnet a test scan was acquired to check the audibility of the auditory stimuli over the scanner-generated noise, and to ensure that the stimuli were presented at a comfortable listening level. Subjects were instructed to watch a movie (without sound) presented on a custom designed LCD screen and to ignore all tone events.
In devising the MMN and control blocks it was essential to match overall sensory stimulation. We took advantage of an MMN paradigm in which two tones are presented equiprobably and the order of their presentation varied (Sams et al., 1983; Giese-Davis et al., 1993). In this paradigm tones are arranged to compose alternating ‘mini-sequences’ of tones 1 and 2. The number of trials in a given mini-sequence varies such that the occurrence of a switch from tone 1 to tone 2 (and vice-versa) is irregular. As such, the switch trial tone is a ‘deviant’ and elicits the MMN, and the repeating tone is the ‘standard’. This basic stimulation paradigm, which has proven to elicit robust MMNs, was used for the MMN condition (Sams et al., 1983; Giese-Davis et al., 1993). For the matched control condition to the aforementioned MMN condition, the same two tones were alternated sequentially to form a regular pattern, such that no MMN would be elicited. In this way, the frequency specific activations of tonotopic cortices produced by the two tones occur in both conditions and activity specific to the MMN generators and MMN related processes can be readily discerned.
During a single fMRI scanning session, five stimulation conditions (‘frequency MMN’, ‘frequency control’, ‘duration MMN’, ‘duration control’ and ‘rest’) were administered in a block fashion during two separate scans. Both MMN blocks consisted of alternating ‘mini-sequences’ of two tones: mini-sequences were 2, 3, or 4 repetitions of a single tone (each sequence length was represented equiprobably) followed by a mini-sequence of the other tone, and so on. The first tone in a sequence was expected to generate a MMN; this was confirmed by an electrophysiological pilot study in which the MMN was elicited using the same stimuli and stimulus paradigm as in the fMRI study. A constant SOA of 500 ms was employed for both the MMN and control conditions. Control blocks consisted of two tones alternating in a regular fashion (e.g. T1, T2, T1, T2, etc.). Figure 1 provides a schematic of the experimental paradigm.
Each subject participated in one ‘frequency scan’ and one ‘duration scan’. Each scan lasted 10.5 minutes. In the frequency scan, MMN-generating mini-sequences (described above) of tone 1 (T1) and tone 2 (T2) (same duration but differing in frequency) were presented in 32 s blocks that were alternated with control blocks consisting of regular sequences of T1 and T2. Each of these blocks was presented five times over the course of a scan and was interspersed with 32 s rest blocks during which no auditory stimulation was delivered through the ear inserts. Siémilarly, during the ‘duration scan’ mini-sequences of T1 and T3 (same frequency but varying in duration) were alternated with control blocks of sequentially alternating presentations of the same two tones and interspersed with rest blocks. Half of the subjects participated in the duration scan first.
Data Acquisition and Analyses
T2*-weighted echo-planar images (EPI/TE/flip angle = 2 s, 64 ms, 90°; voxel size = 4 × 4 × 5 mm; matrix size = 64 × 64) were acquired on a 1.5T Siemens VISION scanner. The TRs were evenly distributed across the acquisition cycle so that auditory stimulation from the scanner was continuous during the scan. This resulted in a streaming into the background of the continuous scanner noise such that it clearly segregated from the auditory stimuli presented over the headphones. During each scan, 320 volumes were acquired in the axial plane, on each of 22 contiguous slices covering the entire brain. The first five images were discarded to allow for stabilization of the blood oxygenation level dependent (BOLD) signal.
For anatomical localization, high-resolution (13 mm) T1-weighted images of the whole brain were acquired using a standard three-dimensional magnetization prepared rapid gradient echo (MPRAGE) pulse sequence. Anatomical images were normalized into Talairach coordinates (Talairach and Tournoux, 1998) and the functional data were registered to the normalized anatomical data.
