The interest in rhythm cognition in non-human animals is motivated by the search for signs of musicality as a means to get an insight in the evolutionary and causal processes underlying human musicality (Trehub, 2003; Honing and Ploeger, 2012; Hoeschele et al., 2015; Honing et al., 2015; Trehub et al., 2015; Honing, 2018b)¹. Most animals show at least some sort of rhythmic behavior, like walking, flying, crawling, or swimming. It is hence not unnatural to think that the perception (and enjoyment) of rhythm might be shared by most animals, as was argued by Darwin (1871) and Patel (2014). While only recently experimental research is finding some support for this claim (Wilson and Cook, 2016), there are also aspects of rhythm cognition that appear to be species-specific (Fitch, 2013), such as the capability to perceive a regular pulse in a varying rhythm (i.e., one level of a metrical structure) and consequently being able to synchronize to it (i.e., rhythmic entrainment), referred to as beat-based timing (Merchant and Honing, 2014).

Beat-based timing in humans is a complex neurocognitive phenomenon that depends on a dynamic interaction between auditory and motor systems in the brain (Grahn and Brett, 2009; Morillon et al., 2014; Patel and Iversen, 2014; Hoeschele et al., 2015; Merchant and Yarrow, 2016; Ross et al., 2017). It is hypothesized to be facilitated by bidirectional, and potentially causal links between the auditory and motor areas in the brain, including the motor cortico-basal-ganglia-thalamo-cortical (mCBGT) circuit, that appear to be more developed in humans as opposed to non-human primates and related species (Patel et al., 2009; Mendoza and Merchant, 2014; Patel and Iversen, 2014; Petkov and Jarvis, 2014; Merchant et al., 2015a; Wilson and Cook, 2016).

These observations lead to the gradual audiomotor evolution (GAE) hypothesis (Merchant and Honing, 2014) that suggests beat-based timing to be gradually developed in primates, peaking in humans but present only with limited properties in other non-human primates, while humans share interval-based timing with all non-human primates and related species. Thus, the GAE hypothesis accommodates the fact that the performance of rhesus monkeys is comparable to humans in single interval tasks—such as categorization, interval reproduction, and interception—, but differs in multiple interval tasks, such as synchronization, continuation, and rhythmic entrainment (Honing and Merchant, 2014; Merchant and Honing, 2014).

In the current paper, we will focus on beat and isochrony perception as two key components of musicality (Merchant et al., 2015b; Honing, 2018a), and provide further evidence for the GAE hypothesis by studying rhythm perception in two rhesus monkeys (Macaca mulatta). For this study we used an existing ERP paradigm that allows for testing and disentangling the contributions of beat perception and isochrony to the temporal predictability of the stimulus (Bouwer et al., 2016).

Most existing animal studies on beat-based-timing and rhythmic entrainment (Wilson and Cook, 2016) have used behavioral methods to probe the presence of beat perception, such as tapping tasks (Zarco et al., 2009; Hasegawa et al., 2011; Hattori et al., 2015) or measuring head bobs (Patel et al., 2009; Schachner et al., 2009; Cook et al., 2013). However, if the production of synchronized movement to sound or music is not observed in certain species, this is no evidence for the absence of beat perception. It could well be that while certain species are not able to synchronize their movements to a regular beat, they may be capable of beat perception (i.e., the ability to perceive a regular pulse in a temporally and/or acoustically varying rhythm; Honing, 2012). With behavioral methods that rely on overt motoric responses it is difficult to separate between the contribution of perception and action. More direct, electrophysiological measures such as auditory event-related brain potentials (ERPs) allow to test for neural correlates of rhythm cognition, including beat perception (Honing et al., 2014).

While the vast majority of previous studies on animals have used implanted electrodes to record electroencephalograms (EEG) (Javitt et al., 1994; Laughlin et al., 1999; Pincze et al., 2001), non-invasive electrophysiological techniques such as scalp recorded evoked potentials (EP) and event-related potentials (ERP) are considered an attractive alternative. Next to being a mandatory requirement for studying some non-human primates such as chimpanzees (Fukushima et al., 2010; Hirata et al., 2013), these methods allow for a direct comparison between human and non-human primates. As such they have contributed to establishing animal models of the human brain and human brain disorders (Godlove et al., 2011; Gil-da-Costa et al., 2013), a better understanding of the neural mechanisms underlying the generation of human evoked EP/ERP components (Fishman and Steinschneider, 2012), as well as delineating cross-species commonalities and differences in brain functions, including rhythm cognition (Ueno et al., 2008, 2010; Fukushima et al., 2010; Reinhart et al., 2012; Hirata et al., 2013; Itoh et al., 2015). We will describe the most relevant ERP components for rhythm perception below.

