Proceedings paper

 

Functional brain imaging of rhythm perception.

Renaud Brochard1, André Dufour2, Carolyn Drake3 and Christian Scheiber4.

 

1 Keele University, Psychology Department, Staffs ST5 5BG, United-Kingdom (psa32@keele.ac.uk).

2 Centre d'études de Physiologie appliquée, Université Louis Pasteur, Strasbourg, France

3 Laboratoire de Psychologie Expérimentale, Université Paris V and CNRS, Boulogne/Seine, France.

4 Institut de Physique Biologique, Hôpital Civil, Strasbourg, France

 

 Very few biological studies have investigated how we perceive musical rhythm. In the present research, we aimed to find the cerebral bases of perceptual and cognitive processes of rhythm using Functional Magnetic Resonance Imaging (fMRI). Many authors (e.g. Lerdahl and Jackendoff, 1983; Povel and Essens, 1985; Parncutt, 1994; Drake, 1998; see Clarke, 1999 for a recent review) have proposed the existence of two types of temporal processes that appear fundamental in the perception of simple rhythmic sequences: the segmentation of an ongoing sequence into groups of events on the basis of their physical characteristics and the extraction of underlying temporal regularities.

The first process ("grouping") is based on gestalt principles and depends, among other physical characteristics, on the relative proximity in time of sound events. The occurrence of a longer gap between two events creates a boundary leading to the perception of two distinct perceptual units. A sound sequence is thus perceived as a succession of rhythmic groups.

The second fundamental process ("meter") may occur in parallel. It corresponds to the extraction of temporal regularities in the sequence in the form of an underlying pulse, often influenced by the alternation of strong and weak beats in the musical sequence (that is the metric structure stricto sensu).

Both grouping and metric processes seem to depend on the complexity and the hierarchical structure of the musical sequences but seem to be functional in early life (see Drake, 1998 for a review). Neuropsychological studies in brain-lesioned or "surgical" patients have shown that the processing of these two rhythmic organizations could be impaired relatively independently from one another, suggesting that processing metric and grouping characteristics may involve different mechanisms and networks in the human brain (Peretz, 1990; Liégois-Chauvel et al, 1998). In order to investigate this problem, we asked subjects to compare pairs of short rhythmic sequences differing in the position of one event moved towards another group ("grouping" condition) or slightly displaced in order to disrupt the underlying pulse ("metric" condition). During these tasks, we measured their brain activity in an fMRI scanner.

  

Method :

Subjects

9 subjects (3 males, 6 females, mean age: 25; right-handed) participated in the experiment. All reported normal hearing and no neurological history.

Stimuli

Our stimuli consisted of pairs of short (3.8 sec) rhythmic sequences separated by a silence of 1.7 sec. Each rhythmic event consisted of a complex percussive sound with a fundamental pitch corresponding to A4 (F0=440Hz). Each sound lasted 50ms and had a level of 90 dB SPL. The sounds were played through headphones via a pneumatic system. The average level of the noise produced by the scanner was reduced to 75dB SPL by the means of sound protecting headphones. Depending on the sequences, the number of events varied from 8 to 12.

Grouping task: In this task, the sequences were irregular, with alternating Inter Onset Intervals (IOI) of 250, 700 and 1300 ms in order to prevent subjects from extracting a regular pulse or metric structure (ratio = 1 :2.8 :5.2). In half of the trials, one event was displaced from one group to the next. An example for one trial is shown in Figure 1.

Metric task: In this task, the sequences were composed of the alternation of 3 IOI sharing integer ratios 1:2:4 (250, 500 and 1000 ms). In half of the trials, one event was delayed or advanced by 70 ms, locally disrupting metric expectancies but still remained in the same group. An example for one trial is shown in Figure 2.

Baseline condition: Both rhythmic tasks alternated with the same control task. In this task, subjects heard a continuous sound with the same pitch and timbre, of a total duration around 10 seconds. We used such a sound in order to remove any activity involving non temporal auditory processing. We decided to use a continuous sound rather than regular filler patterns as other authors did (e.g. Sakai et al, 1999) in order to prevent the subject from performing any cognitive task involving processing of temporal information during the control condition.

Procedure

Before scanning, subject were presented with practice trials in order to check their understanding of the tasks. During scanning, series of several pairs of sequences were presented for both conditions. The subjects had to press the right button of a mouse if both sequences of one pair were identical and the left button if they were different. During the control condition, they pressed alternatively the right or left button as soon as the sound stopped (subjects were informed that no reaction time was measured). Specific instructions were given to the subjects: they were not allowed to inner-sing or tap the rhythm and had to wait until the end of the second sequence to give their answer. They had to focus on the regularity in the metric task and on the succession of events in the grouping task (we did not mention the presence of distinct groups).

