The Princeton Guide to Evolution

VIII.9

Linguistics and the Evolution of Human Language

Mark Pagel

OUTLINE

1. What is language?

2. When did language evolve?

3. Why did language evolve?

4. The evolution of human languages

5. Languages adapt to speakers

6. The future of language evolution

This chapter discusses how human language differs from all other forms of animal communication, when and why it evolved, and how elements of language evolve over long periods of time. It closes with a brief account of how languages might evolve in an increasingly globalized world.

GLOSSARY

Cognate. The term in linguistics analogous to homology in evolution. Two words are cognate if they derive from a common ancestral word (e.g., the English water and the German wasser are cognate words, but neither is cognate to the French eau).

Homology. A term used in genetics and evolution to identify two genes or two traits thought to derive from a common ancestor. Human and chimpanzee hands and fingers are homologous, as are many human and chimpanzee genes. Bat wings, by comparison, are not homologous to bird wings, even though both are used for flying.

Language. A form of communication unique to humans consisting of discrete words formed into sentences composed of subjects, objects, and verbs.

Phylogeny. A tree diagram like a family tree but conventionally depicting the evolutionary relationships among a group of species. The diagram can be generalized to any evolving objects transmitted from one generation to the next, including languages.

Regular Sound Correspondences. A term used in linguistics to describe statistical regularities in the way sounds change in words over long periods of time. One of the best known of these is the regular replacement of a p sound at the beginning of a Latin word with an f sound in the Germanic languages, as in the change from Latin pater to father.

1. WHAT IS LANGUAGE?

We instinctively recognize that human language is unique among all forms of biological communication, but what do we mean by that? Most animals communicate, but humans are the only animals with language. Human language is distinct in having the property of being compositional: we alone communicate in sentences composed of discrete words that take the roles of subjects, objects, and verbs. This makes human language a digital form of communication as compared with the continuously varying signals that typify the grunts, whistles, barks, chest thumping, bleating, odors, colors, chemical signals, chirruping, or roars of the rest of life. Those familiar sights, sounds, and smells might signal an animal’s status or intentions, or indicate its physical prowess; they might tell a predator it has been spotted, or send a message to nearby relatives of an imminent danger. But lacking subjects, verbs, and objects, these acts of communication do not combine and recombine to produce an endless variety of different messages. Thus, your pet dog can tell you it is angry, and even how angry it is, but it cannot recount its life story.

2. WHEN DID LANGUAGE EVOLVE?

No one knows when the capacity to communicate with language evolved, but we can narrow the range of possibilities. Our closest living relatives, the modern chimpanzees, cannot speak, but we can, and so this tells us that language evolved sometime in the 6–7 million years between the time of our common ancestor and the late arrival of modern humans, around 160,000 to 200,000 years ago. For about the last 2.5 million years of that time, the evolutionary lineage of Homo or “human” species that would eventually give rise to modern humans, or Homo sapiens, was evolving in Africa. One of the best known of these Homo ancestors is Homo erectus, which probably evolved about 2 million years ago and had a relatively large brain, although at around 700 to 800 cc, it was still only a little more than half the size of the modern human brain. Homo erectus skulls reveal the impressions of two slightly protruding regions of the brain—known as Broca’s and Wernicke’s areas after their discoverers—that neuroscientists have identified as being involved in speech, at least in humans. This finding had led some researchers to suggest that Homo erectus had the capacity for speech, even if perhaps a rudimentary one. But Broca’s and Wernicke’s areas are also enlarged in some apes, so their presence is not by itself a definitive indicator that a species had language. There is also no evidence to suggest Homo erectus had anything even remotely close to the complex societies, tools, or other artifacts that we recognize as fully human.

