"Scientific" public opinion polls arrived on the American scene in 1936. Examining the work of opinion surveyors George Gallup and Elmo Roper, this essay tracks the early career of a new social scientific technology, one that powerfully shaped conceptions of "the public." Pollsters described their craft as a democratic one that could accurately represent the U.S. populace. Yet, their assumptions about that same public-and the techniques they employed to measure it-undermined such claims, and even risked calling the polling profession into question. To understand why Gallup and Roper fell short of their stated ambitions, one must turn not only to the state of midcentury sampling methods but also to the corporate sponsors and commercial pressures underlying their enterprise.