The Rise And Fall Of The Third Chimpanzee

The gulf between animal vocal communication and human speech has traditionally been viewed as unbridgeable. In fact, recent studies of animal vocalizations show some of them to be far more sophisticated than we had previously suspected. On the other hand, there are dozens of cases in which humans have been forced by exceptional social circumstances to create simplified languages, possibly illustrating two primitive stages in the evolution of human language. Thus, we are beginning to understand how our most unique and important distinction from animals nevertheless arose from animal precursors.

THE MYSTERY OF human language origins is the most crucial in understanding how we became uniquely human. After all, it is language that allows us to communicate with each other far more precisely than any animal can. Language enables us to formulate joint plans, to teach one another, and to learn from what others have experienced elsewhere or in the past. With it, we can store precise representations of the world in our minds, and hence encode and process information far more efficiently than any animal can. Without language we could never have conceived and built Chartres Cathedral – or V-2 rockets. For these reasons, I speculated in Chapter Two that the Great Leap Forward (the stage in human history when innovation and art at last emerged) was made possible by the emergence of spoken language as we know it.

Between human language and the vocalizations of any animal lies a seemingly unbridgeable gulf. It has been clear since the time of Darwin that the mystery of human language origins is an evolutionary problem: how was this unbridgeable gulf nevertheless bridged? If we accept that we evolved from animals lacking human speech, then our language must have evolved and become perfected with time, along with the human pelvis, skull, tools, and art. There must once have been intermediate language-like stages linking monkey grunts to Shakespeare’s sonnets. Darwin diligently kept notebooks on his children’s linguistic development, and reflected on the languages of ‘primitive’ peoples, in the hope of solving this evolutionary mystery.

Unfortunately, the origins of language prove harder to trace than the origins of the human pelvis, skull, tools, and art. All of the latter may persist as fossils that we can recover and date, but the spoken word vanishes in an instant. In frustration, I often dream of a time machine that would let me place tape-recorders in ancient hominoid camps. Perhaps I would discover that australopithecines uttered grunts little different from those of chimpanzees; that early Homo erectus used recognizable single words, progressing after a million years to two-word sentences; that Homo sapiens before the Great Leap Forward became capable of strings of words that were longer but still without much grammar; and that syntax and the full range of modern speech sounds arrived only with the Great Leap.

Alas, we have no such retrospective tape-recorder, and no prospects for ever getting one. How can we hope to trace speech origins without such a magic time machine? Until recently, I would have said that it was hopeless to do more than speculate. In this chapter, however, I shall try to draw on two exploding bodies of knowledge that may allow us to begin building bridges across the gulf between animal and human sounds, by starting from each of its opposite shores.

Sophisticated new studies of wild animal vocalizations, especially those of our primate relatives, constitute the bridgehead on the animal shore of the gulf. It has always been obvious that animal sounds must have been precursors of human speech, but only now are we beginning to sense how far animals have come towards inventing their own ‘languages’. In contrast, it has not been clear where to locate the bridgehead on the human shore, since all existing human languages seem infinitely advanced over animal sounds. Recently, though, it has been argued that a numerous set of human languages neglected by most linguists truly exemplifies two primitive stages on the human side of the causeway.

Many wild animals communicate with each other by sounds, of which bird-songs and the barking of dogs are especially familiar to us. Most of us are within earshot of some calling animal on most days of our lives. Scientists have been studying animal sounds for centuries. Despite this long history of intimate association, our understanding of these ubiquitous and familiar sounds has suddenly expanded because of the application of new techniques: use of modern tape-recorders to record animal calls, electronic analysis of the calls to detect subtle variations imperceptible to the unaided human ear, broadcasting recorded calls back to animals to observe how they react, and observing their reactions to electronically reshuffled calls. These methods are revealing animal vocal communication to be much more like language than anyone would have guessed thirty years ago.

The most sophisticated ‘animal language’ studied to date is that of a common, cat-sized African monkey known as the vervet. Equally at home in trees and on the ground in savannah and rainforest, vervets are among the monkey species that visitors to East African game parks are most likely to see. They must have been familiar to Africans for the hundreds of thousands of years that we have existed as the species Homo sapiens. They may have reached Europe as pets over 3,000 years ago, and they certainly have been familiar to European biologists exploring Africa since the Nineteenth Century. Many laypeople who have never visited Africa are still acquainted with vervets from visits to the zoo.

Like other animals, wild vervets regularly face situations in which efficient communication and representation would help them to survive. About three-quarters of wild vervet deaths are caused by predators. If you are a vervet, it is essential to know the differences between a martial eagle, one of the leading killers of vervets, and a white-backed vulture, an equally large soaring bird that eats carrion and is no danger to live monkeys. It is vital to act appropriately when the eagle appears, and to tell your relatives. If you fail to recognize the eagle, you die; if you fail to tell your relatives, they die, carrying your genes with them; and if you think it is an eagle when it is really just a vulture, you are wasting time on defensive measures while other monkeys are safely out there gathering food.

Besides these problems posed by predators, vervets have complex social relationships with each other. They live in groups and compete for territory with other groups. Hence it is also essential to know the difference between a monkey intruding from another group, an unrelated member of your own group likely to push you, and a close relative in your own group on whose support you can count. Vervets that get into trouble need ways of telling their relatives that they, and not some other monkey, are in trouble. They also need to know and communicate about sources of food: for instance, which of the thousand plant and animal species in the environment are good to eat, which are poisonous, and where and when the edible ones are likely to be found. For all these reasons, vervets would profit from efficient ways of communicating about and representing their world.

Despite these reasons, and despite the long and close association between vervets and humans, we had no appreciation of their complex world knowledge and vocalizations until the mid-1960s. Since then, observations of vervet behaviour have revealed that they make finely graded discriminations among types of predators, and among each other. They adopt quite different defensive measures when threatened by leopards, eagles, and snakes. They respond differently to dominant and subordinate members of their own troop, differently again to dominant and subordinate members of rival troops, differently to members of different rival troops, and differently to their mother, maternal grandmother, sibling, and unrelated members of their own troop. They know who is related to whom: if an infant monkey calls, its mother turns towards it, but other vervet mothers turn instead towards that infant’s mother to see what she will do. It is as if vervets had names for several predator species and several dozen individual monkeys.

The first clue to how vervets communicate this information came from observations that the biologist Thomas Struhsaker made on vervets in Kenya’s Amboseli National Park. He noted that three types of predator triggered different defensive measures by vervets, and also triggered alarm calls sufficiently distinct that Struhsaker could hear the differences even without making any sophisticated electronic analysis. When vervets encounter a leopard or any other species of large wild cat, male monkeys give a loud series of barks, females give a high-pitched chirp, and all monkeys within earshot may run up a tree. The sight of a martial or crowned eagle soaring overhead causes vervets to give a short cough of two syllables, whereupon listening monkeys look up into the air or run into a bush. A monkey who spots a python or other dangerous snake gives a ‘chuttering’ call, and that stimulates other vervets in the vicinity to stand erect on their hind legs and look down (to see where the snake is).

