Abstract
Timing is everything, but our understanding of the neural mechanisms of timing remains limited, particularly for timing of sequences. Temporal sequences can be represented relative to a recurrent beat (beat-based or relative timing), or as a series of absolute durations (non-beat-based or absolute timing). Neuroimaging work suggests involvement of the basal ganglia, supplementary motor area (SMA), the premotor cortices, and the cerebellum in both beat- and non-beat-based timing. Here we examined how beat-based timing and non-beat-based sequence timing were affected by modulating excitability of the supplementary motor area, the right cerebellum, and the bilateral dorsal premotor cortices, using transcranial direct current stimulation (tDCS). Participants were subjected to a sham stimulation session, followed an active stimulation session where anodal or cathodal 2mA tDCS was applied to the SMA, right premotor cortex, left premotor cortex, or the cerebellum. During both sessions, participants discriminated changes in rhythms which differentially engage beat-based or non-beat-based timing. Rhythm discrimination performance was improved by increasing SMA excitability, and impaired by decreasing SMA excitability. This polarity-dependent effect on rhythm discrimination was absent for cerebellar or premotor cortex stimulation, suggesting a crucial role of the SMA and/or its functionally connected networks in rhythmic timing mechanisms.
A fascinating, possibly uniquely human behaviour is the capacity to perceive the beat in sequences of temporal intervals (e.g., in music or speech), even though beats are not necessarily indicated by distinguishing acoustic features. Beat perception, or the ability to sense a beat in rhythms, appears spontaneously in humans, without training (Winkler et al., 2009). Beat perception is thought to engage relative timing mechanisms, in which the temporal intervals of a pattern are coded relative to each other (Essens & Povel, 1985; Yee et al., 1994; Teki et al., 2011a; Teki et al., 2011b). This relative timing is often called ‘beat-based’ timing, because the intervals can be encoded relative to a regular, periodic beat interval. Beat-based timing improves accuracy during discrimination, synchronization, and reproduction of temporal sequences (Essens & Povel, 1985; Yee et al., 1994; Patel et al., 2005; Grahn & Brett, 2007; Chen et al., 2008c). Relative timing stands in contrast to ‘absolute’ timing, also termed duration-based or non-beat-based timing, in which the absolute durations of temporal intervals are encoded individually in a stop-watch like manner (Teki et al., 2011a).
An large body of functional neuroimaging studies have suggested involvement of the supplementary motor area (SMA), the basal ganglia, the premotor cortex, and the cerebellum in timing (e.g., Schubotz & von Cramon, 2001; Ullen et al., 2003; Lewis et al., 2004; Grahn & Brett, 2007; Chen et al., 2008b; Bengtsson et al., 2009; Teki et al., 2011b). Determining the specific role of each area in different timing processes remains an active area of investigation. Neuroimaging studies find that the SMA and basal ganglia respond more to beat-based than non-beat-based than rhythms (Grahn & Brett, 2007; Grahn & Rowe, 2009; Teki et al., 2011b; Geiser et al., 2012; Grahn & Rowe, 2013; Li et al., 2019). This pattern is consistent across tasks, including perceptual judgements (McAuley et al., 2012), discrimination (Grahn & Brett, 2007), or attending to non-temporal aspects of the stimuli such as loudness (Geiser, Notter, & Gabrieli, 2012) and pitch (Grahn & Rowe, 2009). Moreover, neuropsychological work in patients with Parkinson’s disease finds selective deficits in beat-based, but not non-beat-based timing (Grahn & Brett, 2009; Breska & Ivry, 2018). The premotor cortex and cerebellum appear to respond in both beat-based and non-beat-based contexts (Bengtsson et al., 2005; Chen et al., 2006), or respond more to non-beat-based than to beat-based contexts (Grahn & Rowe, 2009; Nozaradan et al., 2017; Teki et al., 2012). In a similar vein, patients with cerebellar degeneration show selective deficits in non-beat-based timing, despite showing intact beat-based timing (Grube et al., 2010a; Breska & Ivry, 2018). It has thus been suggested that a functional network involving the SMA and basal ganglia subserves beat-based timing, whereas a functional network involving the cerebellum subserves absolute timing (e.g., Teki et al., 2011b).
