Facial Expressions and the Evolution of the Speech Rhythm

Both speech and nonhuman primate vocalizations are produced by the coordinated movements of the lungs, larynx (vocal folds), and the supralaryngeal vocal tract (Ghazanfar & Rendall, 2008; Fitch & Hauser, 1995). The vocal tract consists of the pharynx, mouth, and nasal cavity through which a column of air is produced. The shape of this column determines its resonance properties and thus in which frequency bands of the sound produced at the laryngeal source get emphasized or suppressed. During vocal production, the shape of the vocal tract can be changed by moving the various effectors of the face (including the lips, jaw, and tongue) into different positions. The different shapes, along with changes in vocal fold tension and respiratory power, are what give rise to different sounding vocalizations. To put it simply: Different vocalizations (including different speech sounds) are produced in part by making different facial expressions.

Vocal tract motion not only changes the acoustics of vocalizations by changing their resonance frequencies but also results in the predictable deformation of the face around the mouth and other parts of the face (Yehia, Kuratate, & Vatikiotis-Bateson, 2002; Yehia, Rubin, & Vatikiotis-Bateson, 1998; Hauser & Ybarra, 1994; Hauser, Evans, & Marler, 1993). Different macaque monkey (Macaca spp.) vocalizations are produced with unique lip configurations and mandibular positions, and the motion of such articulators influences the acoustics of the signal (Hauser & Ybarra, 1994; Hauser et al., 1993). For example, coo calls, like /u/ in speech, are produced with the lips protruded, whereas screams, like the /i/ in speech, are produced with the lips retracted (Figure 1). Facial motion cues used by humans for speech reading are present during primate vocal production as well. The fact that different vocalizations are produced through different facial expressions and are therefore inherently “multisensory” is typically ignored by theories regarding the evolution of speech/language that focus solely on laryngeal control by the neocortex (or the lack thereof; Arbib, 2005; Jarvis, 2004).

Naturally, any vertebrate organism (from fishes and frogs to birds and dogs) that produces vocalizations will have a simple, concomitant visible motion in the area of the mouth. However, in the primate lineage, both the number and diversity of muscles innervating the face (Burrows, Waller, & Parr, 2009; Huber, 1930a, 1930b) and the amount of neural control related to facial movement (Sherwood, 2005; Sherwood et al., 2005; Sherwood, Holloway, Erwin, & Hof, 2004; Sherwood, Holloway, Erwin, Schleicher, et al., 2004) increased over the course of evolution relative to other taxa. This increase in the number of muscles allowed the production of a greater diversity of facial and vocal expressions in primates (Andrew, 1962). The inextricable link between vocal output and facial expressions allows many nonhuman primates to recognize the correspondence between the visual and auditory components of vocal signals. Macaque monkeys (Macaca mulatta), capuchins (Cebus apella), and chimpanzees (Pan troglodytes) all recognize auditory–visual correspondences between their various vocalizations (Evans, Howell, & Westergaard, 2005; Izumi & Kojima, 2004; Parr, 2004; Ghazanfar & Logothetis, 2003). For example, macaque monkeys, without training or reward, match individual identity and expression types across modalities (Sliwa, Duhamel, Pascalis, & Wirth, 2011; Ghazanfar & Logothetis, 2003), segregate competing voices in noisy conditions using vision (Jordan, Brannon, Logothetis, & Ghazanfar, 2005), and use formant frequencies to estimate the body size of conspecifics (Ghazanfar et al., 2007). More recently, monkeys trained to detect vocalizations in noise demonstrated that seeing concomitant facial motion sped up their RTs in a manner identical to that of humans detecting speech sounds (Chandrasekaran, Lemus, Trubanova, Gondan, & Ghazanfar, 2011).

There are also some very important differences in how humans versus nonhuman primates produce their utterances (Ghazanfar & Rendall, 2008), and these differences further enhance human multisensory communication above and beyond what monkeys can do. One universal feature of speech—typically lacking in monkey vocalizations—is its bisensory rhythm. That is, when humans speak, both the acoustic output and the movements of the mouth are highly rhythmic and tightly correlated with each other (Chandrasekaran, Trubanova, Stillittano, Caplier, & Ghazanfar, 2009).