Beginning in 1977, a husband-and-wife team named Robert Seyfarth and Dorothy Cheney proved that these calls really had the different functions suggested by Struhsaker’s observations. Their experimental procedure was as follows. Firstly, they made a tape-recording of a monkey giving a call whose apparent function Struhsaker had observed (say, the ‘leopard call’). Then, on a later day, after locating the same troop of monkeys, either Cheney or Seyfarth hid the tape and loudspeaker equipment in a bush nearby, while the other started filming the monkeys with a cine or video camera. After fifteen seconds, one of the two scientists broadcast the tape while the other kept filming the monkeys for one minute to see whether the monkeys behaved appropriately for the call’s suspected function (for example, whether the monkeys ran up a tree on hearing a broadcast of the supposed ‘leopard’ call). It turned out that playback of the ‘leopard call’ really did stimulate the monkey to run up a tree, while the ‘eagle call’ and ‘snake call’ similarly stimulated monkeys into behaviour that seemed to be associated with these calls under natural conditions. Thus, the apparent association between the observed behaviour and the calls was not coincidental, and the calls did have the functions suggested by observation.

The three calls that I have mentioned by no means exhaust a vervet’s vocabulary. Besides those loud and frequently given alarm calls, there appear to be at least three fainter alarms that are given less frequently. One, triggered by baboons, causes listening vervets to become more alert. A second, given in response to mammals like jackals and hyenas that prey on vervets only infrequently, causes the monkeys to watch the animal and perhaps move slowly towards a tree. The third faint alarm call is a response to unfamiliar humans and results in the vervets quietly moving towards a bush or the top of a tree. However, the postulated functions of these three fainter alarm calls remain unproven because they have not yet been tested by playback experiments.

Vervets also utter grunt-like calls when interacting with each other. Even to scientists who have spent years listening to vervets, all these social grunts sound the same. When the grunts are recorded and displayed as a frequency spectrum on the screen of a sound-analysing instrument, they look the same. Only when the spectra were measured in elaborate detail could Cheney and Seyfarth detect (sometimes but not always!) average differences between the grunts given in four social contexts: when a monkey approaches a dominant monkey, when it approaches a subordinate monkey, when it watches another monkey, or when it sees a rival troop.

Broadcasts of grunts recorded in these four different contexts caused monkeys to behave in subtly different ways. For example, they looked towards the loudspeaker if the grunt had originally been recorded in the ‘approach dominant monkey’ context, while they looked in the direction towards which the call was being broadcast if it had originally been recorded in the ‘see rival troop’ context. Further observations of the monkeys under natural conditions showed that the natural calls had also been eliciting this subtly different behaviour.

Vervets are much more finely attuned than we are to their calls. Merely listening to and watching vervets, without recording and playing back their calls, gave no hint that they had at least four distinct grunts – and may have many more. As Seyfarth writes, ‘Watching vervets grunt to each other is really very much like watching humans engaged in conversation without being able to hear what they’re saying. There aren’t any obvious reactions or replies to grunts, so the whole system seems very mysterious – mysterious, that is, until you start doing playbacks.’ These discoveries illustrate how easy it is to underestimate the size of an animal’s vocal repertoire.

The vervets of Amboseli have at least ten putative ‘words’: their words for ‘leopard’, ‘eagle’, ‘snake’, ‘baboon’, ‘other predatory mammal’, ‘unfamiliar human’, ‘dominant monkey’, ‘subordinate monkey’, ‘watch other monkey’, and ‘see rival troop’. However, virtually every claim of animal behaviour suggesting elements of human language is greeted with scepticism by many scientists, who are convinced of the linguistic gulf separating us from animals. Such sceptics consider it simpler to assume that humans are unique, and that the burden of proof should be borne by anyone who thinks otherwise. Any claim of language-like elements for animals is considered a more complicated hypothesis, to be dismissed as unnecessary in the absence of positive proof. Yet the alternative hypotheses by which the sceptics instead attempt to explain animal behaviour sometimes strike me as more complicated than the simple, and often plausible, explanation that humans are not unique.

It seems a modest claim to propose that the different calls which vervets give in response to leopards, eagles, and snakes actually refer to these animals or are intended as communications to other monkeys. However, sceptics were disposed to believe that only humans could emit voluntary signals referring to external objects or events. The sceptics proposed that the vervet alarm calls were merely an involuntary expression of the monkey’s emotional state (‘I’m scared out of my wits!’) or of its intent (‘I’m going to run up a tree’). After all, those explanations apply to some of our own ‘calls’. If I saw a leopard coming at me, I too might emit a reflex scream even though there was no one around with whom to communicate. We grunt as a reflex when we throw ourselves into some physical activities, such as lifting a heavy object.

Suppose that zoologists from an advanced civilization in outer space observed me to give a trisyllabic scream, ‘argh, leopard’, and to climb a tree, when I saw a leopard. The zoologists might well doubt that my lowly species could express anything beyond grunts of emotion or intent – certainly not symbolic communications. To test their hypothesis, the zoologists would resort to experiments and detailed observations. If I screamed regardless of whether any other human was in earshot, that would support the theory of a mere expression of emotion or intent. If I screamed only in the presence of another person, and only when approached by a leopard but not by a lion, that would suggest a communication with a specific external referent. And if I gave the scream to my son but remained silent when I saw the leopard stalk a man with whom I had frequently been seen to fight, the visiting zoologist would feel certain that a purposeful communication was involved.

Similar observations convinced earthling zoologists of the communicative role of vervet alarm calls. A solitary vervet chased by a leopard for nearly an hour remained silent throughout the whole ordeal. Mother vervets give more alarm calls when accompanied by their own offspring than by unrelated monkeys. Vervets occasionally give the ‘leopard call’ when no leopard is present but when their troop is fighting with another troop and losing the fight. The fake alarm sends all combatants scrambling for the nearest tree and thereby serves as a deceptive ‘time out’. The call is clearly a voluntary communication, not an automatic expression of fear at the sight of a leopard. Nor is the call a mere reflex grunt given in the act of climbing a tree, since a calling monkey may either climb a tree, jump out of a tree, or do nothing, depending on the circumstances.

The supposition that the call has a well-defined external referent is especially well illustrated by the ‘eagle call’. Among large, broad-winged, soaring hawks, vervets usually respond with the eagle call to the martial eagle and the crowned eagle, their two most dangerous avian predators. They usually do not respond to the tawny eagle, and almost never to the black-chested snake eagle and white-backed vulture, which do not prey on vervets. Seen from below, black-chested snake eagles look rather similar to martial eagles in their shared pale underparts, banded tail, and black head and throat. Hence vervets rate as good bird-watchers. Their lives depend on it!

