The Evolution of Language PDF

10 The Evolution of Language Key Concepts learnability argument ostensive communication Universal Grammar (UG) parameter setting recursion FOXP2 gene social grooming hypothesis social contract hypothesis cultural transmission hypothesis Without language, social...

10 The Evolution of Language Key Concepts learnability argument ostensive communication Universal Grammar (UG) parameter setting recursion FOXP2 gene social grooming hypothesis social contract hypothesis cultural transmission hypothesis Without language, social interaction would be impoverished beyond recognition. It enables us to re- veal our innermost thoughts to others, or, if the mood takes us, to disguise them with misinformation and lies. With language, action can be coordinated so that a group of people can act as one – even if chimpanzees could conceive of a pyramid they still couldn’t build one because they lack the ability to coordinate action through language. Language also, as we shall see in Chapter 14, enables hard- won knowledge to be passed on to others – including our children – enabling culture to proliferate in ways that would not have been possible in our languageless ancestors (in fact, as we shall see, one theory proposes that language evolved to facilitate cultural transmission). When language evolved it was evolutionary dynamite. Not only did it vastly extend the range of things that ancestral hu- mans were capable of, enabling them, perhaps, to outcompete other hominins around at the time, but it is also likely to have had an impact on the evolution of the brain itself. It is unlikely that our languageless ancestors had brains identical to ours but lacking the appropriate language circuitry; it is more likely that the gradual evolution of communicative sophistication led to huge leaps in the way that we interact with others. So great are the advantages of language to our species that surely it must have been the product of natural (or sexual) selection. This chapter begins by outlining the complexity of learning a language and then discusses the extent to which language evolved and is domain-specific. The question of evolution naturally leads to the speculation that there are genes for language, and we discuss current thinking on language genes. Finally, we present some speculation as to the original evolutionary function of language: what was it, in ancestral times, actually for? Is Language Specific to Human Beings? There is no doubt that non-human animals communicate with each other in the sense that they have evolved particular ways to convey information such as their particular psychological state to others (think of a growling dog), but animals communicate other information as well. The work of Austrian ethologist Karl von Frisch on the dances of honeybees is a celebrated example. Von Frisch discovered that after finding a profitable food patch, bees would return to the hive and engage in an elaborate dance to communicate the physical location of the food source and the estimated quantity of nectar to be had (see Figure 10.1). Bees indicate the location of the nectar source to their hive-  Is Language Specific to Human Beings? 251 Figure 10.1 Bees communicate the location of a nectar source using the round dance (the roughly heart-shaped paths) and the waggle dance (the wiggly line through the middle). mates by dancing at an angle to the vertical that is the same as the angle between the sun and the food source. The bee waggles its abdomen with a vigour directly proportional to the richness of the patch, allowing the bees to judge whether it is worth exploiting: a low-quality patch might produce a languid wiggle, a high-quality patch might lead to a full-blooded twerk. Another classic example of complex communication in non-human animals is that of vervet monkeys who have been shown to use different vocalisations to indicate the presence of dif- ferent types of predator such as snakes, cheetahs and eagles (Cheney and Seyfarth, 1982). It is not only primates who appear to be able to do this. Prairie dogs (a ground-living member of the squirrel family; see Figure 10.2) appear to have a range of distinct warning calls made in response to specific stimuli. They will, for example produce different calls to coyote, red-tailed hawk, deer and humans. Perhaps even more surprisingly, when prairie dogs were presented with unfamiliar objects such as a European ferret or a black oval they all gave the same call independently of one another (Slobod- chikoff, 2002) and were even able to incorporate the colour of the predators into their vocalisations (Slobodchikoff et al., 2009). In this case the ‘predators’ were three similar-looking women wearing different coloured clothing. Giving the same call is important because it suggests that their calls are genuinely referring to particular attributes of the stimuli (colour, for example, or shape) rather than just making random vocalisations. It suggests that the prairie dog communication system has rules for generating new ‘words’ and that these rules are shared by the community – something that is essential if you need to be understood. 252 The Evolution of Language Figure 10.2 A prairie dog making an alarm call. The researcher who carried out these studies (Con Slobodchikoff of Northern Arizona University) speculates that prairie dogs have some (possibly innate) rules that enable them to pro- duce calls to novel stimuli, although the details of this are obscure. Why might prairie dogs have evolved such a sophisticated system of alarm calls? One possibility is that prairie dogs are preyed on by many different kinds of other animals including coyotes, badgers, domestic cats and dogs, ferrets, owls, eagles, bobcats, snakes and, of course, humans. With such a variety of predators a flexible system of warning calls would help them to flee from predators. Presumably, as with vervet monkeys, the appropriate response to a particular predator varies as a result of the specific nature of the predator. Whether any of these (and other) instances of communication constitute language is a long-standing debate, although currently the majority opinion is that they do not. (Things do seem to be changing – slowly – here as researchers study more animals in more detail.) With respect to the criteria presented in Box 10.1 no animal communication system yet meets all seven (Aitchison, 1989), the principal difference being that human language is much more creative and flexible than animal communication systems. Honeybees, for example, are geniuses at giving directions to sourc- es of nectar, but appear unable to use their sophisticated dances to communicate other potentially useful information such as the existence and location of a rival hive. Vervet monkeys can signal the presence of predators but do not appear to use their calls to convey information about the quality of potential mates, habitats or sources of food. Human language, however, is used to communicate information relating to every realm of human experience. Box 10.1 What Is Language? Producing a definition for language is not easy but here are some fairly widely accepted crite- ria that help to identify whether a form of communication is or is not a language (Aitchison, 1989).  Is Language Specific to Human Beings? 253 Box 10.1 (cont.) Use of vocal auditory channel. Fairly self-explanatory, a language should use verbal com- munication. In fact, this is probably one of the weakest and most controversial of the criteria, because it rules out the sign language used by the hearing impaired, which is certainly a language by all other criteria (and enables communication as complex as spoken language). It also rules out written communication. As Aitchison points out, this criterion captures the essence of human language rather than being a necessary feature and is of little use for deciding whether animals have language. Arbitrariness. Languages use symbols that need bear no relationship to what is being dis- cussed. The word ‘dog’, for example, does not physically sound like a dog. The word ‘minus- cule’, meaning very small, is a much larger word than is ‘huge’, meaning very big. Similarly, one can shout the word ‘quiet’ with no disruption of its meaning. Some animals do appear to have the property of arbitrariness in their communications. Semanticity. Language means something. When we hear the word ‘chair’, we know that someone is referring to a particular type of object (and we also assume that the person has some intention behind referring to the chair). Cultural transmission. Languages are handed down from generation to generation; they are not usually generated spontaneously (although there is evidence that this has happened with some sign languages). Languages are, among other things, cultural artefacts and as such are jealously guarded by their speakers. Spontaneous usage. Language is spoken freely, not only under duress. Even human babies seem to have an overriding desire to babble; it is not something that their parents need to encour- age in them. Turn taking. Apart from very specific contexts such as a lecture or a soliloquy in a Shake- spearean play, we don’t tend to speak for very long periods at a time, neither do we tend to talk while someone else is speaking; we take it in turns. Again, turn taking is something that happens very early in life, seemingly without tuition (most of the time!). Duality or double articulation. Languages are double-layered. The component sounds of a word such as d-o-g are meaningless by themselves; they only acquire meaning when they are placed in a very specific order. Displacement. Language can refer to things that are not there. Either because they are no longer present (temporal displacement), they are some distance away (spatial displacement) or they are totally hypothetical. When we say ‘If it rains tomorrow, I will stay at home’ we are refer- ring to a hypothetical state of affairs. Structure dependence. All human languages are fussy about how words go together. The meaning of ‘have a nice day’ is dependent on the words occurring in a particular sequence and its meaning is not some average of the meaning of the words. In contrast, many vocalisations made by animals (and humans) require no structure. The howls of pain produced by a person having hit their thumb with a hammer certainly communicate their suffering, but no structure is required for this communication to occur. Creativity. Language is infinitely expressive; we do not have a set number of responses like some dumb artificial intelligence program. When something novel happens, we are able to describe it, or at least attempt to. 254 The Evolution of Language Human Language and the Combinatorial