Origin of speech

Spectrogram of American English vowels [i, u, ɑ] showing the formants f₁ and f₂

The word "language" derives from the Latin lingua, "tongue". Phoneticians agree that the tongue is the most important speech articulator, followed by the lips. A natural language can be viewed as a particular way of using the tongue to express thought.

The human tongue has an unusual shape. In most mammals, it is a long, flat structure contained largely within the mouth. It is attached at the rear to the hyoid bone, situated below the oral level in the pharynx. In humans, the tongue has an almost circular sagittal (midline) contour, much of it lying vertically down an extended pharynx, where it is attached to a hyoid bone in a lowered position. Partly as a result of this, the horizontal (inside-the-mouth) and vertical (down-the-throat) tubes forming the supralaryngeal vocal tract (SVT) are almost equal in length (whereas in other species, the vertical section is shorter). As humans move their jaws up and down, the tongue can vary the cross-sectional area of each tube independently by about 10:1, altering formant frequencies accordingly. That the tubes are joined at a right angle permits pronunciation of the vowels [i], [u] and [a], which nonhuman primates cannot do.[28] Even when not performed particularly accurately, in humans the articulatory gymnastics needed to distinguish these vowels yield consistent, distinctive acoustic results, illustrating the quantal nature of human speech sounds.[29] It may not be coincidental that [i], [u] and [a] are the most common vowels in the world's languages.[30] Human tongues are a lot shorter and thinner than other mammals and are composed of a large number of muscles, which helps shape a variety of sounds within the oral cavity. The diversity of sound production is also increased with the human’s ability to open and close the airway, allowing varying amounts of air to exit through the nose. The fine motor movements associated with the tongue and the airway, make humans more capable of producing a wide range of intricate shapes in order to produce sounds at different rates and intensities.[31]

In humans, the lips are important for the production of stops and fricatives, in addition to vowels. Nothing, however, suggests that the lips evolved for those reasons. During primate evolution, a shift from nocturnal to diurnal activity in tarsiers, monkeys and apes (the haplorhines) brought with it an increased reliance on vision at the expense of olfaction. As a result, the snout became reduced and the rhinarium or "wet nose" was lost. The muscles of the face and lips consequently became less constrained, enabling their co-option to serve purposes of facial expression. The lips also became thicker, and the oral cavity hidden behind became smaller.[31] Hence, according to Ann MacLarnon, "the evolution of mobile, muscular lips, so important to human speech, was the exaptive result of the evolution of diurnality and visual communication in the common ancestor of haplorhines".[32] It is unclear whether human lips have undergone a more recent adaptation to the specific requirements of speech.

Compared with nonhuman primates, humans have significantly enhanced control of breathing, enabling exhalations to be extended and inhalations shortened as we speak. Whilst we are speaking, intercostal and interior abdominal muscles are recruited to expand the thorax and draw air into the lungs, and subsequently to control the release of air as the lungs deflate. The muscles concerned are markedly more innervated in humans than in nonhuman primates.[33] Evidence from fossil hominins suggests that the necessary enlargement of the vertebral canal, and therefore spinal cord dimensions, may not have occurred in Australopithecus or Homo erectus but was present in the Neanderthals and early modern humans.[34][35]

Anatomy of the larynx, anterolateral view

The larynx or voice box is an organ in the neck housing the vocal folds, which are responsible for phonation. In humans, the larynx is descended, it is positioned lower than in other primates. This is because the evolution of humans to an upright position shifted the head directly above the spinal cord, forcing everything else downward. The repositioning of the larynx resulted in a longer cavity called the pharynx, which is responsible for increasing the range and clarity of the sound being produced. Other primates have almost no pharynx; therefore, their vocal power is significantly lower.[31] Humans are not unique in this respect: goats, dogs, pigs and tamarins lower the larynx temporarily, to emit loud calls.[36] Several deer species have a permanently lowered larynx, which may be lowered still further by males during their roaring displays.[37] Lions, jaguars, cheetahs and domestic cats also do this.[38] However, laryngeal descent in nonhumans (according to Philip Lieberman) is not accompanied by descent of the hyoid; hence the tongue remains horizontal in the oral cavity, preventing it from acting as a pharyngeal articulator.[39]

