Distant Horizons: Digital Evidence and Literary Change

As literary historians have learned to compare thousands of volumes at a time, they have stumbled onto century-spanning trends that are not yet fully understood. This book explores some of those trends in English-language literature. It shows, for instance, that patterns of literary judgment are very durable: a model trained on reviewing patterns in the nineteenth century can also predict the choices of twentieth-century reviewers. Chapter 2 traces the consolidation of detective fiction and science fiction; Chapter 4 measures the gradual blurring of boundaries between grammatically masculine and feminine characters. Throughout the argument, emphasis falls on the gradual emergence of a specialized literary language that continues to shape our assumptions about the purpose of poetry and fiction even today. The book also explains the new modes of quantitative analysis that are making these patterns visible. Instead of framing a debate about “digital humanities,” or a conflict between “close” and “distant" reading, the book presents statistical models as interpretive strategies akin to humanistic interpretation. The argument relies especially on the premise that machine learning can be trained on different subsets of evidence, in order to help scholars reason about the differences between historical perspectives.