Importantly however, the theory of distinct neural processes subserving beat-based and non-beat based timing has mostly been supported by correlational neuroimaging evidence(Grahn & Brett, 2007; Grahn & Rowe, 2009; Teki et al., 2011b; Geiser et al., 2012; Grahn & Rowe, 2013; Li et al., 2019), or by neuropsychological work in patient populations (Grahn & Brett, 2009; Grube et al., 2010a; Breska & Ivry, 2018) who can have more global deficits that might not relate directly to the task. Relatively few studies employ perturbational methods in neurotypical humans, and such studies typically perturb only one or two brain areas implicated in beat perception (Malcolm et al., 2008; Grube et al., 2010b; Giovannelli et al., 2014; Ross et al., 2018b). Here, in neurotypical young adults, we examine how beat perception is affected by modulating multiple brain areas implicated in beat-based and non beat-based timing (i.e., supplementary motor area, left and right premotor cortex, and cerebellum), using transcranial direct current stimulation. Transcranial direct current stimulation (tDCS) is thought to modulate spontaneous neural firing and synaptic efficacy of neurons by altering resting membrane potential (e.g., Bindman et al., 1962; Lafon et al., 2017), and has been proposed to have functionally specific effects by modulating activity of task-relevant neuronal networks (Bikson & Rahman, 2013). Given the large individual differences in beat perception ability (Grahn & McAuley, 2009; Grahn & Schuit, 2012; Sowinski & Dalla Bella, 2013), as well as large individual differences in tDCS responsivity (Chew et al., 2015), we employed a within-subjects approach, where participants completed a placebo (sham) tDCS session followed by an active tDCS session whilst discriminating between rhythms which differentially engage beat-based timing. We hypothesize that a functional network involving the SMA plays a primary role in beat-based timing: thus, modulating excitability of the SMA was predicted to affect performance when processing beat-inducing rhythms.
Method
Participants
A total of 121 participants completed the experiment for course credit. Cerebellum anodal (n=14), cerebellum cathodal (n=16), left premotor anodal (n=15), left premotor cathodal (n=16), right premotor anodal (n=15) right premotor cathodal (n=15), SMA anodal (n=15), SMA cathodal (n=15). Participants were not pre-selected for music or dance training. Participants passed a safety screening for tDCS and gave informed consent. The Human Research Ethics Committee at Western University approved the study and all experiments were performed in accordance with relevant guidelines and regulations.
Transcranial Direct Current Stimulation
Transcranial direct current stimulation (tDCS) involves passing a weak current from one area of the brain to another area of the brain. Unlike transcranial electric and transcranial magnetic current stimulation, tDCS does not itself cause neural firing: it modulates spontaneous firing and synaptic efficacy of neurons by altering resting membrane potential. Anodal tDCS increases the likelihood of spontaneous neural firing, whereas cathodal tDCS decreases the likelihood of spontaneous neural firing (Purpura & McMurtry, 1965). tDCS thus modulates neuronal networks activated at the time Hence, despite its lack of spatial specificity, tDCS can be functionally specific (Bikson & Rahman, 2013).
Before behavioural testing, the scalp area overlying the stimulation site was located using the international electroencephalographic 10-20 system. This method of localization is sufficient for tDCS using large electrodes as used here (Fregni et al., 2006). Electrode montages are as follows. SMA: active electrode positioned 2 cm anterior to Cz, reference electrode positioned on contralateral orbit (Vollmann et al., 2013). Cerebellum: active electrode positioned 3 cm right of the inion, reference electrode positioned on the right buccinator muscle (Galea et al., 2009). PMC: active electrode positioned 2 cm anterior and 2 cm to the right of C3 for left PMC, and 2cm anterior and 2cm to the right of C4 for right PMC (Nitsche et al., 2003; Boros et al., 2008), as neuroimaging studies suggest that the dorsal premotor cortex is located about 15–25 mm anterior to the primary motor cortex (C3, C4)(Picard & Strick, 2001). The reference electrode was positioned on contralateral orbit for both right and left PMC. Electrodes were secured using Velcro straps. For the active tDCS conditions, the current was gradually ramped up to the 2 mA level over 30 s upon commencing the rhythm discrimination task. The stimulation remained on during the task for a maximum of 40 minutes and was ramped down at the end of the stimulation period. For the sham tDCS conditions, the stimulation was ramped off over 30 s immediately after reaching 2 mA. This method is sufficient to achieve blinding in stimulation-naïve participants (Ambrus et al., 2012). Stimulation was generated with a Dupel Stimulator (Dupel Ionophoresis System, MN) using two 4 x 6 cm rubber electrodes placed in saline-soaked sponges (current density of 0.04 mA/cm2; 0.9% NaCl) and highly conductive electrode gel (e.g., Signa Gel 40,000 micromhos/cm).