A question that excites great differences of opinion is whether the Neanderthals spoke. This sister species to modern humans last shared a common ancestor with us sometime around 500,000 to 700,000 years ago. Neanderthals had large brains, they could control fire, and they had managed to occupy much of Eurasia by perhaps 300,000 years ago. Speculation that they had language has been fanned by the recent discovery that the Neanderthals had the same variant of a segment of DNA known as FOXP2, possessed by modern humans, that has been implicated, among many other effects, in influencing the fine motor control of facial muscles that is required for the production of speech. FOXP2 is not a gene itself, even though it is often described as one, but rather a short segment of DNA that affects how other genes are expressed. In fact, FOXP2 affects the human brain by causing about 50 genes to be expressed more and another 50 to be expressed less.

But, just as was true of Broca’s and Wernicke’s areas in Homo erectus, having the same variant of FOXP2 as modern humans do doesn’t tell us that the Neanderthals had language—they might have, but we cannot conclude with certainty that they did on the basis of this short segment of DNA. An analogy explains why. Most people’s cars have engines and so do Ferraris. But this doesn’t make every car a Ferrari. Closer to our own case, we know the modern human brain differs greatly from those of the Neanderthals in having a more fully developed and highly interconnected neocortex. This is the evolutionarily new and enlarged uppermost layer of the brain that is implicated in symbolic thinking and language. In simple terms, that thick layer of our cortex is why modern humans have foreheads and the Neanderthals did not. Given these differences, there is no compelling reason to conclude that FOXP2 affected Neanderthals’ brains as it does ours.

Further evidence against the Neanderthals’ having language comes from the archaeological record, which shows they did not produce the range of tools and symbolic objects that we associate with the cognitive complexity of Homo sapiens. For example, by perhaps 60,000 years ago modern humans were producing abstract and realistic art, and jewelry in the form of threaded shell beads, teeth, ivory, and ostrich shells; they used ochre and had tattoos; they had small stone tools in the form of blades and burins; they made artifacts of bone, antler, and ivory, as well as tools for grinding and pounding; and their improved hunting and trapping technology included spear throwers, nets, and possibly even bows. The Neanderthals did not produce any of these things—for instance, there is no evidence they even had needles that would have allowed them to produce sewn clothing. Even where there is some evidence that they might have had jewelry or body decoration, it cannot be ruled out that they acquired these from modern humans who inhabited Europe contemporaneously. And whereas modern humans somehow had the creative capacity to spread out around the entire world, building the technologies that allowed them to inhabit territories from the Arctic to the scorched deserts of Saharan Africa, the Neanderthals never left Eurasia.

This line of reasoning leads to the conclusion that language evolved with the arrival of our own species. Our capacity for language was almost certainly present in our common ancestor, because today all human groups speak, and speak equally well. There are no languages that are superior to others, and no human groups that speak primitive as opposed to advanced languages. If this hypothesis is correct, then language is no more than around 160,000 to 200,000 years old, although some anthropologists think language arose even later, pointing to a sharp increase beginning around 70,000 to 100,000 ago in evidence of symbolic thinking and the complexity of human societies. Even though our species arose prior to 100,000 years ago, it is just possible that all modern humans alive today trace their ancestry to common ancestors who lived around that time, so language evolving at this later date, though unlikely, is possible.

3. WHY DID LANGUAGE EVOLVE?

Evolution does not produce complex adaptations like language in a single moment. Usually, something pulls the trait and its precursors along so that it pays its way by granting some advantage to its bearers. To most people the advantage of having language is obvious—it allows us to communicate. All animals could benefit from being able to communicate as humans do, however, so this does not explain why humans have language and no other species does. Instead, we need to search for ways that having language granted benefits to speakers in our species but not in others. But searching for an advantage to language leads immediately to two evolutionary predicaments. One is that much of what a speaker has to say might benefit someone else, and potentially at a cost to the speaker. Natural selection never promotes naive altruism such as this, and so language would quickly have died a silent death, as selfish people all too willing to listen would have profited but with no intention of returning the speaker’s favor. Alternatively, perhaps language evolved to help humans mislead or trick others. It is in fact a fundamental tenet of communication that animal signals evolve to benefit the signalers. This solves the problem of altruism, because language might help speakers benefit at others’ expense. But this poses the second predicament: if others know that speakers’ acts of communication are designed to benefit themselves, surely this will favor people who don’t listen.