Vervet alarm calls are not an involuntary expression of either fear or intent. They have an external referent that may be quite exact. They are finely targeted communications which are more likely to be given honestly if the caller cares about the listener, and which may also be given dishonestly to enemies.

Sceptics dispute proposed analogies between animal sounds and human speech on the additional grounds that human speech is learned, but that many animals are born with the instinctive ability to utter the sounds characteristic of their species. However, young vervets appear to learn how to utter and respond to sounds appropriately, just as human infants. The grunts of an infant vervet sound different from those of an adult. ‘Pronunciation’ gradually improves with age until it becomes virtually adult at about the age of two years, somewhat less than half the age for vervet puberty. That is like human children attaining adult pronunciation at the age of five years; my sons, who are almost four years old, are still sometimes difficult to understand. Infant vervets do not learn to give reliably the correct response to an adult’s call until the age of six or seven months. Until then, an adult’s snake alarm call may send the infant jumping into a bush, the correct response to an eagle but a suicidal response to a snake. Not until the age of two years does the infant consistently emit each alarm call in the correct context. Before that age, the young vervet may call ‘eagle!’ not only when a martial or crowned eagle goes overhead, but also when any other bird flies over, and even when a leaf flutters down from a tree. Child psychologists refer to such behaviour in our own children as ‘overgeneralizing’ – as when a child greets not just dogs but also cats and pigeons with ‘bow-wow’.

If vervet calls are indeed partly learned rather than entirely instinctive, one might expect vervet populations in different parts of Africa to have developed different ‘dialects’ for the same reason that different human populations have. That is, ‘word’ meanings and pronunciations would gradually change with time, but the changes would develop independently in different areas and would be transmitted by learning, leading first to different dialects and eventually to different languages. This prediction of dialect differences has yet to be tested for vervets, since all the detailed studies of their vocal communication to date have been made in one small area of Kenya. However, song dialects are well developed in some bird species whose young learn the locally correct song from adult birds that they hear around them as they grow up. In a North American songbird called the white-crowned sparrow, such dialect differences are so pronounced that experienced bird-watchers near San Francisco can pinpoint an individual sparrow’s home within ten miles.

So far, I have loosely applied human concepts such as ‘word’ and ‘language’ into vervet vocalization. Let’s now compare human vocalizations and those of subhuman primates more closely. In particular, let’s ask ourselves three questions. Do vervet sounds really constitute ‘words’? How large are animals’ ‘vocabularies’? Do any animal vocalizations involve ‘grammar’ and merit the term ‘language’?

Firstly, on the question of words, it is clear at least that each vervet alarm call refers to a well-defined class of external dangers. That does not imply, of course, that a vervet’s ‘leopard call’ designates the same animals to a vervet as the word ‘leopard’ does to a professional zoologist – namely, members of a single animal species, defined as a collection of potentially interbreeding individuals. We already know that vervets give their leopard alarm in response not just to leopards but also to two other medium-sized cat species (caracals and servals). If the ‘leopard call’ is a word at all, it would not mean ‘leopard’ but instead ‘medium-sized cat that is likely to attack us, hunts in a similar way, and is best avoided by running up a tree’. However, many human words are used in a similar generic sense. For example, most of us other than ichthyologists and ardent fishermen apply the generic word ‘fish’ to any cold-blooded animal with fins and a backbone that swims in the water and might be worth eating.

Instead, the real question is whether the leopard call constitutes a word (‘medium-sized cat that … etc.’), a statement (‘there goes a medium-sized cat’), an exclamation (‘watch out for that medium-sized cat!’) or a proposition (‘let’s run up a tree or take other appropriate action to avoid that medium-sized cat’). At present it is not clear which of those functions the leopard call fills, or whether it fills a combination of them. Similarly, I was excited when my then one-year-old son Max said ‘juice’, which I proudly took to be one of his first words. To Max, though, the syllable ‘juice’ was not just his academically correct identification of a external referent with certain properties, but it also served as a proposition: ‘Give me some juice!’ Only at a later age did Max add more syllables, like ‘gimme juice’, to distinguish propositions from pure words. Vervets show no evidence of having reached that stage.

On the second question of extent of ‘vocabulary’, even the most advanced animals seem, on the basis of present knowledge, to be far behind us. The average human has a daily working vocabulary of around a thousand words; my compact desk dictionary claims to contain 142,000 words; but only ten calls have been distinguished even for vervets, the most intensively studied mammal. Animals and humans surely do differ in vocabulary size, yet the difference may not be as great as these numbers suggest. Remember how slow has been our progress in distinguishing vervet calls. Not until 1967 did anyone realize that these common animals had any calls with distinct meanings. The most experienced observers of vervets still cannot separate some of their calls without machine analysis, and even with machine analysis the distinctness of some of the suspected ten calls remain unproven. Obviously, vervets (and other animals) could have many other calls whose distinctness we have not yet recognized.

There is nothing surprising about our difficulties in distinguishing animal sounds, when one considers our difficulties in distinguishing human sounds. Children devote much of their time for the first several years of their lives to learning how to recognize and reproduce the distinctions in the utterances of adults around them. As adults, we continue to have difficulty distinguishing sounds in unfamiliar human languages. After four years of high-school French between the ages of twelve and sixteen, my problems with understanding spoken French are embarrassing compared to the abilities of any four-year-old French child. But French is easy compared to the Iyau language of New Guinea’s Lakes Plains, in which a single vowel may have eight different meanings depending on its pitch. A slight change in pitch converts the meaning of the Iyau word meaning ‘mother-in-law’ into ‘snake’. Naturally, it would be suicidal for an Iyau man to address his mother-in-law as ‘beloved snake’, and Iyau children learn infallibly to hear and reproduce pitch distinctions that for years confounded even a professional linguist devoted full-time to the study of the Iyau language. Given the problems we have ourselves with unfamiliar human languages, of course we must still be overlooking distinctions within the vervet vocabulary.

However, it is unlikely that any studies on vervets will reveal to us the limits attained by animal vocal communication, because those limits are probably reached by apes rather than by monkeys. While the sounds made by chimps and gorillas seem to our ears to be unsophisticated grunts and shrieks, so did the sounds made by vervet monkeys until they were studied carefully. Even unfamiliar human languages can sound like undifferentiated gibberish to us.