Across all languages studied to date, speech typically exhibits a 3- to 8-Hz rhythm that is, for the most part, related to the rate of syllable production (Chandrasekaran et al., 2009; Greenberg, Carvey, Hitchcock, & Chang, 2003; Crystal & House, 1982; Malecot, Johonson, & Kizziar, 1972; Figure 2A). Both mouth motion and the acoustic envelope of speech are rhythmic. This 3- to 8-Hz rhythm is critical to speech perception. Disrupting the acoustic component of this rhythm significantly reduces intelligibility (Elliot & Theunissen, 2009; Ghitza & Greenberg, 2009; Smith, Delgutte, & Oxenham, 2002; Saberi & Perrott, 1999; Shannon, Zeng, Kamath, Wygonski, & Ekelid, 1995; Drullman, Festen, & Plomp, 1994), as does disrupting the visual components arising from facial movements (Vitkovitch & Barber, 1996). Thus, the speech rhythm parses the signal into basic units from which information on a finer (faster) temporal scale can be extracted (Ghitza, 2011). Given the importance of this rhythm in speech and its underlying neurophysiology (Ghazanfar & Poeppel, in press; Giraud & Poeppel, 2012), understanding how speech evolved requires investigating the origins of its bisensory rhythmic structure.

Oddly enough, macaque monkey vocalizations have a similar acoustic rhythmicity but without the concomitant and temporally correlated rhythmic facial motion (Figure 2B). Macaque vocalizations are typically produced with a single ballistic facial motion. Modulation spectra analyses of the acoustic rhythmicity of macaque monkey vocalizations longer than 400 msec reveals that their rhythmicity is strikingly similar to that of the acoustic envelope for speech (Figure 3A). Both signals fall within the 3- to 8-Hz range (see also Cohen, Theunissen, Russ, & Gill, 2007, for shared low-frequency components of macaque monkey calls and speech). Moreover, examination of a single call category (Figure 3B, top) or an exemplar (Figure 3B, bottom) shows that this rhythmicity is not the result of averaging across call-type categories or averaging within a single call category. Thus, one key evolutionary question is: How did we evolve from a presumptive ancestral unisensory, acoustic-only vocal rhythm (Figure 2B) to the one that is audiovisual, with both mouth movements and acoustics sharing the same rhythmicity (Figure 2A)?

One theory posits that the rhythm of speech evolved through the modification of rhythmic facial movements in ancestral primates (MacNeilage, 1998, 2008). In extant primates, such facial movements are extremely common as visual communicative gestures. Lip smacking, for example, is an affiliative signal observed in many genera of primates (Redican, 1975; Hinde & Rowell, 1962; Van Hooff, 1962), including chimpanzees (Parr, Cohen, & de Waal, 2005). It is characterized by regular cycles of vertical jaw movement, often involving a parting of the lips but sometimes occurring with closed, puckered lips. Although lip smacking by both monkeys and chimpanzees is often produced during grooming interactions, monkeys also exchange lip-smacking bouts during face-to-face interactions (Ferrari, Paukner, Ionica, & Suomi, 2009; Van Hooff, 1962). Moreover, lip-smacking gestures are among the first facial expressions produced by infant monkeys (De Marco & Visalberghi, 2007; Ferrari et al., 2006) and are frequently used during mother–infant interactions (Ferrari et al., 2009). According to MacNeilage (1998, 2008), during the course of speech evolution, such nonvocal rhythmic facial expressions were coupled to vocalizations to produce the audiovisual components of babbling-like (i.e., consonant–vowel-like) speech expressions in the human lineage (Figure 2C).