We employed a within-subjects approach to counter the large individual differences in beat perception ability (Grahn & McAuley, 2009; Grahn & Schuit, 2012; Sowinski & Dalla Bella, 2013), as well as large individual differences in tDCS responsivity (Chew et al., 2015). Each participant first completed a first sham tDCS condition in a first session, followed by active tDCS in the second session. In each session, during stimulation, participants completed the rhythm discrimination task (described below), with different rhythms used for each run to reduce practice effects. Participants completed both sessions on one visit, allowing us to eliminate participant drop-outs between visits. Time constraints made it unfeasible to have the active tDCS condition before the sham tDCS condition within a single-visit design, as effects can persist up to an hour post-stimulation (Nitsche & Paulus, 2000): waiting for tDCS effects to washout would require the visit to last more than 3 hours.
Auditory Stimuli
We used rhythms known to differentially induce beat perception as in previous work (Grahn & Brett, 2007). These rhythms were created using integer ratio and non-integer ratio related sets of intervals. The integer-ratio intervals were related by ratios of 1:2:3:4, and the non-integer ratio intervals were related by 1:1.4:3.5:4.5. The shortest interval (i.e., 1) ranged from 220 to 270 ms, in 10 ms steps. Sine tones (rise/fall times of 8 ms) sounded for the duration of each interval, ending 40 ms before the specified interval length to create a silent gap that demarcated the intervals. The sequences used filled intervals, as opposed to a brief presentation of the stimulus marking interval onset). The other intervals in the rhythm were multiples of the shortest interval. For metric simple rhythms, the intervals were arranged in groups of four units (e.g., in the sequence 211314, an interval onset consistently occurs every four units). The patterns were constructed to induce a perceptual accent at the beginning of each group of four units (Povel & Okkerman, 1981). The perceptual accents cued participants to a regular beat structure, in which the beats coincided with the onset of each group (Essens, 1995). In addition, if participants choose a faster rate for the beat, the sequence can still be measured by that rate (e.g., measured in units of two rather than four). The intervals in the metric complex rhythms were arranged to not be reliably grouped into two, three, or four units (e.g., 341211), and therefore difficult to measure by any unit but the shortest unit. Since there is no regular grouping, and no regular accent occurrence, no beat is induced.
Task Procedure
Rhythms were presented binaurally over Sennheiser headphones. Participants first completed four familiarization trials, where they were asked to judge if the rhythms presented were same or different. Participants completed a first block of 30 trials whilst undergoing sham stimulation with 30 trials, followed by a second block of 30 trials whilst undergoing verum stimulation. The words ‘First time’, ‘Second time’, and ‘Same or different?’ were displayed on the screen during the first, second, and third rhythm presentations, respectively (see Figure 2). Participants indicated whether the third rhythm presentation was the same as or different from the first two presentations by pressing one key for ‘same’ and another key for ‘different’ on a computer keyboard. Between trials, there was a 2s interval of silence. On half of the trials, the rhythm in the third presentation differed (or deviated) from the first two presentations, where some intervals in the sequence were transposed. For example, for the rhythm with the sequence 211413: the 3 interval and the 1 interval were transposed, resulting in a deviant sequence 211431. Following previous work (Grahn & Brett, 2007), we only used deviant sequences that were in the same category as the standard sequences. That is, a metric simple standard sequence could not have a metric complex deviant sequence, and a metric complex standard sequence could not have a metric simple deviant sequence. For example, 43122 cannot have 43212 as a possible deviant sequence, because the onsets would no longer be grouped in units of four and would violate the regular accent structure of the sequence.
Data Analysis
Task performance was quantified using sensitivity (d’ scores) and percent correct scores, consistent with previous work (Grahn & Brett, 2007).
Statistical analysis
Unlike p-values, Bayes factors do not tend to over-estimate the evidence against the null hypothesis (Gelman & Tuerlinckx, 2000; Wetzels et al., 2011). We thus chose to use Bayesian statistics to evaluate evidence for the alternative hypothesis and for the null hypothesis. Analyses were conducted in JASP (Version 0.13.1; JASP Team, 2020). The default Cauchy prior widths (0.707) values in JASP were used to quantify the relative evidence that the data came from the alternative versus a null model.