Unfortunately, vocal communication by wild chimps and other apes has never been studied by the methods applied to vervets, because of logistical problems. The width of a troop’s territory is typically less than 2,000 feet for vervets but is several miles for chimps, making it far harder to carry out playback experiments with video cameras and hidden loudspeakers. These logistical problems cannot be overcome by studying groups of apes caught in the wild and held captive in conveniently-sized zoo cages, because the captives generally constitute an artificial community of individuals caught at different African locations and thrown together in a cage. As I will discuss later in this chapter, humans originally speaking different languages, when captured at different African locations and thrown together as slaves, converse in only the crudest shadow of human language, virtually without any grammar. Similarly, captive apes taken from the wild must be virtually useless for studying the degree of sophistication of a vocal community of wild apes. The problem will remain unsolved until someone works out how to do for wild chimps what Cheney and Seyfarth have done for wild vervets.

Several groups of scientists have nevertheless spent years training captive gorillas, common chimps, and pygmy chimps to understand and use artificial languages based on plastic chips of different sizes and colours, or on hand signs similar to those used by deaf people, or on consoles, like a gigantic typewriter with each key bearing a different symbol. The animals have been reported to learn the meanings of up to several hundred symbols, and a pygmy chimp has recently been reported to understand (but not to utter) a good deal of spoken English. At the least, these studies of trained apes reveal that they possess the intellectual capabilities for mastering large vocabularies, begging the obvious question of whether they have evolved such vocabularies in the wild.

It is suspicious that wild gorilla troops may be seen sitting together for a long time, grunting back and forth in seemingly undifferentiated gibberish, until suddenly all the gorillas get up at the same time and head off in the same direction. One wonders whether there really was a transaction concealed within that gibberish. Because the anatomy of apes’ vocal tracts restricts their ability to produce the variety of vowels and consonants that we can, the vocabulary of wild apes is unlikely to be anywhere as large as our own. Nevertheless, I would be surprised if wild chimp and gorilla vocabularies did not eclipse those reported for vervets and comprise dozens of ‘words’, possibly including names for individual animals. In this exciting field where new knowledge is being rapidly acquired, we should keep an open mind on the exact size of the vocabulary gap between apes and humans.

The last unanswered question concerns whether animal vocal communication involves anything that could be considered grammar or syntax. Humans do not only have vocabularies of thousands of words with different meanings. We also combine those words and vary their forms in ways prescribed by grammatical rules that determine the meaning of the word combinations. Grammar thereby allows us to construct a potentially infinite number of sentences from a finite number of words. To appreciate this point, consider the different meaning of the following two sentences, composed of the same words and endings but with different word order, which constitutes one set of the grammatical rules that specify sentence meaning in the English language:

If human language did not involve grammatical rules, those two sentences would have exactly the same meaning. Most linguists would not dignify an animal’s system of vocal communication with the name of language, no matter how large its vocabulary, unless it also involved grammatical rules.

No hint of syntax has been discovered in the studies of vervets to date. Most of their grunts and alarm calls are single utterances. When a vervet gives a sequence of two or more utterances, all analysed cases have proved to consist of the same utterance repeated, as has also been the case when one vervet has been recorded responding to another vervet’s call. Capuchin monkeys and gibbons do have calls of several elements used only in certain combinations or sequences, but the meanings of these combinations remain to be deciphered (by us humans, that is).

I doubt that any student of primate vocalizations expects even wild chimps to have evolved a grammar remotely approaching the complexity of human grammar, complete with prepositions, verb tenses, and interrogative particles. However, it remains for the present an open question whether any animal has evolved syntax. The necessary studies on the wild animals most likely to use grammar – pygmy or common chimps – simply have not yet been attempted.

In short, while the gulf between animal and human vocal communication is surely large, scientists are rapidly gaining understanding of the causeway that evolved over that gulf from the animal side. Now let’s trace the bridge from the human side. We have already discovered complex animal ‘languages’; do any truly primitive human languages still exist?

To help us recognize what a primitive human language might sound like if there were any, let’s remind ourselves of the ways in which normal human language differs from vervet vocalizations. One difference is that of grammar. Humans, but not vervets, possess grammar, meaning the variations in word order, prefixes, suffixes, and changes in word roots (such as ‘they’, ‘them’, ‘their’) that modulate the sense of the roots. A second difference is that vervet vocalizations, if they constitute words at all, stand only for things that one can point to or act out. One could try to argue that vervet calls do include the equivalents of nouns (‘eagle’) and verbs or verb phrases (‘watch out for the eagle’). Our words clearly include both nouns and verbs that are distinct from each other, as well as adjectives. Those three parts of a speech referring to specific objects, acts, or qualities are termed lexical items. But up to half of the words in typical human speech are purely grammatical items, with no referent that one can point to.

These grammatical words include our prepositions, conjunctions, articles, and auxiliary verbs (words like ‘can’, ‘may’, ‘do’, and ‘should’). It is much harder to understand how grammatical items could evolve than it is for lexical items. Given someone who understands no English, you can point to your nose to explain what that noun means. Apes might similarly come to agree on the meanings of grunts functioning as nouns, verbs, or adjectives. How, though, do you explain the meaning of ‘by’, ‘because’, ‘the’, and ‘did’ to someone who understands no English? How could apes have stumbled on such grammatical terms?

Yet another difference between human and vervet vocalizations is that ours possess a hierarchial structure, such that a modest number of items at each level creates a larger number of items at the next higher level. Our language uses many different syllables, all based on the same set of a few dozen sounds. We assemble those syllables into thousands of words. Those words are not merely strung haphazardly together but are organized into phrases, such as prepositional phrases. Those phrases in turn interlock to form a potentially infinite number of sentences. In contrast, vervet calls cannot be resolved into modular elements and lack even a single stage of hierarchical organization.

As children, we master all of this complex structure of human language without ever learning the explicit rules governing it. We are not forced to formulate the rules unless we study our own language in school or learn a foreign language from books. So complex is our language’s structure that many of the underlying rules currently postulated by professional linguists have been proposed only in recent decades. This gulf between human language and animal vocalizations explains why most linguists never discuss how human language might have evolved from animal precursors. They instead regard that question as unanswerable and therefore unworthy even of speculation.

The earliest written languages of 5,000 years ago were as complex as those of today. Human language must have achieved its modern complexity long before that. Can we at least recognize linguistic missing links by searching for primitive peoples with simple languages that might represent early stages of language evolution? After all, some tribes of hunter-gatherers retain stone tools as simple as those that characterized the whole world tens of thousands of years ago. Nineteenth-century travel books abound with tales of backward tribes who supposedly used only a few hundred words or who lacked articulated sounds, were reduced to saying ‘ugh’, and depended on gestures for their communications. That was Darwin’s first impression of the speech of the Indians in Tierra del Fuego. But all such tales proved to be pure myth. Darwin and other western travellers merely found it as hard to distinguish the unfamiliar sounds of non-western languages as non-westerners found English sounds, or as zoologists find the sounds of vervet monkeys.