Anterolateral view of head and neck

Despite all this, scholars remain divided as to how "special" the human vocal tract really is. It has been shown that the larynx does descend to some extent during development in chimpanzees, followed by hyoidal descent.[40] As against this, Philip Lieberman points out that only humans have evolved permanent and substantial laryngeal descent in association with hyoidal descent, resulting in a curved tongue and two-tube vocal tract with 1:1 proportions. Uniquely in the human case, simple contact between the epiglottis and velum is no longer possible, disrupting the normal mammalian separation of the respiratory and digestive tracts during swallowing. Since this entails substantial costs – increasing the risk of choking whilst swallowing food – we are forced to ask what benefits might have outweighed those costs. Some claim the clear benefit must have been speech, but other contest this. One objection is that humans are in fact not seriously at risk of choking on food: medical statistics indicate that accidents of this kind are extremely rare.[41] Another objection is that in the view of most scholars, speech as we know it emerged relatively late in human evolution, roughly contemporaneously with the emergence of Homo sapiens.[42] A development as complex as the reconfiguration of the human vocal tract would have required much more time, implying an early date of origin. This discrepancy in timescales undermines the idea that human vocal flexibility was initially driven by selection pressures for speech.

At least one orangutan has demonstrated the ability to control the voice box.[43]

To lower the larynx is to increase the length of the vocal tract, in turn lowering formant frequencies so that the voice sounds "deeper" – giving an impression of greater size. John Ohala argued that the function of the lowered larynx in humans, especially males, is probably to enhance threat displays rather than speech itself.[44] Ohala pointed out that if the lowered larynx were an adaptation for speech, we would expect adult human males to be better adapted in this respect than adult females, whose larynx is considerably less low. In fact, females invariably outperform males in verbal tests, falsifying this whole line of reasoning. William Tecumseh Fitch likewise argues that this was the original selective advantage of laryngeal lowering in humans. Although, according to Fitch, the initial lowering of the larynx in humans had nothing to do with speech, the increased range of possible formant patterns was subsequently co-opted for speech. Size exaggeration remains the sole function of the extreme laryngeal descent observed in male deer. Consistent with the size exaggeration hypothesis, a second descent of the larynx occurs at puberty in humans, although only in males. In response to the objection that the larynx is descended in human females, Fitch suggests that mothers vocalising to protect their infants would also have benefited from this ability.[45]

Hyoid bone – anterior surface, enlarged

Most specialists credit the Neanderthals with speech abilities not radically different from those of modern Homo sapiens. An indirect line of argument is that their toolmaking and hunting tactics would have been difficult to learn or execute without some kind of speech.[46] A recent extraction of DNA from Neanderthal bones indicates that Neanderthals had the same version of the FOXP2 gene as modern humans. This gene, mistakenly described as the "grammar gene", plays a role in controlling the orofacial movements which (in modern humans) are involved in speech.[47]

During the 1970s, it was widely believed that the Neanderthals lacked modern speech capacities.[48] It was claimed that they possessed a hyoid bone so high up in the vocal tract as to preclude the possibility of producing certain vowel sounds.

The hyoid bone is present in many mammals. It allows a wide range of tongue, pharyngeal and laryngeal movements by bracing these structures alongside each other in order to produce variation.[49] It is now realised that its lowered position is not unique to Homo sapiens, whilst its relevance to vocal flexibility may have been overstated: although men have a lower larynx, they do not produce a wider range of sounds than women or two-year-old babies. There is no evidence that the larynx position of the Neanderthals impeded the range of vowel sounds they could produce.[50] The discovery of a modern-looking hyoid bone of a Neanderthal man in the Kebara Cave in Israel led its discoverers to argue that the Neanderthals had a descended larynx, and thus human-like speech capabilities.[51][52] However, other researchers have claimed that the morphology of the hyoid is not indicative of the larynx's position.[11] It is necessary to take into consideration the skull base, the mandible, the cervical vertebrae and a cranial reference plane.[53][54]

The morphology of the outer and middle ear of Middle Pleistocene hominins from Atapuerca, Spain, believed to be proto-Neanderthal, suggests they had an auditory sensitivity similar to modern humans and very different from chimpanzees. They were probably able to differentiate between many different speech sounds.[55]

Hypoglossal nerve
Hypoglossal nerve, cervical plexus, and their branches
Details
Identifiers
Latin	nervus hypoglossus
Anatomical terms of neuroanatomy

The hypoglossal nerve plays an important role in controlling movements of the tongue. In 1998, a research team used the size of the hypoglossal canal in the base of fossil skulls in an attempt to estimate the relative number of nerve fibres, claiming on this basis that Middle Pleistocene hominins and Neanderthals had more fine-tuned tongue control than either Australopithecines or apes.[56] Subsequently, however, it was demonstrated that hypoglossal canal size and nerve sizes are not correlated,[57] and it is now accepted that such evidence is uninformative about the timing of human speech evolution.[58]

