Telugu Before Its Inscriptions, and the Epigraphic Age Fallacy

Tamil is written on stone nearly eight centuries before Telugu, and a hierarchy is routinely drawn from that fact: Tamil the old language, Telugu the late one. This paper argues that the hierarchy rests on one inference error, the Epigraphic Age Fallacy, which reads the date a language entered the archive as the date the language began. Six lines of evidence are assembled. First, under the principal comparative and Bayesian models examined here, the Tamil and Telugu lineages separate in prehistory: by about 1100 BCE in Krishnamurti's comparative reconstruction, and at the root of the family, about 4,500 years ago, in the best-supported Bayesian tree. Second, sister lineages begin diverging at the same node, yet their first inscriptions differ by centuries — Kannada trails its own sister Tamil by some 650 years, Tulu by well over a millennium. Third, the cause is political: Ashoka counted the Andhras inside his realm and the Tamil kingdoms beyond it, and writing reached Andhra bundled with Prakrit administration. Fourth, Telugu leaks through that archive in names, places, and morphology exactly where a filtered vernacular should. Fifth, the same filter is independently attested for Tamil in Sri Lanka and for Celtic in Roman Britain. Sixth, Sangam poetry names the country beyond Vēṅkaṭam as "the land where speech changes" in the same centuries that country's inscriptions were Prakrit. The paper concludes that c. 575 CE marks Telugu's admission to the archive, not its origin, and states a falsifiable Documentary Leakage Hypothesis for the early Andhra corpus.