How a Throat Learned to Mean: The Long Descent of Language

Press two fingers gently to the front of your throat and swallow. That small machine sliding under the skin is doing something no other animal’s throat can do, and it nearly killed us to get it. Somewhere in our lineage the larynx, the voice box, dropped lower in the throat than it sits in any ape. It opened a long resonating chamber above the vocal cords, a tube we can shape into a hundred distinct vowels and consonants. It also crowded the airway and the food pipe into the same junction, which is why humans, almost uniquely, can choke to death on a swallowed grape. We pay that price every meal. We paid it, the best guess runs, because the creature that could make finely distinguished sounds could mean finely distinguished things, and meaning was worth dying over.

Nobody was there to record the first word, and nobody ever will be; speech leaves no fossils. We have skulls, a few telling bones, scatterings of pigment and burial and tool, and from these we try to reconstruct the single most consequential invention our species ever made—the thing that let one mind reach into another and rearrange its contents. Take the one gene everyone has heard of. In the 1990s a London family known in the literature as the KE family came to scientists with a striking inherited disorder: across three generations, many of its members struggled badly with speech and grammar, unable to reliably form a past tense or repeat a made-up word. The cause was a single mutated gene, FOXP2. The press crowned it the grammar gene, which it is not. FOXP2 is a regulator that switches other genes on and off; versions of it run through mice and crucially through songbirds, where it governs the learning of song. It is one ancient cog the human line tuned for the fine motor sequencing that speech and song both demand. There is no single gene for language, but language did leave genetic footprints, and we can read a few.

From Cry to Symbol

Many animals signal. A vervet monkey gives one alarm call for a leopard, another for an eagle, another for a snake, and its troop responds correctly to each. That is impressive, and it is not language. The vervet’s calls are a closed list fixed by biology, tied to the here and now; the monkey cannot invent a new call, combine two into a larger meaning, or warn of a leopard that came yesterday or might come tomorrow. The leap that made us human was the leap from a finite set of signals welded to the present to an open, recombinable system that can speak of the absent, the false, the future, the purely imagined. A noise points at a thing. A symbol stands for it, in its absence, and can be slotted into endless new arrangements. We still argue bitterly about how the crossing was made—one camp, the gesture-first theorists such as Michael Corballis, says the hand carried the first symbols, since apes gesture far more flexibly than they vocalize and the deaf invent full sign languages wherever they gather; a rival camp keeps the action in the throat from the start, in song and the call that travels through dark. The honest position is that we do not know, because both the early gesture and the early cry are equally gone.

A noise points at a thing. A symbol stands for it, in its absence—and that one move is the whole crossing.

Behind that quarrel sits the deepest fault line in the field. On one side stands Noam Chomsky, who argued from the 1950s that the human mind comes equipped with an innate language faculty, a universal grammar—a built-in scaffolding that explains how a child, on scrappy evidence, infallibly grows a full grammar in a few short years. Steven Pinker gave this its popular flag, the language instinct, the claim that grammar is as much a part of our biology as the web is of the spider. On the other side stands Michael Tomasello, who studies apes and toddlers and finds the origin not in a grammar organ but in shared intentionality: the peculiarly human hunger to attend to the same thing together, to read another’s intention, to point at something simply so that you and I can both look and know we are both looking. Watch a one-year-old point at the moon and then turn to check your face, to share the moon with you. No ape does this; an ape points to get you to fetch, never to share attention for its own sake. Where Chomsky sees an organ, Tomasello sees a practice—grammar constructed, child by child, from joint attention and the patterns heard in real use. The two camps have spent decades talking past each other, and neither has won; the likeliest truth lives somewhere between the inborn instinct and the shared glance, in the space where a primate first became a speaker.

