Getting the beat: Entrainment of brain activity by musical rhythm and pleasantness

,withgreatermetereffectsduringdissonantthanconsonantmusic.Theseresultsrevealthat 30 thebasalganglia,involvedbothinemotionandrhythmprocessing,criticallycontributetorhythmicentrainmentof 31 subcortical brain circuits by music.

entrainment is a very common phenomenon: Who has not been caught with the foot tapping or the body moving to the music heard in the background?In the present study we directly ask the question how musical rhythm makes our brain act in synchrony with the music, and whether this effect depends on subjective pleasantness or not.To study the nature of rhythmic entrainment and its neural underpinnings, we engaged participants in a functional magnetic resonance imaging (fMRI) paradigm while they performed a visuomotor attentional task in which targets appeared either in or out of synchrony with the music, and manipulated musical pleasantness by using either consonant or dissonant music.
The term entrainment describes a physical principle "whereby two rhythmic processes interact with each other in such a way that they adjust towards and eventually 'lock in' to a common phase and/or periodicity" (Clayton et al., 2005, p. 5).The synchronization 53 of bodily rhythms with music entails entrainment phenomena at dif- In keeping with this view, it has been shown that temporal expectancies can engender cross-modal integrative effects on attentional resources (Lange and Roder, 2006).This implies that if attention is enhanced at a specific moment in time, stimulus processing can be facilitated for all sensory modalities, independently of the task-relevant modality (Teder-Salejarvi et al., 2002).Accordingly, behavioral findings suggest that entrainment induced by an auditory rhythm can influence visual attention (Escoffier et al., 2010) and that listening to classical music can entrain attentional resources in synchrony with the musical meter (Bolger et al., 2013;Tierney and Kraus, 2013).
In addition, it has been suggested that entrainment may constitute a key source of emotions experienced during music listening (Janata et al., 2012;Witek et al., 2014).According to a recent psychological framework proposed by Juslin and colleagues (Juslin et al., 2010), different bodily rhythms may synchronize to those present in the music, consequently generating emotional feelings via proprioceptive feedback mechanisms.However, this framework does not specify which synchronization level is particularly critical, or whether the same principle applies to different levels of the system, including not only bodily and physiological rhythms but also higher cognitive processes such as attention.Moreover, few studies have investigated the neural mechanisms linking musical rhythms with entrainment and emotion.Recent work using transcranial magnetic stimulation (TMS) reported that corticospinal excitability is increased during metrically strong rhythmical sequences (Cameron et al., 2012) or high-groove music (Stupacher et al., 2013).
However, an ideal candidate brain substrate for mediating such links might lie in the basal ganglia, as these structures are implicated in motor control (Jueptner and Weiller, 1998;Turner and Desmurget, 2010), rhythm processing (Grahn and Brett, 2007;Thaut et al., 2008), as well as pleasant emotional experiences (Salimpoor et al., 2011;Trost et al., 2012).The basal ganglia might therefore be well placed for integrating rhythmical information with both cognitive and affective components of musical experience.On the other hand, cross-modal influences on attention and its deployment over time are known to recruit cortical areas in posterior parietal lobule (Coull and Nobre, 1998;Macaluso and Driver, 2001), including for synchronization of motor responses with auditory (non-musical) sequences (Bolger et al., 2014).Therefore, parietal attention systems might also contribute to the effect musical rhythm has on attention and entrainment.
Here, we directly tested how musical meter engenders cross-modal entrainment of visuomotor processes, by obtaining both behavioral and fMRI measures in human volunteers.We also investigated whether entrainment would interact with the affective appreciation of the music, and thus be enhanced by its pleasantness.Based on previous research (Bolger et al., 2013), we expected that an attentional entrainment of visuomotor performance by concomitant music should make response times faster to visual targets appearing simultaneously with strong beats of the musical meter, as compared with targets appearing on weak beats.
In addition, we also tested the affective entrainment hypothesis, according to which there is a link between rhythmic entrainment processes and emotion induction via music (Juslin et al., 2010;Trost and Vuilleumier, 2013).Previous research already suggested that entrainment in terms of sensorimotor synchronization may enhance subjective experience of pleasantness even in non-musical conditions (Fairhurst et al., 2012;Janata et al., 2012).Furthermore, motor or attentional entrainment appears directly linked to musical pleasantness, as rhythmical patterns of a certain complexity range are rated as more pleasant and evoke stronger feelings of groove (Witek et al., 2014).Here, however, we aimed at testing the affective entrainment hypothesis in a reverse causal direction, by determining whether (and how) positive affect elicited by pleasant music would enhance the rhythmic entrainment of attentional processes.Specifically, we examined whether the pleasantness of music would produce a stronger entrainment of visuomotor performance, by comparing such effects during consonant (pleasant) and dissonant (unpleasant) music (Koelsch et al., 2006).On the one hand, due to greater enjoyment of the two consecutive visual targets) with a few long and many short intereach trial, participants were reminded by written instructions to listen attentively to the music, to fixate the fixation cross, and to press a button with the index finger as fast as possible when a circle appeared around the fixation cross.Immediately after the musical piece ended, six questions were presented (one after the other) on a different screen background and probed for the participants' subjective evaluation of the preceding piece.These questions were evaluations of the subjectively felt emotions (level of arousal and valence), the subjective impression of felt entrainment (formulated as the urge to move or dance to the music), and familiarity with the musical stimulus.The evaluations were designed as statements to which the participants could agree or disagree to different degree.The answers were indicated by using a sliding cursor that could be moved (by right or left key presses) on a horizontal scale from −3 to +3 (−3 = (I agree) not at all, +3 = (I agree) absolutely).
The order of questions was constant for all participants.Subjects were instructed to answer spontaneously, but there was no time limit for responses.The last response to the questionnaire automatically triggered the next musical stimulus presentation.Therefore, the overall scanning time of a session varied slightly between subjects (average 573 scans per run, standard deviation 27 scans).However, only the scans during the musical epochs were included in the analyses, which comprised the same amount of scans across subjects.
In the fMRI experiment, auditory stimuli were presented binaurally with an audio system and MRI compatible headphones (CONFON DAP-center mkII and CONFON HP-Pi-US, MR confon GmbH, Germany).
The loudness of the music was adjusted for each participant individually, prior to fMRI scanning.Visual instructions were seen on a screen backprojected on a headcoil-mounted mirror.Responses were recorded with a response button box (HH-1 × 4-CR, Current Designs Inc., USA).The behavioral study was conducted exactly in the same manner, using the same task and musical stimuli as in the fMRI experiment, but took place in a quiet, dimly lit room.

