In the hills of western Nepal, a few dozen people have some command of Kusunda, a language isolate with no demonstrated relatives anywhere. Nepal's 2021 census counted 253 ethnic Kusunda and 23 self-reported native speakers; the one person able to sustain a conversation in it is Kamala Sen Khatri, whose elder sister held that position before her.
The Kusunda were forest foragers, called Ban Raja ("kings of the forest") in Nepali; their own name for themselves is recorded variously as Mihaq, Myahq or Gemyehak. The language was twice presumed dead. After settlement and intermarriage scattered the community, it resurfaced in 2004, when three Kusunda who had travelled to Kathmandu for citizenship papers turned out to include a fluent speaker from Sakhi, in Rolpa district. By 2005 seven or eight fluent speakers were known, the youngest in her thirties. When the speaker Rajamama Kusunda died on 19 April 2018, the language was again presumed extinct, although the sisters Gyani Maiya and Kamala were still alive and fluent; Gyani Maiya Sen Kusunda, who had helped relaunch interest in it, died on 25 January 2020 at 85.
Its isolation has drawn a long list of rejected matches. B. K. Rana argued in 2002 that Kusunda was Tibeto-Burman, like most of its neighbours; Watters's 2005–2006 grammatical description concluded that it is not. Others have tried to attach it to Nihali (Whitehouse 1997), to the Yeniseian languages of central Siberia (Gurov 1989), to Burushaski and the Caucasian families (Reinhard and Toba 1970), and — in a 2014 proposal by George van Driem — to Yeniseian and Burushaski together. None of these has been demonstrated.
Part of the reason is that Kusunda neither sounds nor works like the languages around it. It has uvular consonants, which no other language in the region uses, and it lacks the retroflex consonants that are everywhere in South Asia. Its case system is nominative–accusative with no ergativity, again unlike its Tibeto-Burman neighbours. In a 2014 reanalysis, Mark Donohue, Bhoj Raj Gautam and Madhav Pokharel found that ordinary declarative clauses have no unambiguous way to mark negation: the irrealis suffix -u can carry either a positive or a negative reading, and speakers lean on a separate negative existential to say that something is not so.
Since 2019 the researcher Uday Raj Aaley has run Kusunda classes with Kamala Sen Khatri as resource person and support from Nepal's Language Commission. Four residential sessions — 2019, 2021, 2022 and a final one in March 2023 — came to about 270 hours of teaching; the first drew 20 students in grades four to twelve, three of them not Kusunda at all. Aaley and Timotheus Bodt call the work an awakening: the language had been essentially dormant for 50 to 60 years, and with no living speech community to join, every learner starts from zero.
The obstacles are concrete. There is no settled orthography — the uvular sounds are hard to render in Devanagari — and no textbook, audio or video teaching materials; Aaley is still writing a first book of everyday words and phrases. Everyone in the community now speaks Nepali, and intermarriage means no child is acquiring Kusunda at home. Whether a language learned deliberately, as a heritage subject, can be carried to a generation that speaks it without thinking is the question the classrooms in Dang are now testing.
| Meaning | Watters (2006) | Aaley & Bodt (2019) |
|---|---|---|
| bird | qotu | qo.tow |
| one | qasti | qɐs.ti |
| wind | qaːi | qɐj |
| drink! (command) | qon | qɔn |
| snow, hail | yaq | jaq |
| water, rain (no uvular) | taŋ | tɐŋ |