Experience is processing: a theory of consciousness in one axiom and five definitions, tested on model networks, a worm's wiring and a language model

Abstract

"Consciousness" is a collective word. Process-Identity Theory (PIT, "Erleben = Verarbeitung") sets one axiom — experience is processing, the same thing under two names — and five definitions, and derives forty propositions from them. They separate four things that can be measured apart: whether a system experiences (how long and how richly it carries a perturbation — the stage), where one system ends and the next begins (the boundary), who the system is (its identity, the form its wiring has taken), and what it can report and store (access). The axiom cannot be tested; most of the propositions can, and were, on three legs: 127 tests on small recurrent model networks (T1–T110 with a series of foundations), 18 re-measurements with a calibrated stage measure (N1–N18) that also run on the complete wiring of C. elegans at eight developmental stages (Witvliet et al. 2021) and in both sexes (Cook et al. 2019), and a language-model branch of 38 exploratory and 36 pre-registered experiments on Pythia-410m, published as its own record (doi:10.5281/zenodo.23280536). Of the 127 tests, fifty-four confirmed a proposition and seventy-one came back negative or limited; every failure is reported, and several definitions were revised by their own tests. What stands: experience and identity are two independent axes; identity is the form of the wiring, built by the order of experience, graded and irreversible, and it goes before the stage under damage; one wiring is one stage, and whether two systems are one is decided by the speed of the path between them; the clinical map of anaesthesia, amnesia, dream, psilocybin, seizure and locked-in appears in a language model run as a soft loop in the same order as in humans. What follows for language models is stated as far as it is measured, and no further. The theory is marked as a theory: what would refute each part is listed. Files: the theory paper (version 1.0, English, main text in plain language with the full axiom system, every test and everything that fell in the Supplement); the German canon document "Erleben = Verarbeitung — Ein axiomatisches System zum Bewusstsein" (Fassung 9, with every test entry, every revision dated, and the archived earlier version of the paragraph "Der Fall KI"); the model-network programme as two packages — the main programme (code, pre-registrations, results, logs, 5–9 September 2026) and the calibrated re-measurements (24 September 2026) — each with a README; code under MIT, everything else CC BY 4.0. The connectome inputs are others' published data and are cited, not redistributed. AI involvement: Claude (Anthropic), a large language model, wrote the code, the pre-registrations, the evaluations and the drafts of the paper under the author's decisions, and is named as co-author of the paper; this is declared in every document, and no self-report of the model enters any result. Models: Claude Opus 5.5 (Anthropic) for the model-network programme, the calibrated re-measurements and the first language-model branch (September 2026), and for two external readings of the final drafts; Claude Fable 5.1 (Anthropic) for the pre-registered language-model experiments, the measurement paper, the theory paper and the deposited packages (October 2026).