Why certain languages are vanishing in two generations while others nearly dead are reviving

On July 31, 1882, a boy was born in Jerusalem who had no name for what he was. His father, Eliezer Ben-Yehuda, had arrived in the city the previous year with a plan that everyone around him considered somewhere between quixotic and deranged: he would speak only Hebrew, raise his children in Hebrew, and in doing so resurrect a spoken vernacular after approximately 1,700 years in which the language had served for prayer and scholarship but not for groceries, arguments, or asking to stay up late.

The boy — Itamar Ben-Avi — grew up to be the first native Modern Hebrew speaker in recorded history. Scholars of language revival, most notably Ghil’ad Zuckermann, use the Hebrew case as the single documented instance of a complete vernacular resurrection: a language that ceased to function as a daily spoken tongue for roughly seventy generations, now spoken by approximately seven million people as their first language.

So here is the question this makes necessary. If a dead language can be pulled back from 1,700 years of disuse, why are thousands of languages dying right now — while fluent speakers are still alive? The answer, once you understand what the Hebrew revival actually required, makes the situation for most dying languages look considerably worse, not better.

The direction of travel

Papua New Guinea has approximately 840 languages spoken by a population of around ten million — more than any other country on earth, over 12% of the world’s total. Nobody is suppressing them. The PNG government has no campaign against indigenous language use. No colonial authority currently bans Huli, Kewa, or Melpa.

A 2021 study published in PNAS (Kik et al., PMC8179190) surveyed 6,190 secondary school students across 30 schools, covering speakers of 392 languages — 46% of PNG’s languages and 5.5% of global linguistic diversity. The headline finding: approximately 91% of parents were fluent in their indigenous language. Among the students: 58%. The projected rate for the generation after them: 26%.

Three generations. No banning required.

The drivers the study identified are not mysterious: urbanization, cash-economy participation, mixed-language households (only 16% of which used indigenous languages at home, versus 38% of same-language households), and formal schooling conducted in Tok Pisin and English. A language that doesn’t open doors gets left behind.

Scale outward. Of the world’s approximately 7,168 living languages, around 3,078 — roughly 43% — are classified as endangered according to Ethnologue and UNESCO data. In Brazil, roughly 1,000 indigenous languages existed at Portuguese arrival in 1500. The 2022 IBGE census counted 295 still spoken. Languages continue to disappear through the same urbanization drawing communities into Portuguese-mediated cash economies, with children learning in schools where Portuguese is the language of instruction and advancement. Many of the surviving languages have fewer than fifty speakers. Some have fewer than ten: a single extended family, perhaps, in a riverside settlement where Portuguese is the language of school, of trade, of everything beyond the household threshold.

In Siberia, the Ket language retains approximately 200 fluent speakers among a Ket population of roughly 1,200. Across approximately 41 distinct indigenous Siberian ethnic groups, most languages are in critical or terminal stages of endangerment.

The pattern is identical across places whose political histories could hardly differ more: a post-colonial Pacific state, an independent South American republic, what remains of the Soviet sphere. Economic integration arrives; the dominant language follows the economy; children grow up making different calculations than their grandparents made. The political story varies. The linguistic outcome doesn’t.

The UNESCO endangerment scale

UNESCO classifies living languages on five levels: Vulnerable (most children speak it, though use may be restricted to certain domains), Definitely Endangered (children no longer learn it as mother tongue at home), Severely Endangered (spoken by grandparents; parents may understand but don't pass it to children), Critically Endangered (youngest speakers are grandparents or older, minimal transmission), and Extinct. The PNG trajectory — 91% parental fluency to 58% student fluency to a projected 26% — moves a language community from Definitely Endangered toward Critically Endangered within a single lifetime, without any external political pressure. The Expanded GIDS, used by Ethnologue, adds further granularity at both ends of the scale.

What colonialism actually did

The objection forms immediately: colonialism created these economic conditions. The British empire, Spanish conquest, Soviet Russification — these weren’t neutral market forces; they were coercive restructurings of entire societies designed to make indigenous people economically dependent on the colonizer’s language. True. But the objection, correctly understood, points to the mechanism — it’s economic — rather than rescuing colonialism as a stand-alone explanation for what is happening now. The economic logic persists long after the political authority that created it has gone.

The Welsh case is the most instructive test. The Welsh Not — the wooden plaque hung around children’s necks as punishment for speaking Welsh in school, in use through the mid-19th century — is the image that survives in cultural memory as the symbol of English language suppression. By the 1921 census, Welsh had approximately 37% of the population as speakers. The Welsh Not was long gone. Then came the postwar economic expansion: coal and steel, English-speaking urban employment, the welfare state administered in English. By 1971, Welsh had fallen to approximately 21% of Wales’s population. The Welsh Not is remembered. The Valleys employment boom is not. The suppression happened and the language survived it. Prosperity nearly killed it.