Actually, it turns out that there is no correlation between linguistic and social complexity. Technologically primitive people do not speak primitive languages, as I discovered on my first day among the Foré people in the New Guinea highlands. Foré grammar proved deliciously complex, with postpositions similar to those of the Finnish language, dual as well as singular and plural forms similar to those of Slovenian, and verb tenses and phrase construction unlike any language I had encountered previously. I have already mentioned the eight vowel tones of New Guinea’s Iyau people, whose sound distinctions proved imperceptibly subtle to professional linguists for years. Nor could we reverse Darwin’s prejudice by claiming an inverse correlation between linguistic and social complexity, citing the advanced civilizations of China and England, whose languages are simple in the sense of having little or no word inflection (verb conjugations and noun declensions). French verbs are much more highly inflected than are modern English verbs (nous aimons, vous aimez, ils aiment, etc.), yet the French consider themselves the most highly civilized people.

Thus, while some peoples in the modern world retained primitive tools, none retained primitive languages. Furthermore, Cro-Magnon archaeological sites contain lots of preserved tools but no preserved words. The absence of such linguistic missing links deprives us of what might have been our best evidence about human language origins. We are forced to try more indirect approaches.

One indirect approach is to ask whether some people, deprived of the opportunity to hear any of our fully evolved, modern languages, ever spontaneously invented a primitive language. According to the Greek historian Herodotus, the Egyptian king Psammeticus intentionally carried out such an experiment in the hope of identifying the world’s oldest language. The king assigned two newborn infants to a solitary shepherd to rear in strict silence, with instructions to listen for their first words. The shepherd duly reported that both children, after mouthing nothing but meaningless babble until the age of two, ran up to him and began constantly repeating the word becos. Since that word meant ‘bread’ in the Phrygian language then spoken in central Turkey, Psammeticus supposedly conceded that the Phrygians were the most ancient people.

Unfortunately, Herodotus’s brief account of Psammeticus’s experiment fails to convince sceptics that it was carried out as rigorously as described. It illustrates why some scholars prefer to honour Herodotus as the Father of Lies, rather than as the Father of History. Certainly, solitary infants reared in social isolation, like the famous wolf boy of Aveyron, remain virtually speechless and do not invent or discover a language. However, a variant of the Psammeticus experiment has occurred dozens of times in the modern world. In this variant, whole populations of children heard adults around them speaking a grossly simplified and variable form of language, somewhat similar to that which normal children themselves speak at around the age of two years. The children proceeded unconsciously to evolve their own language, far advanced over vervet communication but simpler than normal human languages. The results were the new languages known as pidgins and creoles, which may provide us with models of two missing links in the evolution of normal human language.

My first experience of a creole was with the New Guinea lingua franca known either as Neo-Melanesian or pidgin English. (The latter name is a confusing misnomer, since Neo-Melanesian is not a pidgin but rather a creole derived from an advanced pidgin – I shall explain the difference later – and it is only one of many independently evolved languages equally misnamed as pidgin English.) Papua New Guinea boasts about 700 native languages within an area similar to that of Sweden, but no single one of those languages is spoken by more than three per cent of the population. Not surprisingly, a lingua franca was needed and it arose after the arrival of English-speaking traders and sailors in the early 1800s. Today, Neo-Melanesian serves in Papua New Guinea as the language not only of much conversation, but also of many schools, newspapers, radios, and parliamentary discussions. The advertisement in the appendix to this chapter (see here) gives a sense of this newly evolved language.

When I arrived in Papua New Guinea and first heard Neo-Melanesian, I was scornful of it. It sounded like long-winded, grammarless baby-talk. On speaking a form of English according to my own notion of baby-talk, I was disturbed to discover that New Guineans did not understand me. My assumption that Neo-Melanesian words meant the same as their English cognates led to spectacular disasters, notably when I tried to apologize to a woman in her husband’s presence for accidentally jostling her, only to find that Neo-Melanesian pushim does not mean ‘push’ but instead means ‘have sexual intercourse with’.

Neo-Melanesian proved to be as strict as English in its grammatical rules. It was a subtle language that let one express anything sayable in English. It even let one make some distinctions that cannot be expressed in English except by means of clumsy circumlocutions. For example, the English pronoun ‘we’ actually lumps together two quite different concepts: ‘I, plus you to whom I am speaking’, and ‘I, plus one or more other people, but not including you to whom I am speaking’. In Neo-Melanesian these two separate meanings are expressed by the words yumi and mipela respectively. After I have been using Neo-Melanesian for a few months and then meet an English-speaker who starts talking about ‘we’, I often find myself wondering, ‘am I included or not in your “we”?’

Neo-Melanesian’s deceptive simplicity and actual suppleness stem partly from its vocabulary, partly from its grammar. Its vocabulary is based on a modest number of core words whose meaning varies with context and becomes extended metaphorically. For instance, while Neo-Melanesian gras can mean English ‘grass’ (whence gras bilong solwara [salt water] means ‘seaweed’), it also can mean ‘hair’ (whence man i no gat gras long head bilong em becomes ‘bald man’).

The derivation of Neo-Melanesian banis bilong susu as the word for ‘bra’ further illustrates the suppleness of the core vocabulary. Banis, derived from the English word ‘fence’, is the Neo-Melanesian word with that meaning, as in the expression banis pik for ‘pigpen’. Susu, taken from Malay as the word for ‘milk’, is extended to mean ‘breast’ as well. That sense in turn provides the expressions for ‘nipple’ (ai [eye] bilong susu), ‘prepubertal girl’ (i no gat susu bilong em), ‘adolescent girl’ (susu i sanap [stand up]), and ‘aging woman’ (susu i pundaun pinis [fall down finish]). Combining these two roots, banis bilong susu denotes a bra as the fence to keep the breasts in, just as banis pik denotes pigpen as the fence to keep pigs in.

Neo-Melanesian grammar appears deceptively simple because of what it lacks or else expresses by circumlocutions. These omissions include such seemingly standard grammatical items as plural and case forms of nouns, inflectional endings of verbs, the passive voice of verbs, and most prepositions and verb tenses. Yet Neo-Melanesian has passed far beyond baby-talk and vervet sounds in many other respects, including its conjunctions and auxiliary verbs and pronouns, and its ways of expressing verb moods and aspects. It is a normal complex language in its hierarchical organization of phonemes, syllables, and words. It lends itself so well to hierarchical organization of phrases and sentences that election speeches by New Guinean politicians rival the German prose of Thomas Mann in their convoluted structure.