All data processing and analysis was conducted using the AFNI image-analysis software package (Cox, 1996). Images were realigned to an image at approximately the mid-point of the time-series acquisition. In all subjects, head motion never exceeded 0.75 mm along any axis. Prior to statistical analyses the images were spatially smoothed (FWHM 6 mm) to compensate for individual variation in cortical anatomy, and normalized into Talairach coordinates (Talairach and Tournoux, 1998). Data acquired from each of the two types of scans (frequency and duration) were analyzed separately. To estimate the BOLD response associated with each condition, regressors representing the timing of each stimulation epoch were convolved with a canonical hemodynamic response function and used in a multiple regression analysis.
Functional Auditory ROI
For both the frequency and duration scans, regression coefficients were modeled separately for the MMN, control, and rest blocks of each scan. For each participant, a general linear test (GLT) contrasting activity associated with the control condition with that associated with rest was performed. The linear coefficient values from these contrasts were entered into a groupwise t-test and compared against the null hypothesis. A single functional auditory region of interest (ROI) was generated by combining all voxels that were significantly activated (P < 0.05) during the control condition (relative to rest) in both scans (frequency and duration).
Change-detection Related Activations
In our principal analysis we sought to compare functional activity in auditory cortex associated with the MMN response (and related auditory change-detection processes) as a function of the acoustic feature of the eliciting deviant. As such, we used linear regression to model the regressor coefficients for MMN, control and rest conditions for each of the two scan types. Regressor coefficients for MMN and control conditions were contrasted, on an individual subject basis, in a GLT. The resulting linear coefficients were entered into separate t-tests and tested, across subjects, against the null hypothesis. These t-tests considered only those voxels falling within the groupwise functional ROI (above). The results from both t-tests were further constrained to include only clusters of four or more neighboring voxels with t values equivalent to P < 0.05.
As a secondary analysis, and to provide a more thorough description of the group data, the linear coefficients from the MMN versus control contrasts (for frequency and duration scans) were entered into two additional t-tests where the ROI constraint was lifted. For this more liberal analysis only significantly activated voxels belonging to clusters of eight or more were considered.
Results
Sensory Auditory Activation: Control Blocks versus Rest Blocks
There was a significant increase in the BOLD response in auditory cortices for the control auditory stimulation blocks when compared with the rest blocks. This activation extended throughout much of the superior temporal plane of the right and left hemispheres, extending anteriorly into the posterior portion of the frontal lobes and posteriorly into the posterior portion of the superior temporal sulcus. This pattern of activation was similar for both the frequency and duration scans. The functional auditory ROI used in the group analysis is shown in Figure 2 shaded in green.
MMN Blocks versus Control Blocks
The ROI analysis revealed a significant increase in the BOLD response for MMN compared with control blocks for both the duration and frequency scans. In the right hemisphere these activations were situated within the superior temporal gyrus (STG), in primary auditory cortex for the frequency scan, and in secondary auditory cortex for the duration scan. Consistent with findings from the majority of ERP and MEG studies (Levänen et al., 1996; Frodl-Bauch et al., 1997; Rosburg, 2003), MMN-related activation in the duration-MMN scan was located posterior to that obtained in the corresponding frequency-MMN scan (see Fig. 2), In the left hemisphere these activations were found in the transverse temporal gyrus (in secondary auditory cortex) for the duration scan, and in the posterior STG for the frequency scan. In contrast to the right hemisphere, the frequency-MMN activation was more posterior than the duration-MMN activation. In both hemispheres, the duration-MMN activations were slightly lateral to the frequency-MMN activations. The data from this analysis clearly show that anatomically distinct regions of auditory cortex are involved in the generation of duration- and frequency-MMNs. Table 1 presents the Talairach coordinates of these activations.
