These elements here represent human-curated examples of each of the features examined from our corpus of manuscripts in the 'Parker on the Web' repository. We determined the most significant components of a manuscript's mise-en-page--its physical layout. We provided training data for our teams of computing engineers, who created an algorithm that permitted a computer program to search for particular information retrieval features on a manuscript page. Our four main foci were Litterae Notabiliores, Enlarged Capitals, Rubrics, and Intertextual Space. Many hundreds of thousands of images were returned, with an excellent non-error rate, demonstrating unequivocally that machine learning works effectively for this kind of study. The implications of this are very significant indeed. Already, we are able to revise current scholarship about the nature of manuscript production in the twelfth century.