To understand how language overcame these two predicaments and to grasp why language evolved, we need to appreciate something that is true of human social behavior but that is virtually absent in the rest of the animal kingdom. Members of the human species are the only ones who routinely cooperate with other unrelated members of their species. Whereas most cooperation elsewhere in nature is limited almost entirely to helping kin or other relatives—that is, to acts of nepotism—humans routinely exchange favors, goods, and services with people other than those in their immediate families. We have an elaborate division of labor, we engage in task sharing, and we have learned to act in coordinated ways, such as when we go to war or simply combine or coordinate our energies to complete some task. But as powerful as this form of social behavior is, it is risky because in each of these situations we run the risk of being taken advantage of by other members of our group who might hold back, enjoying the spoils of cooperation but without having to pay the costs.

Elsewhere I have suggested that the role of language in this complex social behavior—and the reason it evolved—was to act as the conduit for carrying the information needed to make our form of cooperation work and for keeping people’s selfish instincts in check. It is a highly specialized piece of social technology, and so great were the benefits deriving from the cooperation language made possible that it easily paid its way in evolutionary terms. Both evolutionary predicaments are solved because now speakers and listeners benefit from having language. We use it to negotiate exchanges, to make plans, to remember lessons from the past, and to coordinate actions. Persons who try to take advantage of this cooperation, perhaps by failing to return someone else’s goodwill or by not contributing their fair share, can be exposed as “cheats” and their reputation tarnished. These are all complicated social acts that other animals don’t perform, and they require more than grunts, chirrups, odors, and roars. This explains why we and we alone have language: our particular brand of sociality could not exist without it.

4. THE EVOLUTION OF HUMAN LANGUAGES

It is by now widely accepted that modern humans first arose in East or possibly southern Africa, and then later swept “out of Africa” in two waves, one perhaps 120,000 years ago and a second one sometime between 50,000 and 70,000 years ago. Opinions differ about the success of the first wave; some researchers suggest modern humans got no farther than what is known today as the Levant region of the Middle East, where they eventually went extinct. Others believe that people from this first wave survived and spread eastward, establishing the populations that would eventually move into Southeast Asia, Indonesia, Papua New Guinea, and Australia. Whichever scenario is true, by the time of the second wave, modern humans had gone on to establish their presence permanently outside Africa and in a matter of a few tens of thousands of years had occupied nearly every environment on earth.

Why this sudden change in the behavior of a single species? The answer is almost certainly “language.” While all other species are largely confined to the environments to which their genes are adapted, having language and the social systems it made possible granted humans the capacity to adapt at the cultural level by producing new technologies, and it was via successive bouts of innovative cultural adaptation that our ancestors came to occupy the world. Language had granted humans the capacity to share ideas. This simple development meant that later generations could benefit from the accumulated wisdom of the past, and new ideas could quickly spread among the members of a population. Once a species has the ability to cooperate with unrelated individuals, a vast store of knowledge and wisdom is unleashed as the society can then draw on a wider range of talents and skills than would be available to a single individual or even to a family. We therefore expect this style of cooperation to be associated with an explosion of complexity, and this is precisely what the archaeological record reveals.

