How Do I Know What to Say Next Barenholtz’s Autogenerative Theory as an Enrichment of Harrisean Integrationism

Jul 16, 2026

Roy Harris’s Integrationist linguistics challenges referentialism at the core of many computational language approaches, treating language as a situated, bipartite activity that coordinates prospective joint action rather than mapping code-like labels to an external world. Integrationism leaves gaps in the mechanism sustaining sign-driven openness, in the continuity between linguistic and non-linguistic semiosis, and in how archives of past integrations structure new participation. Elan Barenholtz’s LLM-driven autogenerative theory is argued to supply these missing accounts without compromising Integrationism’s core commitments.

LanguageEnglish

CategoryResearch & Report

FormatPDF

Pages15

Reading Time38 min

Abstract

Introduction

Autogenerative theory and prospective openness

Semiotic continuity between linguistic and non-linguistic activity

Archive of past integrations and new participation

Implications for natural language processing and LLM design

arXiv :2607 .0789 1v 1 [ cs .CL] 8 Jul 2026 How Do I Know What to Say Next? Barenholtz’s Autogenerative Theory as an Enrichment of Harrisean Integrationism Prof J. Mark Bishop Emeritus Goldsmiths, University of London Prof Stephen J. Cowley Emeritus University of Southern Denmark June 2026 Abstract Roy Harris’s Integrationist linguistics offers a compelling critique of the referentialist tradition embedded deep at the heart of computational approaches to language, arguing that language is not a code that maps onto a pre-given world but a situated, bipartite activity oriented toward prospective joint action. Yet Integrationism leaves certain explanatory gaps: it does not fully account for the structural mechanism by which signs sustain prospective openness, it undertheorises the continuity between linguistic and non-linguistic semiotic activity, and it offers no detailed account of the structural properties of the accumulated archive of past integrations. This paper argues that Elan Barenholtz’s autogenerative theory of language, developed in response to the behaviour of Large Language Models (LLMs), can fill precisely these gaps, enriching Integrationism without undermining any of its core commitments. Specifically, the autogenerative account provides: a structural mechanism for the prospective openness that Harris identifies as central to bipartite communication; a computational correlate for Harris’s thesis of semiotic continuity between language and other sign-making activity; and a theory of the archive: what the accumulated residue of past integrations looks like and how new participants draw upon it. The synthesis preserves Harris’s ontological primacy of the situated integrative act while adding explanatory content that Integrationism itself does not supply. For practitioners and researchers in natural language processing and large language model design, the argument offers a principled account of what the statistical structure that LLMs so effectively exploit actually is, and of what it cannot, by its nature, provide. 1 Introduction Roy Harris, who served as Professor of General Linguistics at Oxford until his death, developed Integrationist Linguistics over several decades as a sustained critique of what he called the “language myth,” namely the assumption, embedded so deeply in Western thought that it is rarely noticed, that words are fixed labels transparently conveying thoughts about an independent external world. (Harris, 1981) On the “telementational” view Harris attacked, communication succeeds when a thought in one mind is successfully replicated in another: language is merely the transparent conduit. Against this, Harris argued that language is a situated, bipartite activity: signs do not carry fixed meanings between speakers but coordinate what participants will do next, together, in a particular context. Meaning is not transmitted; it is prospectively constructed. In a recent iai essay, cognitive scientist Elan Barenholtz argues that Large Language Models (LLMs) have revealed something equally radical about the nature of language: that it does not work by pointing to or describing reality, but by generating contextually appropriate continuations of itself. (Barenholtz, 2026) Barenholtz calls this the “autogenerative property” of language. Crucially, LLMs operate without any direct exposure to the lived world (the proverbial “deaf, dumb and blind kid, who sure plays a mean pinball,” to borrow a phrase from The Who (TheWho, 1969)) yet nevertheless generate coherent and contextually appropriate language. An LLM knows nothing about the actual redness of red or the physical distance of far; it knows only where these words fall in relation to every other word in a high-dimensional relational space. And yet from this manipulation of what he calls “empty symbols,” coherent, contextually appropriate language emerges. Barenholtz is careful, however, not to claim that the human brain literally i

FAQs

What core claim does Harris’s Integrationist linguistics make about language meaning?

Harris argues that language meaning is not transmitted as fixed labels about an independent world. Instead, signs coordinate participants’ next actions in context, so meaning is prospectively constructed.

What explanatory gaps remain in Integrationism according to the paper?

The paper states that Integrationism does not fully explain the structural mechanism that sustains prospective openness, under-theorizes continuity between linguistic and non-linguistic semiosis, and lacks a detailed account of the archive of past integrations and how newcomers draw on it.

How does Barenholtz’s autogenerative theory enrich Integrationism in response to LLMs?

The paper argues that autogeneration provides a structural mechanism for prospective openness, a computational correlate for semiotic continuity, and a theory of the archive—clarifying what LLMs exploit statistically and what they cannot provide by nature.

Jul 16, 2026

Upload by Clementine