This talk will report the results of research into the last two centuries of Anglo-American literary history. The project has uncovered a range of specific results. For instance, the phenomena we call "genres" seem to come in very different shapes and sizes. Some remain remarkably stable for 150 years or more; others last only a decade. The diversity of this evidence poses a problem for many existing theories of genre. But beyond these specific results, the talk will reflect on the nature of methodological conversation between humanists and scientists. Humanistic insights provided a foundation for this project: we started by assuming, for instance, that descriptions of genre depend on the perspective of historically situated observers. But the sciences contributed much more than number-crunching: machine learning is also a philosophical discourse, and it can help literary historians reflect on what it means to generalize about a host of diverse examples.