Data acquisition and analysis
For the analysis of behavioral performance, reaction times (RTs) were averaged for each of the experimental conditions, after excluding trials where the RT was more than twice the standard deviation away from the mean of each participant.Repeated-measure ANOVAs were performed on the reaction times with the two factors meter (strong versus weak beat) and consonance (consonant versus dissonant version).
The answers of the questionnaire were analyzed with two-sample MRI images were acquired using a 3 T whole body MRI scanner (Trio TIM, Siemens, Germany) with the product 12 channel head coil.A highresolution T1-weighted structural image (0.9 × 0.9 × 0.9 mm 3 ) was obtained using a magnetization-prepared rapid acquisition gradient echo sequence (time repetition [TR] = 1.9 s, time echo [TE] = 2.32 ms, time to inversion [TI] = 900 ms).Functional images were obtained using a continuous-sound echo planar imaging (EPI) sequence (Seifritz et al., 2006) with the following parameters: 36 slices, slice thickness 3. A standard statistical analysis was performed using the general linear model implemented in SPM8.Consonant and dissonant musical epochs were modeled by two separate boxcar regressors, in addition to four event regressors modeling the onsets of visual targets in the four experimental conditions.To account for movement-related variance, we entered realignment parameters into the same model as 6 additional covariates of no interest.For the event-related analyses, we computed (at the first-level) the parameter estimates corresponding to the event-related regressors for the four target onset conditions in a design matrix that also modeled the overall state differences associated with consonant and dissonant music epochs, allowing us to covary out these sustained changes from the modulation of phasic responses to targets.The parameter estimates for each target conditions were subsequently entered for the second-level group analysis (random-effects) using a factorial design ANOVA with the factors meter and consonance and 2 levels each.For all results, we report clusters with a voxel-wise threshold of p b 0.001 (uncorrected) and cluster-size N 3.

