In this paper I attempt to reconcile the apparent definiteness of econometric practice with the vagueness of subjective probabilities assumed in Knightian decision theory. I argue that some standard uses of classical inference are Knightian in spirit, even though the formal justification of...