An expression in focus is standardly taken to evoke a contextually salient set of alternatives to itself (Rooth 1992). In an influential study, Husband and Ferreira (2016) proposed that these alternative sets are derived online from the set of semantic associates primed by processing a word in focus. The authors argue that this process occurs via a selectional mechanism which maintains the activation of alternatives while that of non-alternative primes decays. However, the semantic associates of foci are not always a superset of the contextually salient alternatives. For instance, imagine a mural depicting a tank driving through a meadow. In this context, ‘tank’ is a likely alternative to ‘flower’ in “Silvia only painted the [flower]F on the mural” despite the fact that these words presumably do not prime one another. Such examples present a gap in the literature and a possible challenge to the strictest interpretation of Husband and Ferreira (2016)’s proposal. I hypothesize that focus alternatives are generated through the interaction of discourse context and the question-under-discussion (Roberts 1996). I present a probe task experiment in which the alternative status of a non-associate to focus is modulated by context, but the sentence containing focus itself remains the same. This design helps to isolate the hypothesized effects of discourse/QUD from that of semantic priming.