A Language Dies Every Two Weeks

On the morning of January 21, 2008, Marie Smith Jones died in her apartment in Anchorage, Alaska. She was eighty-nine years old, and she was the last person on Earth who spoke Eyak as a first language. There was no one left to argue with her about a word's correct pronunciation, no one to share a private joke in the tongue she had grown up speaking on the Copper River delta. With her went the only mind in the world that still thought in the syllables of Eyak. A linguist named Michael Krauss had been recording her, on and off, for some forty years, and what he buried with her was not just a friend but an entire grammar, a way of building meaning out of sound that had taken centuries to evolve and would never be heard spoken again.

Eyak was not unusual in its fate, only in how carefully it was documented on the way out. Across the world, languages are vanishing at a pace that has no precedent in recorded history. By common estimates, a language falls silent roughly every two to three months, and within the lifetime of a child born today, linguists fear that a large share of the world's tongues will follow. This article asks a deceptively simple question: how many languages are there, why are so many of them dying, and what exactly do we lose when the last speaker takes their final breath?

Counting the World's Voices, and Watching Them Fade

It is harder than it sounds to count the world's languages, because the line between a language and a dialect is partly political and partly arbitrary. Still, the most widely cited catalogue, the reference work known as Ethnologue, lists roughly 7,151 living languages in its 2024 edition. That number feels reassuringly large, but the headline figure hides an alarming detail: about forty percent of those languages, two in every five, are now considered endangered. At present rates, one of them goes extinct roughly every two to three months.

The bleak projection that launched the modern field came from the same linguist who recorded Marie Smith Jones. In a landmark 1992 article in the journal Language, Michael Krauss argued that the world stood to lose as much as ninety percent of its languages, by becoming either moribund or fully extinct, by the year 2100. A language he called moribund is one that is no longer being learned by children, which means it is alive today but already has no future, a speaker community whose youngest members are middle-aged or older. Krauss's article was less a prediction than an alarm, a deliberate call to document what could be documented and to act where action might still help.

A Map of Diversity That Is Wildly Uneven

The world's linguistic wealth is not spread evenly across its peoples. The single most striking fact about it is how skewed the distribution is. Just twenty-five languages, the familiar giants such as English, Mandarin Chinese, Spanish, Hindi, and Arabic, are spoken by roughly half of all human beings. The other half of humanity is divided among more than seven thousand smaller languages, many with only a few thousand speakers, and a great number with far fewer. It is among these smaller languages that endangerment is concentrated, which is why the loss can feel invisible from inside a major-language city: the languages disappearing are rarely the ones most people have ever heard.

Geography compounds the unevenness. Linguistic diversity clusters in a handful of hotspots. Papua New Guinea alone is home to more than 840 languages, the densest concentration on the planet, a consequence of rugged terrain that kept neighboring valleys isolated for millennia. Indonesia hosts roughly 700, Nigeria around 500, and India about 450. Those four countries together account for nearly forty percent of the world's living languages, packed onto a small fraction of its land. The places richest in languages, in other words, are not the places richest in money or power, and that mismatch turns out to be central to why languages die.

A Five-Level Scale for a Slow Disaster

To track the decline systematically, the cultural agency UNESCO published its Atlas of the World's Languages in Danger, first issued in 1996 and most recently revised in 2010. The atlas sorts languages onto a five-level scale that runs from vulnerable, through definitely endangered and severely endangered, to critically endangered, and finally to extinct. The crucial thing about this scale is what it measures. It is not primarily about how many people speak a language, but about transmission, the question of whether the language is still being passed to the next generation.

This distinction matters more than it first appears. A language with a million speakers, all of them elderly, with no children learning it, is in far graver danger than a language with two thousand speakers who are raising their toddlers in it. The first is severely or critically endangered no matter how large its present community, because the chain of transmission has snapped; the second, though small, is alive in the only sense that ultimately counts. A language is vulnerable when most children still speak it but its use is restricted to certain settings, and it is critically endangered when only the oldest generation, often the grandparents, remembers it, and even they speak it partially and seldom. Extinction arrives when the last fluent speaker, someone like Marie Smith Jones, dies.

Three Pressures That Push a Language Toward Silence

It is tempting to imagine that languages die because their speakers freely abandon them, the way one might give up an old habit. The reality is harsher and less voluntary. Language death is rarely a free choice. Linguists who study the process tend to point to three recurring drivers, distinct in mechanism but often working together.

The first is economic. When speaking a dominant language becomes the price of a job, an education, or a place in the market, families make a rational calculation and raise their children in the language that pays, sacrificing the heritage tongue for the sake of opportunity. The second is the collapse of intergenerational transmission, which is partly a consequence of the first: once parents stop speaking the language to babies, no amount of pride or nostalgia can keep it alive, because a language survives only by being learned anew in infancy. The third, and the darkest, is deliberate state suppression. For much of the nineteenth and twentieth centuries, governments across the world ran boarding schools and residential schools designed expressly to strip Indigenous children of their languages, punishing them for speaking the words their grandparents had taught them. These three forces, economic gravity, the broken chain of transmission, and outright coercion, account for most of the languages now sliding toward extinction.