To evaluate stimulation-induced changes in discrimination performance across different stimulation sites and polarities, Rhythm (metric complex, metric simple) x Stimulation (sham, stimulation) x Polarity (anodal, cathodal) x Site (SMA, cerebellum, right PMC, left PMC) bayesian ANOVAs were run on d’ and percent correct. Analyses estimated the evidence for including each effect across matched models via estimating inclusion Bayes factor (BFincl) for each effect. Analyses also estimated the evidence for excluding each effect by estimating an exclusion Bayes factor (BFexcl) for each effect. Where applicable, simple effects analyses were used to follow-up interactions. Jeffreys’s evidence categories for interpretation, were taken as the standard for evaluation of the reported Bayes Factors, where the size of the Bayes factors were estimated as weak (1–3), substantial (3–10), or strong (>10) evidence for the hypotheses tested.
Results
Effect of tDCS on rhythm discrimination
Rhythm discrimination performance
Replicating previous results (Grahn & Brett, 2009; Grahn, 2012), d’ and percent correct scores were higher for metric simple than for metric complex rhythms, indicating better discrimination performance for metric simple rhythms than for metric complex rhythms [d’: metric complex: 1.16±0.06, metric simple: 1.76±0.08; percent correct: metric complex= 69.1±0.9%, metric simple= 76.6±1.1%; main effect of rhythm type BFincl = 6.267e+28]. Figure 3 and 4 respectively show d’ and percent correct scores and from the sham condition to the stimulation condition. Generally, patterns in percent correct scores appear similar to d’ scores, similar to previous work (Grahn & Brett, 2007; 2009).
Stimulation affected rhythm discrimination performance differently depending on stimulation site and stimulation polarity, as shown by Stimulation x Site x Polarity interactions [d’: BFincl = 3.019, percent correct: BFincl = 2.574]. Follow-up Rhythm (metric complex, metric simple) x Stimulation (Sham, Stimulation) x Polarity (Anodal, Cathodal) were run separately for each stimulation site.
For SMA, there were Stimulation (Sham, Anodal) x Polarity (Anodal, Cathodal) interactions [d’: BFincl = 298.500, percent correct: BFincl = 288.447]. Follow-up simple effects analyses (i.e., Rhythm x Stimulation Bayesian ANOVAs) run separately for the anodal and cathodal conditions showed that anodal SMA stimulation improved discrimination performance compared to sham, particularly for metric simple rhythms (main effect of stimulation for d’: BFincl = 4.060, percent correct: BFincl = 3.679 (see Figure 3 and 4), whereas cathodal SMA stimulation worsened discrimination performance compared to sham (main effect of stimulation for d’: BFincl = 16.732, percent correct: BF incl =17.243).
For cerebellum, there was some evidence suggesting that stimulation worsened discrimination performance, [d’: BFincl = 1.941, percent correct: BFincl = 9.102], which appeared polarity-independent (substantial evidence against the stimulation by polarity interaction [d’: BFexcl = 3.710, percent correct: BFexcl = 2.985].
For the left premotor cortex condition, stimulation appeared to have no effect, as there was substantial evidence against including the main effect of stimulation [d’: BFexcl = 5.175; percent correct: BFexcl = 4.466] and the stimulation x polarity interaction [d’: BFexcl = 3.614, percent correct: BFexcl = 3.913]. Similarly for the right premotor cortex condition, there was evidence to exclude the main effect of stimulation [d’: BFexcl = 2.315; percent correct: BFexcl = 7.347e −8] and the stimulation x polarity interaction [d’: BFexcl = 3.592, percent correct: BFexcl = 3.713].
Discussion
Here, we provide initial evidence suggesting that discrimination of metric simple and metric complex rhythms can be improved by increasing SMA excitability, and impaired by decreasing SMA excitability. This polarity-dependent effect on discrimination of metric simple rhythms was not evident for cerebellar or premotor cortex stimulation, suggesting a functional role of the SMA and/or its functionally connected networks in beat-based timing.
How might the SMA support beat-perception?