Behavioral results
A 2 × 2 repeated-measures ANOVA with the factors meter and consonance on reaction times of participants from the behavioral experiment revealed significant main effects for the two factors, meter (F(1,19) = 10.37,p b 0.005) and consonance (F(1,19) = 7.33, p b 0.014).RTs to visual targets were faster when presented on a strong (1st) relative to a weak (2nd) beat, and faster during consonant than dissonant music, a pattern compatible with an influence of both meter and pleasantness on visuomotor processing.The interaction between the two factors was also significant (p b 0.024).The same analysis on the reaction times from participants in the fMRI experiment also showed significant main effects for both meter (F(1,17) = 11.50, p b 0.004) and consonance (F(1,17) = 11.31,p = 0.015), with a similar facilitation pattern in RTs.The interaction between the two factors was not significant (F(1,17) = 1.65, p = 0.216).However, a similar ANOVA on data from all participants, combining the behavioral and fMRI experiments together, with the additional categorical variable group, did not only confirm the main effects of meter (F(1,36) = 21.08,p b 0.0001) and consonance (F(1,36) = 17.24, p b 0.0002), but also revealed a significant interaction between these two factors (1,36) = 7.07, p b 0.012; Figs.1B and C).There were no interactions of meter or consonance with the factor group, indicating that both groups showed a very similar pattern of reaction times despite the lack of significant meter x consonance in the behavioral experiment.These results therefore accord with our predictions, namely, that response facilitation by rhythmic entrainment should occur on strong beats (relative to the weak beat) regardless of consonance, and that consonance may however modulate the percep-342 tion of meter.Furthermore, as predicted, the strongest entrainment oc-343 curred for visual targets synchronized with a strong beat during 344 pleasant music, whereas the least entrainment occurred for visual tar-345 gets synchronized with a weak beat during unpleasant music (see 346 Figs.1B and C).

347
The analyses of answers to the questionnaire showed that consonant 348 pieces were evaluated as more pleasant, more arousing, more 349 entraining, more familiar, or more natural than the dissonant versions 350 (see Table 1).

Effect of consonant music 353
We first compared the general effect of consonant and dissonant  whereas the effect of meter in the precuneus seems to be primarily driv-400 en by consonant music.

402
The interaction between the two experimental factors was finally 403 verified by directly contrasting the strong versus weak beats in disso-404 nant music against the corresponding beat effect in consonant music.

405
Significant effects were found in bilateral caudate nuclei and right ante-406 rior insula (Table 3, Fig. 4).In other words, the caudate was especially 407 responsive to the difference between strong and weak beats in disso-       to the expectation that pleasantness should boost the entrainment of attention by meter, this activation pattern actually converges with our behavioral data to suggest that meter and consonance produced distinct influences on the synchronization of attentional processes to music.