The mismatch negativity (MMN) is an auditory event-related component that was shown to be sensitive to rhythmic violations in both humans and monkeys (see Honing et al., 2014 for a review). The MMN can be used as an index of a violation of temporal expectation using an oddball paradigm, by identifying a negative peak shortly after the deviant (the “oddball”) that is maximal at fronto-central midline electrode sites and has sources in the auditory cortices and in the inferior frontal gyrus (i.e., not primarily reflecting motor cortex activity; Gil-da-Costa et al., 2013). The larger the violation of rhythmic expectations, the larger is the amplitude of the MMN (Näätänen et al., 2007; Winkler, 2007). The MMN has been shown to be indicative of beat perception in humans, with deviants on the beat within a repeating metrical auditory pattern eliciting a larger MMN than deviants off the beat (Ladinig et al., 2009, 2011; Winkler et al., 2009; Bouwer et al., 2014, 2016; Honing et al., 2014; Bouwer and Honing, 2015; Mathias et al., 2016).

The P3a, thought to reflect the redirection of attention to a deviant stimulus (Polich, 2007) and possibly an index of the conscious perception of a deviant (cf. Mathias et al., 2016; Peretz, 2016), often emerges just after the MMN, and has a latency of 200–250 ms in humans and between 100 and 250 ms in rhesus monkeys (Picton et al., 1974; Polich, 2007). Gil-da-Costa et al. (2013) provided functional evidence that the neural generators of both MMN and P3a may be homologous in humans and monkeys, despite the observed differences in latency (see Table 1 for an overview).

In addition to the ERP responses that reflect the detection of a deviant stimulus, the P1 and N1 responses, two early auditory event-related components, have been shown to be sensitive to the timing of the stimulus presentation (Costa-Faidella et al., 2011; Schwartze et al., 2013; Pereira et al., 2014; Teichert, 2016). In general, both P1 and N1 have an inverse relationship of ERP amplitude with temporal predictability (Javitt et al., 2000; Costa-Faidella et al., 2011; Schwartze et al., 2013). As such they are not indicative of beat perception per se, but of the temporal predictability of the stimulus. For instance, Schwartze et al. (2013) showed the P1 and N1 to be smaller in an isochronous as opposed to a jittered rhythmic sequence. Generally, increasing the predictability of the auditory stimulation (both in stimulus timing and stimulus probability) leads to a pronounced N1 attenuation (Costa-Faidella et al., 2011). Both components are maximal over fronto-central electrodes and there is some consensus on their homolog in rhesus monkeys, most notably on the N1 (Teichert, 2016) (see Table 1 for an overview).

Recently, Honing et al. (2012) were able to show, for the first time, that an MMN-like response can be measured in rhesus monkeys (M. mulatta). [See (Ueno et al., 2008) using a similar method in a chimpanzee (Pan troglodytes), and (Gil-da-Costa et al., 2013) for a recent study comparing humans and macaques (Macaca fascicularis)].

In addition, Honing et al. (2012) showed a sensitivity of the MMN in response to pitch deviants and infrequent omissions, showing that it was possible, in principle, to use an identical paradigm for human and non-human primates to probe beat perception. However, and contrary to what was found in human adults and infants (Winkler et al., 2009; Bouwer et al., 2014), no difference was found in the MMN in response to omissions in beat and offbeat positions. This lead to the conclusion that rhesus monkeys are unable to sense the beat (Honing et al., 2012). In addition, a strong response was found for onsets of rhythmic groups suggesting a sensitivity to rhythmic structure (A similar result was reported in Selezneva et al., 2013 showing large responses to changes in a repeating temporal pattern, while measuring gaze and facial expressions in monkeys).

However, Bouwer et al. (2014) pointed out that the earlier paradigm (Winkler et al., 2009; Honing et al., 2012) needs additional controls to be certain that any effects (or the lack thereof) are due to beat perception, and not, for instance, a result of pattern matching, acoustic variability or sequential learning. While rhesus monkeys have apparently little or no ability to perceive a beat, they are able to detect the regularity of an isochronous visual or auditory metronome (Zarco et al., 2009; Merchant et al., 2015b; Gámez et al., 2018). This suggests a capacity for making temporal predictions which most likely depends on absolute interval perception (Merchant and Honing, 2014). As such monkeys might not have beat perception, but could still be able to sense the regularity in an isochronous stimulus.