Power of Grammar The process of forming a one-to-one mapping between sound and object such as that displayed by vervet monkeys was an important step in the evolution of language. One might even imagine a situation whereby this ability is extended and refined to lead, ultimately, to a human-like vocabulary which contains many thousands of words. The average literate person knows between 50,000 and 100,000 words; however if you remove words that share a common morphological root – such as walk, walking, walked, and compounds made of known words (wristwatch, pencil case) – the aver- age American high school graduate knows around 45,000 words (Nagy and Anderson, 1984). Hav- ing a different word for each concept is fine as far as it goes, but given the large number of things that we might wish to talk about we would soon find ourselves running out of space to store all the words we know and time in which to learn them all. Furthermore, if for each idea there had to be a distinct ‘word’ how would you ever say anything new? You could always invent a new word, of course, but since no one else would know what the word means they wouldn’t be able to understand you. Lan- guage’s second trick is even more impressive than an extensive vocabulary; it gives us a means of obtaining limitless powers of expression from a comparatively small set of words: grammar. The communication structures of other animals, as far as we know, do not have grammar, which is why vervet monkeys and honey bees are stuck within their limited ‘linguistic’ universe, unable to ‘say’ anything that hasn’t already been ‘said’ by others a thousand times before (although prairie dogs might be an exception). Interestingly von Frisch recounts an experiment where he placed sugar water some distance up an antenna only a short distance from the hive. When bees were shown the source they flew back to the hive and duly gave directions through the medium of dance. This led to a host of bees buzzing around the base of the antenna, apparently unable to find the source. Von Frisch concluded that: ‘The bees have no word for ‘up’ in their language. There are no flowers in the clouds’ (von Frisch, 1954, 139). Pinker and Bloom – an Evolutionary Theory of Language Acquisition In their landmark paper ‘Natural language and natural selection’ (1990) the psycholinguists Steven Pinker and Paul Bloom argued that language has many of the characteristics of complex organs such as the human eye or hand. Like the eye, language has positive fitness implications in that it enables information to be transmitted more efficiently than by non-linguistic means, but it also allows differ- ent forms of information to be transmitted. The properties of displacement and creativity (see Box 10.1) mean that we are free from making fixed responses about the here and now. With language we can refer easily to past events, speculate on the future and discuss any issue that might come to mind, again things that are difficult using non-linguistic communication. Pinker and Bloom argue that natural selection is the only known mechanism that can produce the adaptive complexity that we see in language. A particularly difficult problem for non-selectionist theories is how to explain all the costs associated with the specialist hardware that language requires. The comprehension and production of language uses a large amount of costly neural material in dedicated brain regions such as Broca’s area and Wernicke’s area (see Figure 10.3). Furthermore, the design of the vocal tract required to make complex articulation possible also has the unfortunate side effect of making dying from choking a distinct possibility. (Non-hu- man mammals and, indeed, pre-linguistic babies are able to swallow and breathe at the same time.)  Learning Language 255 Parietal lobe Frontal lobe Motor areas related to Occipital lobe speech Auditory cortex Broca’s area Wernicke’s area Temporal lobe Figure 10.3 Regions of the brain involved in language processing showing Broca’s area, Wernicke’s area and the related areas involved in hearing and articulation. If language did not arise by natural selection, it is hard to explain how our ancestors could have borne these costs without them being offset by some very special advantages. In the next few sections we examine some of the evidence for Pinker and Bloom’s argu- ment. First, we discuss the fundamental building blocks of language, the sounds, or phonemes that are used to transmit meaning. Second, we discuss how words are learned and what assumptions chil- dren need to make when learning a word. Finally, we present Chomsky’s work on grammar which suggests that language learning is supported by an innate language ‘organ’. Learning Language Learning the Sounds of Language With few exceptions – such as the sign languages used, and sometimes created, by hearing impaired communities – languages transmit meaning through the medium of sound. Why should this be the case when – as generations of people with hearing impairments have demonstrated – a combination of gestural and facial expressions can serve just as well; what is so special about sound? First, using sound enables multitasking; your hands can get on with other work while you talk. Second, you don’t have to be looking at the person with whom you’re communicating, which means that you can attend visually to other things while communicating and others can easily attract your attention with warning cries for example. Third, you can communicate in the dark – an important attribute in our ancestral environment of equatorial Africa, where it is dark for approximately 12 hours a day every day. Although sound is the medium of choice for all languages not all languages use the same sounds. Each language contains a particular set of minimal acoustic building blocks known as pho- nemes (such as ‘b’, ‘a’ and ‘m’) and languages differ as to how many phonemes they use. English, for instance, uses around 40 different phonemes (depending on dialect), Polynesian, however, uses only 11 phonemes and Khoisan (a language of southern Africa) as many as 140. Some phonemes such as the ‘click’ of Bantu languages as used by the !Kung San (the exclamation mark representing one of those clicks) sound alien to English speakers’ ears, but to speakers of other languages English itself uses sounds that are difficult to pronounce. For example, the ‘th’ sound in English words such 256 The Evolution of Language as ‘thick’ or ‘the’ is difficult for many non-native speakers who often approximate by using ‘s’ or ‘z’. Some phonemes are not only difficult to pronounce for a non-native speaker, they are indiscrim- inable from other phonemes in the language, even after 500 trials of training (Pinker, 1994, 264). Two-month-old infants, however, are able to discriminate between all of the phonemes in all of the world’s languages (Kuhl et al., 1992). This is not so surprising because the very presence of different phonemes in a particular language depends upon the ability of babies to distinguish them in the first place. However, infants learn rapidly and this ability starts to decline around 6 months of age as they home in on their native phonemes, where ‘native’ simply means the sounds that they are exposed to. For instance, older babies in a Spanish-speaking environment will start to treat the phonemes ‘v’ and ‘b’ as the same, just as their parents do, whereas younger babies show a discrimination. So, before a baby utters its first word it has already learned a great deal about the language spoken around it. Babies, it seems, are predisposed to attend to language-like sounds from birth, as we would expect if language were an evolved instinct. But phonemes are just the building blocks of language, they don’t in themselves contain any meaning; for this they need to be combined to produce words and how children learn words is what we turn to next. Learning Words Learning words seems easy. A person points at an object, says its name aloud and the child gradually learns to associate the sound pattern with the object (this is one of the claims made by Skinner’s behaviourism). However, as we shall see, this simple case of word learning (known as ostensive communication or ostension) only appears easy because it is supported by a variety of cognitive mechanisms. Learning by ostension is more complex than it seems and to illustrate this point philos- opher W. V. O. Quine devised the ‘gavagai problem’ (Quine, 1960). Quine invites us to imagine that we are travellers in a foreign land accompanied by a guide; we cannot speak a word of his language, and he cannot speak a word of ours. Suddenly as we rise to the crest of a hill a rabbit dashes out in front of us. The native points at the rabbit and shouts ‘gavagai’. What does he mean? Most people assume that gavagai must be the local word for rabbit, but why do we assume this? Gavagai might be an exclamation of surprise such as ‘blimey’; maybe he is referring to its colour or the way that it is moving, or maybe he is merely shouting ‘there goes my dinner’. There is no way of telling what he means, but we make the assumption that he is naming the whole object. Young children learning words face a similar problem. When, for instance, a parent points at a brown furry animal and calls it a dog there is nothing in this act that tells the child that the parent is referring to the type of animal rather than giving that particular dog’s name (‘Fido’) or passing some comment on its shaggy coat. In the absence of this additional information the child has to provide his own by making assump- tions about the naming process. Constraints on Word Learning Ellen Markman (1989) suggests that the child makes (at least) three assumptions when learning words. First, the child makes the whole object assumption: the child assumes in the absence of any other evidence that when an object is named the person refers to the whole object, not its parts, colour or some other attribute. Second, the child makes the taxonomic assumption, when an object is named the child assumes that the person is referring not to that specific object (‘Fido’) nor to its very general category (mammal, animal or living thing) but at a medium level of generality known as the basic level (Rosch et al., 1976). Basic level categories usually contain  Acquiring Grammar 257 items that resemble each other in crucial ways. For example, ‘table’ is a basic level category whereas ‘furniture’ is at a higher level of abstraction. Although tables differ somewhat in size and shape, they are quite similar and much more similar to each other than the members of the ‘fur- niture’ category which contains tables, beds and wardrobes. Finally, children make the mutual exclusivity