The decisive test is Ireland. British rule suppressed Irish: by 1851, the language had fallen to around 23% of the population. Then independence. Ireland made Irish the first official language of the Republic, mandated it in schools, spent heavily on preservation in the Gaeltacht. The language continued to decline. Parents in the Gaeltacht — not under coercion — shifted toward English because English meant opportunity: within the Republic, and especially in emigration to Britain and America. Reg Hindley’s The Death of the Irish Language (Routledge, 1990) argues that the transmission shift in Gaeltacht communities was organic rather than imposed. The state wanted revival. The economic logic hadn’t changed. Revival failed.

Soviet policies in Siberia represent the genuinely coercive case: explicit Russification through boarding schools and forced settlement. These schools removed children from their communities for months or years at a stretch, ensuring they became functionally fluent in Russian before they’d consolidated their native language at home. The suppression and the economic restructuring were entangled; coercion accelerated what economic pressure was already beginning to accomplish. But the fatal blow, as everywhere else, came when intergenerational transmission failed. Once children stopped being raised in the indigenous language at home, no degree of adult cultural attachment could reverse the process.

Remove colonial political authority entirely, replace it with a government committed to the language’s survival, and if the economic conditions don’t shift, the language still dies. The colonialism frame gets the history right but misidentifies the mechanism.

The threshold

In 1991 the sociolinguist Joshua Fishman published Reversing Language Shift (Multilingual Matters), which introduced the Graded Intergenerational Disruption Scale. The GIDS runs from Stage 8 — a handful of elderly speakers with no community function — down to Stage 1, where a language is fully operational across institutional life. The hinge is Stage 6: intergenerational transmission within the home and family.

Fishman’s central argument, now standard in the field, is that everything else is secondary if Stage 6 fails. A language can have official status, a literary canon, government subsidies, broadcast media, and university departments, and still die if parents don’t speak it to their children. And conversely — the underappreciated half — a language with none of those things survives as long as grandmothers are speaking it to grandchildren in the kitchen.

The PNG data makes Stage 6 failure visible in real time. In the Kik et al. study, 88% of students said they intended to teach their indigenous language to their own children. Eighty-eight percent. The intention is there. The trajectory is 91% to 58% to 26% regardless. The gap between what people say they will do and what the economic logic actually produces is where the mechanism lives: when the costs of not transmitting the dominant language are immediate and concrete — worse opportunities, smaller world, worse life outcomes for your child — and the costs of not transmitting the indigenous language are diffuse and future-facing, the outcome is not really a choice so much as a structural gravity.

The threshold is crossed not in a decisive moment but in an accumulation of ordinary mornings. Someone switches to Tok Pisin or English because this particular conversation requires it, then the next, and the children grow up mostly in one language and partially in another, and by the time the following generation arrives the other tongue has become a heritage artifact rather than a mother tongue. The moment a language loses its last generation of fluent adults who could teach it to children is irreversible. Language acquisition requires fluent-speaker input — a constraint that cannot be funded or legislated away. What makes this particularly hard to address is that the threshold is almost impossible to perceive in real time. The last generation of fluent adults does not know it is the last. The decisions accumulate silently, house by house, morning by morning, and only become legible in retrospect — when the generation that might have acted is already past the window where it was possible.

Fishman's Graded Intergenerational Disruption Scale (GIDS)

Stage 8: Community members remember but no active speakers remain. Stage 7: Elderly speakers exist but have no contact with children — the last window for intervention. Stage 6: Intergenerational transmission within the home: the critical threshold. Stage 5: Literacy tradition in the language among some adults. Stage 4: Language taught in schools but not as medium of instruction. Stage 3: Use in local work contexts below professional level. Stage 2: Active in lower-level government and local media. Stage 1: Full institutional use across education, government, and national media. Fishman's prescription: secure Stage 6 before attempting Stage 1 interventions. The Expanded GIDS (EGIDS), now used by Ethnologue, adds granularity at recovery stages above Stage 1.

The three that came back

Ben-Yehuda understood Stage 6 instinctively, decades before Fishman named it. His solution was to start there — building the Hebrew-speaking household from which everything else would have to grow — then working outward to create the social and institutional context that would make more such households possible, then necessary. The Language Committee (Va’ad HaLashon), founded in 1890, standardized vocabulary and coined terms for modern objects and concepts.

