Resilient and exposure-resilient functions are functions whose output appears random even if some portion of their input is either revealed or fixed. We explore an alternative way of characterizing these objects that ties them explicitly to the theory of randomness extractors and simplifies current proofs of basic results. We also describe the inclusions and separations governing the various classes of resilient and exposure-resilient functions. Using this knowledge, we explore the possibility of improving existing constructions of these functions and prove that one specific method of doing so is impossible.