assumption. If a child already knows the name for an object and it is referred to by another word, then the child assumes that the second word is referring to some other aspect of the object such as its colour, manner of movement, material and so on, rather than assuming that it is a synonym. To demonstrate this, Markman presented three-year-olds with an array of objects which included familiar objects (cups) and unfamiliar objects (tongs). When children were shown a set of tongs and told that it was biff and asked to collect more biffs the children assumed that biff must be the name for tongs (whole object assumption) and collected more, even though the tongs were different in subtle ways (taxonomic assumption). However, when shown a pewter cup and told that that was biff the children, knowing that the object already has the name ‘cup’ assume that it must be something else about the object that is biff rather than its name (mutual exclusivity). Generally they assume that it must be the unusual material that it is made out of and choose other pewter objects as examples of biff. Here we can see mutual exclusivity operating which enables children to infer word meaning from context. Children’s Sensitivity to the Attention of Others in Word Learning The behaviourist model that word learning is solely the formation of an association between two stimuli (object and sound) has been shown to be false; furthermore, word learning by ostension requires special mechanisms if the child is not to drown in a sea of hypotheses about what people mean when they name an object. It also appears that sounds are only associated with objects under particular circumstances. For instance, if an infant is playing with a novel toy and hears a novel word, she only assumes that the word is the name for the toy if someone uses the word while they themselves are attending to the toy (Baldwin et al., 1996). Similarly, if a child is looking at a novel object and an adult produces a novel name, the child will shift attention to the adult’s eyes to see what it is that the adult is attending to. This research shows that learning the meaning of words is rather more than the formation of associations between sound and object and involves assumptions and the sharing of attention that are probably inborn. The importance of innate knowledge is even more prominent when we consider the way that children acquire the grammatical structures of lan- guage, which is what we address next. Acquiring Grammar: Chomsky, Innateness and the Universal Grammar Noam Chomsky is a famous political writer and activist (see Figure 10.4) but some people don’t realise that he is equally famous as a linguist. Chomsky was once asked if he preferred his political work to his work on linguistics. He replied that his political work was done out of a sense of duty and if the world’s problems went away, he would happily devote himself to the pursuit of knowledge for its own sake (Horgan, 2016). We approach Chomsky with some trepidation, not only because his work is closer to maths than most psychologists would be comfortable with, but also because his theory changed considerably over time, as we shall see. But we start at the beginning. 258 The Evolution of Language In the mid 1950s Chomsky (1957) made the outrageous claim that language learning is impossible. What he meant was that given the amount of information that children have to go on, learning was not possible without extra information coming from elsewhere, an argument known as the argument from the poverty of the stimulus or the learnability argument. He reasoned that children learn language so quickly and with so few errors that they cannot be merely learning language by trial and error as behaviourists such as Skinner suggested or by another general learning mechanism such as analogy (see Chomsky, 1959). The only way that language could be learned was if children are given support by innate knowledge which fills in the blanks left by messy human speech. To illustrate this, we will use one of Chomsky’s most celebrated examples, that of turning statements into questions. Suppose you are given the following statement: 1. The man is playing football. We can turn this into a question that asks what the man is doing by simply taking the ‘is’ and moving it to the beginning of the sentence as in example 2: 2. Is the man playing football? Now Chomsky asks us to imagine that the child is attempting to discover (or induce) the linguistic rule that turns statements into questions simply by observing instances of statements and questions as above. Imagine that the child encounters a different sentence to turn into a question: 3. The man who is wearing shorts is playing football. The problem is that the word ‘is’ appears twice; which one should the child take? A child familiar with simple sentences such as those like Sentence 1 might be tempted to take the first ‘is’ to produce *4. Is the man who wearing shorts is playing football? which is ungrammatical (denoted by the asterisk). If children simply worked out grammatical rules by trial and error – as many early theories claimed – then we should expect to see such errors at least some of the time, but we don’t. Why not? Figure 10.4 Linguist and political activist Noam Chomsky.  Acquiring Grammar 259 According to Chomsky language isn’t just a linear stream of words, it is hierarchically structured and from the earliest years, children understand this. Implicitly children break up (the technical term is parse) utterances into higher-level units such as noun phrases and verb phrases. In the first sentence ‘The man’ is a noun phrase, containing information about the subject of the sentence, ‘is playing football’, on the other hand, is a verb phrase which contains information about what the man is doing. In the second sentence, the noun phrase is rather longer ‘the man who is wearing shorts’ because ‘who is wearing shorts’ is saying something about the man (noun) rather than about the action he is performing (verb). Because the question requires the child to ask what the man is doing (the verb phrase), the ‘is’ that is moved has to be the one that relates to the verb phrase rather than that of the noun phrase. (If we wanted to ask a question about the man – whether he is wearing shorts – we would move the ‘is’ from the noun phrase to end up with ‘is the man who is playing football wearing shorts’.) The overwhelming majority of people who can speak a language have probably never even heard of noun and verb phrases, and they are things that do not come nat- urally to students of grammar and linguistics in the first instance, but everyone is implicitly aware of their existence, otherwise they would not be able to speak in grammatical sentences. This is a classic example of what Chomsky calls the learnability argument; the information needed to turn statements into questions is not something that is given to children in the linguistic environment, nor do parents explicitly teach children about verb and noun phrases. But this infor- mation must come from somewhere and Chomsky argues that it is innate. There is one crucial question: do children produce grammatical rather than ungrammatical utterances when turning statements into questions? Crain and Nakayama (1986) conducted an ex- periment in which three-year-olds were required to ask a Jabba the Hut puppet whether, for instance, the dog who is wearing the blue collar likes playing fetch. In the majority of cases, children were able to turn such statements into grammatical questions, and never made errors such as those in sen- tence 4 where the incorrect ‘is’ belonging to the noun phrase is moved to the front of the sentence. Other evidence shows that babies are capable of understanding complex syntactic struc- ture. Katherine Hirsh-Pasek and Roberta Golinkoff (1991) showed that babies between 13 and 15 months of age who were in the one-word stage of language production could nevertheless discrimi- nate between similar sounding but syntactically distinct phrases. The infants were seated in front of two televisions, each of which showed a film in which a pair of adults dressed up as Cookie Mon- ster and Big Bird from television’s Sesame Street performed simple activities. A voice-over said, ‘Oh look! Big Bird is washing Cookie Monster! Find Big Bird washing Cookie Monster!’ (or vice versa). Careful measurement showed that the babies spent reliably longer looking at the screen that matched the action rather than the other (foil) screen that showed a different but closely related ac- tion (for instance Cookie Monster is washing Big Bird). This shows that even before infants are able to string words together, they are able to understand the different roles of subject, verb and object. The Universal Grammar Chomsky argues that language learning is supported by a language organ which contains knowledge of the Universal Grammar which is the abstract specification that underlies all human languages. Languages differ syntactically in many ways, some put the verb between the subject of the sentence and its object (like English, which is therefore known as a Subject Verb Object or SVO language), others put the verb first and then the subject and then the object (such as Gaelic), whereas still others put the subject then the object followed by the verb (German is such an example). Table 10.1 shows the proportion of each of the logically possible six orderings of subject, verb and object in a sample  260 The Evolution of Language Table 10.1 Proportion of languages adopting each of the six logically possible word orderings from a sample of 402 of the world’s languages Word order Number of languages % Languages SOV 180 45 SVO 168 42 VSO 37 9 VOS 12 3 OVS 5 1 OSV 0 0 Total 402 Source: Tomlin (1986). OVS and OSV being extremely rare. Notice also that in 96 per cent of this sample the subject is placed before the object. Languages vary in other ways. English, for instance, is fussy about the order in which words appear in sentences; change the word order and you change the meaning of the sentence. Other languages such as Latin and some Australian aboriginal languages such as Warlpiri are more relaxed about word order, instead who’s doing what to whom is conveyed by the use of special affix- es (sounds that are added on to the end, beginning or in the middle of words). Although languages vary significantly in their syntax there are certain abstract properties that all languages share. Languages differ in the orders of subject, object and verb, but all languag- es use these syntactic components; all languages have structures which perform the function of nouns, verbs, adjectives and prepositions; all languages have other features such as a topic