The mechanism wasn’t persuasion. It was engineering conditions under which Hebrew was the only viable option. The Yishuv — the pre-state Jewish community assembling in Palestine — drew immigrants speaking Yiddish, Ladino, Arabic, Russian, and Polish, with no common vernacular. Hebrew was not one attractive option among several; it was the only language that could function as a shared tongue for this specific community in this specific place. Economic utility wasn’t argued for; it was manufactured from the structure of the situation. By 1922, the British Mandate recognized Hebrew as one of three official languages of Palestine. By 1948, it was the primary language of the new Israeli state.

The Yiddish question

The Hebrew revival was historically contingent in ways the standard account tends to suppress. At the founding of the Yishuv, Yiddish had more speakers than Hebrew and a rich modern literary culture — Isaac Bashevis Singer would win the Nobel Prize for Literature in 1978 for work written in it. Whether Hebrew or Yiddish would become the dominant language of a Jewish state was not a settled question before the Second World War. The Holocaust destroyed the Yiddish-speaking communities of Eastern Europe; approximately six million people who might otherwise have argued, in demographic and practical terms, for Yiddish as a state vernacular were killed. Benjamin Harshav's Language in Time of Revolution (University of California Press, 1993) examines the social conditions of the Hebrew revival and the competing claims of Yiddish in this period. It doesn't diminish the Hebrew revival to note that its conditions included the erasure of the most plausible competitor. It does complicate the idea that it's a cleanly replicable model.

Welsh and Māori achieved something more modest: they slowed the collapse and built conditions under which stabilization might eventually hold. Given what each started with — millions of speakers for Welsh, a far narrower speaker community for Māori — the slowing is the result.

Welsh had been declining since the 1921 census recorded approximately 37% of Wales as Welsh speakers. By the 1970s the trajectory was unambiguous. The response came in waves: S4C, the Welsh-language television channel, in 1982; the Education Reform Act 1988 making Welsh a core subject in Welsh-medium schools; the Welsh Language Act 1993 giving Welsh equal official status with English; the Welsh Language (Wales) Measure 2011 establishing enforceable standards for public-service use. Welsh-medium schools now serve approximately 22–23% of Welsh pupils. The 2021 UK Census counted 538,300 Welsh speakers — 17.8% of the population, down from 562,016 (19%) in 2011. The Welsh government’s Cymraeg 2050 target aims for one million speakers by 2050.

Welsh is in managed decline. The intervention has slowed the rate of loss; it hasn’t reversed it. What Welsh had that most dying languages don’t: millions of speakers, territorial identity, UK Parliament-level political infrastructure, and Welsh-medium institutions that genuinely need Welsh speakers to function.

The Māori case arrived at the same crossroads through different political architecture and with a narrower window. The kōhanga reo — “language nest” — movement in New Zealand began in 1982 with a specific and time-limited logic: there were still fluent kaumātua (elders) available to staff total-immersion programs for infants. Five founding locations expanded to more than 300 within three years. Te reo Māori was recognized as an official language of New Zealand under the Māori Language Act 1987. The Te Kupenga 2018 survey found nearly 48% of Māori adults reporting some ability to speak te reo Māori; the 2021 New Zealand General Social Survey found 7.9% of all New Zealanders speaking te reo Māori at least fairly well, up from 6.1% in 2018.

Had the kōhanga reo launched a decade later, the fluent elders who staffed it would have been gone. The window was that narrow.

Why the model doesn’t travel

Three cases. One complete vernacular resurrection. Two instances of managed decline that may be stabilizing. What each of them required that most of the world’s 3,078 endangered languages simply don’t have:

State-level institutional power. The Zionist movement operated at the scale of a state-in-formation. Welsh had the UK Parliament, the Welsh Assembly, devolved government with real budgets and legislative authority. Māori had Treaty of Waitangi jurisprudence — a legal framework giving Māori communities standing to compel state action. A language with 200 speakers in a PNG highland village has none of this. The Kik et al. study documented 88% of students intending to transmit their indigenous language. That intention, without institutional support, means those students will raise their children in Tok Pisin and English anyway — because that’s what the schools teach, what the economy operates in, what the media speaks.

A critical mass of fluent adult speakers still alive when the intervention starts. Every revival required them. Ben-Yehuda’s generation had studied Hebrew textually and could speak it functionally. Welsh adults in the 1980s were Welsh-medium speakers who could teach, broadcast, and staff institutions. The kōhanga reo worked because fluent kaumātua were still available in 1982. For the hundreds of languages now at Stage 7 or Stage 8 of Fishman’s scale, that window has already closed. There are no fluent adults left who could run an immersion program. Documentation is still possible; revival is not. These are different things, and conflating them is how funding for documentation gets diverted into revival programs that cannot produce results.