How One Tongue Became Thousands

Grant us the first language, however it arose. The next miracle is that it did not stay one. There are some seven thousand languages alive today and untold thousands dead, and they are not a chaos—they fall into families, branching trees with common ancestors, exactly as species do. The discovery began, fittingly, in India. In 1786 Sir William Jones, a British judge in Calcutta who had fallen in love with Sanskrit, told the Asiatic Society that Sanskrit, Greek, and Latin bore a resemblance too strong to be accident—that they must have sprung, in his careful phrase, from some common source which perhaps no longer exists. He was right. That vanished source we now call Proto-Indo-European, the unrecorded mother tongue of a family stretching from Iceland to Bengal.

“sprung from some common source, which, perhaps, no longer exists”—Sir William Jones, Third Anniversary Discourse, 1786

No one ever wrote Proto-Indo-European down, yet linguists can reconstruct hundreds of its words, because sound change is astonishingly regular. Languages do not corrupt randomly; they drift along lawful paths, and the comparative method exploits this: line up the daughters and the regularities expose the parent. Latin pater, Sanskrit pitar, English father, German Vater—the p that became f in the Germanic branch did so across the whole vocabulary at once, a wholesale shift Jacob Grimm catalogued and we still call Grimm’s Law. From such correspondences the lost ancestor swims back into view like a developing photograph. Two of those daughters tower over the rest for the light they shed. Sanskrit and Latin are both gloriously inflected, case-rich languages: a noun changes its ending to announce its role, so word order can dance freely while the grammar holds firm. Sanskrit pushes furthest, with eight cases, a dual number for things that come in pairs, and a system of sound-junction at word boundaries so intricate it has its own name. These are not primitive tongues straining toward expression. They are baroque cathedrals of grammar—and they sit on the same family tree as the plain analytic English you are reading, which long ago shed nearly all its endings and pinned meaning to word order instead.

The Other Antiquity: Dravidian

But the Indo-European tree is not the only forest, and on the Indian subcontinent it grows beside one of equal age and entirely separate root. The Dravidian family—Tamil, Telugu, Kannada, Malayalam, and a scattering of smaller tongues—is not a branch of Indo-European at all. It is its own lineage, with its own reconstructable ancestor, Proto-Dravidian, and a literary antiquity that humbles most of the world’s languages: the oldest Tamil poetry, the Sangam corpus, reaches back roughly two thousand years and is read with pride and real comprehension by Tamil speakers today. To treat the languages of India as a single thing is to miss the most interesting fact about them—two vast, ancient, unrelated families have shared one peninsula for thousands of years, and the sharing has changed them both.

Where Sanskrit inflects, Dravidian agglutinates, and the difference is worth feeling. An inflecting language fuses grammatical information into endings that blur and merge; an agglutinating one strings meaning out in clean, separable beads, suffix after suffix, each doing one job. A single Tamil or Malayalam verb can stack tense, person, number, negation, and politeness into one long transparent train, so that what English needs a whole clause to say arrives as one architected word. Dravidian also gave Indian languages a distinction between an inclusive and an exclusive we—one word for we meaning you and I together, another for we meaning my people but not you—a thing English cannot draw without a sentence of explanation, and which, once you have it, you miss sorely. The two families did not merely coexist; they interpenetrated. Sanskrit poured an enormous vocabulary into the Dravidian languages, words for thought and religion and government. The traffic ran the other way too, and deeper: many linguists hold that the retroflex consonants which give Indian languages their distinctive sound—the curled-tongue t and d a Western ear hears as exotic—entered Sanskrit itself under Dravidian influence, so that the very mouth-feel of classical Sanskrit may carry a Dravidian signature. Two strangers met on a subcontinent and, over millennia, came to speak in each other’s accents.

How a Language Carves the World

Here the question turns inward, from where languages came from to what they are—how a string of sounds holds a meaning at all. The man who reframed this for the modern age was Ferdinand de Saussure, a Swiss linguist whose most influential book he never wrote; it was reconstructed from students’ notes after his death in 1913. Saussure said the sign is two-sided, a marriage of the signifier (the sound-image, the word tree) and the signified (the concept of a tree), and that the bond between them is arbitrary. There is nothing tree-like in the word tree; the Frenchman says arbre, the German Baum, and none is righter than the others. Meaning is not a label pasted onto a pre-cut world. It is generated inside the system, by difference—a word means largely because it is not the other words around it; hot signifies by not being warm, not being cold. Saussure also drew the line that organized a century of study: langue, the shared system every speaker carries, against parole, the fleeting individual act of speaking it.

