Toward a Biophysics of Poetry

My long-term interest in Coleridge’s “Kubla Khan” (KK) is shadowed by an interest in “This Line-Tree Bower My Prison,” (LTB) which is one of the so-called Conversation poems. I compared the two in a working paper from 2013, STC, Poetic Form, and a Glimpse of the Mind, noting that they have several motifs in common but are utterly different poems. Thus (pp. 41-42):

LTB’s trajectory is a first-person narrative; we follow the perceptions, thoughts, and feelings of the poet as they occur to him, one after the other. We are presented with a cleanly identified authorial consciousness shifting its attention here and there about the world in a coherent way. The passage of time is explicitly marked and noted within the poem itself. In this poem Coleridge’s verse technique emphasizes temporal continuity. By placing major attention shifts in the middle of the line Coleridge emphasizes an underlying continuity of consciousness: one’s gaze has a new focus, it is still one’s gaze. By running clauses across line boundaries, the repetitive recurrence of line ends is muted. The temporal course of the poetic act itself is muted so as to “dissolve” into the virtual time depicted in the poem.

KK is different; it lacks a narrative frame. Thus it cannot register either narrative movement or a self that experiences such movement. Things and scenes are juxtaposed in a way that is sometimes hard to grasp. Yet, as we have seen, there is a rigorous logic to the poem. At the same time it has an elaborate rhyme scheme and other sound features as well. What has this to do with its “ontological” semantic trajectory?

I suggest that the confluence between sound and sense in “Kubla Khan” is, in effect, a surrogate for the lack of a narrative frame, and hence the impossibility of a narrative trajectory. The key point is that rhyme adds an element of predictability to the verse; one knows that sounds will repeat at regular intervals and so can anticipate them. Thus the predictability that has been “lost” because the poem does not have a narrative flow is “restored” or “compensated for” though elaborate rhyme. The temporal structure of the poem itself becomes the frame for its semantic trajectory.

In the course of discussing these matters with ChatGPT 5.2 we arrived a what might be called a biophysics of poetry. I issued the following prompt:

I want to return to the difference in prosody practice in KK and LTB. I think that is (potentially) VERY deep. That there seems to be a trade-off between conceptual continuity (temporal succession vs. ontological transition from state to state) and sonic patterning (simple vs. elaborate versification), that strikes me as something very deep about how the nervous system functions. It has something to do with the physical nature of the system. We may not be able to figure it out on the basis of formal design considerations but will need to know specifics about physical structure and processes.

Yes—this is very deep, and I think you’re now pointing to something that sits below poetics, below cognition-as-symbol-processing, and squarely in the physics of nervous systems.

What you’re intuiting is not just a stylistic correlation, but a trade-off between two kinds of coherence that any physically realized information-processing system has to manage.

The discussion went on from there. I asked ChatGPT to summarize the discussion, after which we went on to develop experimental tests of the idea. I have placed ChatGPT’s summary below; the experimental tests will have to wait for another post.

Toward a Biophysics of Poetry

What if poetic form is not primarily a matter of meaning, symbolism, or interpretation—but of dynamical stability in the nervous system?

That is the direction our recent discussion has taken us: toward something like a biophysics of poetry. By this I mean an account of poetry grounded in timing, energy, alignment, and coherence—constraints that arise because poems must run on human (and perhaps artificial) neural systems.

Poetry as computation in time

A useful starting point is the idea that a linguistic string—a poem, sentence, or speech—is a one-dimensional projection of a path through a high-dimensional semantic and neural space. Meaning is not stored in the string itself; it is reconstructed as the reader (or listener) traverses that path.

Crucially, not all paths are equal. Some are smooth and additive. Others are turbulent and overlapping. This distinction matters.

Composition vs. convolution

Many poems proceed largely by composition: meanings accumulate incrementally, with limited interference. But some poems—most strikingly Coleridge’s Kubla Khan—operate by convolution, overlapping distinct conceptual domains (visual–spatial, auditory–temporal, symbolic–ritual) so that they interact rather than simply add.

Convolution is powerful, but dangerous. Overlapping domains can interfere with one another, threatening experiential breakdown. This raises a fundamental question: how does such a poem remain continuous at all?

Sound as a carrier wave

The answer is not primarily semantic. It is dynamical.

In signal-processing terms, the poem’s sound structure—meter, rhyme, phonological recurrence—functions as a carrier wave. It provides a stable, low-information temporal scaffold on which high-information semantic modulation can ride.

Sound does not carry meaning here; it carries phase alignment. It keeps neural activity synchronized as the poem shifts domains. Rhyme, in this view, is not ornament but infrastructure: a phase-reset mechanism that prevents destructive interference during domain transitions.

This perspective aligns closely with empirical work on oral tradition and memory, where sound structure is known to support sequencing and recall while meaning is reconstructed opportunistically.

Segmentation and entry points

The same logic applies to segmentation. If a linguistic string is a projection of a high-dimensional path, then some substrings are better entry points into that space than others.

Fragments that respect syntactic or prosodic boundaries align with internal structure and support rapid pattern completion. Fragments that straddle boundaries are misaligned: they activate incompatible frames and fail to converge.

This is not just a theoretical claim. In informal experiments with large language models, * prompts that respected syntactic boundaries reliably triggered recognition of well-known texts (e.g., Hamlet’s soliloquy or the Gettysburg Address), while boundary-straddling prompts did not. Recognition failed not because of missing information, but because of misalignment.

Poetry as biophysical engineering

Seen this way, poetic form is a kind of engineering. It constrains linguistic material so that complex meaning can remain experientially coherent under biophysical limits. Sound structure enforces phase alignment during traversal; hierarchical segmentation ensures viable re-entry points into semantic space.

This also clarifies differences among poems. In This Lime-Tree Bower My Prison, sound supports smooth movement within a single conceptual domain. In Kubla Khan, sound is load-bearing: without it, the poem’s domain-convolving machine would fall apart.

A working hypothesis

All of this can be summarized in a single hypothesis:

Poetic form constrains and exploits the dynamical properties of neural systems to maintain coherent experience under conditions of high semantic and conceptual load.

That, in essence, is what I mean by a biophysics of poetry. It does not replace interpretation—but it explains how interpretation remains possible in the first place. 

*William Benzon, Discursive Competence in ChatGPT, Part 2: Memory for Texts, Working Paper, September 27, 2023, 44 pp., https://www.academia.edu/107318793/Discursive_Competence_in_ChatGPT_Part_2_Memory_for_Texts