Here, we found that modulating SMA excitability altered performance on the rhythm discrimination task in a polarity-dependent fashion. These findings support the idea that the SMA and the basal ganglia are involved in maintaining an internal representation of beat intervals (Grahn & Brett, 2007; Grahn & Rowe, 2009), facilitating performance on the rhythm discrimination task. Our results are broadly consistent with findings of greater SMA-basal ganglia activation during the processing of metric simple rhythms compared to metric complex rhythms (Grahn & Brett, 2007; Grahn & Rowe, 2009). It has been suggested that the SMA networks help to form forward temporal predictions (Macar et al., 2004). The role of the SMA in beat-maintenance is consistent with evidence in synchronization-continuation tasks where participants are asked to synchronize movements to external stimuli and then continue synchronization upon withdrawal of the external stimuli: here, the SMA tends to be activated during the continuation phase, and not the synchronization phase (Rao et al., 1997; Lewis et al., 2004). Similarly, patients with SMA lesions also show a selective deficit in the continuation phase but not the synchronization phase of the synchronization-continuation task (Halsband et al., 1993). Recent proposals have suggested that the SMA is tuned to anticipate an upcoming beat interval, and sends signals to the dorsal striatum, helping the striatum generate internally generated representations of the beat cycle, which in turn activates new SMA neural subpopulations via the thalamus (Cannon & Patel, 2021). Future studies might test this hypothesis via recordings in SMA and basal ganglia, perhaps in patients with electrocorticography and deep brain stimulator implants.
Effect of premotor cortex tDCS on rhythm discrimination performance
The premotor cortex is thought to be engaged in planning, selection, and control motor programs based on external events (Picard & Strick, 2001). The finding that stimulation of the left or right premotor cortex did not interfere with rhythm discrimination performance is consistent with the idea that the premotor cortex plays a primary role in synchronization and control of motor programs in response to external events, rather than beat perception per se. Although some findings implicate a role of the premotor cortex in beat-based timing (Chen et al., 2006; Chen et al., 2008a; c; 2009), all of these studies involve the synchronization of rhythms using repetitive TMS. Synchronization requires participants to synchronize movements to the onset of each tone of a rhythm, or to each beat in the rhythm. This not only requires participants to encode and maintain the beat interval, but also to produce a synchronized motor response, evaluate the accuracy of that response after each tap, and correct the timing of inaccurate taps. Indeed, the greater difficulty of synchronizing to non-beat rhythms (and consequently greater demands on motor planning, evaluation, and error correction) might have resulted in the increased premotor cortex activation when synchronizing to non-beat-rhythms than beat rhythms (Chen et al., 2008c). The notion that the premotor cortex plays a general role in synchronization of movements to external stimuli is supported by an increasing body of evidence showing that non-invasive stimulation of the premotor cortex modulates synchronization performance, when synchronizing to isochronous cues (Doumas et al., 2005; Del Olmo et al., 2007; Malcolm et al., 2008; Pollok et al., 2008; Bijsterbosch et al., 2011; Ruspantini et al., 2011), when adjusting for changes in cue onsets (Bijsterbosch et al., 2011; Kornysheva & Schubotz, 2011a; Ruspantini et al., 2011), or when tapping to rhythms which differentially engage beat perception (Kornysheva & Schubotz, 2011b; Giovannelli et al., 2014). These findings show distinct effects of stimulation on aspects of synchronization (e.g., tempo-matching versus phase-matching), and different effects on dorsal versus ventral premotor cortex stimulation, or left versus right premotor cortex stimulation. A detailed discussion of these findings is outside the scope of this study. To the best of our knowledge, there are no papers published in peer-reviewed journals demonstrating effects of stimulating the premotor cortex on beat perception tasks that do not require motor synchronization to external stimuli in humans. One study currently published on a preprint server has examined how stimulation of premotor cortex affects capacity to perceive changes in tempo in music (Ross et al., 2018a). However, music contains many redundant cues that signal beat onsets which aid the perception of tempo changes: effects of premotor cortex stimulation in this study might thus not be directly related to beat perception in temporal intervals without redundant cues. To elucidate the role of the premotor cortex in perceiving and synchronizing to the beat, future studies should examine effects of modulating premotor cortex excitability on the processing of beat and non-beat-based rhythms using both perceptual tasks and motor synchronization within the same subjects.