Meaning is not a label pasted onto a pre-cut world. It is generated inside the system, by difference.

Two great rivals answered Saussure. Chomsky’s generativism looks beneath the surface variety of grammars for a single deep universal structure common to all human language and, he argues, to the mind itself—the differences between Tamil and Latin mere settings on one underlying machine. Michael Halliday’s functionalism turns the telescope around: language is the way it is, he held, because of what it has to do; grammar is shaped by use, and you understand a form by asking what job it performs, not what abstract rule generates it. And behind all three lenses lies the vertiginous question: if each language carves the world differently, does the tongue you happen to speak shape the reality you can think? This is linguistic relativity, the Sapir-Whorf hypothesis. In its strong form—that language imprisons thought, that you cannot conceive what your tongue has no word for—it is almost certainly false; we coin words daily and translate across the deepest gulfs. But in its modest form it keeps proving true in startling ways. Speakers of languages that orient by compass directions rather than left and right keep a flawless internal sense of north. Whether your grammar forces you to mark an object’s gender, the source of your knowledge, or the rank of the person you address changes what you must notice to speak at all. Language may not cage thought, but it does decide what you cannot help attending to. Many Indic and East Asian languages oblige their speakers to track what English lets slide: evidentiality, a marker showing whether you saw a thing, inferred it, or merely heard it; dense honorific systems with several words for you sorted by intimacy and respect; aspect, the inner shape of an action, carved with a precision English handles only clumsily. None of this makes one language superior. It makes each a different instrument, attuned to register a different set of frequencies in the same shared world.

The Grammar That Behaved Like Mathematics

No account of how language encodes reality can omit the man who, more than two millennia before Saussure, treated his own tongue as a formal system and very nearly perfected it. Panini, working in the northwest of the subcontinent around the fourth century BCE, composed a grammar of Sanskrit, the Ashtadhyayi, of roughly four thousand terse rules. It is not a grammar in the gentle schoolbook sense; it is closer to an algorithm—an ordered, recursive machine that, fed a root and a set of conditions, generates the correct form of any word, with meta-rules governing which rules fire when several apply at once. Twentieth-century linguists building formal grammars for computers were startled to find that a Bronze-Age Indian had got there first, anticipating ideas about generative rules and ordered derivation that the West reinvented only with Chomsky and the theory of computation. This precision was not idle cleverness. It grew from a culture that believed sound itself was sacred and that a mispronounced syllable in a mantra could undo its power, so the exact form of every utterance had to be fixed and transmitted without loss across centuries. Where the West eventually trusted writing, the Indian tradition built an oral technology of staggering accuracy, encoding each verse several ways as a checksum against error, and Panini’s grammar was its keystone.

Set that against the structuralist’s arbitrary sign and you have the field’s whole tension in a single frame. To Saussure, the bond between word and thing is accidental, a convention that could have fallen any other way. To the Sanskrit grammarians, sound was load-bearing, the syllable a real force in the world. They cannot both be wholly right, and the argument between them is no quaint antique curiosity—it is the live question of whether language merely points at reality from outside or somehow participates in it. So return your two fingers to your throat and speak a sentence, any sentence. The dropped larynx that shapes it was bought with the risk of choking. The motor control that sequences it owes something to a gene you share with finches. The grammar that orders it is either an organ you were born with or a practice you assembled as a pointing child, and serious people still come to blows over which. The very words are worn cognates of sounds made by people dead five thousand years, carved into families by laws as regular as physics, and they cut your shared world along seams a Tamil speaker or a Latin poet would never have chosen. Every time you mean something to another person and are understood, you set in motion the longest, strangest, least finished invention our species has ever made—a throat that learned, against all odds, to mean.