The Question Linguists Banned: Where Language Came From, and Why There Are 7,000

In August and September 1977 two spacecraft left Earth with a gold-plated copper disc bolted to the outside of each. The Voyager Golden Record carries photographs, music, the sound of surf and thunder, and spoken greetings in fifty-five languages, from Akkadian, which nobody has spoken in thousands of years, to Wu Chinese. If anything ever finds it, it will be handed fifty-five separate solutions to the same problem.

Which is a peculiar thing for a species that cannot explain how it arrived at the first one.

The question of where language came from is genuinely open. It is so open that in 1866 the Linguistic Society of Paris, the most serious body of its kind in Europe, banned papers on the subject. Not because it was unimportant. Because there was no evidence to argue from, so every submission was somebody's story, and the stories could not be told apart. That ban shaped the field for the better part of a century.

Why there is almost nothing to go on

Speech does not fossilise. The tongue, the larynx, the soft tissue of the throat and the wiring of the brain all rot, and what survives is bone. You can measure a skull and argue about what its owner could physically have produced, but a vocal tract capable of speech is not proof of speech, in the same way that having hands is not proof of writing.

Then there is the gap. Writing is roughly five thousand years old. Language is very probably older by an order of magnitude, with most estimates putting the emergence of modern language capacity somewhere in Africa within the last two hundred thousand years, and some scholars arguing for far earlier. That means almost the entire history of human language happened before anyone could record any of it.

Genetics has helped a little and been oversold a lot. The gene FOXP2 turns up constantly in coverage as "the language gene", which it is not. It was identified through a British family, several of whose members had a severe inherited difficulty with speech and grammar, and it does appear to matter for the fine motor control speech requires. But it is one gene among many, it does considerable work in other organs and in other species, and no single gene builds something as layered as syntax. Knowing it is involved tells you roughly as much about language as knowing that muscles are involved tells you about football.

No language anywhere is primitive

The Victorians who first tried this assumed a ladder, with their own languages near the top and the languages of people they had colonised somewhere near the bottom, closer to whatever the original had been.

Every bit of that was wrong. There are roughly seven thousand languages currently spoken and not one of them is a simplified version of anything. Every one has a full grammar, a system for marking who did what to whom, ways to talk about things that have not happened and might not, and the machinery to coin a word for anything its speakers encounter. Languages spoken by a few hundred people in one valley are frequently more grammatically elaborate than English, not less. Some pack into a single verb what English needs a clause for.

So there is no living fossil to examine. Nobody speaks the early version. The oldest thing in the room is the same age as everything else in it.

How one language becomes thousands

The second half of the question is easier, because this part can be watched.

Language changes constantly and slightly. Vowels drift, consonants soften, meanings narrow or wander, and a generation of children picks up a version marginally different from their parents'. Nobody notices, because the change per generation is smaller than the range of ordinary variation.

Cut two groups of speakers off from each other and both keep drifting, in different directions, forever. Give it a few centuries and they can no longer understand one another. Give it a few thousand years and only a specialist can tell they were ever the same thing.

The evidence for this is unusually solid, because the drift follows rules. In 1786 William Jones told the Asiatic Society that Sanskrit, Greek and Latin resembled each other too closely and too systematically for the resemblance to be chance, and proposed a common ancestor nobody had ever recorded. In 1822 Jacob Grimm, the same Grimm who collected the fairy tales, described the regular sound shift that separated the Germanic languages from the rest of that family, which is why Latin pater corresponds to English father and Latin piscis to English fish. The p turns to f, reliably, across the whole vocabulary.

Latin is the clean demonstration. It was one language with one army behind it, and about fifteen hundred years after the empire stopped enforcing it, its descendants are Portuguese, Spanish, French, Italian and Romanian, which are separate languages by any test you like.

Why the map is so lumpy

If drift alone explained the count, languages would be spread fairly evenly over people. They are not, and the pattern tells you what actually holds a language together.

