Archaeology of Information in Language

Archaeology of Information in Language (AIL) is a versioned research programme investigating whether language can preserve measurable traces of its own historical reorganizations and, under stricter conditions, information about past social, technical, or procedural processes. The programme separates three analytical levels. AIL-OBS0 evaluates whether corpora from successive periods are genuinely comparable with respect to region, genre, text function, archive, medium, scribal tradition, and preservation. AIL-M1 measures the path of change between documented language states only after corpus comparability has been established or the data have been standardized to a common reference distribution. AIL-M2 is a proposed generative framework for reconstructing deeper, partly unobserved history using independent archaeological, geographical, population, and historical evidence. Version 1.0 presents the epistemic contract of the programme, the formal AIL-M1 metric, corpus-selection and causal models, an identifiability analysis, falsification criteria, and a blind-validation protocol. A calibration based on Middle English negation shows that, under a fixed four-period partition, comparison of the initial and final states alone omits approximately 16.9% of the measured trajectory. This result demonstrates the importance of intermediate states, but it does not establish a universal linguistic clock or measure the absolute age of a language. Prior-based pilots involving Polish, Mandarin, and Hindi-Hindustani, together with a sixteen-language simulation, are presented as sensitivity and scenario analyses rather than empirical rankings. A methodological design trial using Sumerian material shows how changes in the surviving archive can imitate linguistic change, motivating the mandatory AIL-OBS0 corpus-comparability gate. AIL does not infer lost civilizations from grammatical complexity and does not date languages without independent chronological anchors. Its next decisive test is frozen, blind, multilingual validation across independent constructions and corpora.