At first, I ignorantly assumed that Neo-Melanesian was a delightful aberration among the world’s languages. It had obviously arisen in the 170 years since English ships started visiting New Guinea, but I supposed that it had somehow developed from baby-talk that colonists spoke to the natives they believed incapable of learning English. Only when I began working in Indonesia and learned the Indonesian language did I sense that Neo-Melanesian origins exemplified a much broader phenomenon. On the surface, Indonesian is incomprehensible to an English speaker and totally unrelated to Neo-Melanesian, because its vocabulary is largely Malay. Nevertheless, Indonesian reminded me of Neo-Melanesian in its word use and in the grammatical items that it possessed or lacked.

As it turns out, dozens of other languages resemble Neo-Melanesian and Indonesian in structure. They have arisen independently around the globe, with vocabularies variously derived largely from English, French, Dutch, Spanish, Portuguese, Malay, or Arabic. They appeared especially in and around plantations, forts, and trading posts, where populations speaking different languages came into contact and needed to communicate, but where social circumstances impeded the usual solution of each group learning the other’s language. Many cases throughout the tropical Americas and Australia, and on tropical islands of the Caribbean, Pacific, and Indian Oceans, involved European colonists importing workers who came from afar and spoke many different tongues. Other European colonists set up forts or trading posts in already densely populated areas of China, Indonesia, or Africa.

Strong social barriers between the dominant colonists and the imported workers or local populations made the former unwilling, the latter unable, to learn the other’s language. Usually the colonists scorned the local people, but in China the scorn was mutual: when English traders set up a post at Canton in 1664, the Chinese would no more abase thenselves by learning the foreign devils’ language or teaching them Chinese than would the English learn from or teach the heathen Chinese. Even if those social barriers had not existed, the workers would have had few opportunities to learn the colonists’ tongue, because workers so greatly outnumbered colonists. Conversely, the colonists would also have found it difficult to learn ‘the’ workers’ tongue, because so many different languages were often spoken among the workers.

Out of the temporary linguistic chaos that followed the plantations’ or forts’ founding, simplified but stabilized new languages emerged. Consider the evolution of Neo-Melanesian as an example. After English ships began to visit Melanesian islands just east of New Guinea in about the year 1820, the English also took islanders to work on the sugar plantations of Queensland and Samoa, where workers of many language groups were thrown together. From this Babel sprung the Neo-Melanesian language, of which the vocabulary is eighty per cent English, fifteen per cent Tolai (the Melanesian group that furnished many of the workers), and the rest derived from Malay, German, and other languages. The German element was added when German traders arrived in the 1870s and took over northeastern New Guinea as a colony in 1884. Germans must often have shouted the expletive heraus! (‘get out!’) at New Guineans, thereby inspiring the Neo-Melanesian words raus for ‘get out’ and rausim for ‘throw out’, as well as metaphorical extensions such as rausim bol for ‘castrate’.

Linguists distinguish two stages in the emergence of the new languages: initially, the crude languages (termed pidgins), then later the more complex ones, referred to as creoles. Pidgins arise as a second language for colonists and workers who speak differing native (first) languages and need to communicate with each other. Each group (colonists or workers) retains its native language for use within its own group; each group uses the pidgin to communicate with the other group, and in addition workers on a polyglot plantation may use pidgin to communicate with other groups of workers. An illustration of how quickly pidgins may arise is given by my own experience soon after I first arrived in Indonesia. An Indonesian worker and I were dropped together by helicopter in an uninhabited mountain range to survey birds. We had no Indonesian/English dictionary, knew nothing of each other’s language, and could teach each other words only by pointing. Within a week we had evolved a crude pidgin, based solely on Indonesian nouns, to communicate about camp chores: for instance rice fire meant ‘to cook rice’, while bird binoculars meant ‘to watch birds’.

Compared to normal languages, pidgins are greatly impoverished in their sounds, vocabulary, and syntax. A pidgin’s sounds are generally only those common to the two or more native languages thrown together. For example, many New Guineans find it hard to pronounce our consonants ƒ and v, but I and other native English speakers find it hard to pronounce the vowel tones and nasalized sounds rampant in many New Guinean languages. Such sounds became largely excluded from New Guinean pidgins and then from the Neo-Melanesian creole that developed from them. Words of early-stage pidgins consist largely of nouns, verbs, and adjectives, with few or no articles, auxiliary verbs, conjunctions, prepositions, or pronouns. As for grammar, early-stage pidgin discourse typically consists of short strings of words with little phrase construction, no regularity in word order, no subordinate clauses, and no inflectional endings on words. Together with that impoverishment, variability of speech within and between individuals is a hallmark of early-stage pidgins, which approximate an anarchic linguistic free-for-all.

Pidgins that are used only casually by adults who otherwise retain their own separate native languages persist at this rudimentary level. For example, a pidgin known as Russonorsk grew up to facilitate barter between Russian and Norwegian fishermen who encountered each other in the Arctic. That lingua franca persisted throughout the Nineteenth Century but never developed further, as it was used only to transact simple business during brief visits. Both those groups of fishermen spent most of their time speaking Russian or Norwegian with their compatriots. In New Guinea, on the other hand, the pidgin gradually became more regular and complex over many generations because it was used intensively on a daily basis, but most children of New Guinean workers continued to learn their parents’ native languages as their first language until after the Second World War.

However, pidgins evolve rapidly into creoles when a generation of one of the groups contributing to a pidgin begins to adopt the pidgin itself as its native language. That generation then finds itself using pidgin for all social purposes, not only for discussing plantation tasks or bartering. Compared to pidgins, creoles have a larger vocabulary, much more complex grammar, and consistency within and between individuals. Creoles can express virtually any thought expressible in a normal language, whereas trying to say anything even slightly complex is a desperate struggle in pidgin. Somehow, without any equivalent of the Académie Francaise to lay down explicit rules, a pidgin expands and stabilizes to become a uniform and more sophisticated language.

This process of creolization is a natural experiment in language evolution that has unfolded independently dozens of times in the modern world. The sites for the experiment have ranged from mainland South America and Africa to Pacific islands; the labourers, from Africans and Portuguese to Chinese and New Guineans; the dominant colonists, from English and Spaniards to other Africans and Portuguese; and the century, from at least the Seventeenth to the Twentieth. What is striking is that the linguistic outcomes of all these independent natural experiments share so many similarities, both in what they lack and in what they possess. On the negative side, creoles are simpler than normal languages in that they usually lack conjugations of verbs for tense and person, declensions of nouns for case and number, most prepositions, distinctions between events in the past and present, and agreement of words for gender. On the positive side, creoles are advanced over pidgins in many respects: consistent word order; singular and plural pronouns for the first, second, and third persons; relative clauses; indications of the anterior tense (describing actions occurring before the time under discussion, whether or not that time is the present); and particles or auxiliary verbs preceding the main verb and indicating negation, anterior tense, conditional mood, and continuing as opposed to completed actions. Furthermore, most creoles agree in placing a sentence’s subject, verb, and object in that particular order, and also agree in the order of particles or auxiliaries preceding the main verb.