. | Brain region . | Center of mass Talairach coordinates: x, y, z . | Volume, Ml . | T(1,19) . | P . |
---|---|---|---|---|---|
Duration | R superior temporal gyrus | 64, −26, 15 | 320 | 2.13 | 0.047 |
L transverse temporal gyrus | −57, −14, 11 | 320 | 2.70 | 0.010 | |
Frequency | R superior temporal gyrus | 56, −17, 6 | 256 | 2.45 | 0.025 |
L superior temporal gyrus | −45, −44, 19 | 256 | 2.38 | 0.028 |
. | Brain region . | Center of mass Talairach coordinates: x, y, z . | Volume, Ml . | T(1,19) . | P . |
---|---|---|---|---|---|
Duration | R superior temporal gyrus | 64, −26, 15 | 320 | 2.13 | 0.047 |
L transverse temporal gyrus | −57, −14, 11 | 320 | 2.70 | 0.010 | |
Frequency | R superior temporal gyrus | 56, −17, 6 | 256 | 2.45 | 0.025 |
L superior temporal gyrus | −45, −44, 19 | 256 | 2.38 | 0.028 |
. | Brain region . | Center of mass Talairach coordinates: x, y, z . | Volume, Ml . | T(1,19) . | P . |
---|---|---|---|---|---|
Duration | R superior temporal gyrus | 64, −26, 15 | 320 | 2.13 | 0.047 |
L transverse temporal gyrus | −57, −14, 11 | 320 | 2.70 | 0.010 | |
Frequency | R superior temporal gyrus | 56, −17, 6 | 256 | 2.45 | 0.025 |
L superior temporal gyrus | −45, −44, 19 | 256 | 2.38 | 0.028 |
. | Brain region . | Center of mass Talairach coordinates: x, y, z . | Volume, Ml . | T(1,19) . | P . |
---|---|---|---|---|---|
Duration | R superior temporal gyrus | 64, −26, 15 | 320 | 2.13 | 0.047 |
L transverse temporal gyrus | −57, −14, 11 | 320 | 2.70 | 0.010 | |
Frequency | R superior temporal gyrus | 56, −17, 6 | 256 | 2.45 | 0.025 |
L superior temporal gyrus | −45, −44, 19 | 256 | 2.38 | 0.028 |
A second statistical analysis, in which the entire brain volume was considered, revealed additional extra-auditory cortical activations related to change-detection processes, mainly in frontal and parietal regions (see Fig. 3). These were found for both the duration and frequency scans. For the duration MMN versus control contrast there were significant bilateral increases in superior frontal cortices, the middle temporal lobes, and the inferior and precuneus regions of the parietal lobes. There were right hemisphere unilateral MMN-related activations of the middle frontal gyrus (including the supplementary motor area and right dorsal premotor cortex), the middle temporal gyrus, and the anterior cingulate; and in the left hemisphere of the inferior frontal gyrus and the post central gyrus of the parietal lobe. Comparing the frequency MMN condition to its corresponding control condition yielded significant enhancements in the BOLD signal within the inferior and middle frontal gyrus, inferior parietal lobule, and middle occipital gyrus of the right hemisphere. In the left hemisphere only the superior frontal gyrus contained any significant clusters of activation. Table 2 presents Talairach coordinates of these activations.
