We highlight challenges to extant formulations of Q-equivalence approaches to ellipsis licensing, which converge on the idea that sluices must be anaphoric to a question the antecedent makes salient. We argue that antecedents themselves are not responsible for raising any particular issue; instead, we propose that antecedents make salient a proposition — namely, a set of worlds — which may be partitioned into potential question meanings "about" the antecedent. This set is definedas the set of non-singleton covers of the antecedent proposition. The sluice must be a member of this set. We show how this proposal solves known puzzles in the literature, including Dayal & Schwarzschild's (2010) Antecedent Correlate Harmony generalization and Barker's (2013) 'Answer Ban'. We additionally argue that this account more naturally explains the prevalence of sprouting in natural language than previous accounts.