Science and emotions are typically juxtaposed: science is considered rational and unattached to outcomes, whereas emotions are considered irrational and harmful to science. Ethnographic studies of the daily lives of scientists have problematized this opposition, focusing on the emotional experiences of scientists as they go about their work, but they reveal little about disciplinary differences. We build on these studies by analyzing Citation Classics: accounts about the making of influential science. We document how highly cited scientists retrospectively describe emotional aspects of their research and assess variation in these narratives across six diverse disciplines: Chemistry; Clinical Medicine; Neurobiology; Physics; Plant and Animal Science; and Psychology and Psychiatry. Using correspondence analysis, we develop a multidimensional model to explain disciplinary variation in scientists’ accounts of emotions and link this variation to internal, external, and material aspects of the disciplines. We find differences in norms of appropriate emotional expression, or "feeling rules," between the "hard" and "soft" sciences, the basic and applied sciences, and the sciences that study living organisms versus those that study organs, cells, or atoms. By comparing accounts across disciplines and elaborating the structuring principles underlying these patterns, we integrate knowledge from varied case studies into an integrative and multifaceted model.