Cultural conceptions of science in 100 billion tokens of text

Public trust in science is high, but extends unevenly to scientists and their work. Recent accounts of this phenomenon refer to scienciness, suggesting that more prototypical examples and practitioners of science benefit more from the public’s general trust in science. However, current understandings of the public concept of science are too limited to yield definitive predictions about which scientists, and which scientific work, are perceived as more sciency than others. Here, we analyse patterns of linguistic associations in over 100 billion words to characterise the semantics of the public concept of science. These analyses show that sources, occupations, institutions, and social outcomes vary systematically in the strength of their association with science, and that this variation reliably predicts the extent to which they are associated with trust. These results shed light on the public concept of science, confirm sciency exemplars of science are more trusted, and help understand why the public trust science in general but also reject scientists and their work.