. | Brain region . | Center of mass Talairach coordinates: x, y, z . | Volume, Ml . | T(1,19) . | P . |
---|---|---|---|---|---|
Duration | L inferior frontal gyrus | −52, 35, −4 | 1856 | 2.80 | 0.012 |
L superior frontal gyrus | −6, 60, 23 | 1600 | 2.83 | 0.011 | |
−41, 39, 28 | 512 | 2.31 | 0.033 | ||
R inferior parietal lobe | 42, −51, 50 | 3904 | 2.50 | 0.022 | |
R parietal lobe, precunius | 18, −72, 40 | 768 | 2.47 | 0.024 | |
R middle frontal gyrus | 49, 33, 26 | 896 | 2.42 | 0.026 | |
33, 1, 62 | 512 | 2.46 | 0.024 | ||
R frontal lobe, medial frontal gyrus | 1, −13, 70 | 1856 | 2.33 | 0.031 | |
R superior frontal gyrus | 14, 50, 33 | 1024 | 2.25 | 0.037 | |
8, 12, 57 | 576 | 2.48 | 0.023 | ||
L parietal lobe, precunius | −1, −50, 39 | 3328 | 2.92 | 0.009 | |
L parietal lobe, postcentral gyrus | −46, −25, 42 | 1920 | 2.65 | 0.016 | |
L inferior parietal lobe | −51, −59, 44 | 2368 | 2.82 | 0.010 | |
L parietal lobe, postcentral gyrus | −2, −50, 68 | 512 | 2.42 | 0.026 | |
R anterior cingulate | 1, 40, −4 | 576 | 2.6 | 0.018 | |
R middle temporal gyrus | 56, −50, −5 | 1216 | 2.53 | 0.021 | |
L middle temporal gyrus | −56, −53, 6 | 512 | 2.44 | 0.025 | |
Frequency | R middle frontal gyrus | 33, 39, 23 | 1344 | 2.67 | 0.016 |
33, 12, 26 | 640 | 2.44 | 0.026 | ||
42, 2, 53 | 512 | 2.30 | 0.033 | ||
R inferior frontal gyrus | 52, 10, 18 | 960 | 2.19 | 0.042 | |
L superior frontal gyrus | −20, 23, 51 | 512 | 2.73 | 0.014 | |
R inferior parietal lobule | 53, −38, 45 | 1152 | 2.74 | 0.014 | |
R middle occipital gyrus | 41, −75, 6 | 832 | 2.28 | 0.035 |
. | Brain region . | Center of mass Talairach coordinates: x, y, z . | Volume, Ml . | T(1,19) . | P . |
---|---|---|---|---|---|
Duration | L inferior frontal gyrus | −52, 35, −4 | 1856 | 2.80 | 0.012 |
L superior frontal gyrus | −6, 60, 23 | 1600 | 2.83 | 0.011 | |
−41, 39, 28 | 512 | 2.31 | 0.033 | ||
R inferior parietal lobe | 42, −51, 50 | 3904 | 2.50 | 0.022 | |
R parietal lobe, precunius | 18, −72, 40 | 768 | 2.47 | 0.024 | |
R middle frontal gyrus | 49, 33, 26 | 896 | 2.42 | 0.026 | |
33, 1, 62 | 512 | 2.46 | 0.024 | ||
R frontal lobe, medial frontal gyrus | 1, −13, 70 | 1856 | 2.33 | 0.031 | |
R superior frontal gyrus | 14, 50, 33 | 1024 | 2.25 | 0.037 | |
8, 12, 57 | 576 | 2.48 | 0.023 | ||
L parietal lobe, precunius | −1, −50, 39 | 3328 | 2.92 | 0.009 | |
L parietal lobe, postcentral gyrus | −46, −25, 42 | 1920 | 2.65 | 0.016 | |
L inferior parietal lobe | −51, −59, 44 | 2368 | 2.82 | 0.010 | |
L parietal lobe, postcentral gyrus | −2, −50, 68 | 512 | 2.42 | 0.026 | |
R anterior cingulate | 1, 40, −4 | 576 | 2.6 | 0.018 | |
R middle temporal gyrus | 56, −50, −5 | 1216 | 2.53 | 0.021 | |
L middle temporal gyrus | −56, −53, 6 | 512 | 2.44 | 0.025 | |
Frequency | R middle frontal gyrus | 33, 39, 23 | 1344 | 2.67 | 0.016 |
33, 12, 26 | 640 | 2.44 | 0.026 | ||
42, 2, 53 | 512 | 2.30 | 0.033 | ||
R inferior frontal gyrus | 52, 10, 18 | 960 | 2.19 | 0.042 | |
L superior frontal gyrus | −20, 23, 51 | 512 | 2.73 | 0.014 | |