The factors responsible for this remarkable convergence are still controversial among linguists. It is as if you drew a dozen cards fifty times from well-shuffled decks and almost always ended up with no hearts or diamonds, but with one king, a jack, and two aces. The interpretation I find most convincing is that of linguist Derek Bickerton, who views many of the similarities among creoles as a result of a human genetic blueprint for language.

Bickerton derived his view from his studies of creolization in Hawaii, where sugar planters imported workers from China, the Philippines, Japan, Korea, Portugal, and Puerto Rico in the late Nineteenth Century. Out of that linguistic chaos, and following Hawaii’s annexation by the US in 1898, a pidgin based on English developed into a fully fledged creole. The immigrant workers themselves retained their original native language. They also learned the pidgin that they heard, but they did not improve on it, despite its gross deficiencies as a medium of communication. That, however, posed a big problem for the immigrants’ Hawaiian-born children. Even if the children were lucky enough to hear a normal language at home because both mother and father were from the same ethnic group, that normal language was useless for communicating with children and adults from other ethnic groups. Many children were less fortunate and heard nothing but pidgin even at home, when mother and father came from different ethnic groups. The children also did not have adequate opportunities to learn English because of the social barriers isolating them and their worker parents from the English-speaking plantation owners. Presented with an inconsistent and impoverished model of human language in the form of pidgin, Hawaiian labourers’ children spontaneously ‘expanded’ pidgin into a consistent and complex creole within a generation.

In the mid-1970s Bickerton was still able to trace the history of this creolization by interviewing working-class people born in Hawaii between 1900 and 1920. Like all of us, those children soaked up language skills in their early years but then became fixed in their ways, so that their speech in their old age continued to reflect the language spoken around them in their youth. (My children too will soon be wondering why their father persists in saying ‘icebox’ rather than ‘refrigerator’, decades after the iceboxes of my parents’ own childhood disappeared.) Hence the elderly adults of various ages, whom Bickerton interviewed in the 1970s, provided him with virtually frozen snapshots of various stages in Hawaii’s pidgin-to-creole transition, depending on the subjects’ year of birth. In that way, Bickerton was able to conclude that creolization had begun by 1900, was complete by 1920, and was accomplished by children in the process of acquiring the ability to speak.

In effect, the Hawaiian children lived out a modified version of the Psammeticus experiment. Unlike the Psammeticus children, the Hawaiian children did hear adults speaking and were able to learn words. Unlike normal children, however, the Hawaiian children heard little grammar, and what they did hear was inconsistent and rudimentary. Instead, they created their own grammar. That they did indeed create it, rather than somehow borrowing grammar from the language of Chinese labourers or English plantation owners, is clear from the many features of Hawaiian creole that differ from English or from the workers’ languages. The same is true for Neo-Melanesian: its vocabulary is largely English, but its grammar includes many features absent from English.

I do not want to exaggerate the grammatical similarities among creoles by implying that they are all essentially the same. Creoles do vary depending on the social history surrounding creolization – especially on the initial ratio between the numbers of plantation owners (or colonists) and workers, how quickly and to what extent that ratio changed, and for how many generations the early-stage pidgin could gradually borrow more complexity from existing languages. Yet many similarities remain, particularly among those creoles that quickly arose from early-stage pidgins. How did each creole’s children come so quickly to agree on a grammar, and why did the children of different creoles tend to reinvent the same grammatical features again and again?

It was not because they did it in the easiest or sole way possible to devise a language. For instance, creoles use prepositions (short words preceding nouns), as do English and some other languages, but there are other languages that dispense with prepositions in favour of postpositions following nouns, or else noun case endings. Again, creoles happen to resemble English in placing subject, verb, and object in that order, but the borrowing from English could not account for creole grammar, because creoles derived from languages with a different word order still use the subject-verb-object order.

These similarities among creoles seem likely to stem from a genetic blueprint that the human brain possesses for learning language during childhood. Such a blueprint has been widely assumed ever since the linguist Noam Chomsky argued that the structure of human language is far too complex for a child to learn within just a few years, in the absence of any hard-wired instructions. For example, at the age of two my twin sons were just beginning to use single words. As I write this paragraph a bare twenty months later, still several months short of their fourth birthday, they have already mastered most of the rules of basic English grammar that people who immigrate to English-speaking countries as adults often fail to master after decades. Even before the age of two, my children had learned to make sense of the initially incomprehensible babble of adult sound coming at them, to recognize groupings of syllables into words, and to realize which groupings constituted underlying words despite variations of pronunciation within and between adult speakers.

Such difficulties convinced Chomsky that children learning their first language would face an impossible task unless much of language’s structure were already pre-programmed into them. Hence Chomsky reasoned that we are born with a ‘universal grammar’ already wired into our brains to give us a spectrum of grammatical models encompassing the range of grammars in actual languages. This pre-wired universal grammar would be like a set of switches, each with various alternative positions. The switch positions would then become fixed to match the grammar of the local language that the growing child hears.

However, Bickerton goes further than Chomsky and concludes that we are pre-programmed not just to a universal grammar with adjustable switches, but to a particular set of switch settings: the settings that surface again and again in creole grammars. The pre-programmed settings can be overridden if they turn out to conflict with what a child hears in the local language around it. But if a child hears no local switch settings at all because it grows up amidst the structureless anarchy of pidgin language, the creole settings can persist.

If Bickerton is correct in that we really are pre-programmed at birth with creole settings that can be overridden by later experience, then one would expect children to learn creole-like features of their local language earlier and more easily than features conflicting with creole grammar. This reasoning might explain the notorious difficulty of English-speaking children in learning how to express negatives: they insist on creole-like double negatives such as ‘Nobody don’t have this’. The same reasoning could explain the difficulties of English-speaking children with word order in questions.

To pursue the latter example, English happens to be among the languages that uses the creole word order of subject, verb, and object for statements: for instance, ‘I want juice’. Many languages, including creoles, preserve this word order in questions, which are merely distinguished by altered tone of voice (‘You want juice?’). However, the English language does not treat questions in this way. Instead, our questions deviate from creole word order by inverting the subject and verb (‘Where are you?’, not ‘Where you are?’), or by placing the subject between an auxiliary verb (such as ‘do’) and the main verb (‘Do you want juice?’). My wife and I have been barraging my sons from early infancy onwards with grammatically correct English questions as well as statements. My sons quickly picked up the correct order for statements, but both of them are still persisting in the incorrect creole-like order for questions, despite the hundreds of correct examples that my wife and I utter for them every day. Today’s samples from Max and Joshua include ‘Where it is?’, ‘What that letter is?’, ‘What the handle can do?’, and ‘What you did with it?’. It is as if they are not yet ready to accept the evidence of their ears, because they are still convinced that their pre-programmed creole-like rules are correct.