IPA: Vowels
Front Central Back Close i y ɨ ʉ ɯ u Near-close ɪ ʏ ʊ Close-mid e ø ɘ ɵ ɤ o Mid e̞ ø̞ ə ɤ̞ o̞ Open-mid ɛ œ ɜ ɞ ʌ ɔ Near-open æ ɐ Open a ɶ ä ɑ ɒ
IPA help  audio full chart template Legend: unrounded • rounded

According to one influential school,[59][60] the human vocal apparatus is intrinsically digital on the model of a keyboard or digital computer (see below). Nothing about a chimpanzee's vocal apparatus suggests a digital keyboard, notwithstanding the anatomical and physiological similarities. This poses the question as to when and how, during the course of human evolution, the transition from analog to digital structure and function occurred.

The human supralaryngeal tract is said to be digital in the sense that it is an arrangement of moveable toggles or switches, each of which, at any one time, must be in one state or another. The vocal cords, for example, are either vibrating (producing a sound) or not vibrating (in silent mode). By virtue of simple physics, the corresponding distinctive feature – in this case, "voicing" – cannot be somewhere in between. The options are limited to "off" and "on". Equally digital is the feature known as "nasalisation". At any given moment the soft palate or velum either allows or does not allow sound to resonate in the nasal chamber. In the case of lip and tongue positions, more than two digital states may be allowed.

The theory that speech sounds are composite entities constituted by complexes of binary phonetic features was first advanced in 1938 by the Russian linguist Roman Jakobson.[61] A prominent early supporter of this approach was Noam Chomsky, who went on to extend it from phonology to language more generally, in particular to the study of syntax and semantics.[62][63][64] In his 1965 book, Aspects of the Theory of Syntax,[65] Chomsky treated semantic concepts as combinations of binary-digital atomic elements explicitly on the model of distinctive features theory. The lexical item "bachelor", on this basis, would be expressed as [+ Human], [+ Male], [- Married].

Supporters of this approach view the vowels and consonants recognised by speakers of a particular language or dialect at a particular time as cultural entities of little scientific interest. From a natural science standpoint, the units which matter are those common to Homo sapiens by virtue of biological nature. By combining the atomic elements or "features" with which all humans are innately equipped, anyone may in principle generate the entire range of vowels and consonants to be found in any of the world's languages, whether past, present or future. The distinctive features are in this sense atomic components of a universal language.

In recent years, the notion of an innate "universal grammar" underlying phonological variation has been called into question. The most comprehensive monograph ever written about speech sounds, The Sounds of the World's Languages, by Peter Ladefoged and Ian Maddieson,[30] found virtually no basis for the postulation of some small number of fixed, discrete, universal phonetic features. Examining 305 languages, for example, they encountered vowels that were positioned basically everywhere along the articulatory and acoustic continuum. Ladefoged concluded that phonological features are not determined by human nature: "Phonological features are best regarded as artifacts that linguists have devised in order to describe linguistic systems".[66]

Birds flocking, an example of self-organization in biology

Self-organisation characterises systems where macroscopic structures are spontaneously formed out of local interactions between the many components of the system.[67] In self-organised systems, global organisational properties are not to be found at the local level. In colloquial terms, self-organisation is roughly captured by the idea of "bottom-up" (as opposed to "top-down") organisation. Examples of self-organised systems range from ice crystals to galaxy spirals in the inorganic world.

A termite mound (Macrotermitinae) in the Okavango Delta just outside Maun, Botswana

According to many phoneticians, the sounds of language arrange and re-arrange themselves through self-organisation.[67][68][69] Speech sounds have both perceptual (how one hears them) and articulatory (how one produces them) properties, all with continuous values. Speakers tend to minimise effort, favouring ease of articulation over clarity. Listeners do the opposite, favouring sounds that are easy to distinguish even if difficult to pronounce. Since speakers and listeners are constantly switching roles, the syllable systems actually found in the world's languages turn out to be a compromise between acoustic distinctiveness on the one hand, and articulatory ease on the other.