R inferior parietal lobule | 53, −38, 45 | 1152 | 2.74 | 0.014 | |
R middle occipital gyrus | 41, −75, 6 | 832 | 2.28 | 0.035 |
. | Brain region . | Center of mass Talairach coordinates: x, y, z . | Volume, Ml . | T(1,19) . | P . |
---|---|---|---|---|---|
Duration | L inferior frontal gyrus | −52, 35, −4 | 1856 | 2.80 | 0.012 |
L superior frontal gyrus | −6, 60, 23 | 1600 | 2.83 | 0.011 | |
−41, 39, 28 | 512 | 2.31 | 0.033 | ||
R inferior parietal lobe | 42, −51, 50 | 3904 | 2.50 | 0.022 | |
R parietal lobe, precunius | 18, −72, 40 | 768 | 2.47 | 0.024 | |
R middle frontal gyrus | 49, 33, 26 | 896 | 2.42 | 0.026 | |
33, 1, 62 | 512 | 2.46 | 0.024 | ||
R frontal lobe, medial frontal gyrus | 1, −13, 70 | 1856 | 2.33 | 0.031 | |
R superior frontal gyrus | 14, 50, 33 | 1024 | 2.25 | 0.037 | |
8, 12, 57 | 576 | 2.48 | 0.023 | ||
L parietal lobe, precunius | −1, −50, 39 | 3328 | 2.92 | 0.009 | |
L parietal lobe, postcentral gyrus | −46, −25, 42 | 1920 | 2.65 | 0.016 | |
L inferior parietal lobe | −51, −59, 44 | 2368 | 2.82 | 0.010 | |
L parietal lobe, postcentral gyrus | −2, −50, 68 | 512 | 2.42 | 0.026 | |
R anterior cingulate | 1, 40, −4 | 576 | 2.6 | 0.018 | |
R middle temporal gyrus | 56, −50, −5 | 1216 | 2.53 | 0.021 | |
L middle temporal gyrus | −56, −53, 6 | 512 | 2.44 | 0.025 | |
Frequency | R middle frontal gyrus | 33, 39, 23 | 1344 | 2.67 | 0.016 |
33, 12, 26 | 640 | 2.44 | 0.026 | ||
42, 2, 53 | 512 | 2.30 | 0.033 | ||
R inferior frontal gyrus | 52, 10, 18 | 960 | 2.19 | 0.042 | |
L superior frontal gyrus | −20, 23, 51 | 512 | 2.73 | 0.014 | |
R inferior parietal lobule | 53, −38, 45 | 1152 | 2.74 | 0.014 | |
R middle occipital gyrus | 41, −75, 6 | 832 | 2.28 | 0.035 |
. | Brain region . | Center of mass Talairach coordinates: x, y, z . | Volume, Ml . | T(1,19) . | P . |
---|---|---|---|---|---|
Duration | L inferior frontal gyrus | −52, 35, −4 | 1856 | 2.80 | 0.012 |
L superior frontal gyrus | −6, 60, 23 | 1600 | 2.83 | 0.011 | |
−41, 39, 28 | 512 | 2.31 | 0.033 | ||
R inferior parietal lobe | 42, −51, 50 | 3904 | 2.50 | 0.022 | |
R parietal lobe, precunius | 18, −72, 40 | 768 | 2.47 | 0.024 | |
R middle frontal gyrus | 49, 33, 26 | 896 | 2.42 | 0.026 | |
33, 1, 62 | 512 | 2.46 | 0.024 | ||
R frontal lobe, medial frontal gyrus | 1, −13, 70 | 1856 | 2.33 | 0.031 | |
R superior frontal gyrus | 14, 50, 33 | 1024 | 2.25 | 0.037 | |
8, 12, 57 | 576 | 2.48 | 0.023 | ||
L parietal lobe, precunius | −1, −50, 39 | 3328 | 2.92 | 0.009 | |
L parietal lobe, postcentral gyrus | −46, −25, 42 | 1920 | 2.65 | 0.016 | |
L inferior parietal lobe | −51, −59, 44 | 2368 | 2.82 | 0.010 | |
L parietal lobe, postcentral gyrus | −2, −50, 68 | 512 | 2.42 | 0.026 | |
R anterior cingulate | 1, 40, −4 | 576 | 2.6 | 0.018 | |
R middle temporal gyrus | 56, −50, −5 | 1216 | 2.53 | 0.021 | |
L middle temporal gyrus | −56, −53, 6 | 512 | 2.44 | 0.025 | |
Frequency | R middle frontal gyrus | 33, 39, 23 | 1344 | 2.67 | 0.016 |
33, 12, 26 | 640 | 2.44 | 0.026 | ||
