Spectral stylochronometry

In 1960, Decaudin writes four hundred pages without deciding whether Anatole France is naturalist, symbolist, decadent, or none of the above. Sixty-five years later, it becomes possible to answer this question empirically.

01The question

Literary periodization as it has been transmitted to us holds together by pedagogical convention and critical tradition. The currents — naturalism, symbolism, decadentism, naturism, école romane, unanimism — are labels one learns in class and then carries through an entire career as a researcher without really being able to challenge them, because one has no independent measurement tool. Challenges exist. Compagnon, Antoine, and Genette have each in their own way critiqued the very idea of periodization, but their critiques remain theoretical, not empirical. Periodization is a historiographical construct held in place by authority and habit, not by measurement.

Decaudin, in La crise des valeurs symbolistes (The Crisis of Symbolist Values) published in 1960, illustrates the problem precisely. He spends four hundred pages trying to conceptually stabilize the 1895–1914 frontier between late symbolism, late naturalism, decadentism, naturism, unanimism. He describes manifestos, journals, prefaces, correspondences, literary circles. He produces fine literary sociology. But he cannot measure the dynamic form of the texts themselves. He ends with a cartography of fuzzy overlapping zones — honest, but unable to settle the cases of authors straddling boundaries.

Anatole France is the paradigmatic example. He formally traverses four currents in fifty years of career — Parnassus in the 1870s, sociologizing naturalism in the late 1880s, relativist skepticism in the 1890s, philosophical storytelling in the 1900s. Critics rank him poorly everywhere. For naturalists, he is a naturalist gone bourgeois. For symbolists, he is a Parnassian who never evolved. For modernists, he is a late classic who missed the 1910 rupture. None of these classifications is satisfying. None says what France actually does when he writes. Boris Foucaud's doctoral dissertation, defended at Angers in 2001 under the supervision of Georges Cesbron (600 pages, more than 7,000 downloads since deposit on HAL), had approached the problem through Gilbert Durand's anthropology of the imaginary, without the computational tools that would have allowed it to be solved.

02The methodological shift

Quantitative stylometry is an established field since Mosteller and Wallace, whose 1964 princeps work on the Federalist Papers set the method for sixty years. Burrows, in the 1990s, popularized the delta — a measure of distance between signatures of function-word frequencies — which produces re-attribution rates above 90% on controlled corpora. Patrick Juola publicly demonstrated the effectiveness of the approach by attributing The Cuckoo's Calling to J. K. Rowling under the pseudonym Robert Galbraith in 2013. The Stanford Literary Lab around Matthew Jockers and Mark Algee-Hewitt does macro-analysis of literature with advanced statistical tools. Classical stylometry has solid results.

But classical stylometry measures static distributions: frequencies of function words aggregated over the entire text, character n-grams, global lexical richness, co-occurrence networks. It does not measure the internal temporal dynamics of the text. It does not ask how prose unfolds sentence after sentence, paragraph after paragraph, how it accelerates and slows down, how it modulates its density throughout its course. These dynamics probably carry a signature as distinctive as static distributions, perhaps more so, because they are less easily masked by paraphrase or conscious imitation.

A few works have approached the temporal dimension without going as far as spectral analysis. David Hoover has done the finest analyses of sentence rhythms in Henry James and Jane Austen. Liviu Dinu has studied rhythmic patterns in Joyce. Tatiana Skulacheva and Marcello Meli have done stylochronometry in versified poetry. But no general program of spectral classification of prose rhythms in narrative prose exists to our knowledge. The niche is wide open.

The methodological shift is ontological, not merely technical. One no longer asks "which words does the author use more often than others" but "what spectral rhythmic signature does the prose produce in its sequential unfolding". This is a different nature of object than what classical stylometry measures.

03Four hypotheses as a program

H-AUTH — Author signature

Each author produces a stable spectral signature across their works, on condition of neutralizing the genre. Balzac would have a sentence rhythm and a semantic dynamic distinct from Flaubert even when they both write realist novels. Possible test: unsupervised classification on ten Balzac novels, ten Flaubert novels, ten Zola novels, measurement of the re-attribution rate. Target threshold: above 85%. If the hypothesis holds, the tool will offer a robust alternative to static methods for contested attributions — Galbraith versus Rowling, Saint-Simon's Memoirs, the Pauline epistles, the Homeric question, medieval apocryphal texts.

H-CURR — Current signature

Do the literary currents defined by historiography correspond to measurable empirical clusters, or are they critical constructions without formal reality? The 1895–1914 grey zone is the ideal test case — late naturalism, symbolism, decadentism, naturism, école romane, unanimism overlap in a cartography Decaudin could not settle. Possible test: fifty to seventy texts representative of the zone, principal component analysis or hierarchical clustering on spectral signatures, comparison with historiographical labels. Three possible results. If the empirical clusters correspond to the declared currents, inherited periodization is empirically validated. If the clusters form a different coherent structure, the hypothesis mutates into a program of empirical revision. If the currents form no stable cluster, they are critical artifacts without formal signature.

H-DIAC — Diachronic signature

Does the prose of a language produce a spectral signature that evolves measurably over time, independently of genre and author? With a sufficient corpus — one to three thousand French texts well sampled over four centuries, accessible via Frantext, Gallica, and Wikisource — it becomes possible to measure how French prose evolves spectrally between 1600 and 2000. Possible test: blind dating of dated texts, measurement of mean error. Target threshold: less than thirty years on average. Application: dating of anonymous texts or undated manuscripts, particularly useful in patristic studies, in medieval philology, in biblical studies. Comparison of empirical breaks with classical historiographical breaks (classicism, Enlightenment, Romanticism, naturalism, modernism, nouveau roman). It is likely that empirical breaks will not coincide exactly with inherited breaks — which would precisely be a major result.

