Anthropological Science
Online ISSN : 1348-8570
Print ISSN : 0918-7960
ISSN-L : 0918-7960
Sexual selection and the loss of laryngeal air sacs during the evolution of speech
ジャーナル フリー HTML

2018 年 126 巻 1 号 p. 29-34


One noteworthy, but unexplained, aspect of the evolution of human speech is the loss of laryngeal air sacs during hominin evolution. Very little is known about the adaptive significance of this curious trait, or the selection pressures that may have driven the evolution of air sacs among primates, and their later loss in Homo. Here, I review the literature on the loss of laryngeal air sacs during the evolution of speech, and argue that sexual selection may have been a key factor. Although air sacs do not fossilize, the presence or absence of air sacs appears to be correlated with the anatomy of the hyoid bone, and fossil hyoid evidence suggests that air sacs were lost in hominins between 3.3 million and 530000 years ago. Air sacs are hypothesized to have an acoustic function, and some authors have argued that hominins may have lost their air sacs because they would make speech less clear. In other primates, such as gorillas and howler monkeys, air sacs appear to play a role in acoustic size exaggeration and may be linked to reproductive competition. I explore the hypothesis that changes in social organization and mating system towards reduced male–male competition may have relaxed the selection pressure maintaining loud, low-frequency calls in hominins, making air sacs obsolete. While much of the above will remain hypothetical until more concrete data are gathered, we can speculate by saying that air sacs may not have been necessary for the type of quiet vocal interaction that typifies human communication. Perhaps more recent Homo species, with lower levels of sexual dimorphism and increased social tolerance and complexity, began to communicate in a more complex way, eventually leading to spoken language.


Speech is one of the most distinguishing characteristics of our species. Yet, our understanding of the evolution of this quintessentially human trait is far from complete. The dominant approach for the last few decades has been the search for fossil evidence. However, most anatomical traits associated with speech are soft tissues, which do not fossilize, making it difficult to establish when speech first appeared in our evolutionary history (Fitch, 2000a, 2010; Nishimura, 2008; Hauser et al., 2014). This line of enquiry has been very limited in success and has diverted attention away from alternative questions that are equally interesting and more accessible empirically. In particular, it is not only important to ask when speech may have evolved, but it is equally important to ask how and why (Fitch, 2000a; Hauser et al., 2002; Nishimura, 2008). Such questions may be addressed by applying the comparative model, which uses data from extant species to shed light on the evolution and adaptive function of traits (Darwin, 1859; Hauser et al., 2002; Fitch, 2010).

One significant change in vocal anatomy during hominin evolution was the loss of laryngeal air sacs (Fitch, 2000a; de Boer, 2007, 2012; Perlman and Salmi, 2017). All of the great apes, and many other non-human primates, have large, inflatable, soft-walled air sacs that connect to the larynx (Starck and Schneider, 1960; Gautier, 1971; Fitch and Hauser, 1995; Hewitt et al., 2002). However, humans lack this feature (with the exception of the occasional pathological appearance, called laryngoceles (Michaels, 1984)). Thus, comparative evidence suggests that air sacs were an ancestral condition, lost during hominin evolution. Several authors have argued that understanding this change in vocal anatomy may provide important clues as to the drivers and timing of the evolution of speech (Fitch and Hauser, 1998; Fitch, 2000a; Hewitt et al., 2002; de Boer, 2007, 2008, 2012; Nishimura et al., 2007; Riede et al., 2008; Perlman and Salmi, 2017).

Here, I review the literature surrounding the loss of laryngeal air sacs, and argue that sexual selection may have been a key factor in their evolution in some primate species, and their loss in humans. Such conclusions have important implications for our understanding of how changes in social organization and mating system during human evolution may relate to the evolution of speech.

The loss of laryngeal air sacs during hominin evolution: evidence from changes in hyoid morphology

Considerable variation exists in the anatomy of air sacs among mammals, as well as in the location at which the air sacs connect to the vocal tract (Negus, 1949; Harrison, 1995; Fitch and Hauser, 1998; Nishimura, 2003; Frey et al., 2007; Fitch, 2010). It is therefore likely that air sacs evolved independently in multiple lineages. Among primates, at least five types of air sacs are known to exist: lateral ventricular, lateral superior, subhyoidal, infraglottal, and dorsal (Starck and Schneider, 1960; Hayama, 1970; Hewitt et al., 2002; Riede et al., 2008). However, ape air sacs are all lateral ventricular (Hayama, 1970; Hewitt et al., 2002), providing support that this was the ancestral condition for hominins.