An economic argument that operates in daily life. Hebrew speakers in the early Yishuv couldn’t function without Hebrew — it was the only language their colleagues shared. Welsh speakers in Wales exist in a market where some employers require Welsh, particularly in public service and Welsh-medium education, and where Welsh identity carries real institutional weight. Māori has been embedded in New Zealand government services, media, and schooling. The cultural argument for a language’s survival — that it carries irreplaceable knowledge, embodies a unique way of making sense of the world — is true. It has never, in any documented case, reversed language shift at scale on its own. The argument addresses the wrong actor. It speaks to guilt and conscience — to the linguist, the policymaker, the journalist writing the elegy. The people making the actual transmission decision are parents calculating what language will give their child a viable future, and they can see that future clearly.

The PNG data states it plainly: three generations, 91% to 58% to a projected 26%, in a country where the state is not hostile to indigenous languages, where communities say they want to preserve them, where 88% of students said they intended to pass them on. The mechanism proceeds regardless, because it is economic and does not require hostility or malice to operate.

For most of the world’s 3,078 endangered languages, the conditions for revival do not exist and will not exist. That is what the evidence shows.

Somewhere right now, a fluent speaker of a language with two hundred remaining speakers is talking to a grandchild in the dominant regional tongue. Not because she was told to. Not because she forgot her own language. Because the grandchild’s future is going to require the dominant tongue, and she loves the grandchild, and that’s what you do when you love someone — you give them what their life will need.

Nothing marks the moment. No ceremony, no formal decision. The language just doesn’t get passed down in this household this morning, or the next, or the mornings after that. Eventually the window closes. When it does, the question of revival becomes moot — there are no fluent speakers left who could run the program, staff the school, be the grandmother in the kitchen.

The Kik et al. projection — 26% for the generation after the students surveyed — is not a warning. It’s a positional marker: evidence of how far through that threshold process entire linguistic communities have already traveled, in a country where nobody is suppressing anything, where communities want preservation, where 88% of young people said they’d pass the language on. The 26% is the gap between intention and economic structure, measured across a generation.

Hebrew was pulled back. For most of the languages dying now, the conditions that made even the partial recoveries of Welsh and Māori achievable are already foreclosed. What remains is documentation, and the honest acknowledgment that most of what is being lost will not come back.

Gen AI Disclaimer

Some contents of this page were generated and/or edited with the help of a Generative AI.

Media

Square Jehan-Rictus with the plaque with je t’aime = eu te iubesc in 311 laguages – Wikipedia

Key Sources and References

Kik, Alfred, et al., “Language and Ethnobiological Skills Decline Precipitously in Papua New Guinea, the World’s Most Linguistically Diverse Nation,” Proceedings of the National Academy of Sciences, 2021, PMC8179190.

Reshef, Yael, “Revival of Hebrew: Sociolinguistic Dimension,” in Khan, G. et al. (eds.), Encyclopedia of Hebrew Language and Linguistics, Brill, Leiden/Boston, 2013, vol. 3, pp. 408–415.

Zuckermann, Ghil’ad, Revivalistics: From the Genesis of Israeli to Language Reclamation in Australia and Beyond, Oxford University Press, New York, 2020.

Fishman, Joshua A., Reversing Language Shift: Theoretical and Empirical Foundations of Assistance to Threatened Languages, Multilingual Matters, Clevedon, 1991.

Hindley, Reg, The Death of the Irish Language: A Qualified Obituary, Routledge, London, 1990.

Harshav, Benjamin, Language in Time of Revolution, University of California Press, Berkeley, 1993.

Office for National Statistics, “Language, England and Wales: Census 2021,” ONS, 2022.

Office for National Statistics, “Language in England and Wales: Census 2011,” ONS, 2013.

Welsh Government, Cymraeg 2050: A Million Welsh Speakers, gov.wales, 2017.

Stats NZ, Te Kupenga 2018, Statistics New Zealand, 2019.

Stats NZ, New Zealand General Social Survey 2021, Statistics New Zealand, 2022.

Ethnologue: Languages of the World, 26th edition, SIL International, 2023.

UNESCO, Atlas of the World’s Languages in Danger, 3rd edition, UNESCO Publishing, Paris, 2010.

IBGE (Instituto Brasileiro de Geografia e Estatística), Censo 2022: Indígenas, IBGE, Rio de Janeiro, 2023.

Russian Federal State Statistics Service (Rosstat), All-Russian Population Census 2010, Rosstat, Moscow, 2012.

Central Statistics Office Ireland, Irish Language Historical Census Data 1851–2016, CSO, Dublin.