As these newly talkative people moved around the world they evolved different languages. They could do so because, unlike with genes, whose precise sequence of nucleotides determines the proteins they produce, there is no necessary connection—save perhaps for the onomatopoeic words—between a word or sound and its meaning. Currently there are approximately 7000 different languages spoken, or 7000 mutually incomprehensible systems of communication around the world. This makes humans unique—and somewhat bizarrely so—among animals in not being able to communicate with other members of their own species. As it happens, the 7000 different languages are not evenly distributed around the world. Instead, language density, or number of different languages found in a given geographic area, follows almost exactly the pattern of biological species diversity. Both show strong latitudinal gradients, such that both the number of different languages and number of different species in a given area are small in the northernmost regions of the world, and both become more tightly packed in tropical regions (figure 1). In some tropical regions, languages can be so densely packed that a different language can be spoken every few kilometers! For instance, the Vanuatu island of Gaua in the south Pacific Ocean covers 342 km², and like so many of the islands in this region, it is the roughly circular remnant plug of an ancient volcano. Gaua is just 19–21 km in diameter, but this speck of an island supports five distinct languages—Lakon or Vuré, Olrat, Koro, Dorig, and Nume.

Figure 1. Relationships between languages and species. Numbers of North American human language–cultural groups (before European contact) and mammal species are distributed similarly across degrees of latitude. (A) Numbers of languages and numbers of mammal species at each degree of north latitude in North America. The trends reflect the shape of the continent, being narrow in the south regions and growing wider at higher latitudes. Both trends peak at approximately 40°N, where North America is ~4800 km wide. (B) Densities of languages and mammal species, calculated as the number of each found at the specific latitude divided by the area of the continent for a 1° latitudinal slice at each latitude. (Figure and data are reproduced, with permission, from Nature Reviews Genetics [Pagel 2009].)

It is as if human language groups act like distinct biological species, and this linguistic isolation is clearly evident in the ways that languages evolve. Most of us learn our language from our parents and those immediately around us. This dominant form of language transmission from parent to offspring or from older generation to younger generation means that language evolution shares a number of features with genetic evolution (table 1). Thus genes and words are both discrete elements that evolve by a process of descent with modification as they are passed over many generations: genes can acquire mutations, and words can acquire sound changes. For example, the Old English brōthor evolved into the modern English brother. Words can also be borrowed from a donor language and incorporated into a recipient language—the English word beef is borrowed from the French boeuf—just as genes can sometimes be transmitted from one species to another, such as when two closely related species form hybrids.

These and the other parallels between genetic and linguistic evolution mean that when people spread out over an area, and new languages evolve, those languages share many features and form what we recognize as families of languages. Indeed, in The Descent of Man (1871), Darwin noted that “the formation of different languages and of distinct species … [is] curiously parallel….” This parallel had in fact been recognized in the late eighteenth century, nearly 100 years before Darwin, by an English judge, Sir William Jones, working in colonial India during the rule of the English King George III. To process court papers Sir William found it necessary to learn Sanskrit, and in doing so he became aware of curious similarities among Sanskrit, Latin, and Greek. Jones described these to a meeting of the Asiatic Society in Calcutta in 1786, saying:

The Sanskrit language … [bears] … a stronger affinity … [to Greek and Latin] … both in the roots of verbs and in the forms of grammar, than could possibly have been produced by accident; so strong, indeed, that no philologer could examine them all three, without believing them to have sprung from some common source, which, perhaps, no longer exists.” Sir William Jones, Calcutta, February 2, 1786

Jones had identified what linguists would later recognize as the Indo-European language family, which arose around 8000 to 9000 years ago with the advent of farming in the Fertile Crescent, or what is roughly present-day Turkey and Iraq. Farmers and farming technology then spread out from that region, seeding the languages currently spoken all over Europe, parts of central Asia, and the Indian subcontinent. These include the Germanic languages German, Dutch, and English; Romance languages that derive from Latin including French, Spanish, and Italian; Slavic languages; and even Persian, Hindi, and Punjabi.