Although soft tissues do not fossilize, making it impossible to study air sacs in extinct hominins directly, the hyoid bone may provide evidence of the timing of the loss of air sacs in the hominin lineage. The hyoid is the only bone in the vocal tract of humans and non-human primates. In humans, it is a horseshoe-shaped bone situated in the anterior midline of the neck (Standring, 2015), which plays an important role in tongue movement during speech production (Hiiemae et al., 2002; Matsuo and Palmer, 2010). The hyoid is the only bone in the body that does not connect with any other in anthropoids, including humans, and it is rarely discovered during excavations, making it one of the least represented skeletal elements in the fossil record. However, evidence suggests that there was a shift in hyoid morphology during hominin evolution and, particularly, a reduction in the size of the hyoid bulla, the thin-walled cup-shaped shell that forms the anterior part of the hyoid, and into which the air sacs extend (Figure 1). With the exception of orangutans (which seemingly possess a derived hyoid (Brandes, 1932; Brown and Ward, 1988)), great apes all have a bullate hyoid (Fitch, 2009). This appears to be the primitive hominin condition, and its presence in ‘Dikika Girl’ (an Australopithecus afarensis skeleton discovered in Dikika, Ethiopia) confirms that this morphology was present until at least 3.3 million years ago (Alemseged et al., 2006). Hyoid bones are also known from ancestral members of the genus Homo, including Homo neanderthalensis (Arensburg, 1989; Rodriguez et al., 2003), Homo heidelbergensis (Martínez et al., 2008), and Homo erectus (Capasso et al., 2008), all of which indicate that the flat, hyoid morphology found in modern humans (Figure 1) is a derived feature of the Homo lineage. Thus, the shape of the hyoid appears to have changed significantly between 3.3 million (Alemseged et al., 2006) and 530000 (Martínez et al., 2008) years ago, and the loss of the bullate shape, and the transition to a flatter hyoid, is thought to relate to the loss of laryngeal air sacs (Hayama et al., 1982; de Boer, 2012).

Figure 1

Human hyoids (A) lack the large and distinctive bulla of the chimpanzee hyoid (B), into which the air sacs extend. The morphology of the chimpanzee hyoid is typical of that found in great apes. Figure from D’Anastasio et al. (2013), used under the Creative Commons Attribution (CC BY) license. For a more detailed exploration of the chimpanzee hyoid, see

Given that air sacs can become infected (airsacculitis) (Lawson et al., 2006; Kumar et al., 2012), potentially causing disease and reducing fitness, it is argued that air sacs cannot be functionless, and that they would almost certainly be selected against unless they serve a current biological function (de Boer, 2012). Several hypotheses have been proposed regarding the function of laryngeal air sacs, including acting as visual signals, allowing the rebreathing of air to aid respiration, and aiding brachiation by stiffening the thorax (Negus, 1949; Starck and Schneider, 1960; Harrison, 1995). However, these hypotheses have received little support, and many authors now agree that air sacs probably evolved to serve a vocal function (MacLarnon and Hewitt, 1999; Fitch, 2000a, 2009a, 2010; Hewitt et al., 2002; de Boer, 2012). Although it seems clear that air sacs might play an important role in vocal communication, the question remains as to why our ancestors lost their laryngeal air sacs. Was the transition related to positive selection for the loss of air sacs, or was there a relaxation of the selection pressures maintaining them (i.e. did they become obsolete)?

Modelling the acoustic effects of air sacs in human speech

Speech production is based on the source–filter model (Chiba and Kajiyama, 1941; Fant, 1960; Titze, 1994), whereby sound waves are generated by the vocal folds of the larynx (source) and are filtered by the supralaryngeal vocal tract (the filter). The perceptual quality of speech is mainly determined by formant frequencies, which depend on the resonance frequencies of the filter (Chiba and Kajiyama, 1941; Fant, 1960; Titze, 1994; Fitch, 2010; Fitch et al., 2016). Some authors have argued that hominins may have lost their air sacs because they would make speech less clear (de Boer, 2007, 2012; Lieberman, 2011).