and a head. These similarities exist, according to Chomsky, because they are part of the newborn’s innate knowledge of language contained in the Universal Grammar. How do children learn the particulars of the language spoken around them? Chomsky pro- poses that the child has a group of mental switches – called parameters – that are ‘flicked’ as a result of linguistic experience and it is this parameter setting that makes language learning so fast. As a simple example, a child’s universal grammar tells them that there are subjects and objects, but there is a free parameter which allows SVO, VSO or SOV (for example). If the child’s linguistic environment were English then, as a result of exposure to language, the parameter would be set to SVO; on the other hand, if the child were growing up in a German-speaking environment, it would be set to SOV. Chomsky and his followers argue that even when children make grammatical errors many of these are not simply random. It is proposed that many errors made by English children would be perfectly grammatical in some other language. Some errors made by young children in English have been shown to be perfectly grammatical in German. What is happening here? Remember that the Universal Grammar contains the abstract specification for all human languages that exist, have existed and (presumably) some that have never existed as well. The Universal Grammar enables the child, based on impoverished input, to induce the correct rules of the particular language that he or she is hearing and thereby become a member of the linguistic community. The above is a much-simplified account of a complicated theory and there is much debate, even among Chomskyians, as to how the process of parameter setting actually works, and to what extent parameters can be unset (say if the child moves to another linguistic environment). In fact, it appears that children are quite flexible in this regard and are able to acquire new languages with  Acquiring Grammar 261 Chomsky and Evolution Although Chomsky argues that language learning is supported by innate psychological mecha- nisms, he was originally reluctant to accept that these mechanisms were produced by natural se- lection. Rather, he suggested that the language faculty evolved for some other purpose and was co-opted for its current purpose. More recently he has relented and begun to address which parts of the language faculty might be specific to language and thus evolved to support language alone, and those which might have been co-opted (we discussed the notion of domain specificity in Chapter 5). Chomsky and co-workers (Hauser et al., 2002; 2014) expand upon this idea by delineating the ‘faculty of language in the broad sense’ from ‘the faculty of language in the narrow sense’. They refer to these using the initials FLB and FLN but as abbreviations often get in the way of under- standing we will refer to them as ‘language in the broad sense’, and ‘language in the narrow sense’, and variations thereof. Language in the broad sense refers to competencies and processes that are involved in lan- guage but are not necessarily specific to language and may even be shared with other species. These include perception and articulation for both auditory and signed language processing and ways of de- riving meaning from these perceptions. As we have seen, prairie dogs and vervet monkeys have some of these abilities. Furthermore, other non-linguistic aspects of cognition, such as memory, and vision in humans, might also make use of processes that are part of the broad sense of the language faculty. The narrow sense of the language faculty includes only those processes that are specific to humans and specific to language. One possible candidate for a cognitive operation that is specific to language is recursion. Recursion in its simplest sense is a way of embedding one item within anoth- er as many times as one wishes. If you have ever stood between two mirrors facing each other, you will be familiar with something similar. One mirror reflects your image; this image is then reflected by the second mirror which is then reflected by the first again and so on and so forth to infinity. Images of yourself are being ‘embedded’ within larger images of yourself. A similar thing is shown Figure 10.5 A Mandelbrot set. Each of the smaller patterns around the edges is a smaller version of the central shape. As you zoom in more, the pattern keeps repeating. 262 The Evolution of Language in the Mandelbrot set (Figure 10.5) were we to zoom into the image we simply see the larger image repeating itself (it might be worth searching the internet for a video to get the full effect). Recursion occurs in language too. For example, we can embed clauses within other clauses (or comprehend) embedded clauses. We could say: ‘Throw the spear at the mammoth’ Or, to be more specific, if there is more than one mammoth: ‘Throw the spear at the mammoth with the large tusks’ Or, even more specifically, e.g. when many have large tusks: ‘Throw the spear at the mammoth with the large tusks that is running towards you’. Each additional clause (e.g. ‘with the large tusks’) adds information as to which mammoth one should throw the spear at. This is one of the important combinatorial features of grammar, as we saw above. Some books and research articles will tell you that Chomsky’s suggestion that recursion is central and specific to language has been disproven. The linguist Daniel Everett reported that a people in the Amazon region of Brazil, known as the Pirahã, have a language which does not show recursion. Whether this is true or not, it does not actually refute Chomsky’s theory as he was trying to explain how some languages can have recursion, he was not suggesting that recursion is obliga- tory. If the Pirahã language can get away without using recursion, then that’s fine. In fact, those who have studied the Pirahã language in more detail have found that it does show recursion. In just one example (Salles, 2016), a Pirahã speaker was heard to say: ‘Kapoogo’s canoe’s motor is big’ in which a possessive (‘canoe’s’) is embedded within another possessive (‘Kapoogo’s’) showing recursion, just like the way ‘big tusks’ was embedded within ‘mammoth’ above. Hauser et al. leave it as an open question whether recursion is the only feature of the narrow sense of the language faculty (there could be others), and also whether recursion is really specific only to language. They raise as a possibility that recursion could have been developed for spatial navigation and was co-opted for language use. So far there is no evidence that this is the case, although if homologous (same feature by common descent) examples of recursion can be found in non-human animals it would provide evidence for the notion that recursion was co-opted. If this were the case would it suggest that language did not evolve but was merely co-opted? In a review of Hauser et al.’s paper, Jackendoff and Pinker (2005) argue that even if it were true that all the underlying cognitive operations of language were co-opted, it would still not deny that language evolved. Natural selection usually works by modifying what is already present rather than creating new structures. Perhaps the most famous example is that the ossicles of the mammalian ear (the hammer, anvil and stirrup) evolved from bones that originally functioned as part of the reptilian jawbone. But the current adaptive function of the ossicles is clearly to transmit sound vibrations in order to enable auditory perception, the fact that they might have, at one point, served a different function is largely irrelevant to the debate. Whether co-opted or not, many now accept that language involves a complex design of interacting parts that surely could not have evolved by accident. But if language is ‘in the genes’, where are these genes? What progress has been made in identifying how the language faculty is specified in the genome? We will return to this point after first evaluating Chomskyian theory.  Alternative Evolutionary Accounts of Language Acquisition 263 Evaluation of Chomskyian Theory In evaluating Chomsky’s contribution to the evolution of language, we must carefully delineate sev- eral claims. The first claim is that language acquisition would be impossible if it were to simply rely on species-general learning mechanisms (like those proposed by Skinnerian mechanisms) such as the forming of associations, reinforcement and punishment; this is the learnability argument outlined above. The second claim is that extra information is provided to assist the process of language learn- ing in the form of species-specific structures and processes: a universal grammar of some sort. The third is that language acquisition is facilitated by the specific principles and parameters proposed in Chomskyian theory such as the exotically named ‘trace erasure principle’, the ‘ergative case param- eter’ and the ‘nominal mapping parameter’ (none of which you need to worry about, unless you wish to study this in more depth). Whereas practically all psychologists and linguists agree with the first claim – the evidence both empirical and logical is overwhelming – there is, however, more debate with regard to the specific nature of this extra information. Whether, for example, our minds contain an ab- stract representation of grammar that is gradually fleshed out by a process of parameter switching, or, as some have argued, language is acquired by statistical extraction of linguistic regularities (see next). One of the great problems with evaluating Chomsky’s theory, is deciding on which theory to evaluate as Chomsky has changed his theory many times since the 1950s. The most prominent feature of these changes is the gradual stripping away of mechanisms deemed specific only to lan- guage. As we saw in the previous section, there is now maybe only one process specific to the aforementioned recursion. As in many areas of human inquiry, the language acquisition camp tend to cleave into those who consider themselves to be Chomskyian and those who are anti-Chomskyians. This is unfortu- nate. Scientific theory should not be treated as dogma, one should not be forced to either accept it or reject it wholesale the way one might do with a religion or a political ideology. On the contrary, we should explicitly reject such a temptation as unscientific; we should accept those parts of the theory that explain the relevant phenomena and are supported by the evidence and reject those parts of it that do not and are not. Alternative Evolutionary Accounts of Language Acquisition There are a number of theories that propose that language evolved through a process of natural selection but eschew the