Although direct tests of such evolutionary hypotheses are difficult, we can use the 3- to 8-Hz rhythmic signature of speech as a foundation to explore its veracity. There are now many lines of evidence that demonstrate that the production of lip smacking in macaque monkeys is similar to the orofacial rhythms produced during speech. First and foremost, lip smacking exhibits a speech-like rhythm in the 3- to 8-Hz frequency range (Ghazanfar, Chandrasekaran, & Morrill, 2010). This rhythmic frequency range is distinct from that chewing and teeth grinding (an anxiety-driven expression), although all three rhythmic orofacial motions use the same effectors. Yet, it still may be that the 3- to 8-Hz range is large enough that the correspondence between the speech rhythm and the lip-smacking rhythm is coincidental. Below, we provide evidence from development, x-ray cineradiography, EMG, and perception that suggests otherwise.

If the underlying mechanisms that produce the rhythm in monkey lip smacking and human speech are homologous, then their developmental trajectories should be similar (Gottlieb, 1992; Schneirla, 1949). Moreover, this common trajectory should be distinct from the developmental trajectory of other rhythmic mouth movements. In humans, the earliest form of rhythmic and voluntary vocal behavior occurs some time after 6 months of age, when vocal babbling abruptly emerges (Preuschoff, Quartz, & Bossaerts, 2008; Smith & Zelaznik, 2004; Locke, 1993). Babbling is characterized by the production of canonical syllables that have acoustic characteristics similar to adult speech. Their production involves rhythmic sequences of a mouth close–open alternation (Oller, 2000; Lindblom, Krull, & Stark, 1996; Davis & MacNeilage, 1995). This close–open alternation results in a consonant–vowel syllable representing the only syllable type present in all the world's languages (Bell & Hooper, 1978). However, babbling does not emerge with the same rhythmic structure as adult speech, but rather, there is a sequence of structural changes in the rhythm. There are two main aspects to these changes: frequency and variability. In adults, the speech rhythm is ∼5 Hz (Chandrasekaran et al., 2009; Dolata, Davis, & MacNeilage, 2008; Greenberg et al., 2003; Crystal & House, 1982; Malecot et al., 1972), whereas in infant babbling, the rhythm is considerably slower. Infants produce speech-like sounds at a slower rate of roughly 2.8–3.4 Hz (Dolata et al., 2008; Nathani, Oller, & Cobo-Lewis, 2003; Lynch, Oller, Steffens, & Buder, 1995; Levitt & Wang, 1991). In addition to differences in the rhythmic frequency between adults and infants, there are differences in their variability. Infants produce highly variable vocal rhythms (Dolata et al., 2008) that do not become fully adult-like until postpubescence (Smith & Zelaznik, 2004). Importantly, this developmental trajectory from babbling to speech is distinct from that of another cyclical mouth movement—chewing. The frequency of chewing movements in humans is highly stereotyped: It is slow in frequency and remains virtually unchanged from early infancy into adulthood (Green et al., 1997; Kiliaridis, Karlsson, & Kjellberge, 1991). Chewing movements are often used as reference movement in speech production studies because both movements use the very same effectors.

By measuring the rhythmic frequency and variability of lip smacking in macaque monkeys across neonatal, juvenile, and adult age groups, the hypothesis that lip smacking develops in the same way as the speech rhythm was tested (Morrill, Paukner, Ferrari, & Ghazanfar, 2012). There were at least three possible outcomes. First, given the differences in the size of the orofacial structures between macaques and humans (Ross et al., 2009), it is possible that lip-smacking and speech rhythms do not converge on the same ∼5-Hz rhythm. Second, because of the precocial neocortical development of macaque monkeys relative to humans (Malkova, Heuer, & Saunders, 2006; Gibson, 1991), the lip-smacking rhythm could remain stable from birth onwards and show no changes in frequency and/or variability (much like the chewing rhythm in humans; Thelen, 1981). Finally, lip-smacking dynamics may undergo the same developmental trajectory as the human speech rhythm: decreasing variability, with increasing frequency converging onto a ∼5-Hz rhythm.

