The Question

An elderly speaker of an endangered language sitting with a young child in a rural setting

A language does not usually die with a bang. It fades. One generation speaks it at home, the next understands it but answers in the national tongue, and the one after that knows only a few words. When the last fluent elder passes, an entire way of naming the world — its plants, its stories, its jokes, its map of kinship — goes quiet, often without anyone noticing the exact moment it happened.

There are about 7,000 living languages on Earth, but they are distributed with astonishing unevenness. A few dozen dominant tongues are spoken by billions, while thousands of others have only hundreds or dozens of speakers left. According to Ethnologue, roughly 44% are endangered, while the world's 20 largest languages are the native tongue of more than 3.7 billion people. The question is stark: how many of these will survive the century, and what does it mean for human knowledge when so much of it lives only in the memory of ageing speakers who have no one to pass it to?

What the Evidence Shows

The trajectory is not speculative; it is being measured. Catalogues of the world's languages classify around 40% of them as threatened, and studies of speaker demographics find that a language with only elderly speakers and no child learners is, in practical terms, already at the end of its life. When the youngest fluent speaker is sixty, the clock is simply running out. Linguists have long warned that half or more of all languages could disappear this century, an estimate popularised in the 1990s. The most detailed recent modelling is more cautious. A study led by Lindell Bromham at the Australian National University, published in Nature Ecology & Evolution, analysed 6,511 languages against 51 possible pressures and found that without intervention, the rate of loss could triple within 40 years and about 1,500 languages could cease to be spoken by 2100. Surprisingly, contact with other Indigenous languages was not a driver of loss, while road density and more years of schooling — usually in a dominant language — were linked to greater endangerment.

The pressures are consistent across every continent. Urbanisation pulls young people from villages where minority languages survive into cities where a dominant language is the currency of work and schooling. National education systems teach and test in a single official tongue. Media, the internet, and now translation technology reward the biggest languages and quietly starve the smallest. A parent who wants a child to prosper often chooses, understandably, to raise them in the language of opportunity.

"We found that without immediate intervention, language loss could triple in the next 40 years. And by the end of this century, 1,500 languages could cease to be spoken."

— Professor Lindell Bromham, Australian National University, co-author of the study in Nature Ecology & Evolution

Yet the same century that threatens languages also offers unprecedented tools to record and even revive them. Digital archives can capture a fluent speaker's grammar and vocabulary in weeks. Apps teach endangered tongues to new learners. And there are real revival stories — communities that pulled languages back from the brink through schools, media, and stubborn intergenerational effort. Documentation is not the same as survival, but it changes what "extinction" means when a language can be brought back from a recording.

When the last person who dreams in a language closes their eyes, a whole way of seeing the world closes with them.

Why This Is Happening

Economics favours the big languages. A child's future income, education, and mobility usually depend on fluency in a national or global tongue. Faced with that reality, families across the world make the rational choice to prioritise the dominant language, and the minority one slips out of daily use within a generation or two. No villain is required; the market for opportunity does the work on its own.

Transmission breaks at the home, not the school. A language survives only if children learn it naturally from those around them. Once a generation stops speaking it to their infants, formal classes rarely fill the gap. The critical failure point is the everyday household, which is exactly where migration, mixed marriages, and screen-based media in dominant languages exert the most quiet pressure.

The distribution was fragile to begin with. Most of the world's languages never had many speakers — hundreds or a few thousand, often in remote communities. A tongue that small can be tipped into decline by a single road, a school, or a wave of emigration. The sheer top-heaviness of global language, where a handful dominate and thousands cling on, means the long tail was always exposed.


What Could Happen

Half of today's languages fall silent by 2100 Less likely

The pressures of urbanisation, dominant-language schooling, and broken home transmission continue, and thousands of small languages lose their last fluent speakers over the coming decades. Many are documented before they go, preserved in archives even as living use ends. The world's spoken diversity contracts sharply toward a few hundred robust languages.

Heavy losses, but well short of half Most likely

Cheap documentation, translation tools that finally include small languages, and community-led revival movements keep more tongues in active use than expected. Extinctions still occur, but fewer than half are lost as digital tools lower the cost of learning and speaking minority languages. Diversity shrinks but less drastically.

Loss accelerates beyond current estimates Unlikely

Rapid urbanisation, conflict, and climate displacement scatter small language communities faster than anyone can record them, and more than half vanish well before 2100. Documentation cannot keep pace with the disruption. This grim path is possible but less likely than a steady, partly-mitigated decline.

Our Assessment
We assign 15% probability — unlikely that by 2100, at least half of today's roughly 7,000 languages will be extinct or effectively silent. We have lowered this forecast from 66%: the most detailed recent modelling projects about 1,500 languages lost by 2100 without intervention — roughly a fifth — and many languages counted as endangered still have fluent speakers who could pass them on. The loss is still severe: losing even a fifth of the world's languages would mean the disappearance of thousands of years of knowledge, and the rate of loss could triple within decades unless communities get support.

What Can We Do

A researcher recording an elderly speaker with audio equipment to document an endangered language

Language loss can feel like weather — vast and impersonal. But languages live in specific families and communities, and that is exactly where they can be defended.

Speak the small language at home. The single most powerful act of preservation is raising children with a heritage language in daily life, even alongside a dominant one. Bilingual children lose nothing and gain a whole inheritance. Everyday use in the household matters more than any classroom or app.

Support documentation while speakers are living. Recording fluent elders — their stories, grammar, and vocabulary — captures a language before it is lost and gives future revivals something to build on. Back the archives, universities, and community projects doing this work. A recorded language is never fully gone.

Demand small languages in technology. Translation tools, keyboards, and voice systems increasingly decide which languages feel usable in modern life. Push the companies building them to include minority tongues. Visibility in the digital world can turn a dying language into a living, texting, searchable one again.

Treat linguistic diversity as heritage worth funding. Just as nations protect historic buildings and rare species, they can fund bilingual schooling, media, and community programmes for endangered languages. Support the policies that make this possible. What is preserved deliberately tends to survive; what is left to chance usually does not.