argument that language is acquired through the action of innate, domain- specific, mental modules espoused by Chomsky, Pinker and others. Evolutionist Michael Tomasello (see Tomasello, 1999) proposes that language is like any other cultural artefact handed down by our ancestors and needs no domain-specific language organ in order to explain its acquisition. Instead, he proposes that species-specific cognitive, social cognitive and cultural learning processes can account for language learning (see Chapter 14). Unlike other domain-general theories of language learning, such as that proposed by behaviourists, in which children passively acquire language through experience, Tomasello proposes that language learning is the result of children actively attempting to understand adult communication in a context of attention sharing. Although grounded in evolutionary theory, Tomasello argues that the principal difference between humans and non-hu- mans is humans’ ability to identify with their conspecifics (i.e. theory of mind; see Chapter 5), and it is this that enables language learning. Tomasello argues that Chomsky and Pinker have underplayed the role of imitation in lan- guage learning, and overplayed children’s ability to form general rules aided by innate knowledge 264 The Evolution of Language such as the Universal Grammar (Tomasello, 2005). Central to this view is the creativity enabled by inductive processes and rule formation. As we shall see when we discuss the Wug test, three-year- olds are able to inflect novel nouns to make them plural. Some evidence by Akhtar and Tomasello (1997) suggests there are severe boundaries to this creativity. Given a sentence like ‘The ball is getting dacked by Ernie’ (where ‘dacked’ is a novel verb) and asked ‘What is Ernie doing?’ children find it difficult to give the correct answer (dacking); instead they tend to stay close to the form of the verb that they heard. Tomasello argues that this is because children’s language is a lot less to do with abstract rules than Chomsky and others believe. He suggests that children acquire schemas such as those called verb islands which consist of a verb familiar to the child and one or more slots that can take nouns. So a child might have a representation ‘ ____ kicked ____’ with the slots on either side able to take a number of nouns with which the child is familiar. This endows flexibility, but – at least for young children – the verb at the centre of the construction cannot be changed. Tomasello presents some interesting evidence for his theory, but so far it has not yet been used to explain many of the linguistic anomalies that led Chomsky to propose a language-specific learning mechanism (such as the ease with which children turn statements into questions mentioned above). The idea is intriguing but it needs a thorough examination before becoming mainstream. Consistent with Tomasello’s theory is that people with autism – who have impaired theory of mind (see Chapter 5) – show language delay and frequently never develop language to the level of clin- ically normal individuals. However, sufferers from the related condition Asperger’s syndrome may also show substantial impairments in theory of mind but have few difficulties in their ability to learn language. Conclusion: Theories of Language Acquisition We have spent a lot of words discussing Chomsky for two reasons. First, his theory (in fact that should be theories; as we saw above, there is more than one) is probably the most familiar to psy- chology students. And second, it is the most explicit and detailed theory of language learning (which is by no means to say it is correct). Tomasello’s theory sees language emerging out of more general social abilities such as the- ory of mind and exchange relationships and is consistent with the way children learn some aspects of verb use. What is needed is for someone to put these theories to the test by creating sophisticated computational models to demonstrate their workability. So far this has only happened to a limited extent in narrow areas of language. For example, Steven Pinker wrote a 400+ book on the topic, focusing on verb syntax alone. Like many areas of psychology, much of the really hard work has yet to be done. The Search for Language Genes Recent research on a language disorder known as specific language impairment (SLI) has been widely touted as evidence for the genetic basis of the language organ. Specific language impair- ment, as the name implies, is a disorder that targets certain aspects of language production – particu- larly certain aspects of grammar – with no obvious profound sensory or neurological impairment (Bishop et al., 1995). Speech is effortful and often unintelligible with many word order and other grammatical errors evident. The KE family from Birmingham in the British Midlands were studied in the late 1980s by psycholinguist Myrna Gopnik and colleagues (e.g. Gopnik, 1994; Gopnik and  The Search for Language Genes 265 Crago, 1991; Crago and Gopnik, 1994). Sixteen out of 30 members of the extended family were affected by a particularly severe form of SLI leading them to produce sentences such as: ‘It’s a flying finches they are’. ‘Carol is cry in the church’. ‘A Patrick is naughty’. As the quotations above emphasise, SLI sufferers seem to have particular problems with inflectional morphology. Inflectional morphology is the process whereby words are modified to indicate gram- matical features such as number (add ‘s’ to form a plural noun), and tense (add ‘ed’ to make a past tense verb), and so on. Gopnik (1994) tested sufferers on a test known as the ‘Wug’ test. We know from extensive research that children of 3 and 4 years of age are able to inflect novel nouns for (among other things) plurality (Berko, 1958). In the Wug test the child is given a picture of an unfamiliar object that is given a novel name such as Wug, and then shown a further picture representing two such objects and told ‘Now there are two of them, there are two___’ and the child is encouraged to inflect the name to produce the plural, Wugs (see Figure 10.7). Box 10.2 Can Non-human Animals Be Taught Language? Some researchers have tried to teach animals something akin to human language: an area of re- search that has excited much debate and controversy. The research has normally focused on the attempt to teach non-human primates, usually common chimps but also bonobos (pygmy chimps; see Chapter 4) and gorillas, non-verbal languages such as sign language or specially developed languages using pictographic tokens. Spoken languages were abandoned after it was realised that the vocal systems of non-human primates are incapable of making the range of sounds necessary for a properly verbal language. What is at stake here is a number of theoretical issues but most importantly for this chapter, whether Chomsky is correct in his arguments that language is the result of an innate language organ (animal language researcher Herbert Terrace even provoca- tively named one of his chimps Nim Chimpsky). If other animals can be taught language, then the most likely conclusion would be that innate knowledge is not essential for language learning. Given that non-human primates don’t normally use language, it would be somewhat wasteful for them to have a language organ kicking around doing nothing (or waiting for human scientists to attempt to teach them language), and natural selection punishes profligacy mercilessly. Early attempts to teach primates language include Sarah (Premack and Premack, 1972), who com- municated using plastic chips containing symbols, and Washoe (Fouts et al., 1984), who was taught a version of American Sign Language (ASL). Sarah and Washoe were quite successful in comprehending and producing ‘utterances’. For example, Washoe produced utterances such as ‘baby in shoe’ and ‘open hurry’ (when standing outside a door). However, these examples of creative language use are often outnumbered by meaningless utterances and word-salad. Human infants make mistakes too, but far fewer than would be expected if they were merely combining words at random (Brown, 1973). More recent attempts have met with more success. In particular a team led by Sue Savage-Rumbaugh claim to have successfully taught non-human primates the rudiments of language. Of particular interest is Kanzi, a bonobo whose adoptive mother Matata 266 The Evolution of Language Box 10.2 (cont.) was undergoing language training using a keyboard consisting of symbols known as lexigrams. Kanzi showed little interest in the keyboard during Matata’s training but later, after Matata had been sent off for a period of time, Kanzi showed that he could use most of the ten lexigrams that had been on Matata’s keyboard. This was surprising, since usually it takes many hours of special training for apes to learn the lexigrams. Kanzi proved himself a prolific learner and can now produce over 200 words and is able to ‘comprehend’ over 500. He is also able to carry out instructions correctly such as ‘give the pine needles to Kelly’ (Savage-Rumbaugh et al., 1993). The animal language controversy polarises researchers like few other debates in psychology. Its supporters claim that although the ‘languages’ learned by non-human primates are nothing like as complex as those spoken even by young children, the difference is one of degree not type. Many other linguists and cognitive scientists argue that impressive though the animal feats are, they are really little more than clever parlour tricks and bear little resemblance to real languages (Pinker, 1994). Figure 10.6 Sue Savage-Rumbaugh and Kanzi. Here Kanzi is learning to make a campfire and is using lexi- grams to communicate. Suppose, for a moment, that this research was successful and animals were trained to use language; would this undermine Chomsky’s position? Successfully training an animal to use  The Search for Language Genes 267 Box 10.2 (cont.) language might be taken as evidence that a language organ was unnecessary for language learn- ing. Presumably natural selection has not equipped languageless animals with such a faculty in the chance that it will be needed at some point in the future. However, it is worth bearing in mind that apes are typically subjected to an exhaustive regime of language training, each being present- ed with trail after trail of carefully constructed paired associates with teams of researchers taking it in turns to give them instruction. This is very different from the messy and incomplete language curriculum human children experience. Under such circumstances the learnability argument does not apply since language is not being learned in the same way that children learn language. How should evolutionists