The developmental trajectory of monkey lip smacking parallels speech development (Morrill et al., 2012; Locke, 2008). Measurements of the rhythmic frequency and variability of lip smacking across individuals in three different age groups (neonates, juveniles, and adults) revealed that young individuals produce slower, more variable mouth movements, and as they get older, these movements become faster and less variable (Morrill et al., 2012)—this is exactly as speech develops, from babbling to adult consonant–vowel production (Dolata et al., 2008). The developmental trajectory for lip smacking was distinct from that of chewing (Morrill et al., 2012). As in humans (Green et al., 1997; Kiliaridis et al., 1991), chewing had the same slow frequency and consistent low variability across age groups. Thus, these differences in developmental trajectories between lip smacking and chewing are identical to those reported in humans for speech and chewing (Steeve, 2010; Steeve, Moore, Green, Reilly, & McMurtrey, 2008; Moore & Ruark, 1996).

If the human speech and monkey lip smacking have a shared neural basis, one would expect commonalities in the coordination of the effectors involved. One piece of evidence for this comes from motor control. During speech, different sounds are produced through the coordination between key vocal tract anatomical structures: the jaw/lips, tongue, and hyoid. [The hyoid is a bony structure to which the laryngeal muscles attach.] These effectors are more loosely coupled during speech movements than during chewing movements (Matsuo & Palmer, 2010; Hiiemae & Palmer, 2003; Hiiemae et al., 2002; Ostry & Munhall, 1994; Moore, Smith, & Ringel, 1988). X-ray cineradiography (x-ray movies) used to visualize the internal dynamics of the macaque monkey vocal tract during lip smacking and chewing revealed that lips, tongue, and hyoid move during lip smacks (as in speech) and do so with a speech-like 3- to 8-Hz rhythm (Figure 4A and B). Relative to lip smacking, movements during chewing were significantly slower for each of these structures (Figure 4B). The temporal coordination of these structures was distinct for each behavior (Figure 4C). Partial directed coherence measures—an analysis that measures to what extent one time series can predict another (Takahashi, Baccala, & Sameshima, 2010)—revealed that, although the hyoid moves continuously during lip smacking, there is no coupling of the hyoid with lip and tongue movements; whereas during chewing, we observed more coordination between the three structures. These patterns are consistent with what is observed during human speech and chewing (Hiiemae & Palmer, 2003; Hiiemae et al., 2002): The effectors are more loosely coupled during lip smacking than during chewing. Furthermore, the spatial displacement of the lips, tongue, and hyoid is greater during chewing than during lip smacking (Ghazanfar, Takahashi, Mathur, & Fitch, 2012), again similar to what is observed in human speech versus chewing (Hiiemae et al., 2002).

Facial EMG studies of muscle coordination during lip smacking and chewing also revealed very distinct activity patterns associated with each behavior (Shepherd, Lanzilotto, & Ghazanfar, 2012). The coordination of monkeys' orofacial musculature during lip smacking and chewing was measured using EMG electrodes targeting five muscles—three in the lower face (zygomaticus, orbicularis oris, and mentalis) and two in the upper face (frontalis and auricularis; Figure 5A and B). Muscle coordination was evident in both lip-smacking and chewing behavior, but the coordination of the perioral muscles was stronger and more stereotyped during lip smacking than during chewing. Whereas lip smacking is characterized by coherent movements of perioral mimetic muscles, chewing exhibits inconsistent perioral coordination in these muscles despite strong coordination of signal at the auricular and zygomatic sites (Figure 5C). These data suggest that lip smacking (like speech) has a distinct motor program. It is not simply a ritualization of feeding behavior although it may have evolved from such a behavior following a reorganization in the motor program. Reorganization of central pattern generators underlying rhythmic behaviors is not uncommon in nervous system evolution (Newcomb, Sakurai, Lillvis, Gunaratne, & Katz, 2012).