The role of the basal ganglia in rhythmic entrainment
In keeping with the notion that the basal ganglia are involved in the coordination of motor actions and in the perception of rhythmic structures, our novel results point to the caudate nucleus as a key structure that encodes musical meter.In previous imaging studies of rhythm on meter processing, participants had to subsequently reproduce, compare or categorize short rhythmical sequences, requiring explicit attention and top-down internal generation of the rhythm (Grahn and Rowe, 2009;Iversen et al., 2009;Chapin et al., 2010).In contrast, in our paradigm, entrainment to the meter occurred unintentionally and without voluntary effort, in a stimulus-driven manner solely determined by task-irrelevant music played in the background.This is probably also the reason why we did not observe a differential involvement of other brain structures such as the cerebellum or premotor cortex, which are often reported in studies on explicit rhythm perception and production (Molinari et al., 2003;Grahn and Brett, 2007;Chen et al., 2008;Merchant et al., 2013).
The caudate is classically related not only to motor planning, but also to error prediction and reward (Bayer and Glimcher, 2005;Asaad and Eskandar, 2011), and thus constitutes the most "cognitive" portion of the basal ganglia (Grahn et al., 2008).The caudate has previously been reported to be involved in rhythm processing (Bengtsson and Ullen, 2006;Grahn and Brett, 2007) and seems especially engaged when a clear beat is perceived in rhythmical patterns (Chapin et al., 2010) or when sensorimotor synchronization to a beat is easy (Kokal et al., 2011).In neuroimaging studies on music, caudate activity was reported to be sensitive to emotional arousal (Trost et al., 2012), correlate with the anticipation of chills (Salimpoor et al., 2011), and even vary according to musical syntax (Koelsch et al., 2008).Based on these results, we hypothesized that pleasant music would modulate activity particularly in the ventral striatum.Because this portion of the basal ganglia is known to play an important role in reward processing and pleasure (Salimpoor et al., 2011;Trost et al., 2012), we expected stronger entrainment effects in ventral striatum during consonant pleasant music.
However, we did not find this pattern of response; instead we found that event-related activation to targets in the caudate head was most influenced by meter during dissonant music.Nonetheless, a sustained activation in the ventral part of the right caudate was significant in our contrast of consonant versus dissonant music epochs (Fig. 2 and Table 2).Taken together, this suggests that consonant music produced globally higher activation levels in the ventral caudate, over and above the event-related response associated with visual target detection.
This result accords with the fact that the consonant pieces were evaluated as more pleasant, in line with other findings that positive emotions recruit ventral striatal regions (Katsyri et al., 2012;Koelsch and Skouras, 2013).This sustained right caudate activity together with concomitant increases in motor and somatosensory cortical areas (Table 3) could reflect the subjective apprehension of more pleasant consonant music epochs as being more arousing and more entraining (Table 1).Indeed, caudate activity correlates with felt arousal induced by music as well as its valence (Trost et al., 2012) and rewarding value (Salimpoor et al., 2013), being typically more active during pleasant or joyful than during unpleasant or sad music (Koelsch and Skouras, 2013).Moreover, it has also been shown that caudate activity is associated with joint drumming in synchrony and subsequent prosocial behavior (Kokal et al., 2011).
Interestingly, a recent study (Bolger et al., 2014)  Here we used a manipulation of dissonance to modulate the level of 572 pleasantness, similar to other studies (Peretz et al., 2001;Koelsch et al., 2009).Moreover, precuneus activity has been proposed to be modulated by dopaminergic inputs from striatum (Lou et al., 2005).Interestingly, a study by Fairhurst et al. (2012) found that precuneus activation was also associated with high synchronicity, when participants performed a tapping task in synchrony with a virtual partner.Interpersonal synchrony in sensorimotor tasks is known to represent a pleasant state that increases the feelings of affiliation and prosocial behavior (Hove and Risen, 2009;Valdesolo and Desteno, 2011;Launay et al., 2013).Being in synchrony with a partner also means however that there are no conflicting or unexpected events that disturb performance.Pleasant consonant music might thus enhance the facilitation of attention to visual targets by increasing concentration on music and reducing interference by distracting events or thoughts.
In support of such an attention effect, we found that consonance produced a general speeding of RTs in the visual detection task (Fig. 2).Several studies have shown that visual attention can be influenced by the affective state of a person (Ashby et al., 1999;Olivers and Nieuwenhuis, 2006).There is even evidence that visual neglect after parietal lobule stroke is reduced when patients are listening to their preferred versus non-preferred music (Soto et al., 2009).The "broaden-and-built" theory formulated by Fredrickson (2001) proposes that positive emotions have a beneficial effect on various cognitive functions, including in particular a broadening of attentional resources.The global behavioral benefit of consonance in our study could be interpreted in this framework, as an effect of broadened attention induced by pleasant music could have a more global facilitating impact on visuomotor performance and attentional orienting.This could enhance target detection even when these appear in less attended moments in the music (i.e., weak beats).Likewise, research on visual attention has shown that positive affect primes can increase perceptual flexibility and allows switching more rapidly from a local to a global focus in detection tasks (e.g., Tan et al., 2009).In the auditory domain, Olivers and Nieuwenhuis (2005) reported that listening to natural music can abolish the attentional blink effect.In keeping with these data, our results for the pleasant music condition corroborate the notion that listening to music may particularly affect the temporal aspects of attentional processing.Our fMRI results for the effect of consonance on event-related response to visual targets (across meter conditions) accord with this interpretation as we found differential activations in a set of cortical areas encompassing superior parietal lobule, dorsal ACC, and dorsolateral PFC that partly overlapped with the attentional network (Behrmann et al., 2004).

The affective entrainment hypothesis
What support does our study bring to the DAT and the affective entrainment hypothesis (Juslin et al., 2010;Trost and Vuilleumier, 2013)?
On the one hand, we show evidence for time-locking of visuomotor performance with musical meter, associated with highly selective engagement of the bilateral caudateconsistent with rhythmic entrainment.On the other hand, our results reveal that the metrical hierarchy of the music does not become more salient or effective with consonant music, in terms of the facilitation of RTs between metrical conditions.Rather, we find that targets presented with weak beats are equally fast detected as targets presented with strong beats when music is more pleasant.According to DAT, all kinds of metrical music should entrain attentional processes, and therefore strong and weak beats should also produce different attentional levels during pleasant music.However, our results might not necessarily contradict this notion.Our study specifically aims at testing whether the emotional valence of music modulate the entraining effect of meter, a factor which was not explicitly predicted or addressed by DAT.In addition, the facilitation induced by pleasantness might produce ceiling effects in the present paradigm and prevent detection of entrainment in that condition.Further, for design simplicity, here we only compared weak and strong beats but it remains possible that differential entrainment by meter would be better evidenced during expectancies of the visual stimuli themselves, we 201 generated a distribution of inter-trial-intervals (ITI; i.e. time between 202

Fig. 1 .
Fig. 1.A) Experimental design.Participants have to detect visual targets which are presented time-locked to the meter of the music played in the background, either on the first or the second beat.B) Behavioral reaction time results for participants in the fMRI experiment (n = 18).C) Behavioral reaction time results for participants in the behavioral experiment (n = 20).* indicates significant post-hoc tests (Tukey) between the conditions.Error bars indicate standard error.

