Many varieties of Romance show more than one intonation contour available for polar question (PQ) marking. Understanding the pragmatic licensing conditions for these contours is no easy task. Experimental work has tended to account for the variation in terms of dichotomies like information-seeking vs. confirmation-seeking or neutral vs. biased . In this paper I use production data to argue that different languages and dialects will encode different types of information intonationally in PQs, but that the type of information that we find encoded through intonation is quite similar to the type of information encoded through sentence-final particles in languages like Cantonese or Lao. These meanings lie on an epistemic gradient, and the points on the gradient that are encoded linguistically through intonation are language-specific (i.e. language X encodes meaning A, but language Y might encode meaning B, or meanings A & B, etc.). I explore three contours in Puerto Rican Spanish, their phonetic implementations, and their meanings with respect to this epistemic gradient. I argue that we should keep in mind the range of possible meanings of SFPs in other languages in order to refine our methodology in a way that allows us to make better predictions about the pragmatic division of labor among intonation contours, specifically for PQs.