Papua New Guinea has more than eight hundred languages among a population smaller than many single cities. The reasons are mountains, dense forest, deep valleys, small communities, and until recently no state with the reach to impose one language on all of them. What produces linguistic variety is not isolation from the world so much as isolation from your neighbours.

Now look at the opposite. Large contiguous states with roads, conscript armies, national schooling, printing and broadcasting have spent the last few centuries flattening the variety inside their borders, sometimes deliberately and often by punishing children for speaking the wrong thing at school. The result is a handful of very large languages and a very long tail of small ones, with a substantial share of that tail now spoken only by the elderly. Roughly half the seven thousand are considered endangered, which means the count is currently falling rather than holding steady.

Somebody watched a language get built

The closest anyone has come to seeing the original event happened recently, in Nicaragua, and it happened because of a school.

Before the late 1970s deaf Nicaraguans were mostly isolated from one another, each family improvising its own home signs. In 1977 a special education centre in Managua enrolled fifty deaf children. By 1979 there were a hundred. A second school opened in 1980, and by 1983 more than four hundred deaf pupils were enrolled across the two.

The teaching was in Spanish, using lip-reading and finger-spelling, and it did not work. What did work was the playground and the school bus, where several hundred children with several hundred private sign systems had to find common ground. They pooled it. What came out was a shared system, usable, but rough, and closer to a pidgin than a language.

Then the younger children arrived, learned it from the older ones, and did something nobody instructed them to do. They regularised it. They introduced consistent verb agreement, used the space in front of the body systematically to mark who was doing what to whom, and produced a fully grammatical language out of what they had been handed. The linguists who studied it, among them Judy Kegl, who was contacted by the education ministry in 1986, and later Ann Senghas and Marie Coppola, found the crucial thing: the youngest learners signed a more complex grammar than the people who taught them. They changed the language as they acquired it.

Adults did not do this. Children did, and they did it in about a decade, and no one had to tell them what a grammar was.

What that leaves us with

It does not tell us how the first language began. Those children had human brains, human company and a reason to communicate, which is a long way from whatever the situation was in Africa a hundred thousand years ago.

What it does show is that a language does not have to be invented by anyone in particular. Put children together with something rough to work from and the structure assembles itself, because that is apparently what human children do with a signal. Which makes the seven thousand look less like seven thousand inventions and more like one capacity, running for a very long time in a great many separate rooms.

The Golden Record will keep going either way. It carries fifty-five ways of saying hello, sent by a species that has never worked out how it learned to say anything.

Sources

• Wikipedia, "Voyager Golden Record" (the 1977 launches, the gold-plated copper disc, and the spoken greetings in fifty-five languages from Akkadian to Wu Chinese).
• Wikipedia, "Origin of language" (the 1866 ban by the Linguistic Society of Paris for want of evidence, the absence of any fossil record for speech, and the range of estimates for when modern language capacity emerged).
• Wikipedia, "FOXP2" (the gene's identification through a family with an inherited speech and language disorder, its role in motor control, and why it is not a single gene for language).
• Wikipedia, "Nicaraguan Sign Language" (the 1977 Managua centre with fifty children, a hundred by 1979, the second school in 1980 and more than 400 pupils by 1983, the shift from the older cohort's rougher system to the younger cohort's full grammar, Judy Kegl's contact from the education ministry in 1986, and the Senghas and Coppola finding that children changed the language as they learned it).
• Wikipedia, "Grimm's law" (Jacob Grimm's 1822 account of the regular Germanic sound shift, including the p to f correspondence).
• Wikipedia, "Proto-Indo-European language" (William Jones in 1786 on the shared source of Sanskrit, Greek and Latin, and the comparative method built on it).
• Wikipedia, "Languages of Papua New Guinea" (more than eight hundred languages, and the terrain and community size behind that density).
• Ethnologue, "How many languages are there in the world?" (the count of roughly seven thousand living languages and the share now endangered).