Attempts have been made to create synthetic and computer models of the acoustic effects that air sacs would have on humans speech (Riede et al., 2008; de Boer, 2012). These models suggest that air sacs may: (i) increase the amplitude of vocalizations; (ii) induce non-linear source–filter interaction effects; and (iii) change the resonance properties of vocalizations (Riede et al., 2008). Through the use of hearing experiments, de Boer has also shown that air sacs would be likely to affect the perception of speech, by reducing the perceptual effect of vowel-like articulations (de Boer, 2012). If air sacs were lost because they interfered with speech, then the evidence from hyoids would suggest that highly complex vocalizations must have been present by more than 500000 years ago (Martínez et al., 2008). However, as recently noted by Perlman and Salmi (2017), this figure significantly pre-dates some estimates about the timing of the evolution of speech, as some authors have argued the gesture may have been the principal form of linguistic communication up until 150000 (Tomasello, 2008), or even 50000 years ago (Corballis, 2002). It is, of course, possible that laryngeal air sacs were not lost because they interfered with speech, but because they became obsolete and the selection pressure maintaining them was no longer acting. In order to gain further insight into this question, we can analyse the function that air sacs may have in closely related taxa that do not use speech.

The function of air sacs in primate vocal communication

Although most authors agree that air sacs are likely to play a role in vocal communication in non-human primates, questions remain over their exact function. Firstly, by increasing the amplitude of vocalizations, air sacs may allow individuals to produce louder calls that can cover greater distances (Riede et al., 2008). Secondly, air sacs may play an important role in increasing call rate and/or duration, by allowing animals to vocalize for longer or more frequently without hyperventilating (Hewitt et al., 2002). Thirdly, through increasing the variability of vocal tract impendence, air sacs might lead to strong non-linear phenomena in vocalizations (Fitch et al., 2002; Riede et al., 2008). Finally, air sacs have been predicted to affect formant frequencies and spacing (Fitch and Hauser, 1995; Fitch, 2000b; Frey et al., 2007; Riede et al., 2008; Dunn et al., 2015), as: (i) air sacs may introduce new resonance frequencies near the resonance frequency of the air sac in isolation, which would be expected to be lower frequency than the resonances of the vocal tract (and would depend on the size and shape of the air sac); and (ii) the original resonances of the vocal tract would be shifted to higher frequencies and closer together (de Boer, 2008, 2009, 2012; Riede et al., 2008; Dunn et al., 2015). As a consequence of these acoustic effects, vocalizations produced with air sacs are likely to be louder, more chaotic (Fitch et al., 2002), contain more resonances and more energy at lower frequencies, and, potentially, be longer and more frequent (de Boer, 2012). All of these traits may serve important adaptive functions. Most obviously, perhaps, in situations of competition over access to resources, including mates (Wich and Nunn, 2002; Dunn et al., 2015; Charlton and Reby, 2016).

However, these insights have been almost exclusively gained from synthetic and computer modelling. Experimental data testing these hypotheses directly on real air sacs are currently lacking. One possible area of research, which is likely to prove fruitful, is ex-vivo investigation of the effects of air sacs on call acoustics in excised larynx experiments (see Titze et al., 2010; Klemuk et al., 2011; Herbst et al., 2012, 2013; Bowling et al., 2017; Garcia et al., 2017). Another exciting area, which would provide data on vocal-fold vibration during natural vocalizations (thus accounting for the ‘source’ of the sound, and allowing for the ‘filter’ to be analysed—Fant, 1960; Chiba and Kajiyama, 1941), is in-vivo electroglottography (Brown and Cannito, 1995; Herbst and Dunn, 2018). Basic knowledge of the anatomy of air sacs, and other components of the vocal production system, is still lacking for many species of non-human primates, making acoustic modelling challenging. Thus, detailed anatomical work, including computed tomography and magnetic resonance imaging, will provide important insights into the function of air sacs. Finally, high-quality recordings are still lacking for many species. New technology, such as wireless microphones placed on subjects (Ordóñez-Gómez et al., 2015), and improved recording equipment and protocols (Maciej et al., 2011; Fischer et al., 2013), will allow for better acoustic analyses to be carried out, based on anatomical knowledge.

Sexual selection and the size exaggeration hypothesis

The size exaggeration hypothesis posits that some taxa evolve to produce vocal signals with much lower fundamental frequency (f0) and/or formant frequency spacing (Δf) than expected for their size (Fitch, 2000a; Charlton and Reby, 2016). This is thought to be a way of exaggerating body size relative to other species, as lower-frequency sounds are generally associated with bigger body size across species (Morton, 1975; Hauser, 1993; Fitch and Hauser, 1998; Fitch, 2010; Charlton and Reby, 2016). By lowering formant frequencies, and formant dispersion, air sacs are likely to play a key role in size exaggeration.