Table 1. Some analogies between biological and linguistic evolution

Biological evolution	Language evolution
Discrete heritable units (e.g., nucleotides, amino acids, genes, morphology, behavior)	Discrete heritable units (e.g., words, syntax, and grammar)
Mechanisms of replication	Teaching, learning, imitation
Mutation—various mechanisms yielding genetic alterations	Innovation (e.g., formant variation, mistakes, sound changes, introduced sounds and words)
Homology (genes that are related by descent from a common ancestral gene and are now found in different species [e.g., ribosomal genes in nearly every species])	Cognates (words that are related by descent from a common ancestral word and are now found in different languages [e.g., father and pater])
Natural selection	Social selection, trends
Drift	Drift
Cladogenesis (e.g., allopatric speciation [geographic separation] and sympatric speciation [ecological/reproductive separation])	Lineage splits (e.g., geographic separation and social separation)
Anagenesis	Linguistic change without split
Horizontal gene transfer	Borrowing
Hybridization (e.g., horse, zebra, wheat, strawberry)	Language Creoles (e.g., Surinamese)
Correlated genotypes/phenotypes (e.g., allometry, pleiotropy)	Correlated cultural terms (e.g., five and hand)
Geographic clines	Dialects/dialect chains
Fossils	Ancient texts
Extinction	Language death

By comparing the similarities and differences among a group of languages such as those that form the Indo-European family, it is possible to construct evolutionary or phylogenetic trees depicting the probable course of evolution of those languages from their common ancestral or protolanguage. Figure 2 displays just such a tree for a selection of Indo-European languages, including some of its main branches. The most popular and widely used approach to inferring linguistic phylogenies draws on the close parallels between linguistic and genetic evolution such as are listed in table 1. In particular, whereas evolutionary biologists infer phylogenetic trees of species by studying homologous genes, evolutionary linguists infer linguistic trees from sets of cognate words. Just as slight differences in the sequence of a gene can identify sets of closely related species, slight differences in words can identify sets of closely related languages.

Figure 2. The Indo-European language family. Evolutionary tree showing the relationships among the major branches of this language family. Celtic languages include Irish, Breton, and Welsh; Germanic languages include German, Dutch, and English; Romance or Latinate languages include French, Italian, and Spanish; Slavic languages include Russian, Czech, and Polish; Indo-Iranian languages include Persian, Afghan, Hindi, and Punjabi. Hittite and Tocharian are two extinct languages. The base or root of the tree represents the proto-Indo-European language that might have existed 8000 to 9000 years ago.

For example, the word madre means mother in both Spanish and Italian, and we instinctively recognize that both derive from the Latin mater. Likewise, we recognize that the English mother is similar to the German mutter, and again both of these seem to derive from mater. In fact, these comparisons identify English as part of the Germanic branch of the Indo-European languages and identify Spanish and Italian as part of the Romance or Latinate branch of this same family (figure 2).

Not all such comparisons are this simple, but linguists make use of what they call regular sound correspondences to help them identify whether two words are cognate and how closely related they are. One of the best known of these regular sound changes is the replacement of a p sound at the beginning of a word with an f sound, as in the Latin pes or ped, which becomes foot, and pater becomes father. Other regular sound correspondences reveal to linguists that the English five is closely related to the German fünf, that the French cinq is related to the Spanish cinco, and less obviously that all four of these words derive from the Latin quinque. Of course, not all words are cognate. The Spanish agua is cognate to the Italian acqua, but neither is cognate to the English water or to the German wasser. This lack of cognacy serves to reinforce that the Romance languages are a distinct group from the Germanic languages, even though both branches trace their ancestry to a pre-Latin language that might have been spoken 5000 years ago or more.

