Which issues are discussed by candidates in an election campaign? Why are some issues never discussed? Model tractability is lost quickly when dealing with these questions, partly because of the multidimensional voting inherent in models of multiple issues. Our model features two candidates for office who can talk about any subset of issues, allowing uncertainty both on the part of voters and candidates, and taking candidates to be office motivated. Candidates move first and simultaneously, announcing any positions they choose on any issues. To us, salience is simply the discussion of an issue in a campaign. If both candidates and voters are expected utility maximizers, we find salience results, in that candidates typically want to talk about everything (or they are indifferent between talking and nonsalience). Leaving the expected utility framework, we present an example using "Knightian uncertainty" or "maxmin expected utility with multiple priors" of Gilboa-Schmeidler to illustrate how robust nonsalience and salience of issues might be generated.