In speech, disrupting the auditory or visual component of the 3- to 8-Hz rhythm significantly reduces intelligibility (Elliot & Theunissen, 2009; Smith et al., 2002; Saberi & Perrott, 1999; Vitkovitch & Barber, 1996; Shannon et al., 1995; Drullman et al., 1994). To test whether monkeys were differentially sensitive to lip smacking produced with a rhythmic frequency in the species-typical range (mean = 4–6 Hz; Ghazanfar et al., 2010, 2012; Morrill et al., 2012), a preferential-looking procedure was used (Ghazanfar, Morrill, & Kayser, 2013). Computer-generated monkey avatars were used to produce stimuli varying in lip-smacking frequency within (6 Hz) and outside (3 and 10 Hz) the species-typical range but with otherwise identical features (Chandrasekaran et al., 2011; Steckenfinger & Ghazanfar, 2009; Figure 6A). The use of avatar faces allowed control of additional factors that could potentially influence attention, such as head and eye movements and lighting conditions for face and background. Each of two unique avatar faces was generated to produce the three different lip-smacking rhythms.

Measuring looking times to one or the other avatar assessed preferential looking. There were at least five possible outcomes. First, monkeys could show no preference at all, suggesting that they either did not find the avatars salient, that they failed to discriminate the different frequencies, or that they preferred one of the avatar identities (as opposed to the lip-smacking rhythm) over the others. Second, they could show a preference for slower lip-smacking rhythms (3 Hz > 6 Hz > 10 Hz). Third, they could prefer faster rhythms (3 Hz < 6 Hz < 10 Hz; Lewkowicz, 1985). Fourth, they could show avoidance of the 6-Hz lip smacking, preferring the unnatural 3- and 10-Hz rhythms to the natural lip-smacking rhythm. This may arise if monkeys find the naturalistic 6-Hz lip smacking disturbing (perhaps uncanny; Steckenfinger & Ghazanfar, 2009) or too arousing (Zangenehpour, Ghazanfar, Lewkowicz, & Zatorre, 2009). Finally, monkeys could show a preference for the 6-Hz lip smacking over the 3 and 10 Hz, perhaps because such a rhythm is concordant with the rhythmic activity patterns in the neocortex (Giraud & Poeppel, 2012; Karmel, Lester, McCarvill, Brown, & Hofmann, 1977). Monkeys showed an overall preference for the natural 6-Hz rhythm when compared with the perturbed rhythms (Figure 6B). This lends behavioral support for the hypothesis that their perceptual processes are similarly tuned to the natural frequencies of communication signals as they are for the speech rhythm in humans.

These multisensory, developmental, bio-mechanical, and perceptual homologies between human speech and monkey lip smacking suggest that their underlying neural mechanisms for perception and production of communication signals may also be homologous. On the sensory-perception side, neurophysiological work in the inferior temporal lobe demonstrated that neurons in that area are “face sensitive” (Gross, Rocha-Miranda, & Bender, 1972; Gross, Bender, & Rocha-Miranda, 1969) and multisensory (Bruce, Desimone, & Gross, 1981), and it is areas within this region, along with the auditory cortex, that are activated by audiovisual speech in humans (Arnal, Morillon, Kell, & Giraud, 2009; von Kriegstein et al., 2008; van Wassenhove, Grant, & Poeppel, 2005; Callan et al., 2003; Calvert et al., 1999). Similarly, in macaque monkeys, neurons in the STS, pFC, and auditory cortex are driven and modulated by species-specific audiovisual communication signals (Ghazanfar et al., 2010; Chandrasekaran & Ghazanfar, 2009; Ghazanfar, Chandrasekaran, & Logothetis, 2008; Sugihara, Diltz, Averbeck, & Romanski, 2006; Barraclough, Xiao, Baker, Oram, & Perrett, 2005; Ghazanfar, Maier, Hoffman, & Logothetis, 2005). Neurons in the STS are also very sensitive to dynamic faces, including lip-smacking gestures (Ghazanfar et al., 2010). Much of this comparative multisensory work has been reviewed elsewhere (Ghazanfar & Chandrasekaran, 2012) as has the relationship between the speech rhythm, speech perception, and auditory cortical oscillations (Giraud & Poeppel, 2012).