treat this research? With some ambivalence. At one level evolu- tionary psychologists are interested in the kinds of things that organisms do in the wild. This is one of the great differences between an evolutionarily inspired discipline such as ethology and behaviourism which pays little lip service to Darwinism. It is surely interesting that chimps and gorillas can show aspects of linguistic competence (assuming that they can – which many believe is a big assumption) but the training regime required to instil this provides good evidence that such animals were not ‘designed’ for language (or at least not a language like human language). As psycholinguist and philosopher Jerry Fodor points out: ‘That a dog can be trained to walk on its hind legs does not prejudice the claim that bipedal gait is genetically coded in humans. The fact that we can learn to whistle like a lark does not prejudice the species-specificity of birdsong’ (Fodor et al., 1974). Gopnik’s research demonstrated that even adult sufferers of SLI find the Wug test demand- ing. Not only do they frequently fail to use the correct suffix, when they do, it is frequently mispro- nounced to produce words such as ‘Wugs’ (where the final phoneme is an unvoiced ‘s’ as in ‘buss’ rather than a voiced ‘z’). Figure 10.7 A sample question from the Wug test. 268 The Evolution of Language FOXP2: The Language Gene? Certain forms of SLI appear to be genetic in that they run in families (as is the case for the KE family). The pattern of inheritance suggests that it is the result of a single dominant gene on an autosomal (i.e. non-sex) chromosome (see Chapter 2). More recent research suggests that the gene responsible for SLI lies on chromosome 7 (the same chromosome where the Williams syndrome deletions occur; see Chapter 5). Specifically it appears that the damage is in region 7q31 in an area comprising 70 different genes that was called SPCH1 (Fisher et al., 1998). Further research (Lai et al., 2001) suggests that the crucial gene seems to be one known as FOXP2. FOXP2 is not unique to humans but the specific version present in humans is subtly different from those present in other animals. Comparative genetic research by Enard et al. (2002) suggests that the gene reached its current state 100,000–200,000 years ago. Even in non-humans different FOXP2 variants still seem to play a role in communication. Mice who have had FOXP2 removed (known as ‘knock out’ mice) show a decrease in ultrasonic vocalisations (Castellucci, McGinley and McCormick, 2016). It has also been implicated in song learning in zebra finches (Heston and White, 2015). More recent research has capitalised on the ability to use DNA samples from extinct spe- cies, specifically Homo neanderthalensis or Neanderthals for short (see Chapter 3). This research is particularly exciting as it enables us to better understand our closest relative and ourselves. The research suggests that Neanderthals possessed an almost identical variant of FOXP2 as modern humans, further reinforcing the notion that Neanderthals possessed language of some form. Other research suggests that rather than FOXP2 reaching its current state before the anatomically modern human and Neanderthal lineage split from one another, the genetic evidence can be better explained by assuming gene flow between humans and Neanderthals: in other words the two groups interbred (Coop et al., 2008; see also Chapter 3). If this is the case, and it is by no means certain that it is, then it suggests that Neanderthals ‘got’ the FOXP2 gene from humans rather than the other way around as Neanderthal DNA is not present in indigenous Africans but FOXP2, and of course language, is. All of this raises the question – is FOXP2 the language gene? That’s an easy one: it is not. We have known for a long time that complex traits are likely to be caused by many genes, that is they are polygenic. We used to think maybe a few tens or hundreds of genes, but genome-wide associa- tion studies (see Chapters 6 and 13) are now suggesting that complex traits are the result of many thousands or even tens of thousands of genes (other candidates for language include the inscrutably named CNTNAP2, ASPM, MCPH1, PCDH11X and PCDH11Y). So FOXP2 isn’t the language gene, but it is a language gene, right? Well again, not really, because this assumes the genome is like a blueprint and you can trace a particular behaviour back to the gene that produced it in a straightforward way. As we saw in Chapter 2, a better metaphor than the genome-as-blueprint is to think of it as a recipe for a sponge cake in which many ingredients interact with each other and the environment (e.g. the process of mixing the cake and cooking it). In this metaphor you cannot trace one part of the cake back to an individual ingredient. It is a better metaphor, but still not quite right. It is possible to trace certain qualities of the cake back to individ- ual ingredients. You can attribute the sweetness of the cake to the sugar, its richness to the butter, its lightness to the baking powder. This is not the case in genetics: the more we learn the more complex and convoluted the path between genotype and phenotype becomes. As an example, a gene is still identified as a region of DNA that codes for a particular protein (see Chapter 2). However, we now know that the same stretch of DNA can code for different kinds of protein depending on (among other things) which kind of cell it is in; a process called alternative splicing. FOXP2 is also a special  The Search for Language Genes 269 kind of gene which controls whether or not other genes are expressed, and its effects are felt in many parts of the body. The brain, obviously, but also the lungs and gut lining. So to return to our question. FOXP2 is not the language gene. It is somehow something to do with language, or at least speech. It is certainly implicated in language, maybe importantly im- plicated but then so are many, many other thousands of genes. And all of these genes are responsible for many other things too, many of them nothing to do with language. What genetics is teaching us is that we need to change our expectations about the link between genes and behaviour. When Did Language Evolve? Language doesn’t fossilise so it is hard to say with complete certainty whether a particular ancestral hominin did or did not possess language, and to what degree of sophistication. However, evidence from the anatomy of early hominins and the types of cultural artefacts that they produced goes some way towards helping us to answer this question. It used to be thought, for instance, that Neanderthals (discussed earlier) present in Europe around 200,000 years ago did not have language, despite them having larger brains than modern humans. The reasoning for this was that a crucial bone, the hyoid bone, was located too high up the vocal tract to enable certain vowel sounds to be produced. (The hyoid bone acts as a brace for the tongue and larynx giving a greater range of vocal expression than would be possible were it not present.) However, a recent discovery found a Neanderthal with the hyoid bone in a suitably low position to enable such vowels to be produced. This and the recent find- ing that Neanderthals possessed a similar variant of the FOXP2 gene (see above) as modern humans suggests, perhaps, that Neanderthals did possess language. Figure 10.8 The hyoid bone (in red) is important for speech. 270 The Evolution of Language Box 10.3 Language Development and Life History Approach The development of language is traditionally seen as being one of a progression towards lin- guistic competence: children’s vocalisations start off as prelinguistic coos and wails, progress through language-like babbles to one-word then two-word utterances, and then as vocabulary expands and grammatical complexity increases their speech becomes ever more adult-like. Viewing children as universal novices, and childhood as merely an apprenticeship to adult- hood, has had a long history in developmental psychology, but has recently been challenged by some evolutionists. In Chapter 6 we saw how some researchers have adopted a life history ap- proach to development. These researchers argued that children’s behaviour should be understood in two ways: (1) as an attempt to maximise their own survival and (2) as an attempt to increase their reproductive fitness when adults. We gave the example of a caterpillar which has its own unique morphology compared to that of the adult (meaning we cannot merely study caterpillars as miniature butterflies), but are nonetheless trying to maximise their reproductive fitness when adults (by piling on the calories and so becoming a fit and healthy butterfly). A recent paper by Locke and Bogin (2006) suggests that we might likewise view language development in the same way. Rather than seeing child language as merely a lesser form of adult language they ask what a child’s language does for the child and how this might influence their reproductive chances as adults. Locke and Bogin argue that Homo sapiens has a more complex life history involving more stages than our primate relatives. Whereas chimpanzees have infancy, a juvenile and an adult period we have infancy, childhood, a juvenile, an adolescent period and an adult period. They explain these two extra periods (childhood and adolescence) as specific adaptations, which are part and parcel of the evolution of language. In a sense they are claiming that these two extra periods are there to support language – in another sense they are claiming that language is the very reason we have all of these stages. The graph below compares the length of each developmental stage for humans and their ancestors; for example the average age at which the eruption of the first permanent molar is shown (this is taken as marking the end of childhood and the beginning of juvenility in hu- mans). It can be seen that humans have a relatively long period of childhood and a period of adolescence. Locke and Bogin claim that one reason for these two extra stages is to allow for linguistic competence to develop. In terms of language, childhood is a time of engaging in new extrafa- milial friendships. By age six the child is physically able to produce the adult range of vowel sounds. This is a period when complex games are learned and there is verbal competition. There is also a growth in verbal creativity and sex differences emerge in how language is used. Boys tend to speak more assertively and get more attention; girls tend to speak more softly and become more involved in interpersonal relationships. Perhaps the most important development is that of displacement – i.e. the ability to talk about things that are not present. So