Almost everything known about Shompen comes from thin, unreliable sources. The first word list dates to 1875, compiled by a Danish colonial officer named De Roepstorff. E.H. Man added scattered notes in 1886 and a comparative list in 1889, observing that very few Shompen words bore any resemblance to Nicobarese, the language of their coastal neighbors. Then, silence. An entire century passed before new data appeared: seventy words published in 1995, followed by a more extensive collection in 2003. But linguists Roger Blench and Paul Sidwell found the 2003 book riddled with anomalies and inconsistencies -- conventions reversed, transcriptions that contradicted standard practice. They concluded it had likely been copied from an earlier, possibly colonial-era source. George van Driem, another specialist, found the material too difficult to work with at all.

Where Shompen fits in the tree of human languages remains genuinely uncertain. Convention groups it with the Nicobarese languages, a branch of the Austroasiatic family that includes languages spoken in mainland Southeast Asia. But the evidence for that grouping was thin throughout the twentieth century, resting more on geographic proximity than on demonstrated linguistic kinship. Man noted in 1886 that words often differed entirely between the two Shompen groups he encountered -- "to bathe" was something like pugoihoap in one community and hohom in another. Some of these differences may reflect borrowed versus native vocabulary, since koi for "head" appears to be a Nicobarese loan. Glottolog, the major linguistic database, has treated Shompen as a language isolate. Van Driem, working from 1997 data, concluded it was indeed Nicobarese. The debate is unresolved.

What little is known about Shompen phonology reveals intriguing patterns. The language appears to have eight vowel qualities, each of which can be nasalized or lengthened, producing a rich system of vowel contrasts. Its consonant inventory includes aspirated stops, a glottal stop, and both velar and palatal nasals. Blench and Sidwell identified a striking feature: many Austroasiatic roots that end in nasal consonants show up in Shompen with oral stops instead. This pattern resembles what happened in the Aslian and especially Jahaic languages of the Malay Peninsula, where historical nasal endings became fully oral over time. In Shompen, the process may have gone further -- oral stops were apparently lost first, then reacquired as former nasals shifted. Other sound changes include the breaking of long vowels into diphthongs and the loss of word-final h and s. There is no standard orthography. The Shompen have no written tradition.

The Shompen live in the hilly hinterland of the Great Nicobar Biosphere Reserve, protected from outside contact by Indian law. That protection, meant to preserve their way of life, also makes linguistic fieldwork nearly impossible. Villages documented in 1977 -- Dakade, with its fifteen residents and four huts ten kilometers from the coast, Puithey, Tataiya -- represent fleeting glimpses of a people who have largely avoided sustained contact with the outside world. The Dogmar River group had relocated between 1960 and 1977, a reminder that these communities are not static even in isolation. With roughly 400 speakers and no writing system, Shompen ranks among the world's most endangered languages. The proposed Great Nicobar Development Plan, which envisions bringing 650,000 new residents to an island of 8,500, has raised alarm among both indigenous rights advocates and linguists. If the Shompen are displaced or absorbed, their language -- or languages -- will vanish with them, taking its unresolved mysteries into permanent silence.