Data Acquisition and Analyses

A Bruker (Karlsruhe, Germany) 2.0 T system equipped with a 30mT.m-1 gradient coil set for echo planar imaging (EPI) was used to perform all studies. Measures were averaged over four different repetitions for both conditions. The acquisition consists of 32 transaxial gradient-echo planar (GE-EPI) 64*64 brain isotropic (4 mm) slices, so-called volume, and repeated each 5500ms (repetition time), resulting in 120 scans in total. Before statistical analysis, pre-processing of the images had been performed, namely realignment, normalisation and smoothing according to the procedures proposed by Friston et al. and as implemented in the Statistical Parametric Mapping (SPM99b) software (Wellcome Department of Cognitive Neurology; Friston et al., 1995. Statistical analysis was performed: the single condition paradigm (two tasks) was modeled using delayed (5.5 s) boxcar hemodynamic model function in the context of the general linear model (Friston et al., 1995), resulting in a t statistic for each and every voxel. These t statistics were transformed to z statistics. Voxels that survived the statistical criteria of significance (p < 0.01 one tailed) corrected for multiple comparisons constitute a statistical parametric map (SPM). Anatomical identification of activated areas was performed individually by mapping areas onto the subject's own anatomical normalized (T1) images (T1 images on T1 template). Following individual anatomical identification of activated areas for each subject, the identified activated areas from multiple subjects were mapped onto the best fitted area of the normalized template T1 image in the Talairach (Talairach and Tournoux, 1988) reference coordinate system.

 

Results:

Several brain regions revealed significant activation during grouping and metric processing relative to the baseline (control non rhythmical task). Table 1 presents the areas significantly activated for both conditions. Figure 3 and 4 report the activation obtained in the grouping and the metric conditions (N=9) respectively, on a standardised brain. We will review in turn the cerebral regions activated and identify the brain regions active in only one of the conditions.

 

 

 

Figure 3: Statistical parametric analysis (SPM{z}) for 9 subjects in the grouping condition. The foci of signicantly increased activities (shown in red) were rendered on the surface template of a standard brain as implemented in SPM99b (Wellcome Department of Cognitive Neurology, London, UK). The dimmer the color, the deeper the activation.

 

Figure 4: Statistical parametric analysis (SPM{z}) for 9 subjects in the metric condition (see legend of Figure 3).

 

Temporal lobes

The temporal lobes (especially Brodmann areas -BA-s 22 and 40), were bilaterally activated in both tasks. This confirms the idea that musical rhythm is processed in associative auditory areas in the temporal lobes. However, there was a spread of activation in the anterior part of the superior and medial left temporal gyri (BA 22 and 38) in the metric task. The spread of activation in the temporal regions was more posterior (towards BA 40) in the grouping than in the metric condition. These results partly confirm the anterior/posterior dissociation observed in metric and "rhythmic" contrasting tasks observed recently by Liegeois-Chauvel (1998).

Frontal and Prefrontal lobes

A broad bilateral activation in the SMA (supplementary motor area) and some activation in the PMC (premotor cortex, BA 6) were obtained in both grouping and metric tasks. This activation was spread towards more anterior regions in the grouping task. The activation in motor-associated areas may seem surprising since any motor activity (in our case pressing a button to give the answer) should have been eliminated by the motor activity in the control condition. However, activation in these areas has already been reported during the processing of visual and auditory temporal patterns (Tracy et al., 1999; Sakai et al., 1999; Schubotz et al., 2000), but all these tasks were assessed by the reproduction (tapping) of the rhythmic stimuli. To our knowledge, this is the first report of the activation of motor-related areas in a purely perceptual rhythmic task. This is also a strong evidence against previous assumptions that motor areas would only be involved in the programming of the reproduction of rhythmic sequences.

The frontal opercular areas (including Broca's area and its equivalent in the right hemisphere) were activated bilaterally in both tasks. This pattern had already been observed in nonverbal rhythmic tasks (Schubotz et al., 2000). Since part of the motor system (SMA and PMC) was involved in our tasks, it was not possible to assess if the activity in this frontal area was related to inner singing, linked with the articulation of verbal sounds or if it was related to a broad nonverbal time processing system. However after single subject analyses revealed that activation in such regions did not perfectly overlap between the two tasks. A right/left asymmetry of frontal opercular activity was observed for each subject. However, there was no consistency between subjects regarding the preferential use of the right or left side depending on the task.

Middle frontal areas (including BA 45, 46 and 9) were also activated in most subjects, with a larger spread of activation on the right side in the grouping task. This could be related to the memorization of the sequences during the comparison since These brain regions have classically been proposed to mediate attentional processes and working memory when listeners perceive melodies (Zatorre et al., 1994). However, selective attention to the time intervals could be a better explanation since our inter-sequence delay was too short for the participants to completely rehearse the first sequence of each pair.

Cerebellum

We found cerebellar activation in both rhythmic tasks. This also supports the involvement of motor areas in the time-based activities (Tracy et al., 1999; Sakai et al., 1999; Schubotz et al., 2000). However the cerebellum is considered to play a role in perceptual timekeeping tasks. The bilateral cerebellar activity was more lateral in the grouping task and more medial in the metric task (including the vermis). We did not find any evidence of a posterior/anterior opposition as reported by Sakai et al. (1999) in the case of reproduction of metric and non metric sequences.