Once a set of comparisons is made among a group of languages, identifying cognate and noncognate words for a large number of different meanings (a meaning being what a word refers to), the resulting data can be used along with formal methods to infer the phylogenetic tree. A variety of different methods are commonly used to infer trees, including parsimony, distance, and likelihood methods, and these are broadly similar whether applied to genetic or linguistic data. Parsimony methods seek a tree that minimizes the number of evolutionary events along its branches. Thus, parsimony methods would favor a tree that put English with German, and French with Italian or Spanish, over one, for example, that showed German as more closely related to Spanish than to English. To put German next to Spanish would suggest that the word water or wasser had somehow evolved twice. Distance methods, as their name implies, seek to define a distance between all pairs of languages, and those distances are then used to construct a tree. Likelihood methods use formal statistical models to estimate the probability of changes in words through time. A tree is constructed that makes the observed set of changes most probable, given the model of evolution. These methods are all described further in chapters II.1 – II.3.

Trees such as the one depicted in figure 2 have now also been produced for the Austronesian languages, the Bantu languages of Africa, the Arawak languages of South America, the Semitic languages, the Uralic languages of Northern Europe, and some Melanesian languages, and this is an important and growing area of the field of evolutionary linguistics. The existence of sets of languages that comprise families of related languages shows that at least some elements of language evolve slowly enough to preserve signals of their ancestry dating back thousands of years. For example, linguists recognize that the word for two of something is probably derived in all Indo-European languages from a shared ancestral sound that has been conserved for many thousands of years. Thus, in Spanish the word is dos, it is twee in Dutch, deux in French, due (doo-ay) in Italian, dois in Portuguese, duo (δύο) in Greek, di in Albanian, and do in Hindi and Punjabi; Julius Caesar would have said duo. This conservation leads to the proposal that the original or proto-Indo-European word that was spoken perhaps 9000 years ago was also “two”—as it sounds—and indeed, some scholars suggest it was duwo or duoh.

A handful of other words, including three, five, who, I, and you, are also highly conserved. For example, the English word three is tre in Swedish and Danish, drei in German, tre in Italian and tres in Spanish, tria in Greek, teen in Hindi and tin in Panjabi, and tri in Czech, leading to the suggestion that the proto-Indo-European word for three might have been trei. These conserved words are closely followed by pronouns such as he and she, and by the what, where, and why words, all of which often show a striking degree of similarity among many Indo-European languages. Other words, however, can vary considerably across these same languages, meaning that they have evolved or changed at far higher rates. The English word bird, for example, is vogel in German, oiseau in French, and pajaro in Spanish.

When long lists of words are studied for their frequency of change among the languages of a language family such as Indo-European, it is possible to derive estimates of their rates of change using statistical methods. It turns out that most words can be expected to last somewhere around 1500 to 2500 years, with some lasting far longer and others for shorter amounts of time (figure 3). It is possible from these rates of change to calculate a word’s linguistic half-life, defined as the amount of time it takes for there to be a 50 percent chance a word will be replaced by some new, unrelated word (figure 3). This approach shows that some words might be expected to last more than 10,000 years. The linguist Merritt Ruhlen has even proposed a list of “global etymologies” that he thinks are signals leftover from the last common ancestor to all human languages, or our mother tongue. Ruhlen’s list includes words for who, what, two, water, finger, and one. The mere existence of a mother tongue is controversial, and many linguists dispute the list, but Ruhlen cites evidence that traces of these words are found in many language families from all over the world.

Figure 3. Rates of lexical replacement. (A) Counts of the rate at which a word comes to be replaced by a new unrelated word in the Indo-European languages for 200 common vocabulary words, measured in units of numbers of new words per 1000 years of evolution. Fastest to slowest rate represents more than a 100-fold difference. The average = 0.3 ± 0.18 new words per 1000 years (or roughly one every 3000 years), median = 0.27, range = 0.009 to 0.93. (B) Counts of the word half-life estimates as derived from the rates of lexical replacement for the same 200 words. The half-life measures the expected amount of time before a word has a 50 percent chance of being replaced by a new, unrelated word. The average half-life is 5300 years, with a median of 2500 years, and ranges from 750 to a theoretical value of 76,000. (Figure and data are reproduced, with permission, from Nature Reviews Genetics [Pagel 2009].)

