Endangered Languages: The Race to Document the World's Disappearing Tongues

A Language Dies Every Two Weeks

Linguists estimate that between 50 and 90 percent of the world's approximately 7,000 languages will cease to be spoken by the end of the twenty-first century. The loss rate—roughly one language every two weeks—surpasses by orders of magnitude the extinction rate of known biological species. Unlike a species extinction, language death leaves no physical artifact. When the last speaker of a language dies without recorded documentation, the unique phonology, grammar, and lexicon—encoding millennia of ecological knowledge, history, and cognitive diversity—disappears irrecoverably. The numbers are unambiguous.

The causes of language shift are not mysterious. Dominant national or colonial languages provide access to education, employment, and government services. Parents who want their children to succeed often choose not to transmit a minority language. The decision is rational from an individual standpoint and catastrophic for the language community collectively.

UNESCO's Endangerment Framework

The UNESCO Atlas of the World's Languages in Danger (third edition, 2010) categorizes endangerment across six levels based on the intergenerational transmission and overall vitality of each language:

Why Languages Die: Structural Factors

Language death follows predictable socioeconomic patterns. Geographic concentration of speakers in remote or economically marginal areas initially protects a language through isolation but makes the community vulnerable to displacement or economic marginalization. Contact with a dominant language—through schooling, migration, media, or military occupation—triggers language shift when the dominant language offers material advantages the heritage language cannot provide.

The mechanisms of shift are gradual:

• Parents adopt the dominant language with children while maintaining the heritage language with elders
• Young adults use the heritage language passively but prefer the dominant language in active production
• A generation of incomplete learners acquires a simplified, reduced variety that linguists call an obsolescent variety
• The final speaker generation has no interlocutors and the language loses its primary function as a communication tool

Language death is rarely abrupt. The typical trajectory from full vitality to extinction spans 100 to 200 years—long enough for the process to be invisible generation by generation, short enough to be irreversible within a single lifetime of inaction.

Ainu: Two Fluent Speakers in 2020

Ainu, the indigenous language of the Ainu people of Hokkaido, Sakhalin, and the Kuril Islands, was estimated to have only 2 to 10 fluent speakers in 2020, with a broader group of partial speakers and language learners potentially numbering in the hundreds. The Japanese government did not officially recognize the Ainu as an indigenous people until 2019. Decades of assimilation policies prohibited Ainu cultural and linguistic expression in public schools and suppressed transmission across generations.

Ainu is a language isolate—no demonstrated genealogical relationship to any other language has been established, though proposed links to Japanese, Austronesian, and various Siberian language groups have been advanced without scholarly consensus. Its typological features (verb-final with polysynthetic morphology, elaborate evidentiality system) are documented in recordings, grammars, and dictionaries compiled over the twentieth century. The Upopoy National Ainu Museum, opened in Hokkaido in 2020, incorporates Ainu language revitalization as part of its mission, but the gap between two fluent elderly speakers and a community of young learners is enormous.

Cornish: A Language Raised from the Dead

Dolly Pentreath died in 1777, widely cited as the last native speaker of Cornish—the Celtic language of Cornwall in southwestern England. By the early twentieth century, Cornish was extinct as a community language. Yet in 2021, the UK government recognized Cornish as a living language under the European Charter for Regional or Minority Languages, citing a revival community of several thousand speakers and hundreds of children receiving Cornish-medium instruction.

The Cornish revival is built on a reconstructed standard drawn from medieval manuscripts, primarily the Cornish mystery plays of the fifteenth and sixteenth centuries. Three competing orthographic standards—Unified Cornish, Kernewek Kemmyn, and Revived Late Cornish—produced decades of rivalry before the Standard Written Form was adopted in 2008. The revived language is not identical to historical Cornish—no living variety could be—but it provides a vehicle for community identity and cultural continuity. Whether it will achieve full intergenerational transmission remains an open question.

Documentation: Racing the Clock

Language documentation is the systematic creation of a lasting, multifaceted record of a language—its grammar, lexicon, and use in natural social contexts. The Endangered Language Documentation Programme (ELDP), based at SOAS University of London, funded over 200 documentation projects between 2002 and 2019 before merging into the Endangered Languages Archive (ELAR). ELAR holds over 700 collections documenting hundreds of endangered languages in audio, video, and text form.

Documentary linguists face practical constraints that have no equivalent in other scientific fields:

• Speaker communities may distrust outsiders with recording equipment, especially in post-colonial contexts where language documentation was previously associated with government surveillance
• Elders who hold the most complete knowledge of a language may be physically frail or geographically remote
• The communities themselves must determine what is shared publicly and what is restricted—sacred knowledge, ceremonial language, and gendered speech registers are often off-limits to general documentation
• Funding timelines (typically 2–3 years) are poorly matched to the decade-scale commitment that serious community-based documentation requires

Language Nests: The Proven Model for Revitalization

The language nest (kōhanga reo in Māori) model, pioneered in New Zealand beginning in 1982, immerses young children (birth to school age) in the target language during the critical period for language acquisition. Adults who are fluent in the heritage language run the program entirely in that language, creating an environment of natural, comprehensible input before schooling begins in the dominant national language.

The model spread from Māori to Hawaiian (pūnana leo), Welsh, Gaelic, and dozens of other communities worldwide. Evaluations of Māori kōhanga reo graduates show significantly higher rates of Māori proficiency than non-participants. The model works—under one condition: a sufficient number of fluent adult speakers must be available to staff the programs. For languages with very few remaining speakers, the model cannot be implemented without first training a new generation of speakers from recordings and grammatical documentation.

Digital Tools and the Preservation Frontier

Modern technology has transformed what is possible in language documentation and revitalization:

• Mobile phone applications (Aikuma, Lig-Aikuma) enable community members to create high-quality recordings without linguist involvement
• Machine learning and forced alignment tools automatically segment and time-stamp recordings, dramatically accelerating transcription
• Online learning platforms enable diaspora speakers to maintain heritage language competence from anywhere
• Text-to-speech synthesis has been built for languages including Māori and Welsh, enabling digital assistants and audiobooks
• Wikimedia projects host content in over 50 endangered and minority languages, providing both a learning resource and a venue for community-generated written production

Digital preservation solves the archival problem but not the transmission problem. A language stored on servers but not spoken by children is a linguistic museum exhibit, not a living language. The distinction between documentation (recording what exists) and revitalization (restoring intergenerational transmission) is the central tension in contemporary endangered language work—and the one that most urgently requires community decision-making, not just academic expertise.