Language Has Two Parameters - Narrative-Induced Semantic Plasticity and Phase-Sensitive Interpretation

Sep 15, 2026

Uploaded by Aria

Reading fiction or encountering narrative does more than add information; it reshapes the reader. The paper argues that encounters alter persistent relations among simultaneously active meanings, yielding individual and shared reception histories that population-trained language models may not retain. A model can reproduce an encounter’s consequences without actually being changed when the relational state remains context-bound. The work formalizes the missing relational state as phase, derives testable predictions on encounter order, quotation, and suppressed meanings, and discusses semantic poisoning for AI agents.

LanguageEnglish

CategoryResearch & Report

FormatPDF

Pages23

Reading Time58 min

AccessFree to download

The problem

Phase vs amplitude in meaning

Fiction, reception history, and representational reorganization

Example: The Terminator and suppressed meanings

arXiv : 2608 . 1804 1v2 [cs .CL] 20 Aug 2026 Language Has Two Parameters: Narrative-Induced Semantic Plasticity and Phase-Sensitive Interpretation Hollis Robbins, University of Utah [Hollis.Robbins@utah.edu](Hollis.Robbins@utah.edu) Version 2 August 19, 2026 Abstract. Reading fiction or encountering narrative generally does not merely add information. The encounter changes the reader. This paper proposes that encounters alter persistent relations among simultaneously active meanings, producing individual and shared histories that population-trained language models do not necessarily retain. A model may be told of an encounter and reproduce its consequences while the history remains in context; this is not the same as being changed by the encounter. This paper formalizes this missing relational state as phase, derives testable predictions about encounter order, quotation, and suppressed meanings, and argues that future AI agents will need persistent semantic states indexed to particular individuals and relationships. The matching risk is semantic poisoning: an attack that re-signs relations among meanings already present. Keywords: attention; narrative; linguistic form; semantic plasticity; quantum cognition; phase; monosemanticity; language models; agent memory; reception history 1. The problem Long before the transformer era, poetry was described as the careful configuration of words, each of which has a constellation of associations.1 Line lengths, rhyme, and form also carry constellations of associations. The coexistence of these constellations causes amplitude and phase shifts in meaning. Every word carries a cloud of associations in orbit around it, which poets refer to as a constellation: meanings that exert a changing gravitational pull on other words depending on words nearby. The difficulty many people have with poetry is that meaning is hard to discern and poetry can mean different things to different people. Words, vectors, and cosine similarity are the ordinary machinery. All of it estimates the first parameter. This paper is about the second, and what it would take for a language model to represent it. The two parameters are different kinds of quantity; neither can be computed from the other. The first parameter, amplitude, is about the strength and value of meaning by itself. The second parameter, phase, is about the relations of meanings to other meanings: how two or more meanings can be live simultaneously, can combine, can reinforce or cancel one another. Phase can change while the strengths of meaning stay the same; a meaning can be present at full strength with its contribution reversed. You can think of all fiction generally—including poetry, novels, film, drama—as featuring not only language but also characters, genres, plots, histories, retrospective reinterpretations, countless other features. Fictional characters develop their own constellations and become embedded in individual and cultural consciousnesses, gathering their own constellations of associations and entering the constellations of subsequent fiction and poetry. Genres and plots have this property. Encountering fiction reorganizes the representational capacities of readers and it affects the statistical properties of literary language. As an example: midway through The Terminator (1984), James Cameron’s film about a cyborg assassin (played by Arnold Schwarzenegger) sent back from the future to kill a specific target, the cyborg sits in a rented room repairing its wounds.2 The decaying organic matter smells. A janitor knocks, asking if there is a dead cat in there. The film cuts to the cyborg’s 1 I. A. Richards, The Philosophy of Rhetoric (New York: Oxford University Press, 1936) claims words are laden with meanings from past contextsand animate each other in combination. William Empson, Seven Types of Ambiguity (London: Chatto and Windus, 1930), also claims that multiple meanings are live simultaneously; the poetic fun

FAQs

What are the two parameters of language proposed in the paper?

The paper distinguishes amplitude (strength and value of meaning by itself) from phase (how meanings relate to other concurrently active meanings, including reinforcement or cancellation). Neither parameter can be computed from the other.

How does the paper explain how encounters change meaning beyond stored context?

It proposes that encounters change persistent relations among simultaneously active meanings, producing individualized and shared histories. A model can mimic consequences without the relational state being altered in the way an encounter changes a reader.

What predictions does the paper derive regarding narrative encounters?

It formalizes phase as a missing relational state and derives testable predictions about encounter order, quotation, and suppressed meanings—highlighting cases where the same form can mean differently depending on prior narrative exposure.

What is semantic poisoning in the paper’s framework?

Semantic poisoning is an attack that re-signs relations among meanings already present, targeting the persistent semantic states that future AI agents would need for correct individual- and relationship-specific interpretation.

Sep 15, 2026

Uploaded by Aria