The Two Centuries That Silenced the Most

If you wanted to find the most catastrophic loss of linguistic diversity in human history, you would look at the roughly two hundred years between 1800 and 2000. That period saw a linguistic mass extinction driven by colonial expansion, settlement, and the assimilation policies that came with them. The Indigenous languages of Tasmania were effectively silenced by the 1830s, within a single brutal generation of European settlement. In mainland Australia, about 150 Indigenous languages are now extinct or, in the term linguists prefer, sleeping, meaning no fluent speakers remain but documentation might one day allow a revival. The toll in the Americas is even larger; careful estimates put the number of languages lost there since European contact at around 1,500.

The word sleeping rather than dead is not mere euphemism. It reflects a real and important hope: that a language with no living speakers but a surviving written or recorded record is not necessarily gone forever. Whether that hope is warranted depends entirely on what happens next, which brings us to the most encouraging part of the story.

Languages That Came Back From the Edge

Endangerment is not destiny, and a small number of cases prove it. Three are cited so often that they have become the standard counter-examples to the assumption that language death is inevitable.

The most dramatic is Hebrew. For roughly two thousand years it survived as a language of liturgy, scholarship, and prayer, but essentially no one spoke it at home as a native tongue. Beginning in 1882, a determined activist named Eliezer Ben-Yehuda set out to change that, insisting on raising his own son entirely in Hebrew and campaigning relentlessly to expand its vocabulary for everyday modern life. Within a few generations, Hebrew had become the living, native, daily language of millions, the only fully successful revival of a language that had no native speakers at all. Welsh offers a quieter, partial success: long suppressed and shrinking, it has recovered considerable ground since the early 1990s, with roughly thirty percent of the population of Wales now able to speak it, many of them as a learned second language supported by schools and broadcasting. The third case is the most remarkable for its starting point. Wampanoag, also known as Wôpanâak, had no speakers at all for several generations, but beginning in 1996 a linguist named Jessie Little Doe Baird reconstructed it from seventeenth-century missionary texts and documents written in the language itself, and her own daughter became the first native speaker of Wampanoag in more than a century.

These successes share a common thread, and also a sobering lesson. Each took decades of patient, deliberate work, and each depended on a community that wanted the language back badly enough to teach it to its children. Revival is possible, but it is slow, fragile, and never automatic.

What Vanishes When the Last Speaker Dies

It is easy to think of a language as a swappable code for the same set of meanings, as if losing one were no worse than losing one currency when others remain. But this misunderstands what a language is. The anthropologist Franz Boas, a founding figure of modern American anthropology, argued that every language is a complete classificatory system, a particular way of carving the flow of experience into categories. Languages differ not just in their words but in what distinctions they force their speakers to make, in how they bundle space, time, kinship, color, and cause into the units of thought. When a language dies, what is lost is not merely a vocabulary but an entire cosmology, a way of organizing the world that no other language reproduces.

This is also where the limits of documentation come into focus. A documented language is a recovered grammar, a recorded corpus of speech, and a dictionary, and these are genuinely valuable; they are what made the revival of Wampanoag conceivable. But a documented language is not a living one. A living language is a community that speaks it to its babies, that argues and jokes and grieves and bargains in it across every domain of life. Documentation can make revitalization possible, and that is no small thing, yet only transmission, the act of passing the language to a new generation in infancy, actually keeps it alive. The recordings Michael Krauss made of Marie Smith Jones preserve a precious record of Eyak, but they cannot, on their own, bring it back to a single dinner table.

Key Takeaways

The world today holds roughly 7,151 living languages, but the figure is misleading: about half of humanity speaks just twenty-five dominant tongues, while the rest is spread across more than seven thousand smaller languages, two in five of which are now endangered, with one going extinct roughly every two to three months and Michael Krauss's 1992 projection warning of up to ninety percent loss by 2100. Diversity clusters in a few hotspots, with Papua New Guinea, Indonesia, Nigeria, and India holding nearly forty percent of all languages, and UNESCO's five-level scale tracks decline not by speaker count but by whether children are still learning the language. Languages rarely die by free choice; they fall to economic pressure, the collapse of intergenerational transmission, and deliberate state suppression, forces that drove a linguistic mass extinction between 1800 and 2000 in which Tasmanian languages, around 150 Australian languages, and some 1,500 languages of the Americas were lost. Revival is possible, as Hebrew, Welsh, and Wampanoag show, but it is slow and depends on a community willing to teach the language to its young; and what is finally at stake, as Franz Boas understood, is not a vocabulary but a whole way of categorizing the world, which documentation can record yet only living transmission can keep alive.