42, 2, 53 | 512 | 2.30 | 0.033 | ||
R inferior frontal gyrus | 52, 10, 18 | 960 | 2.19 | 0.042 | |
L superior frontal gyrus | −20, 23, 51 | 512 | 2.73 | 0.014 | |
R inferior parietal lobule | 53, −38, 45 | 1152 | 2.74 | 0.014 | |
R middle occipital gyrus | 41, −75, 6 | 832 | 2.28 | 0.035 |
Discussion
Here we show that the neural generators underlying the auditory pre-attentive change-detection system vary as a function of the characteristics of the eliciting stimuli: infrequent changes in the frequency of a repeating tone resulted in differential patterns of MMN-related activation in auditory cortex compared with infrequent changes in the duration of a repeating tone. Thus the MMN system indicates not only that a change has occurred, but also the nature of the change. Our finding supports a model of the MMN in which it functions to update feature-specific sensory memories of acoustic regularities in the environment. Information about the nature of the change may help with the rapid evaluation of the significance of the change.
Our finding of MMN-related activation in the STG is generally compatible with those seen in previous imaging studies (Celsis et al., 1999; Opitz et al., 1999, 2002; Dittmann-Balçar et al., 2001; Wible et al., 2001; Mathiak et al., 2002; Müller et al., 2002; Sevostianov et al., 2002; Doeller et al., 2003; Liebenthal et al., 2003; Schall et al., 2003), even though in the majority of these studies the comparison condition did not fully account for sensory stimulation. This is not surprising since at the cortical level, MMN generation is largely achieved in the auditory cortices, and the basic processing of auditory stimuli at the cortical level is, of course, also achieved in the auditory cortices. Thus the inability to separate activations related to basic auditory processing from MMN-related processing was not expected to result in an altogether different pattern of activation. One would predict more extensive ‘MMN-activations’ in studies where overall stimulation was not matched across conditions. It is not surprising then that most imaging studies that have compared MMN and control (standard only) blocks, or used an event-related design in which the response to the standard is compared with the response to the frequency-deviant, have indeed shown activations in auditory cortex that are more extensive than we find here. It is also worth noting that there has been a relatively high degree of variance in the cortical location of the frequency MMN across previous studies, most of which used fixed-effects designs and are therefore not generalizable to the population at large (Desmond and Glover, 2002). On the other hand, a remarkably consistent location is found in right auditory cortex between the three studies to date — including this one — that have employed large subject populations and random-effects designs (Opitz et al., 2002; Doeller et al., 2003, in which there was no corresponding activation in the left STG; the present study).