5. LANGUAGES ADAPT TO SPEAKERS

But why should words last as long as they do? Given all the opportunities for neighboring language groups to borrow words from one another, and all the ways that words can change, language changes far more slowly that we might expect; that is, from the standpoint of effective communication, words do not need to last much more than around three generations—the time span covered in a typical population of speakers. If they changed faster than that, then we might not be able to talk to our grandparents. But most words last somewhere around 1500 to 2500 years, so why is there such fidelity in language?

One answer comes from thinking of elements of languages as “replicators” that must adapt to the environment of the human mind. Words are sounds that compete for attention in the mind of speakers. So, to be successful words must get themselves replicated by being transmitted to someone else’s mind. It follows that those words that are easiest to remember and say will tend to be those retained. We can see the current competition for space in our mind in common words like sofa and couch, or living room, sitting room, reception room, and parlor. Which of these forms will win? There are no simple answers, but precisely because there is seldom any necessary connection between a sound and its meaning, the competition often focuses on characteristics of the sounds themselves, and thus we can expect the competition to be more intense the more often a word is used. There is a huge disparity in how often different words get used; in fact, so great is the disparity that about 25 percent of all human speech is made up from a mere 25 words. According to the Oxford English Dictionary, the English language’s top 25 include the, I, you, he, this, that, have, to be, for, and and (while they, we, say, and she make it into the top 30). Not surprisingly, perhaps, the words used most frequently have common characteristics: they are short, often monosyllabic, distinct, and easy to pronounce. And this begins to explain why some words last longer than perhaps they need to. It is not that these frequently used words are somehow more important, but that they might be stable over long periods because they have become so highly adapted to our minds that it becomes difficult for a new form to arise and outcompete or dislodge them.

6. THE FUTURE OF LANGUAGE EVOLUTION

Our native language is, perhaps, one of our most intimate traits, being the voice of the “I” or “me” that defines our conscious self. It is the language of our thinking, and it is the code in which many of our memories are stored. Thus it is not surprising that one of the greatest personal losses a people can suffer is the loss of their native language. And yet, currently somewhere about 15 to 30 languages go extinct every year as small traditional societies dwindle in numbers or get overwhelmed by larger neighbors, and younger generations choose to learn the languages of larger and politically dominant societies. Whatever the true numbers of languages going extinct, the loss of languages greatly exceeds the loss of biological species as a proportion of their respective totals.

Some projections say that only a handful of languages will see out this century. This raises the question of which language will win if ever a single language should succeed all others on earth. Currently three languages are spoken by a far greater number of people than any of their competitors. About 1.2 billion people speak Mandarin, followed by around 400 million each for Spanish and English, and these are closely followed by Bengali and Hindi. It is not that these languages are better than their rivals; it is that they have had the fortune of being linked to demographically prosperous cultures. On these counts Mandarin might look like the leader in the race to be the world’s language, but this ignores the fact that vastly more people learn English as a second language—including many people in China—than any other. Already it is apparent that if there is a worldwide lingua franca it is English.

Still, English itself might be transformed as it is bombarded by the influences of such large numbers of nonnative English speakers who bring along their accents, grammar, and words to English when they speak it. This ability of English to take in so-called foreign words has been the key to its adaptability for at least the millennium since the Norman conquest of the English in 1066 brought an influx of Norman French vocabulary. Just as words must adapt to be competitive in the struggle to gain access to our mind, languages have to adapt as a whole to remain useful to their speakers, and those that do so will be the survivors. Self-appointed human “minders” in the form of reactionary grammarians, sticklers for spelling, or those who deliberately try to exclude some words and phrases will succeed in controlling the rate at which their languages naturally change, but in doing so they might consign these languages to the backwaters of international communication. This might already be happening to French and German, as both governments have ministries devoted to language “purity.” The alternative to this control is not the free-for-all that some might fear. If communication is important, languages will never change at rates that imperil the very reason for which they exist.