354
music epochs (t-test contrast), reflecting sustained modulation of 355 brain activity during the whole duration of musical pieces.To this aim, 356 we compared activations modeled by the boxcar regressors for conso-357 nant and for dissonant music pieces in the first-level analysis, in 358 which the transient changes due to target processing were covaried 359 out by separate event-related regressors.Consonant relative to disso-360 nant music produced higher activations not only in the right ventral 361 caudate nucleus, a region of basal ganglia at the interface of affective 362 and cognitive processes, but also in somatosensory and primary motor 363 cortices (Table 2, Fig. 2).The opposite contrast did not show any signif-364 icant voxels above threshold.However, our main analysis and predic-365 tions concerned event-related responses to visual targets appearing in 366 different music conditions, as detailed below.367 Effect of consonance on visual detection 368 Using an ANOVA for the event-related analyses of responses to visu-369 al targets, we first performed a whole-brain SPM contrast to identify any 370 differential activation evoked during consonant versus dissonant music 371 (regardless of synchronization with strong or weak beats).Significant 372 increases were observed in premotor cortex, superior parietal lobule, 373 and anterior cingulate cortex (see Table 2, Fig. S1).This suggests that 374 consonant music modulated the brain response to visual targets by en-375 hancing cortical networks associated with attention and motor prepara-376 tion.The opposite contrast comparing visual targets presented during 377 dissonant versus consonant music showed significant voxels in bilateral 378 superior occipital gyri (Table2), suggesting that visual perceptual pro-379 cesses were more solicited when music was dissonant.380 Main effect of meter on visual detection 381 The next, most crucial comparison concerned visual targets present-382 ed during strong versus weak beats in the music (regardless of conso-383 nance).This contrast revealed significant activations in bilateral 384 caudate nuclei and the right precuneus (Table 3, Fig. 3A), converging 385 with our predictions that parts of the basal ganglia should be critically 386 involved in rhythmic entrainment.The opposite contrasts of weak ver-387 sus strong beats did not reveal any significant clusters.When further t1:1

408
nant music.No region passed our statistical threshold in the inverse in-409 teraction testing for stronger meter effects in consonant relative to 410 dissonant music (precuneus: p b 0.05).Parameters estimates of activity 411 corresponding to these regions are illustrated for each experimental 412 condition in Fig. 4. 413 Discussion 414 We used a novel cross-modal paradigm to study the effect of en-415 trainment by musical rhythm on visuomotor performance and its mod-416 ulation by affective appreciation.Based on the DAT (Jones, 1987), which 417 proposes that attentional orienting may become synchronized to strong t2:1

Fig. 2 .
Fig. 2. Main effect of consonant music.Contrast between the blocks of consonant and dissonant music.Effects significant at p b 0.001 (uncorrected) are shown in yellow, and effects significant at p b 0.01 (uncorrected) are shown in red for illustrative reasons.Coordinates are according to the MNI space (in millimeter resolution).The left panel shows a coronal slice at the level of y = 12, whereas the right panel shows a sagittal slice at the level of x = −8.

Fig. 3 .
Fig. 3. Main effect of meter.Event-related analysis of the visual targets appearing simultaneously with a strong versus a weak beat of the music.

F
consonance manipulation did indeed influence attentional entrainment 481 processes, as distinct brain circuits were preferentially engaged by rhythmic entrainment in the presence of consonant versus dissonant music.Although the neural modulation may at first sight appear opposite

Table 1 t1 : 2
Behavioral evaluations of the consonant and dissonant versions of the musical pieces.
389rately, we found that targets presented with strong versus weak beats of meter (strong vs weak beat) did not reach statistical threshold for 393 this condition in the basal ganglia (right putamen: p = 0.009, right cau-394 date: p b 0.05).Conversely, in dissonant music, the contrast of strong 395 versus weak beats revealed significant and symmetric activations in bi-396 lateral caudate nuclei, plus left superior temporal sulcus and superior 397 temporal gyrus (Table3, Fig.3B).These results indicate that the effect 398 of meter in the caudate is predominating during dissonant music,