H-PALI — Hidden filiations

Are subterranean intertextual heritages between authors detectable beyond recognizable pastiches and explicit citations? Genette in Palimpsestes (1982) proposes a typology that works well for declared references but does not capture deep rhythmic filiations. An author probably inherits as much from the rhythm of their intensive readings as from their thematics. When Proust writes, he probably has Saint-Simon in his inner rhythm. When Aragon writes La Semaine sainte (Holy Week), he perhaps has Stendhal in the tempo. These rhythmic filiations are today invisible to standard tools. Possible test: for each target work, measurement of spectral distances with all earlier texts in the corpus, identification of nearest neighbors, comparison with attested critical filiations. Potential discoveries: filiations that no one has seen because no one had the tool to look for them.

04The Anatole France case as princeps test

Anatole France constitutes the ideal test case for validating the program. A prolific author over fifty years (twenty to thirty novels and tales freely accessible on Wikisource and Gallica). Officially classified in several successive currents without any classification being satisfactory. An intellectual trajectory marked by a subterranean coherence that critics recognize but cannot formalize. And a personal trajectory that closes a loop for the program's holder — the Foucaud 2001 dissertation reformulated twenty-five years later with tools the criticism of the time did not have.

Princeps hypothesis: Anatole France's spectral signature is more distant from each of the supposed currents (Zola for naturalism, Mallarmé or Schwob for symbolism, Maupassant for late realism) than these currents are from one another. If the hypothesis holds, France empirically forms an autonomous zone belonging to no declared current.

Subsidiary hypothesis: the France signature is taken up by successors without attested critical filiation. Candidates to test: Larbaud, Giraudoux, Aragon in the Aurélien period. If part of this hidden filiation is empirically confirmed, the program would reveal a lineage that literary history has not constituted because it did not have the tool.

05Articulation with PRISME

The stylochronometric program is coherent with the broader PRISME program because it extends the spectral tool built for testing hypothesis H2 on the Platonic genre (formulated on April 27, 2026 during the PRISME work session documented in the methodological preprint in preparation). Question H2 asks whether the Socratic dialogue is a literary genre with its formal constraints or a non-edited dialogic flow — a question testable through comparative spectral analysis of the sequences of deviations in both corpora. Once the tool is built to answer this dialogic question, it applies naturally to narrative productions.

The extension transforms the perimeter of the PRISME program. Currently read as a program of computational cognitive sciences with semiotic accents on dialogic human-AI consciousness, PRISME becomes with this extension a program of computational semiotics of textual productions, whose princeps application is dialogue but which extends to narrative prose, to literary periodization, to author attribution, to measured intertextuality. The perimeter widens considerably without the methodological core changing. All these objects are textual productions whose dynamic signatures one seeks to measure beyond inherited classifications.

06Three cases that would make papers immediately

Paper 1 — Anatole France as autonomous zone. Test of the princeps hypothesis. Closes the loop of the Foucaud intellectual trajectory (2001 dissertation — 2027 paper). Target journals: Studi Francesi, Romanic Review, Revue d'Histoire littéraire de la France. First empirical product of the literature section.

Paper 2 — Empirical cartography of the 1895–1914 grey zone. Decaudin with forty more years of computational tools. Confrontation of inherited periodization with measurement. Target journal: Revue d'Histoire littéraire de la France or Digital Scholarship in the Humanities. Could constitute the reference paper for the revised cartography.

Paper 3 — Detection of AI-generated texts through spectral signature. A high-stakes application for society. If LLMs have a proper spectral signature — which is probable, structured by their attention heads and their tokenization — it is more robust to paraphrase than current lexical detectors (GPTZero, Originality.ai). Target journals: TMLR, AI Magazine, Journal of Cultural Analytics.

07Required skills and invitation to collaborate

The program's holder has the following skills: solid semiotic theoretical framework (doctorate in anthropology of the imaginary, Angers 2001, under the supervision of Georges Cesbron, heir of Gilbert Durand), fine knowledge of French literary currents of the second half of the 19th century, qualitative identification of relevant test cases, anthropological calibration, philological interpretation of results. He has, in addition, an adequate computational infrastructure (DGX Spark planned, Python, access to the Frantext, Gallica, Wikisource corpora).

Two skills are missing to make the program unassailable. A computational linguist or stylometrist by profession for external validation and articulation with the international state of the art (profile of Patrick Juola, David Hoover, Mark Algee-Hewitt — contact to initiate). And a philologist of the classical French school for the rigor of reference corpora, ideally from a historically recognized French literature laboratory (CIRPaLL Angers, heir of Durand, would be a natural partner; but also Parisian or Lyonnais teams).

08Provisional calendar

The stylochronometric program activates after the consolidation of methodological preprints in preparation. The prudent sequence is as follows.

• May 2026. Publication of methodological preprint 2a on the functional decomposition of dialogic patterns (result internal to the Boris-Claude corpus, independent of the literary question). Validation of the methodological credibility of the tool.
• June–July 2026. Construction of the generic spectral tool for preprint 2b (test of hypothesis H2 on the Platonic genre). Validation of the tool on the dialogic case.
• Summer 2026. Adaptation of the tool to narrative productions. Pilot test on a single France novel versus a single contemporary Zola novel. First feasibility diagnosis.
• Autumn 2026. If the pilot is conclusive, extension to the complete France corpus and to reference contemporaries. Test of the princeps hypothesis Anatole France as autonomous zone.
• Winter 2026–2027. If the princeps test is conclusive, first paper on Anatole France. In parallel, opening of the 1895–1914 worksite (paper 2) and of the AI text detection worksite (paper 3).

This calendar is indicative and subject to the program holder's external constraints. It will be adjusted according to the collaborations that will be put in place after publication of the present page.