three

The ancestor of three is the Proto-Indo-European root treyes, reconstructed from the agreement of dozens of daughter languages separated by thousands of years and thousands of miles. Sanskrit has tri (त्रि), Greek has treis (τρεῖς), Latin has tres, Armenian has erek, Welsh has tri, Lithuanian has trys, and Russian has tri. This convergence points to a single origin in the language spoken on the Eurasian steppe around 4000 BCE, long before writing, long before any of those languages diverged.

In Old English, the word was þrēo (for feminine and neuter nouns) or þrī (for masculine), and it already carried the essential consonant shape: the th sound from the PIE initial t-, and the vowel pattern that would yield modern three. The Anglo-Saxon Gospels of the late tenth century use þrēo in familiar contexts: three days, three disciples, three denials. The grammatical gender distinctions that once required different forms collapsed during the Middle English period.

The Middle English form was thre or three, and by Chaucer's time in the 1380s the spelling had nearly settled. The Great Vowel Shift of the fifteenth and sixteenth centuries affected many English words, but numerals were somewhat insulated by their frequency: words used constantly tend to resist drastic phonological change. Geoffrey Chaucer uses thre throughout the Canterbury Tales without variation; by Shakespeare's time, the modern spelling was standard in print.

Three is among the handful of English words most resistant to borrowing from other languages. Across the world's languages, the lowest numerals are the last to be replaced by loanwords, precisely because they are in constant use. The word three has roughly the same history as the concept it names: ancient, stable, and found in virtually every language descended from the common ancestor spoken on the steppe six thousand years ago.