I have discussed creoles as if they appeared only with the rise of colonialism in the past 500 years. In fact, the social conditions that produced modern creoles have arisen repeatedly during thousands of years of documented human history, and probably long before that. Hence some of the world’s ‘normal’ languages may have passed through stages of creolization and gradually re-evolved a more complex grammar. The possible example closest to home is the language of these pages. There has been a long controversy among linguists over the history of the Germanic language family that includes English, and that presumably arose in the area of the Baltic Sea. As I shall discuss in Chapter Fifteen, Germanic languages belong to a wider grouping of languages termed Indo-European. All Indo-European languages clearly derived much of their vocabulary and grammar from an ancestral language known as proto-Indo-European, which may have been spoken in southern Russia 5,000 years ago and then spread west across Europe. However, the Germanic languages also include many word roots and grammatical features unique to them, and absent from all other Indo-European families. Familiar examples include the English words ‘house’, ‘wife’, and ‘hand’, close to the modern German words Haus, Weib, and Hand. The shores of the Baltic are the source of prized amber that was traded to southern Europe and Russia thousands of years ago, just as it is still traded around the world today. Could the Germanic languages have arisen as a creole, when proto-Indo-European traders settled among proto-Germanic tribes of the Baltic to buy amber in exchange for pottery, battle-axes, and horses?

Now let’s pull together all these animal and human studies to try to form a coherent picture of how our ancestors progressed from grunts to Shakespeare’s sonnets. A well-studied early stage is represented by vervet monkeys, with at least ten different calls that are under voluntary control, are used for communication, and have external referents. The calls may function as words, explanations, propositions, or as all of those things simultaneously. Scientists’ difficulties in identifying those ten calls have been such that more surely await identification, but we still do not know how large the vervet vocabulary really is. We also do not know how far other animals may have progressed beyond vervets, because the vocal communications of the species most likely to have eclipsed vervets, the common and the pygmy chimp, have yet to be studied carefully in the wild. At least in the laboratory, chimps can master hundreds of symbols that we teach them, suggesting that they have the necessary intellectual equipment to master symbols of their own.

The single words of young toddlers, like ‘juice’ as uttered by my son Max, constitute a next stage beyond animal grunts. Like vervet calls, Max’s ‘juice’ may have functioned as some combination of a word, an explanation, and a proposition. But Max has made a decisive advance on vervets by assembling his ‘juice’ word from the smaller units of vowels and consonants, thereby scaling the lowest level of modular linguistic organization. A few dozen such phonetic units can be reshuffled to produce a very large number of words, such as the 142,000 words in my English desk dictionary. That principle of modular organization lets us recognize far more distinctions than can vervets. For example, they name only six types of animals, whereas we name nearly two million.

A further step towards Shakespeare is exemplified by two-year-old children, who in all human societies proceed spontaneously from a one-word to a two-word stage and then to a multi-word one. But those multi-word utterances are still mere word strings with little grammar, and their words are still nouns, verbs, and adjectives with concrete referents. As Bickerton points out, those word strings are rather like the pidgin languages that human adults spontaneously reinvent when necessary. They also resemble the strings of symbols produced by captive apes whom we have instructed in the use of those symbols.

From pidgins to creoles, or from the word strings of two-year-olds to the complete sentences of four-year-olds, is another giant step. In that step were added words lacking external referents and serving purely grammatical functions; elements of grammar such as word order, prefixes and suffixes, and word root variation; and more levels of hierarchical organization to produce phrases and sentences. Perhaps that step is what triggered the Great Leap Forward discussed in Chapter Two. Nevertheless, creole languages reinvented in modern times still give us clues to how these advances arose, through the creoles’ circumlocutions to express prepositions and other grammatical elements. As another illustration of how this might have happened, my Indonesian colleague and I were just in the process of reinventing prepositions when the helicopter picked us up and terminated our experiment in pidgin evolution. We had begun to assemble word strings that functioned as locative prepositional phrases but were still composed solely of nouns with concrete referents – strings such as ‘spoon top plate’ and ‘spoon bottom plate’, to mean that the spoon was on or under the plate. Many virtual prepositions in Neo-Melanesian, Indonesian, and other creoles are similarly constructed.

If you compare the Neo-Melanesian advertisement on pages 150–51 with a Shakespearean sonnet, you might conclude that a huge gap still exists. In fact, I would argue that, with an advertisement like Kam insait long stua bilong mipela, we have come 99.9% of the way from vervet calls to Shakespeare. Creoles already constitute expressive complex languages. For example, Indonesian, which arose as a creole to become the language of conversation and government for the world’s fifth most populous country, is also a vehicle for serious literature.

Animal communication and human language once seemed to be separated by an unbridgeable gulf. Now, we have identified not only parts of the bridges starting from both shores, but also a series of islands and bridge segments spaced across the gulf. We are beginning to understand in broad outline how the most unique and important attribute that distinguishes us from animals arose from animal precursors.

Appendix

NEO-MELANESIAN, IN ONE EASY LESSON

Try to understand this Neo-Melanesian advertisement for a department store:

Kam insait long stua bilong mipela – stua bilong salim olgeta samting – mipela i-ken helpim yu long kisim wanem samting yu laikim bikpela na liklik long gutpela prais. I-gat gutpela kain kago long baiim na i-gat stap long helpim yu na lukautim yu long taim yu kam insait long dispela stua.

If some of the words look strangely familiar but do not quite make sense, read the advertisement aloud to yourself, concentrate on the sounds, and ignore the strange spelling. As the next step, here is the same advertisement rewritten with English spelling:

Come inside long store belong me-fellow – store belong sellim altogether something – me-fellow can helpim you long catchim what-name something you likim, big-fellow na liklik, long good-fellow price. He-got good-fellow kind cargo long buyim, na he-got staff long helpim you na lookoutim you long time you come inside long this-fellow store.

A few explanations should help you make sense of the remaining strangenesses. Almost all the words in this sample of Neo-Melanesian are derived from English, except for the word liklik for ‘little’, derived from a New Guinean language (Tolai). Neo-Melanesian has only two pure prepositions: bilong, meaning ‘of’ or ‘in order to’, and long, meaning almost any other English preposition. The English consonant f becomes p in Neo-Melanesian, as in stap for ‘staff’, and pela for ‘fellow’. The suffix -pela is added to monosyllabic adjectives (hence gutpela for ‘good’, bikpela for ‘big’), and also makes the singular pronouns ‘me’ and ‘you’ into plural ones (for ‘we’ and ‘you’ -plural). Na means ‘and’. So the advertisement means:

Come into our store – a store for selling everything – we can help you get whatever you want, big and small, at a good price. There are good types of goods for sale, and staff to help you and look after you when you visit the store.