Effect of cerebellar tDCS on rhythm discrimination performance
Cerebellar tDCS resulted in lower discrimination accuracy (i.e., percent correct scores) regardless of stimulation polarity, with both beat and non-beat rhythms, although this effect appeared more robust for percent accuracy than discrimination sensitivity (d’). The polarity-independent effect on discrimination accuracy is perhaps unsurprising given the increasing numbers of studies that show polarity-independent effects of anodal and cathodal cerebellar tDCS on working memory (Ferrucci et al., 2008; Van Wessel et al., 2016), motor control and learning (Shah et al., 2013; Verhage et al., 2017), motor memory retention (Taubert et al., 2016), and conditioned eyeblink responses (Beyer et al., 2017). Indeed, one recent meta-analysis found no evidence for a polarity-dependent effect of cerebellar tDCS (Oldrati & Schutter, 2018). Polarity-dependent effects of tDCS on cortex do not necessarily generalize to the cerebellum, as the organization of cerebellar neurons differs fundamentally from that of the cortex (van Dun et al., 2016; Woods et al., 2016).
A few interpretations are possible for how cerebellar tDCS impaired discrimination accuracy here. First, cerebellar tDCS might have induced impairments in working memory, similar to previous findings (Van Wessel et al., 2016)(Ferrucci et al., 2008). In the rhythm discrimination task used here, working memory is required to remember and compare the first rhythms with the test rhythms. Second, cerebellar tDCS might have impaired absolute timing processes, consistent with previous findings of worsened absolute timing (Grube et al., 2010b). Both interpretations may be true. Judicious experimental designs which explicitly manipulate working memory load and/or stimuli which differentially engage absolute timing might provide evidence for or against these two possible interpretations.
Limitations
An important limitation in the current work is the lack of counterbalancing of the order of sham and active tDCS: the sham tDCS session always preceded the active tDCS session. Thus, the results of tDCS shown here might be interpreted as affecting learning to discriminate rhythms rather than performance in discriminating rhythms. Importantly however, only the group receiving anodal SMA tDCS improved rhythm discrimination performance from the first sham tDCS session to the second active tDCS session. No improvement in discrimination performance was evident in any other group. A general learning effect could only be true if stimulation for all the other sites (left PMC, right PMC, cerebellum) modifies the learning effect. Furthermore, the cathodal SMA tDCS group also showed worse rhythm discrimination in their second active tDCS session than in their first sham tDCS session. Furthermore, previous studies which test the effect of SMA stimulation do not show effects of SMA tDCS on learning (Foerster et al., 2013). It thus seems more likely that SMA tDCS modulated capacity to discriminate rhythms, rather than learning to discriminate rhythms.
A second limitation is the difficulty in blinding tDCS administration. Blinding in tDCS can be challenging even using double-blind designs where one experimenter conducts electrode placement and initiates stimulation and a different experimenter administers the behavioural task (O’connell et al., 2012). Here we used a single-blind design, and did not assess participant or experimenter blinding. Our protocols of employing sham tDCS first followed by active tDCS could have unblinded participants (Turi et al., 2019). Participants however, likely remained blinded to stimulation polarity, and it seems unlikely that participant expectations could have resulted in polarity-dependent effect of stimulation the SMA. We also did not predict the effect of cerebellar tDCS impairing rhythm discrimination performance regardless of stimulation polarity: this effect seems unlikely to have resulted from participant or experimenter expectations. Finally, the selective polarity-dependent effect on SMA but not for right and nor left premotor cortex conditions suggests a genuine effect of SMA stimulation on rhythm discrimination performance. However, it is clear that tDCS effects can be variable between individuals (Chew et al., 2015) as well as within individuals (e.g., Chew et al., 2015; López-Alonso et al., 2015), with factors such as cortical morphology (e.g.,Filmer et al., 2020), neurochemical concentrations (e.g., Filmer et al., 2020), genetics, time of day, gender, age, hormone levels (for a review, see Ridding & Ziemann, 2010). This work presents initial evidence suggesting that SMA tDCS can modulate rhythm discrimination performance. Replication and extension of this work will be necessary to increase confidence in these results.
Summary
Neuroimaging and neuropsychological evidence implicate a role of a functional network encompassing the supplementary motor area and the putamen in beat perception. Here, we show that non-invasive stimulation of the supplementary motor area can have polarity-dependent effects on discrimination of auditory rhythm. Although the current evidence implicates a role of the supplementary motor area in beat perception, exactly how the SMA interacts with other brain areas during beat perception remains unclear. Exploring this question by combining perturbational methods such as brain stimulation with methods that afford both high temporal and spatial resolution (e.g., magnetoencephalography) could elucidate the mechanisms through which regions implicated in processing time interact and contribute to beat perception.