Where and when have investigators aspired to study human culture with more trustworthy or 'objective' methods borrowed from the natural sciences? This book revisits the dominant scientific method, 'coding,' with which investigators from sociology to literary criticism have sampled texts and catalogued their cultural messages. By returning to the primary documents used in canonical examples of such research, the author demonstrates that the celebrated hard outputs rest on misleading samples and on unfeasible classifying of the texts' meanings. The problems are so consistent, they reveal why it is a contradiction in principle to try to create 'data' out of such complex texts. As an alternative, the author illustrates how a purely humanistic reliance on families of symptomatic exemplars sponsors research that is retraceable and more accurate and transparent in its conclusions. The book addresses scholars in the field of science studies as well as practitioners across the social sciences who investigate culture.