To examine the perception of isochrony, several studies have compared the responses to temporally regular, isochronous sequences with the responses to temporally irregular, jittered sequences (Schwartze et al., 2011; Teki et al., 2011; Fujioka et al., 2012). The prediction of events in jittered sequences has been suggested to rely on absolute interval perception, while the prediction of events in isochronous sequences has been suggested to be based on beat perception (Schwartze et al., 2011; Fujioka et al., 2012). However, it is possible to predict events in isochronous sequences on the basis of absolute interval perception alone, which may explain why macaques, with little or no ability to perceive a beat (Honing et al., 2012; Merchant and Honing, 2014), respond more accurately to temporally regular than jittered sequences (Zarco et al., 2009), based on their isochrony, rather than on beat-based perception (Merchant et al., 2015a; Merchant and Bartolo, 2017).

Based on these and related neurobiological observations (e.g., Zarco et al., 2009; Merchant et al., 2011) the GAE hypothesis was proposed (Merchant and Honing, 2014), arguing that the integration of sensorimotor information throughout the mCBGT circuit and other brain areas during the perception or execution of single intervals is similar in human and non-human primates, but different in the processing of multiple intervals. While the mCBGT circuit was shown to be also involved in beat-based mechanisms in imaging studies (e.g., Teki et al., 2011), direct projections from the medial premotor cortex (MPC) to the primary auditory cortex (A1) via the inferior parietal lobe (IPL) that is involved in sensory and cognitive functions such as attention and spatial sense (see Figure 1), may be the underpinning of beat-based timing as found in humans, and possibly apes. The GAE hypothesis suggests beat-based timing to be more developed in humans as opposed to apes and monkeys, and that it evolved through a gradual chain of anatomical and functional changes to the interval-based mechanism to generate an additional beat-based mechanism, instead of claiming a categorical jump from single-interval to multiple-interval abilities (i.e., rhythmic entrainment; Patel, 2006; Patel and Iversen, 2014). As such, the GAE hypothesis suggests that beat perception and entrainment have emerged gradually in primate order. This observation is in line with Rauschecker and Scott (2009) earlier suggestion that “the privileged access of the humans' auditory system to the sequential and temporal machinery of the mCBGT circuit emerged gradually in the course of evolution from precursors of the great ape lineage.” Some recent behavioral studies support such a gradual interpretation (Hattori et al., 2015; Large and Gray, 2015) suggesting at least some beat-based timing capabilities in apes that are absent in rhesus monkeys. Finally, the GAE hypothesis is in line with Patel and Iversen (2014), that argue for a causal link between auditory and motor planning regions needed for human beat perception. However, the GAE hypothesis differs from the latter proposal in that it (a) does not claim the neural circuit that is engaged in beat-based timing to be deeply linked to vocal learning, perception, and production, even if some explicit overlap between these neural circuits exists, and (b) that it gradually evolved in primates, instead of being solely present in humans as the only primate capable of vocal learning (Honing and Merchant, 2014).

However, in the current study we will not make claims about the underlying neural mechanisms, nor will we present a systematic comparative study (this is a topic of ongoing research). Instead, we will focus on whether rhesus monkeys are able to sense isochrony and/or the beat in a rhythmic stimulus.

In the current study we adapted an auditory oddball paradigm that was previously used in humans (Bouwer et al., 2016) and that allows for testing and dissociating the contributions of beat perception and isochrony to the temporal predictability of the stimulus.

We presented two rhesus monkeys (M. mulatta) a rhythmic sequence that was made up of a pattern of loud and soft percussive sounds such that the acoustic stimulus could induce a simple binary metrical structure (duple meter), with an accented beat on every other metrical position (see combinations of S1 and S2 in Figure 2). This rhythmic sequence was presented in two conditions: an isochronous condition, in which the sounds were presented in an isochronous fashion, using an inter-onset interval (IOI) of 225 ms, allowing a beat to be induced (i.e., one metrical level of a duple meter). And a jittered condition, in which the IOIs were randomly selected between 150 and 300 ms, as such disabling the perception of a beat. Furthermore, we used intensity decrements to be able to compare the ERP response to deviants in both the isochronous and jittered conditions (since omissions, as used in Honing et al., 2012, wouldn't be recognized as deviants in the jittered condition). By introducing unexpected intensity decrements (i.e., deviants) on both on the beat and offbeat positions, in both the isochronous and jittered conditions, we could probe the effect of metrical position (beat vs. offbeat), as well as the effect of isochrony (isochronous vs. jittered) on the amplitude of the MMN and the P3a. Additionally, we examined the effects of metrical position and isochrony on P1 and N1 responses to standard sounds.