Species with greater levels of precopulatory male–male competition and reduced levels of sperm competition have been shown to possess adaptations that enable them to produce exaggerated calls, suggesting that sexual selection plays a critical role in regulating vocal communication (Delgado, 2006; Charlton and Reby, 2016). Given this, air sacs are likely to be more common, and perhaps larger, in species under strong sexual selection. For example, recent evidence suggests that gorillas may use their laryngeal air sacs for whinny-type vocalizations and male display (Perlman and Salmi, 2017). Similarly, reindeer show pronounced sexual dimorphism in the size of the air sac, and males use air sacs when vocalizing during the rutting period (Frey et al., 2007). However, data on the comparative anatomy of air sacs and hyoids across species are currently very limited.

Variation in hyoid volume among howler monkey species is predicted by mating system

A good model for this kind of approach is the howler monkey (Alouatta spp.), as this primate genus offers the most extreme variation in hyoid and air sac morphology among primates (Schön, 1971; Schön-Ybarra, 1992; Dunn et al., 2015). The members of this genus are also widely distributed, from Mexico to Argentina, and show significant variation in social organization and mating system. Therefore, we would expect the underlying factors related to this variation to be more evident than in other, less widely distributed, taxa. In howler monkeys, the hyoid bulla is greatly enlarged and specialized (Figure 2A), forming a resonating chamber with which they produce some of the loudest vocalizations made by any terrestrial animal. The size of the hyoid bulla, and subhyoidal air sac, varies significantly between sexes and, even more remarkably, vastly among species (Dunn et al., 2015) (Figure 2B). Dunn et al. (2015) found that variation in hyoid volume and testes volume were both predicted by the number of males in the group (i.e. mating system), with more polygynous species being found to have the largest hyoids and smallest testes, and multi-male species being found to have the smallest hyoids and largest testes. Hyoid volume was also found to be negatively correlated with formant dispersion across species (Dunn et al., 2015). This provided the first evidence for evolutionary trade-offs between vocal characteristics related to size exaggeration and sperm production, which were later reported to be relatively common among other mammals (Charlton and Reby, 2016).

Figure 2

Hyoid morphology among howler monkey species. (A) An adult male Guyanan red howler monkey (Alouatta macconnelli) skull, mandible and hyoid, showing the incredible large relative size of the hyoid bone. (B) A phylogeny of the howler monkey (Alouatta) species studied by Dunn et al. (2015), with Ateles fusciceps as an outgroup. Numbers at the nodes indicate the estimated dates for splitting events (Ma), where known (Cortés-Ortiz et al., 2003). 3D models show the variation in size and shape of average hyoids in males (left) and females (right), corresponding to the species in the tree (hyoids were not available for Alouatta pigra females).

Might sexual selection explain the loss of laryngeal air sacs in Homo?

Perhaps, then, variation in the size and shape of air sacs can be linked to variation in reproductive competition among species. If a clear transition in the shape of the hyoid is observed between Australopithecus afarensis 3 million years ago and Homo erectus 530000 years ago, what do we know about changes in mating system of hominins during this time?

Although debated owing to the limited availability of data (Larsen, 2003; Reno et al., 2003), the traditional consensus is that there has been a decrease in sexual dimorphism during hominin evolution (McHenry, 1992). Australopithecus males have been reported to have been significantly larger than females (45 kg and 29 kg, respectively in A. afarensis; Plavcan and van Schaik, 1997; McHenry, 1992), while the appearance of Homo has been associated with larger bodies, longer limbs, bigger brains, and lower magnitudes of sexual dimorphism (Garvin et al., 2017).

While much of the above will remain hypothetical until more concrete data are gathered, we can speculate by saying that air sacs are probably not necessary for the type of quiet vocal interaction that typifies human communication. They may have interfered with speech and/or become obsolete with the changing social dynamics of larger, more tolerant social groups. As grunts and growls turned into gossip and gab, perhaps more recent Homo species did not need to compete so directly over resources and began to communicate in a more complex way. Indeed, observations from a wide range of taxa indicate that as social groups grow, and sociality becomes more complex, vocal complexity would also be expected increase (Blumstein and Armitage, 1997; Wilkinson, 2003; McComb and Semple, 2005; Freeberg, 2006). Thus, the loss of human air sacs may reflect changes in sexual selection, with increased social tolerance and social complexity in multi-male, multi-female groups paving the way for more complex communication and, ultimately, speech.

© 2018 The Anthropological Society of Nippon