Locke and Bogin see childhood as an important development in a number of ways that allows greater linguistic ability to become apparent and ultimately increases both direct and indirect fitness. Later on, adoles- cence, a period where individuals are sexually mature but still not fully adult-like in physical stat- ure and socioemotional competence, is seen by the authors as a stage where language is used for  The Search for Language Genes 271 Box 10.3 (cont.) intra-sex competition (male–male, female–female) and courtship in preparation for adulthood. In a nutshell, by slowing down development and adding two stages to it, Homo sapiens achieved greater verbal competence and hence a selective advantage. Table 10.2 shows the key character- istics of language at each of these stages and the proposed function. The different stages of development for a number of hominins. P/A = Pan and Australo- pithecus aferensis (e.g. Lucy), Aa = Australopithecus africanus, Hh = Homo Habilis, He1 early Homo erectus, He2 late Homo erectus, Hs = Homo sapiens (modern humans). M1 erupt is the average age at which the second set of teeth erupt. Table 10.2 The different stages in human development according to Locke and Bogin Stage Approximate Starts at Language Function duration characteristics (there are considerable sex differences) Infancy 0–36 months birth coos and babbles, first to increase parental words investment by engaging parent Childhood 36 months to eruption of fluency increases, creating and 6 years deciduous teeth self-referential maintaining independence from parents Juvenility 6–10 years eruption of first pragmatic advances: competition between permanent molar gossip, storytelling. group members; verbal ‘duels’, fostering of alliances particularly in males and friendships Adolescence 10–17 years adrenarche increase in complexity; develop social (maturation of use language to networks and adrenal glands, compete with others, explore romantic associated with impress the opposite relationships onset of puberty) sex (esp. for males) Adulthood 17 years + adult stature much of the above still develop and and social skills present maintain networks, develops foster romantic relationships, instruct children The different stages in human development according to Locke and Bogin (2006). 272 The Evolution of Language Box 10.3 (cont.) Brain size (cc) 400 442 610 826 983 1350 20 18 Infancy 16 Childhood 14 Age in years 12 Juvenile 10 Adolescent 8 6 Adult 4 M1 erupt 2 3.1 3.1 3.8 4.5 5.0 6.2 0 P/A Aa Hh He1 He2 Hs Figure 10.9 The different stages of development for a number of hominins. P/A = Pan and Australopithecus aferensis (e.g. Lucy), Aa = Australopithecus africanus, Hh = Homo Habilis, He1 early Homo erectus, He2 late Homo erectus, Hs = Homo sapiens (modern humans). M1 erupt is the average age at which the second set of teeth erupts. Human brain size began to expand relative to body size some 2 million years ago (Boyd and Silk, 2000; Deacon, 1997; see also Chapter 2) and it is possible that this expansion coincid- ed with the evolution of language. Some have argued that Homo erectus – the ancestors of the Neanderthals and Homo sapiens – which appeared in Africa nearly 2 million years ago – might have had the capacity for some language. Wynn (1998) observes that endocasts (casts taken of the inside of the cranium) of Homo erectus suggest that two brain regions thought to be crucial for language – Wernicke’s area and Broca’s area (see Figure 10.3) – are similar to those in modern humans. Other evidence, however, suggests that Homo erectus might not have had sufficiently fine control over breathing to enable a full spoken language (Wynn, 1998; MacLarnon and Hewitt, 1999). Exactly when language evolved is still uncertain, but even if Homo erectus were not ca- pable of a language as sophisticated as that of modern humans, it remains a possibility that they had primitive language skills. The wings of an archaeopteryx might not have enabled the delicately controlled flight of a modern bird, but they permitted a form of gliding flight good enough to in- crease their fitness. Maybe the language of Homo erectus was primitive but good enough to confer an advantage over conspecifics with poorer or non-existent language.  The Search for Language Genes 273 The Evolution of Languages At the time of writing there are 7117 languages in use today (‘in use’ rather than ‘spoken’ because this number includes 144 sign languages). But this 7117 is precarious: in the same way that there are many threats to biodiversity there are threats to linguistic diversity. For example, it has been estimated that, on average, a language becomes extinct every 14 days with approximately 40 per cent of languages currently under threat (Rymer, 2012). Given these statistics, it is likely that in the recent past the number of languages may have been much higher. It is perhaps astonishing to consider that all this diversity is the result of a single act of creation by one of our ancestors. Or put more precisely, a particular combination of genes that began the story of language. What Steven Pinker has called, the big bang. Pinker was speaking metaphori- cally, of course; as a good evolutionist he knows that natural selection seldom, if never, proceeds in giant leaps, baby steps tending to be the order of the day. In a similar way that paleoanthropologists have made attempts to construct the family tree of Homo sapiens (see Chapter 2), so comparative linguists have attempted to recreate the evolution of the languages spoken today. Figure 10.9 shows a schematic representation of thinking in this area. This is only part of a tree representing the Indo-European language family which includes most of the languages traditionally spoken in Europe and the Asian sub-continent. This reconstruction was made possible by a discovery in 1786 by Sir William Jones, a British judge stationed in India. Jones was studying Sanskrit, the ancient language of India, and noticed that there were marked similarities be- tween this language and ancient Greek and Latin (and also other modern languages). See Table 10.3. Jones’ conclusion was that Sanskrit, Latin and Greek must have all had an even older common ancestor, a language that became known as Proto-Indo-European (PIE). It seems that the speakers of PIE spread throughout most of Europe and Asia, leaving perhaps only a few islands of non-Indo-Euro- peans (Basque and Finnish, for example, are not Indo-European languages). British archaeologist Colin Renfrew (1987) proposed that the Indo-Europeans originated in Anatolia (Modern Turkey) in the Fertile Crescent around 7000 BC. They were farmers (the Fertile Crescent is thought to be the birthplace of ag- riculture) and therefore were able to reproduce more rapidly than the hunter-gatherers who presumably inhabited Asia and Europe at the time (see Chapter 14). It is also more than likely that the Indo-Europe- ans mated with these hunter-gatherers, absorbing them into their advanced culture (see Figure 10.10). Table 10.3 Sanskrit compared to other Indo-European languages ancient and modern English mother three me brother Sanskrit matar tri me bhrator Latin mater tres me frater Italian madre tre me fra Spanish madre tres me hermano French mère trois moi frère Greek meter treis me phrater Dutch moeder drie mij broeder German mutter drei mich bruder Norwegian mor tre meg bror Lithuanian mater tri manen brothar Celtic mathair tri me brathair  274 The Evolution of Language PROTO-INDO-EUROPEAN Old Prussian Tocharian A Lithuanian BALTIC TOCHARIAN Tocharian B Latvian Dard Wendish SLAVIC Dardic INDO-IRANIAN Polish Sindhi West Slavic East Slavic Romany Slovac Russian Indic Czech South Slavic ANATOLIAN Urdu Byelorussian Hindi Ukrainian Bihari Slovene Old Church Slavonic Sanskrit Assamese Serbo–Croatian Bulgarian GERMANIC Hittite Bengali Macedonian North Luwian Marathi ARMENIAN Icelandic Germanic Lydian Gujarati Faroese Old Norse Lycian Punjabi Norwegian Iranian Sinhalese Swedish Old Swedish West Danish Old Danish Germanic English Pashto Old English East HELLENIC Frisian Old Frisian Baluchi Germanic Dutch Middle Dutch Old Dutch ITALIC PHRYGIAN Avestan Kurdish Flemish Gothic CELTIC Sogdian Old Low German ALBANIAN Greek Old Persian Afrikaans Low German Old High German Pahlevi German Goidelic Persian Yiddish Irish Gaelic Brythonic Scottish Gaelic Welsh Manx Cornish Osco-Umbrian Breton Gaulish Oscan Umbrian Latino-Faliscan Faliscan Uncertain Placement: Portuguese Latin ILLYRIAN Spanish THRACIAN Catalan Provencal Romanian French Italian Rhaeto-Romance Figure 10.10 The descent of language, the Indo-European family tree. What about all the other language families? Opinion differs as to the identity of the other language families, but some claim that there are seven families in addition to Indo-European. These are listed below: 1. Niger-Congo (1526 languages, 550 million speakers) languages spoken in central and southern Africa including Xhosa, Zulu and Swahili; 2. Austronesian (1227 languages, 326 million speakers) includes languages spoken in Indonesia, Malaysia, New Zealand and Taiwan; 3. Trans-New Guinea (477 languages, 650,000 speakers) languages spoken in Papua-New Guinea and parts of Indonesia; 4. Afro-Asiatic (366 languages, 500 million speakers) languages spoken in northern and central Africa including Ethiopian languages, Hebrew, Maltese, and the many forms of Arabic; 5. Sino-Tibetan (455 languages, 1.4 billion speakers) including languages spoken in Myanmar, In- dia, Nepal and China. Historical linguists such as the late Joseph Greenberg (1987) attempt to reconstruct the ancestors of all this diversity. By carefully analysing the words used in different languages he grouped together all languages apart from Sino-Tibetan and Oceanic into an ancestor language he called ‘Nostratic’  Why Did Language Evolve? 