In addition to MMN activations in auditory cortex, there were frontal and parietal activations related to change-detection processes for both the frequency and duration scans (see Fig. 3). Numerous studies have demonstrated that elicitation of the MMN draws attention to the deviant stimulus (Schröger, 1996; Escera et al., 1998; Schröger and Wolff, 1998), and it is likely that this frontal and parietal activity is in part related to attention switching mechanisms that follow elicitation of the MMN. This activity may be related to generation of the electrophysiological component, the P3a, which often follows the MMN, has neural sources in frontal and parietal cortices (Soltani and Knight, 2000) and is considered to be related to the involuntary switching of attention to the deviant (Escera et al., 1998; Schröger and Wolff, 1998). In addition, a right frontal MMN generator has been hypothesized by Giard et al. (1990) based on scalp current density maps of the MMN (see also Deouell et al., 1998) and receives some support from human lesion studies (Alho et al., 1994b; Alain et al., 1998) and imaging studies (Celsis et al., 1999; Dittmann-Balçar et al., 2001; Müller et al., 2002; Opitz et al., 2002; Schall et al., 2003). Thus frontal activations may also be partially related to MMN generation. Such a frontal generator has been interpreted as the part of the MMN system that initiates the involuntary switching of attention following the initial MMN response to the deviant (Giard et al., 1990; Näätänen, 1992), or alternatively as an enhancing mechanism for small/difficult to detect changes (Opitz et al., 2002; Doeller et al., 2003). Some studies have localized the frontal activity to the inferior frontal gyrus (Opitz et al., 2002). In the present study activation of the right IFG was seen in the frequency condition, whereas there was activation of the left IFG in the duration condition.
While frontal and parietal change-detection activations were seen in both the duration and frequency scans, there were notable differences in terms of extensiveness and hemispheric dominance. There was a bias to the right hemisphere for frequency activations, whereas duration activations were bilateral or biased to the left hemisphere. These differences in lateralization correspond well with the notion that the right hemisphere is more involved in the processing of tonal information (frequency differences) while the left hemisphere is more involved in the processing of temporal information (duration differences) (Zatorre et al., 1994; Fiez et al., 1996; Platel et al., 1997; Belin et al., 1998; Coull and Nobre, 1998; Nobre, 2001; Zatorre and Belin, 2001). In the inferior parietal lobe, duration change-detection activations were bilateral, whereas frequency change-detection activations were limited to the right hemisphere. In frontal regions, change-detection activations were more extensive for duration than for frequency, with duration showing activations in both the right and left hemispheres, and frequency change-detection activity seen mostly in the right hemisphere. Different regions of the IFG, which has previously been associated with MMN processes, were sensitive to change-detection processing in the duration and frequency scans. In this study, duration-related activation was limited to the left hemisphere and was seen in the frontal operculum. The location of this activation was very similar to that in a PET study by Dittmann-Balçar et al. (2001) that looked at duration MMN activations, except that the latter were seen in the right instead of the left hemisphere. The frontal operculum has been associated with temporal perception in a number of studies (Schubotz and von Cramon, 2001; Lewis and Miall, 2003; Coull et al., 2004). In contrast, activation of the IFG for frequency change-detection was specific to the right hemisphere, in a region consistent with the MMN-related activations observed in Opitz et al. (2002) and posterior to that seen in the duration scan of the present study. Frontally there was also activation of right dorsal premotor cortex for duration change-detection (Tailarach coordinates: 33, 1, 62), a region that has been shown to correlate with attention to a temporal judgment task (Coull et al., 2004), and the supplementary motor area (Tailarach coordinates: 1, −13, 70), which is also associated with temporal processing (Macar et al., 2002).
To conclude, using an experimental paradigm that controlled for stimulation across comparison conditions, we found that MMN generators in auditory cortex vary as a function of the nature of the deviating stimulus. Thus the MMN system exhibits feature specificity. Such specificity suggests that the MMN response indicates not only that a change has occurred, but also the nature of that change. The activation of frontal and parietal regions by change-detection processes in both the frequency and duration scans was consistent with the recruitment of attention switching processes.
We thank Deirdre Foxe and Patrick O'Donnell for technical assistance in collecting these data. Work supported by grants from the NIMH (MH65350 and MH63434 to J.J.F.) and the NINDS (NS30029 to W.R.).
