Abstract

British fiction and poetry explodes with textual visuality in the early twentieth century: color, shape, and form, as manifested in description, impression, and image. This dissertation computationally models that visuality, using the eye as a governing metaphor: retinal cones are modeled by inferring textual color, and retinal rods are modeled through object-detection via word sense disambiguation and categorization. Findings include a 93% increase in color expressions across the nineteenth and early twentieth centuries, a 15% increase in the proportions of object and artifacts, and revealing correlations along lines of literary genre, subject heading, and more. These correlate with historical materialities such a dye manufacture, trends in the visual arts such as post-impressionism, and movements in literature such as imagism. A model of literary description, meanwhile, finds that, while visuality increases over time, proportions of description decrease, suggesting structural decompositions in fiction, occurring in parallel with disseminations of vision.

This dissertation is a full-color, interactive document, originally formatted in HTML, and made freely available on the Internet. It contains interactive plots, hyperlinks, and other dynamic elements. It is a social document, annotated by the community, and open for further annotation. If you are reading this in another document format (PDF, DOCX, or paper), you are missing these social, interactive, and dynamic elements. Please visit dissertation.jonreeve.com for the canonical edition of this text, and github.com/JonathanReeve/dissertation for its source code.

Table of Contents

• Introduction
• Chapter 1: Color
• Chapter 2: Shape
• Chapter 3: Image

List of Charts, Graphs, Illustrations

• Introduction
• Figure 1: Richards’s Optical Process of Reading
• Figure 2: Cross-section of an optic nerve. Credit: Laboratoires Servier. Creative Commons Attribution-Share Alike 3.0.

• Chapter 1: Color
• Figure 1: Base color proportions, by author date of birth, smoothed with a five-year average
• Figure 2: “Post-Impressions of the Post-Impressionists” (The Bystander 375)
• Figure 3: Proportions of the color yellow in novels and poems, according to date of publication.
• Figure 4: The cover of the inaugural issue of The Yellow Book
• Figure 5: Proportions of the color mauve in the Google Books corpus of English Fiction, in the nineteenth century
• Figure 6: Walter Sickert, Ennui, 1913
• Figure 7: Colors in To the Lighthouse, plotted in narrative time. For an interactive plot, refer to this explorer.
• Figure 8: Colors in Portrait of the Artist as a Young Man, plotted in narrative time.
• Figure 9: \(CM_X\) “-ish” Word Matrix
• Figure 10: Color Associations Word Clouds. Click on a word cloud to visualize categorizations of these words.
• Figure 11: Hypernym tree map for red, using WordNet hypernyms
• Figure 12: A page from Ridgway’s 1912 Color Standards and Color Nomenclature
• Figure 13: A page from Saccardo’s 1894 Chromotaxia
• Figure 14: Color map average vector similarity with selected words
• Figure 15: Syntax dependency parse of “slopes of bright green grassland lay above them”
• Figure 16: Color inferences for grass, from \(M_P\).
• Figure 17: Color inferences for sheep, from \(M_P\).
• Figure 18: Color inferences for apples, from \(M_P\).
• Figure 19: Image-based model inferences, with example images.
• Figure 20: \(M_I\) Sample Matrix
• Figure 21: HSL Projection of To the Lighthouse Nouns. Please find the interactive version here. Due to the size of the data contained, it could take up to 30 seconds to load.
• Figure 22: Data flow chart for the imagination model
• Figure 23: Color proportions in literature, by date of publication
• Box plot of individual color words
• Figure 24: Base Color Proportions by LCSH
• Figure 25: Base Color Proportions By Genre

• Chapter 2: Shape
• Figure 1: Object proportions over time, \(C_{PG2}\)
• Figure 2: living_thing.n.01 proportions over time, \(C_{PG2}\)
• Figure 3: natural_object.n.01 proportions over time, \(C_{PG2}\)
• Figure 4: artifact.n.01 proportions over time, \(C_{PG2}\)
• Figure 5: Body parts by year of publication, \(C_{PG2}\)
• Figure 6: Body part sums, \(C_{PG2}\)
• Figure 7: Body Part Sums, Body Projection, \(C_{PG2}\)
• Figure 8: Sensory homonculus illustration. Source: OpenStax College; Wikimedia Commons; CC-BY

• Chapter 3: Image
• Figure 1: Moulton’s “cardinal points” of literary form
• Figure 2: Object proportions over time, \(C_{PG2}\)
• Figure 3: J. G. Bartholemew, Isochrone Map, 1914

Acknowledgments

This dissertation would not have been possible without the advice and support of a large network of scholars, at Columbia University and elsewhere. I’d like to thank my dissertation committee: Sarah Cole, Matthew Hart, and Dennis Tenen; the Literary Modeling and Visualization Lab, and the Group for Experimental Methods in the Humanities at Columbia University; Explosion AI, who generously donated a research license to Prodigy, their proprietary human-in-the-loop AI model training software; and the members of the 20/21st Century Colloquium at Columbia University, who provided useful feedback on early drafts of the first chapter.