Other areas

The superior parietal gyrus (BA 7) was actived on the right side of the brain in both tasks, which had been observed in other rhythmic tasks (Platel et al., 1997; Sakai et al.; 1999). This associative multimodal cortex is supposed to take part in timing mechanisms during both perceptual and motor tasks (Sakai et al., 1999). This region has been proposed to be a component of a general time-keeping system (Maquet et al, 1996; Sakai et al, 1999), sometimes associated with the attentional binding of sequential events across time (Posner and Dehaene, 1994).

Finally, in 6 out of 9 participants, left occipital regions (BA 19) were activated in the grouping task only. This could be explained by the use of mental imagery as a strategy to visualise the rhythmic groups and the displacement of an event from one group to another. In fact, such a strategy was reported by most subjects during debriefing. The activation of the visual cortex had already been observed for pitch discrimination in musical sequences (Platel et al., 1997), but was not observed in any rhythmic task.

 

Conclusions:

Our results show that common areas are used by subjects in rhythmic tasks based on two distinct cognitive processes. The brain networks involved in these tasks comprised auditory and motor associated areas. The latter have already been shown to be part of a general time processing system. Our study is the first to report an activation of motor structures during a perceptual task using musical stimuli. Most brain imaging studies of rhythm used motor reproduction to assess subjects' performance. The common areas activated in the metric and grouping tasks included both right and left superior and middle temporal cortices, prefrontal and right superior parietal areas.

We found a large overlap of the areas activated in both tasks for our 9 subjects. Only little activation was specific to one of the two tasks. Even if a double dissociation between metric and nonmetric rhythmic processes has clearly been evidenced in neuropsychological studies (Mavlov, 1980; Peretz, 1990; Liégeois-Chauvel, 1998), very little research has been carried out on the interaction between these processes. Hence, there are still discrepancies about the exact nature of each of these processes and their independence from each other. Further research may show, as already suggested by Liégeois-Chauvel et al. (1998) that these two processes would be less independent than has been observed, especially since they occur in parallel. This could explain why so much overlap was found between metric and grouping task in our study. Since grouping and metric features in musical sequences depend mainly on their hierarchical structures, we envisage in our next brain imaging study to focus on the relation between basic and hierarchical levels of rhythmical sequences.

However, a few areas were activated in only one task. In the metric condition, we observed a large spread of activity in the anterior part of the left temporal lobe and in the left medial frontal gyrus (premotor cortex). In the grouping condition, a larger anterior activation was found in the right prefrontal cortex. Moreover, cerebellar activity showed distinct patterns between the two tasks. Thus, although our results confirm the existence of a general time processing system involved in both of our rhythmic conditions, there are specific regions related to only one of the two distinct types of cognitive processes. However, there was a high variability between subjects and no clear distinctive activation profiles were shared by all subjects. Hence, preliminary analyses after grouping subjects according to their musical expertise suggest that the cerebral networks used by expert musicians to process rhythm seems to be spread over more cerebral structures than those of nonmusicians. Further analyses will be carried out to confirm this view.

 

References:

Drake, C. (1998) Psychological processes involved in the temporal organisation of complex auditory sequences: universal and acquired processes. Music Perception, 16, 11-26.

Clarke, E. F. (1999) Rhythm and Timing in Music. In The Psychology of Music, 2nd Edition. D. Deutsch, Ed. New-york, Academic press, 473-500.

Lerdahl, F., & Jackendoff, R. A. (1983). A generative theory of tonal music. MIT Press. Cambridge, MA.

Liegeois-Chauvel C, Peretz I, Babai M, Laguitton V, Chauvel P (1998) Contribution of different cortical areas in the temporal lobes to music processing. Brain;121 (10):1853-67.

Maquet P, Lejeune H, Pouthas V, et al. (1996) Brain activation induced by estimation of duration : a PET study. NeuroImage 3: (2) 119-126.

Mavlov, L. (1980) Amusia due to rhythm agnosia in a musician with left hemisphere damage : a non-auditory supramodal defect. Cortex, 16, 331-338.

Parncutt, R. (1994) A perceptual model of pulse salience and metrical accent in musical rhythms Music Perception 11: (4) 409-464

Peretz, I. (1990) Processing of local and global musical information in unilateral brain-damaged patients. Brain, 113, 1185-1205.

Platel H., Price C., Baron J-C., Wise R., Lambert J., Frackowiak R.S.J., Lechevalier B., Eustache F., (1997) The structural components of music perception : A functional anatomic study. Brain, 120, 229-243.

Posner, M. I. and Dehaene, S. (1994) Attentional networks. Trends in Neuroscience, 17, 75-79.

Talairach, J. and Tournoux, P. (1988) Co-planar stereotaxic atlas of the human brain: 3-dimensional proportional system: an approach to cerebral imaging. Stuttgart: Thieme.

Tracy, JI, Faro, SH, Mohamed, FB, Pinsk, M and Pinus, A. (2000) Functional localization of a "time-keeper" function separated from attentional resources and task strategy. NeuroImage 11, 228-242.

Zatorre, RJ., Evans, CE. and Meyer, E. (1994) Neural mechanisms underlying cerebral melodic perception and memory for pitch. Journal of Neuroscience, 14, 1908-1919.

 

 Back to index