References
Alain C, Woods DL, Knight RT (
Alho K, Woods DL, Algazi A (
Alho K, Woods DL, Algazi A, Knight RT, Näätänen R (
Atienza M, Cantero JL (
Atienza M, Cantero JL, Dominguez-Marin E (
Belin P, Zilbovicius M, Crozier S, Thivard L, Fontaine A, Masure MC, Samson Y (
Celsis P, Boulanouar K, Doyon B, Ranjeva JP, Berry I, Nespoulous JL, Chollet F (
Clarke S, Rivier F (
Coull JT, Nobre AC (
Coull JT, Vidal F, Nazarian B, Macar F (
Cowan N, Winkler I, Teder W, Näätänen R (
Cox RW (
Csépe V, Karmos G, Molnar M (
Deouell LY, Bentin S, Giard MH (
Desmond JE, Glover GH (
Dittmann-Balçar A, Juptner M, Jentzen W, Schall U (
Doeller CF, Opitz B, Mecklinger A, Krick C, Reith W, Schroger E (
Escera C, Alho K, Winkler I, Näätänen R (
Fiez JA, Raichle ME, Balota DA, Tallal P, Petersen SE (
Frodl-Bauch T, Kathmann N, Moller HJ, Hegerl U (
Giard MH, Lavaikainen J, Reinikainen K, Perrin F, Bertrand O, Pernier J, Näätänen R (
Giard MH, Perrin F, Pernier J, Bouchet P (
Giese-Davis JE, Miller GA, Knight RA (
Javitt DC, Schroeder CE, Steinschneider M, Arezzo JC, Vaughan HG, Jr. (
Javitt DC, Steinschneider M, Schroeder CE, Vaughan HG Jr, Arezzo JC (
Kaas JH, Hackett TA (
Levänen S, Ahonen A, Hari R, McEvoy L, Sams M (
Lewis PA, Miall RC (
Liebenthal E, Ellingson ML, Spanaki MV, Prieto TE, Ropella KM, Binder JR (
Macar F, Lejeune H, Bonnet M, Ferrara A, Pouthas V, Vidal F, Maquet P (
Mathiak K, Rapp A, Kircher TT, Grodd W, Hertrich I, Weiskopf N, Lutzenberger W, Ackermann H (
Merzinich MM, Brugge JF (
Müller BW, Juptner M, Jentzen W, Muller SP (
Opitz B, Mecklinger A, von Cramon DY, Kruggel F (
Opitz B, Rinne T, Mecklinger A, von Cramon DY, Schroger E (
Paavilainen P, Alho K, Reinikainen K, Sams M, Näätänen R (
Platel H, Price C, Baron JC, Wise R, Lambert J, Frackowiak RS, Lechevalier B, Eustache F (
Rosburg T (
Sams M, Alho K, Näätänen R (
Sams M, Hari R (
Sams M, Kaukoranta E, Hamalainen M, Näätänen R (
Schairer KS, Gould HJ, Pousson MA (
Schall U, Johnston P, Todd J, Ward PB, Michie PT (
Scherg M, Berg P (
Schreiner CE, Mendelson JR, Sutter ML (
Schröger E (
Schröger E, Wolff C (
Schubotz RI, von Cramon DY (
Sevostianov A, Fromm S, Nechaev V, Horwitz B, Braun A (
Talairach J, Tournoux P (
Wible CG, Kubicki M, Yoo SS, Kacher DF, Salisbury DF, Anderson MC, Shenton ME, Hirayasu Y, Kikinis R, Jolesz FA, McCarley RW (
Winkler I, Karmos G, Näätänen R (
Woldorff MG, Hillyard SA, Gallen CC, Hampson SR, Bloom FE (
Zatorre RJ, Belin P (
Author notes
1The Cognitive Neurophysiology Laboratory, Program in Cognitive Neuroscience and Schizophrenia, Nathan Kline Institute for Psychiatric Research, 140 Old Orangeburg Road, Orangeburg, New York 10962, USA and 2Departments of Neuroscience & Psychiatry, Albert Einstein College of Medicine, Bronx, New York 10462, USA