275 Box 10.4 Were Early Languages Signed Rather than Spoken? It is interesting to reflect that when many of us think of language, we tend to think of it as some- thing that is spoken. An obvious exception are the sign languages used by people with hearing impairments who preferentially use gestural communication. A common way of thinking about this is that the spoken tradition is the more natural way, and we only fall back on gestures when speech fails us. To some extent this is true. If children can hear speech they learn to speak in re- turn, if they cannot, and are given the opportunity to, they will learn to sign with equal deftness. For Michael Corballis (2020) this is not an accident, he proposes that the earliest human languages were, in fact not spoken but mimed. As evidence for this he presents the fact that the most successful attempts to teach non-human primates language have involved gestural commu- nication rather than speech (see Box 10.2). Outside the lab, although chimpanzees and bonobos make a number of distinct vocalisations, intentional communication seems to depend more on gestures. Chimpanzees have been seen to use 66 gestures conveying 19 distinct meanings such as ‘stop that’, ‘move away’, ‘initiate grooming’, ‘follow me’ and ‘move closer’ (Hobaiter and Byrne, 2014). There is further evidence from neurobiology too. The primate brain has an area in the parietal and temporal regions which responds to intentional hand movements. This is known as a ‘mirror system’ as it responds in exactly the same way if the primate being studied moves its own hand. A system in which input and output are represented in the same way is an ideal candidate for communication: if a third-person gesture meaning ‘food’ is represented in the same way as the first-person gesture would be, we can see how this might enable us to share simple thoughts. In humans these brain regions are now occupied by Broca’s and Wernicke’s areas (see Figure 10.3). There is some evidence, therefore, that the common ancestor of the chimp and the human may have principally communicated using gestures and only more recently, for reasons discussed earlier in this chapter, did humans evolve to preferentially use spoken language. (the word is from the Russian meaning ‘our language’). The remaining languages were grouped under the banner ‘Dene Caucasian’ which is the ancestor of Chinese, Tibetan, Burmese and several other languages of South and East Asia. The ultimate goal is to construct the common ancestor to these two language super-phyla. The problem with this enterprise is the sheer time scales involved. Attempts to recreate Proto-Indo-European around 6000 years ago are controversial enough but as we go further and fur- ther back in time to the earlier languages (past 10,000 years ago) tiny errors in our calculations can lead to us being way off the mark (Hurford, 2012). It is like throwing a ball at a target, if you’re close enough then a small error won’t make much odds. If you move progressively further away the same error means a miss by a mile. And of course, with ancient languages, there is no way of checking how good your aim was. Why Did Language Evolve? If language evolved as a result of natural selection, one obvious question is why did it evolve? What were the particular pressures that led to our linguistic ancestors being favoured over their 276 The Evolution of Language languageless conspecifics? First of all, it is important to re-emphasise that language, like the human eye, almost certainly did not evolve as a result of one huge mutation (a macromutation) that led us from having no language to having a full-blown language as we know it today. More likely, language evolved from non-linguistic vocalisations similar to those we see in non-human primates, gradually becoming more complex over the generations. In a way, the question is difficult because there are so many candidates, as biological an- thropologist Terrence Deacon (1997) notes: From the perspective of hindsight, almost everything looks as though it might be relevant for explaining the language adaptation. Looking for the adaptive benefits of language is like picking only one dessert in your favorite bakery: there are too many compelling op- tions to choose from. What aspect of human social organization and adaptation wouldn’t benefit from the evolution of language? (377) The aspect of language that language researchers have tended to focus on is information transfer, specifically in its use in coordinating hunting or for teaching other people new skills. Recently, how- ever, this view of what language might have been for originally has been challenged by a number of evolutionists, each of whom argues that language evolved for primarily social reasons. We discuss four such theories of the social origins of language next. Gossip and Language Evolution Evolutionist Robin Dunbar (1993; 1996) of Oxford University proposes that information transfer – although important – might not have been evolution’s original motivation for producing language. He argues that language evolved as a means of maintaining social bonds through gossip. Social gossip, he suggests, fulfils in humans a similar role to that provided by grooming in other primates: as a means of forming bonds between members of a community. This is part of a more general claim that intelligence evolved in order to maintain large group sizes; the more indi- viduals you live with the more demands are placed on you to keep track of each individual and of the relationships among them. To meet these cognitive demands, he suggests, there needs to be an increase in brain size. In support of this general claim, Dunbar (1993) plotted the ratio of neocortex (the part of the brain believed to be used for higher cognitive abilities, including language) to total brain volume against the mean size of their group for a range of non-human primates. Humans were not included at this point, since estimating the group size for humans is difficult – for instance it could range from a few tens of people in an extended family group to several million in a large city. The graph is shown in Figure 10.11 and it can be seen that there is a strong positive correlation between group size and neocortex ratio, suggesting that the social brain hypothesis might have some credibility. If the value for human neocortex size is plotted on the same graph we find that the predicted mean group size for humans is around 150 individuals. Dunbar provides some evidence that 150 is a good estimate of human group size. Many hunter-gatherers, for example, live in bands of between 100 and 200 (with a mean of around 150). (The size of hunter-gatherer groups seems, in fact, to be trimodally distributed with overnight camps of a few tens of individuals, and tribes being a few hundred to a couple of thousand individuals; bands are an intermediate grouping of people.) He also suggests that people in the Western world are able to call on an average of 150 individuals for help (if, for example, they needed to borrow money).  Why Did Language Evolve? 277 1.5 Striate cortex (V1) Non–V1 neocortex Linear (Striate cortex (V1)) 1 Linear (Non–V1 neocortex) Mean group size contrasts 0.5 0 –0.5 –1 –0.2 –0.15 –0.1 –0.05 0 0.05 0.1 0.15 0.2 Non-V1 neocortex and visual component contrasts Figure 10.11 Ratio of neocortex to group size in a number of non-human primate communities. So Dunbar’s argument is that one of the pressures that led to the development of the brain is the need to remember all of the people that we know and interact with them in an efficient manner. But what has this got to do with language? Well, it has been suggested by Dunbar and others (see Dunbar, 1993) that in non-human primates the social glue that holds a group together is provided by grooming. Many primates such as baboons and chimpanzees will groom each other putatively to remove ticks and other parasites from their bodies, but also as a means of forming and maintaining relationships. The amount of grooming that a primate needs increases in direct relationship to the size of the group of which the individual is a member, so much so that some primates can spend as much as 20 per cent of their daytime engaged in grooming activities. It has been estimated that if humans were to maintain group sizes of 150 individuals by using one-on-one grooming then they would need to spend around 50 per cent of their time groom- ing. This figure is too high for us to attend to all of the other daytime chores, so Dunbar argues that our ancestors developed a new form of social grooming – language. Language enables us to engage more than one person in conversation and we can also ‘groom’ while performing other activities such as hunting and gathering. Given that the average group size for non-human primates is 50 and that for humans is 150, it follows that if humans are to spend no more than 20 per cent of their time engaged in social grooming, language must be three times more efficient than one-to-one grooming. Research by Dunbar et al. (1995) indicates that the maximum number of conversational partners that is typically possible is around three. Dunbar presents some interesting evidence for his language-as-social-grooming hypothe- sis. Studies of language use suggest that it is most often used for gossiping and other social chitchat rather than for other means (such as giving instructions or information transfer for example). Dun- bar et al. eavesdropped on people’s conversations and calculated that between 60 and 70 per cent of the time conversations were of a social rather than pedagogical nature. Interestingly women were no more likely to engage in social conversation than were men. 278 The Evolution of Language EVALUATION OF THE SOCIAL GROOMING HYPOTHESIS Dunbar proposes that language evolved for two reasons (1) to exchange social information and (2) to enable individuals to interact with more people at any one time. Many linguists and psychologists would accept the first proposition that language evolved for primarily social reasons rather than, say discussing hunting techniques or transmitting information about the physical world. The second proposition is, however, more contentious. It may be the case that larger groups lead to more surviving offspring but it remains an open question as to whether it was the ability to form larger groups that originally gave our linguis- tic ancestors an advantage over non-linguistic hominins, or whether group size was merely a side effect of the evolution of language. In support of this last point, Derek Bickerton (2007) asks why it was necessary for something as complex and costly as language to replace grooming – which is, at root, a fairly simple activity. What is the purpose of syntax, semantics and the like, and couldn’t one have simply got away with making soothing, cooing vocalisations to facilitate social bonding? In other words, the selective pressure (the need for social bonding) does not specify the particular design features of language. The Social Contract Hypothesis In contrast to Chomsky and Pinker, who have tended to focus on syntax as the hallmark of human language, Deacon (1997) focuses on another aspect of language – that of reference. He argues that the key difference between human language and animal communication is that human languages use symbols (words) to refer to objects. He argues, therefore, that the main phenomenon that needs explaining is the shift from non-symbolic to symbolic communication. He suggests that the mechanism for such a shift is likely to be sexual selection rather than natural selection, ob