Agent-based computer models take the perspective of self-organisation at the level of the speech community or population. The two main paradigms are (1) the iterated learning model and (2) the language game model. Iterated learning focuses on transmission from generation to generation, typically with just one agent in each generation.[70] In the language game model, a whole population of agents simultaneously produce, perceive and learn language, inventing novel forms when the need arises.[71][72]

Several models have shown how relatively simple peer-to-peer vocal interactions, such as imitation, can spontaneously self-organise a system of sounds shared by the whole population, and different in different populations. For example, models elaborated by Berrah et al. (1996)[73] and de Boer (2000),[74] and recently reformulated using Bayesian theory,[75] showed how a group of individuals playing imitation games can self-organise repertoires of vowel sounds which share substantial properties with human vowel systems. For example, in de Boer's model, initially vowels are generated randomly, but agents learn from each other as they interact repeatedly over time. Agent A chooses a vowel from her repertoire and produces it, inevitably with some noise. Agent B hears this vowel and chooses the closest equivalent from her own repertoire. To check whether this truly matches the original, B produces the vowel she thinks she has heard, whereupon A refers once again to her own repertoire to find the closest equivalent. If this matches the one she initially selected, the game is successful, otherwise, it has failed. "Through repeated interactions", according to de Boer, "vowel systems emerge that are very much like the ones found in human languages".[76]

In a different model, the phonetician Björn Lindblom[77] was able to predict, on self-organisational grounds, the favoured choices of vowel systems ranging from three to nine vowels on the basis of a principle of optimal perceptual differentiation.

Further models studied the role of self-organisation in the origins of phonemic coding and combinatoriality, which is the existence of phonemes and their systematic reuse to build structured syllables. Pierre-Yves Oudeyer developed models which showed that basic neural equipment for adaptive holistic vocal imitation, coupling directly motor and perceptual representations in the brain, can generate spontaneously shared combinatorial systems of vocalisations, including phonotactic patterns, in a society of babbling individuals.[67][78] These models also characterised how morphological and physiological innate constraints can interact with these self-organised mechanisms to account for both the formation of statistical regularities and diversity in vocalisation systems.

The gestural theory states that speech was a relatively late development, evolving by degrees from a system that was originally gestural. Human ancestors were unable to control their vocalisation at the time when gestures were used to communicate; however, as they slowly began to control their vocalisations, spoken language began to evolve.

Three types of evidence support this theory:

1. Gestural language and vocal language depend on similar neural systems. The regions on the cortex that are responsible for mouth and hand movements border each other.
2. Nonhuman primates minimise vocal signals in favour of manual, facial and other visible gestures in order to express simple concepts and communicative intentions in the wild. Some of these gestures resemble those of humans, such as the "begging posture", with the hands stretched out, which humans share with chimpanzees.[79]
3. Mirror Neurons

Research has found strong support for the idea that spoken language and signing depend on similar neural structures. Patients who used sign language, and who suffered from a left-hemisphere lesion, showed the same disorders with their sign language as vocal patients did with their oral language.[80] Other researchers found that the same left-hemisphere brain regions were active during sign language as during the use of vocal or written language.[81]

Humans spontaneously use hand and facial gestures when formulating ideas to be conveyed in speech.[82][83] There are also, of course, many sign languages in existence, commonly associated with deaf communities; as noted above, these are equal in complexity, sophistication, and expressive power, to any oral language. The main difference is that the "phonemes" are produced on the outside of the body, articulated with hands, body, and facial expression, rather than inside the body articulated with tongue, teeth, lips, and breathing.

Many psychologists and scientists have looked into the mirror system in the brain to answer this theory as well as other behavioural theories. Evidence to support mirror neurons as a factor in the evolution of speech includes mirror neurons in primates, the success of teaching apes to communicate gesturally, and pointing/gesturing to teach young children language. Fogassi and Ferrari (2014) monitored motor cortex activity in monkeys, specifically area F5 in the Broca’s area, where mirror neurons are located. They observed changes in electrical activity in this area when the monkey executed or observed different hand actions performed by someone else. Broca’s area is a region in the frontal lobe responsible for language production and processing. The discovery of mirror neurons in this region, which fire when an action is done or observed specifically with the hand, strongly supports the belief that communication was once accomplished with gestures. The same is true when teaching young children language. When one points at a specific object or location, mirror neurons in the child fire as though they were doing the action, which results in long-term learning[84]

Critics note that for mammals in general, sound turns out to be the best medium in which to encode information for transmission over distances at speed. Given the probability that this applied also to early humans, it is hard to see why they should have abandoned this efficient method in favour of more costly and cumbersome systems of visual gesturing – only to return to sound at a later stage.[85]

By way of explanation, it has been proposed that at a relatively late stage in human evolution, hands became so much in demand for making and using tools that the competing demands of manual gesturing became a hindrance. The transition to spoken language is said to have occurred only at that point.[86] Since humans throughout evolution have been making and using tools, however, most scholars remain unconvinced by this argument. (For a different approach to this issue – one setting out from considerations of signal reliability and trust – see "from pantomime to speech" below).