This design allows for testing several hypotheses. First of all, we expected an MMN and P3a for all deviants in both the isochronous and the jittered conditions, irrespective of isochrony or metrical position. This to make sure that the auditory system of monkeys is sensitive to unexpected amplitude decrements (deviants) in a rhythmic stream. For this we predicted an effect of Type (standard vs. deviant).

Second, we did not expect to find evidence for beat perception, in line with earlier findings (Honing et al., 2012; Merchant and Honing, 2014). As such, we predicted no interaction between Position (Beat vs. Offbeat) and Isochrony (Isochronous vs. Jittered). To show beat perception, the difference between the MMN responses to deviants on the beat and offbeat in the isochronous condition should be more pronounced than the difference between the MMN responses to deviants on beat and offbeat positions in the jittered condition, in which beat perception is disabled. Without such an interaction between metrical position and isochrony, the differences between the MMN responses to different metrical positions should be interpreted as a result of sequential learning instead of beat perception (see Bouwer et al., 2016).

Third, we did expect the MMN and the P3a to be affected by the isochrony of the stimulus, with both having a higher amplitude in the isochronous condition (isochronous rhythm) as compared to the jittered condition (random rhythm). When monkeys are sensitive to the temporal regularity of the isochronous stimulus (Merchant et al., 2015b; Gámez et al., 2018) this should help in predicting the next event (i.e., increasing its temporal predictability) and enhancing its processing (Schwartze et al., 2011; Bouwer et al., 2014). Hence, it can be expected that the amplitude of the MMN and P3a in response to deviants in the isochronous condition is larger than in the jittered condition. As such, we predicted an interaction between Isochrony (Isochronous vs. Jittered) and Type (Standard vs. Deviant). In addition, we expected an inverse relationship between the amplitude of the P1 and N1 and temporal predictability. When the amplitude of the N1 and P1 is attenuated by the isochrony of the stimulus this can be used as additional evidence for isochrony perception (as was shown for humans in Schwartze et al., 2013).

In the current study we examined the role of beat perception and isochrony perception in two rhesus monkeys using the same stimuli in an oddball paradigm that was previously used in humans (Bouwer et al., 2016). Similar to humans, we found MMNs to all deviants as opposed to standards in both the isochronous and jittered conditions. From this we conclude that the monkey brain is sensitive to unexpected amplitude decrements (deviants). This is in line with earlier studies that used either omissions (Honing et al., 2012) or intensity deviants (Gil-da-Costa et al., 2013).

However, and contrary to what was found in humans (Bouwer et al., 2016), there are no significant differences in the MMN between deviants in beat positions as opposed to deviants in an offbeat position, neither in the isochronous condition nor the jittered condition (i.e., no effect of metrical position, nor an interaction with regularity; see Figure 8). So while in humans the beat appears to modulate the amplitude of the MMN, in monkeys there was not such an effect. This suggests the absence of beat perception in monkeys, in line with an earlier study (Honing et al., 2012).

With regard to the P3a, Monkey A did show an interaction between Type and Isochrony, but this was small and in the opposite direction as compared to humans (see Figure 8). This makes the latter result difficult to interpret. The effect might be a result of overlapping components (e.g., an interference with the strong response to the next onset; marked with dashed lines in Figures 5, 6), as such affecting the overall amplitude of the ERP signal. It could also be a result of attentional fluctuations that are modulating the P3a, as was shown in human studies (cf. Bouwer et al., 2016). Future work should manipulate attention to make sure whether and how attention influences rhythm perception.

By contrast, both monkeys appear to be sensitive to the isochrony of the stimulus, as is supported by an overall larger MMN response to deviants in the isochronous as opposed to the jittered condition (see Figure 8). Note however, that we cannot make the claim that monkeys detect isochrony (hence the use of the term “sense”). It could well be that the monkey brain is able to sense isochrony, but that the monkey is not aware of this, as was recently shown in two beat-deaf humans (Mathias et al., 2016). The latter study suggested the P3a to be associated with conscious perception or “awareness,” comparable to what is observed in tone deaf humans (Peretz, 2016).

With regard to the N1 and P1 the results are mixed and not consistent between Monkey A and Monkey B. While the attenuation of the N1 of Monkey A is in support for isochrony perception, the P1 is in opposite directions for the two monkeys. This difference could again be caused by attention, which was not controlled for in this study. As such we have to be cautious in our interpretation of the P1 and N1 in the standards, and will base our conclusions mainly on the MMN in response to the deviants (see summary in Table 5 and main results in Figure 8).

