What was classification as it first took modern form in the eighteenth century, how did it work, and how did it relate to earlier describing and ordering? We offer new answers to these questions by considering an example less well known than that of botany or zoology, namely medicine, and by reconstructing practice on paper. The first and best-known disease classification is the “nosology” of the Montpellier physician François Boissier de Sauvages de Lacroix. Its several editions, we show, were less products than process: published tools for building a classification system. The disorder of a hitherto unstudied notebook that Boissier de Sauvages kept throughout this process provided a way of breaking with the topical order of earlier physicians’ humanistic commonplace books of disease observation while sustaining the paper practices those earlier physicians—and Sauvages himself as a student—had used to order disease. This suggests a different picture of historical change than that of a scholarly world of ordered words giving way to a scientific one of ordered things. Classification, in the case of Sauvages’ nosology, arose through an incomplete break with, and intensified practice of, a past way of ordering the described world. The humanist paper practice that had made observationes, differently applied, now made species. Classification into genera and species by similarity and difference, which Sauvages’ nosology shared with botany, was an algorithm of paper and ink practice—in its operation more machine-like than humanist textual practice yet in its effects more creative and re-creative of categories and questions of relationship. Thus a new empiricism of generalizations (species) arose out of the older, Renaissance empiricism of particulars (observations, facts).