Much less is known about the neural mechanisms underlying the production of rhythmic communication signals in human and nonhuman primates. The mandibular movements shared by chewing, lip smacking, vocalizations, and speech all require the coordination of muscles controlling the jaw, face, tongue, and respiration. Their foundational rhythms are likely produced by homologous central pattern generators in the pons and medulla of the brainstem (Lund & Kolta, 2006). These circuits are present in all mammals, are operational early in life, and are modulated by feedback from peripheral sensory receptors. Beyond peripheral sensory feedback, the neocortex is an additional source influencing how differences (e.g., frequency and variability) between orofacial movements may arise (Lund & Kolta, 2006; MacNeilage, 1998). Whereas chewing movements may be largely independent of cortical control (Lund & Kolta, 2006), lip smacking and speech production are both modulated by the neocortex, in accord with social context and communication goals (Caruana, Jezzini, Sbriscia-Fioretti, Rizzolatti, & Gallese, 2011; Bohland & Guenther, 2006). Thus, one hypothesis for the developmental changes in the frequency and variability of lip smacking and speech is that they are a reflection of the development of neocortical circuits influencing brainstem central pattern generators.

One important neocortical node likely to be involved in this circuit is the insula. The human insula is involved in multiple processes related to communication, including feelings of empathy (Keysers & Gazzola, 2006) and learning in highly dynamic social environments (Preuschoff et al., 2008). Importantly, the human insula is also involved in speech production (Bohland & Guenther, 2006; Ackermann & Riecker, 2004; Catrin Blank, Scott, Murphy, Warburton, & Wise, 2002; Dronkers, 1996). Consistent with an evolutionary link between lip smacking and speech, the insula also plays a role in generating monkey lip smacking (Caruana et al., 2011). Electrical stimulation of the insula elicits lip smacking in monkeys, but only when those monkeys are making eye contact (i.e., are face-to-face) with another individual. This demonstrates that the insula is a social sensory-motor node for lip-smacking production. Thus, it is conceivable that, for both monkey lip smacking and human speech, the increase in rhythmic frequency and decrease in variability are, in part at least, because of the socially guided development of the insula. Another possible cortical node in this network is the premotor cortex in which neurons respond to seeing and producing lip-smacking expressions (Ferrari, Gallese, Rizzolatti, & Fogassi, 2003).

A neural mechanism is needed to link lip-smacking-like facial expressions to concomitant vocal output (the laryngeal source). This separate origin of laryngeal control remains a mystery. A plausible scenario is that the cortical control of the brainstem's nucleus ambiguus, which innervates the laryngeal muscles, is absent in all primates save humans (Deacon, 1997).

Two core features of speech production—its rhythmical structure and temporal coordination of vocal tract effectors—are shared with lip smacking. Yet, there are striking differences between the two modes of expression, the most obvious of which is that lip smacking lacks a vocal (laryngeal) component. Although a quiet consonant-like bilabial plosive or /p/ sound is produced when the lips smack together, no sound is generated by the larynx. Thus, the capacity to produce vocalizations during rhythmic vocal tract movements seen in speech seems to be a human adaptation. How can lip smacking be related to speech if there is no vocal component? In human and nonhuman primates, the basic mechanisms of voice production are broadly similar and consist of two distinct components: the laryngeal source and the vocal tract filter (Ghazanfar & Rendall, 2008; Fitch & Hauser, 1995; Fant, 1970). Voice production involves (1) a sound generated by air pushed by the lungs through the vibrating vocal folds within the larynx (the source) and (2) the modification through linear filtering of this sound by the vocal tract airways above the larynx (the filter). The filter consists of the nasal and oral cavities whose shapes can be changed by movements of the jaw, tongue, hyoid, and lips. These two basic components of the vocal apparatus behave and interact in complex ways to generate a wide range of sounds. The lip-smacking hypothesis for the evolution of rhythmic speech and the data that support it only address the evolution of vocal tract movements (the filter component) involved in speech production.