In short, the ERPs of both monkeys appear to be influenced by isochrony, but not by the induced beat. Where humans show a clear interaction between metric position and isochrony (Bouwer et al., 2016), and as such show evidence for beat perception, the current results provide no evidence for beat perception in rhesus monkeys. Hence, the hypothesis that beat-based timing is common to all animals (Darwin, 1871; Wilson and Cook, 2016) is not supported by these results.

While the underlying neural mechanism (be it interval-based timing or otherwise; Teki et al., 2011, 2016) is as yet unclear, we take this result as further support for the GAE hypothesis (Merchant and Honing, 2014). The GAE hypothesis suggests beat-based timing to be a result of bidirectional bottom-up and top-down interactions between the auditory and motor areas of the brain, including the mCBGT circuit and parietal areas such as the IPL (see Figure 1), connections that are quite developed and efficient in humans and that emerged gradually in the course of evolution from precursors of the great ape lineage (Rauschecker and Scott, 2009; Honing and Merchant, 2014; Merchant and Honing, 2014; Morillon et al., 2014; Merchant and Yarrow, 2016).

Parts of the audiomotor circuit, including areas such as the putamen in the basal ganglia or the parietal cortex, also process information from the dorsal stream of visual processing (Kimura, 1992; Merchant et al., 2004). Therefore, it could well be that the areas involved in the strong visuomotor coupling in monkeys partially overlap with the beat-entrainment audiomotor system, in line with the predictive and entrainment abilities of monkeys with visual stimuli (Takeya et al., 2017; Gámez et al., 2018). Thus, where humans show a preference for auditory metronomes, monkeys have a clear preference for visual stimuli (Gámez et al., 2018). Applying the current paradigm to the visual modality (e.g., with different intensity flashes rather than different intensity tones) might be able to show beat-entrainment for visual stimuli.

Gámez et al. (2018) also provides evidence that monkeys can predictively entrain to an isochronous metronome, even when it accelerates or decelerates. These findings suggest that the beat-based mechanisms of macaques might not be as restricted as previously thought (Patel, 2014). Thus, it is crucial that future experiments focus on finding the limits in beat perception and entrainment capabilities of monkeys, using gradually more complex levels of metrical periodicity in their stimuli (Schwartze and Kotz, 2015; Bouwer et al., 2016).

Finally, and as predicted by the GAE hypothesis, we expect beat-based timing in the auditory modality to be present in some rudimentary form in apes (not monkeys). Two recent behavioral studies support this interpretation (Hattori et al., 2015; Large and Gray, 2015) for a chimpanzee (P. troglodytes) and a bonobo (Pan paniscus). Applying the same oddball paradigm to a chimpanzee (cf. Ueno et al., 2008) might find further support for the GAE hypothesis (Merchant and Honing, 2014). In addition, the paradigm could be extended to other species, for example, pigeons, rodents, cats and dogs (e.g., Nelken and Ulanovsky, 2007; Howell et al., 2012; Schall et al., 2015; Harms et al., 2016), where the MMN can be measured to probe the potentially shared capabilities of isochrony perception and/or beat perception.

While musicality is likely made up of many components, it appears to be good strategy to start with a focus on core aspects, like beat perception (cf. Honing, 2018b). The core aspects of musicality are well suited for comparative studies, both cross-cultural and cross-species, and the nature and extent of their presence in non-human animals have attracted considerable debate in the recent literature. These recent discussions, combined with the availability of suitable experimental techniques for tracking these phenomena in human and non-human animals, make this a timely and feasible enterprise. Of course, we need to remain cautious about making claims on music-specific modes of processing until more general accounts have been ruled out. It still has to be demonstrated that the constituent components of musicality, when identified, are indeed domain specific. In contrast, the argument that music is a human invention (Patel, 2018) depends on the demonstration that the components of musicality are not domain specific, but each cognitively linked to some non-musical mental ability. So while there might be quite some evidence that components of musicality overlap with non-musical cognitive features (Patel, 2018), this is in itself no evidence against musicality as an evolved biological trait or set of traits. As in language, musicality could have evolved from existing elements that are brought together in unique ways, and that system may still have emerged as a biological product through evolutionary processes, such as natural or sexual selection. As such there is no need for musicality to be modular or show a modular structure. Alternatively, based on the converging evidence for music-specific responses along specific neural pathways, it could be that brain networks that support musicality are partly recycled for language, thus predicting more overlap than segregation of cognitive functions. In fact, this is one possible route to test the Darwin-inspired conjecture that musicality precedes music and language (Honing, 2018a).

HH and FB conceived and designed the experiments. HM and LP performed the experiments. FB and HH analyzed the data. HH, FB, and HM wrote the paper.