Finding the beat: a neural perspective across humans and non-human primates

Abstract

Humans possess an ability to perceive and synchronize movements to the beat in music ('beat perception and synchronization'), and recent neuroscientific data have offered new insights into this beat-finding capacity at multiple neural levels. Here, we review and compare behavioural and neural data on temporal and sequential processing during beat perception and entrainment tasks in macaques (including direct neural recording and local field potential (LFP)) and humans (including fMRI, EEG and MEG). These abilities rest upon a distributed set of circuits that include the motor cortico-basal-ganglia-thalamo-cortical (mCBGT) circuit, where the supplementary motor cortex (SMA) and the putamen are critical cortical and subcortical nodes, respectively. In addition, a cortical loop between motor and auditory areas, connected through delta and beta oscillatory activity, is deeply involved in these behaviours, with motor regions providing the predictive timing needed for the perception of, and entrainment to, musical rhythms. The neural discharge rate and the LFP oscillatory activity in the gamma- and beta-bands in the putamen and SMA of monkeys are tuned to the duration of intervals produced during a beat synchronization-continuation task (SCT). Hence, the tempo during beat synchronization is represented by different interval-tuned cells that are activated depending on the produced interval. In addition, cells in these areas are tuned to the serial-order elements of the SCT. Thus, the underpinnings of beat synchronization are intrinsically linked to the dynamics of cell populations tuned for duration and serial order throughout the mCBGT. We suggest that a cross-species comparison of behaviours and the neural circuits supporting them sets the stage for a new generation of neurally grounded computational models for beat perception and synchronization.

Keywords: beat perception; beat synchronization; modelling; neurophysiology.

Figure 1.

Brain areas commonly activated in functional magnetic resonance studies of rhythm perception (cut-out to show mid-sagittal and horizontal planes, tinted purple). Besides auditory cortex and the thalamus, many of the brain areas of the rhythm network are traditionally thought to be part of the motor system. SMA, supplementary motor area; PMC, premotor cortex. (Online version in colour.)

Figure 2.

Induced neuromagnetic responses to isochronous beat sequences at three different tempos. (a) Time-frequency plots of induced oscillatory activity in right auditory cortex between 10 and 40 Hz in response to a fast (390 ms onset-to-onset; upper plot), moderate (585 ms; middle plot) and slow (780 ms; lower plot) tempo (n = 12). (b) The time courses of oscillatory activity in the beta-band (20–22 Hz) for the three tempos, showing beta desynchronization immediately after stimulus onset (shown by red arrows), followed by a rebound with timing predictive of the onset of the next beat. The dashed horizontal lines indicate the 99% confidence limits for the group mean. (c) The time at which the beta desynchronization reaches half power (squares) and minimum power (triangles), and the time at which the rebound (resynchronization) reaches half power (circles). The timing of the desynchronization is similar across the three stimulus tempos, but the time of the rebound resynchronization depends on the stimulus tempo in a predictive manner. (d) Areas across the brain in which beta-activity is modulated by the auditory stimulus, showing involvement of both auditory cortices and a number of motor regions. (Adapted from Fujioka et al. [71].)

Figure 3.

Time-varying modulations in the beta power of an LFP signal that show selectivity to interval duration. (a) Normalized (z-scores, colour-code) spectrograms in the beta-band. Each plot corresponds to the target interval indicated on the left. The horizontal axis is the time during the performance of the task. The times at which the monkey tapped on the push-button are indicated by black vertical bars, and all the spectrograms are aligned to the last tap of the synchronization phase (grey vertical bar). Light-blue and grey horizontal bars at the top of the panel represent the synchronization and continuation phases, respectively. (b) Plots of the integrated power time-series for each target duration. Tap times are indicated by grey triangles below the time axis. Green dots correspond to power modulations above the 1 s.d. threshold (black solid line) for a minimum of 50 ms across trials. The vertical dotted line indicates the last tap of the synchronization phase. (c) Interval tuning in the integrated beta power. Dots are the mean ± s.e.m. and lines correspond to the fitted Gaussian functions. Tuning functions were calculated for each of the six elements of the task sequence (S1–S3 for synchronization and C1–C3 for continuation) and are colour-coded (see inset). (d) Distribution of preferred interval durations for all the recorded LFPs in the beta-band with significant tuning for the auditory condition. (Adapted from Bartolo et al. [98].)

Figure 4.

Multiple layers of neuronal representation for duration, serial order and context in which the synchronization–continuation task is performed. (a). LFP activity that switches from gamma-activity during sensory driven (bottom-up) beat synchronization, to beta-activity during the internally generated (top-down) continuation phase. (b) Dynamic representation of the sequential and temporal structure of the SCT. Small ensembles of interconnected cells are activated in a consecutive chain of neural events. (c) A duration/serial-order tuning curve for a cell that is part of the dynamically activated ensemble on top. Dotted horizontal line links the tuning curve of the cell with the neural chain of events in (b). (Adapted from Merchant et al. [,–110].)