Other differences between lip smacking and speech are the coupling of the lips with the tongue and the range of hyoid movements (Ghazanfar et al., 2012). The coupling of the lips and tongue during lip smacking (Figure 4C) is unlikely to occur during human speech where the independence of these effectors allows for the production of a wide range of sounds (although this has not been tested explicitly). With regard to the range of hyoid movements, the hyoid occupies the same active space during lip smacking and chewing, whereas cineradiography studies of human speech versus chewing show a dichotomy in hyoid movement patterns (Hiiemae et al., 2002). These movement range differences of the hyoid in humans versus macaques could be because of functional differences in suprahyoid muscle length, the degree of neural control over this muscle group, and/or by species differences in hyoid position. During human development, the position of the hyoid relative to the mandible and tongue changes (Lieberman, McCarthy, Hiiemae, & Palmer, 2001). This change allows for an increase in the range of tongue movements and, possibly, hyoid movements, relative to what is observed in nonhuman primates. Movements of either or both effectors could influence the active space of the hyoid thereby increasing the range of possible vocal tract shapes.

How easy would it be to link vocalizations to a rhythmic facial expression during the course of evolution? Recent work on gelada baboons (Theropithecus gelada) proves to be illuminating. Geladas are a highly specialized type of baboon. Their social structure and habitat are unique among baboons and other Old World primates as are a few of their vocalizations (Gustison, le Roux, & Bergman, 2012). One of those unique vocalizations, known as a “wobble,” is produced only by males of this species and during close, affiliative interactions with females. Wobbles are essentially lip-smacking expressions produced concurrently with vocalization (Bergman, 2013). Moreover, their rhythmicity falls within the range of the speech rhythm and lip smacking by macaque monkeys. Given that gelada baboons are very closely related to yellow baboons (their taxa are separated by 4 million years) who do not produce anything like wobble vocalizations, it appears that linking rhythmic facial expressions like lip smacking to vocal output is quite plausible. How geladas achieved this feat at the level of neural circuits is unknown, but finding out could reveal what was critical for the human transition to rhythmic audiovisual vocal output during the course of our evolution.

Human speech is not uniquely multisensory—visible facial motion is inextricably linked to acoustics. The default mode of communication in many primates is also multisensory. Apes and monkeys recognize the correspondence between vocalizations and the facial postures associated with them. One striking dissimilarity between monkey vocalizations and human speech is that the latter has a unique bisensory rhythmic structure in that both the acoustic output and the movements of the mouth are rhythmic and tightly correlated. According to one hypothesis, this bimodal speech rhythm evolved through the rhythmic facial expressions of ancestral primates. Developmental, cineradiographic, EMG, and perceptual data from macaque monkeys all support the notion that a rhythmic facial expression common among many primate species—lip smacking—may have been one such ancestral expression. Further explorations of this hypothesis must include a broader comparative sample, especially investigations of the temporal dynamics of facial and vocal expressions in the great apes. Understanding the neural basis of both lip smacking and speech production—their similarities and differences—would also be illuminating.

A. A. G. dedicates this review to Charlie Gross. Charlie has long been a rewarding influence on me. First, as an undergraduate, when his work on the history of neuroscience captured my imagination, and later, when his pioneering studies on face-sensitive and multisensory neurons in the temporal lobe helped drive the direction of my postdoctoral research. He remains an ongoing influence on me as my close colleague, friend, and unofficial mentor at Princeton University. We thank Adrian Bartlett for the analyses shown in Figure 3 and Jeremy Borjon, Diego Cordero, and Lauren Kelly for their comments on an early version of this manuscript. This work was supported by NIH R01NS054898 (A. A. G.), the James S. McDonnell Scholar Award (A. A. G.), a Pew Latin American Fellowship (D. Y. T.), and a Brazilian Science without Borders Fellowship (D. Y. T.).

Reprint requests should be sent to Asif A. Ghazanfar, Neuroscience Institute, Department of Psychology, Princeton University, Princeton, NJ 08540, or via e-mail: asifg@princeton.edu.