In spatial theory a central concept is salience, or the relative importance of issues in a voter’s mind in evaluating candidates’ platforms. Traditional, self-reported measures of salience have either been national in breadth (“which issues are most important to the nation as a whole?”) or personal (“which issues do you care most about personally?”). In the former case, the subjects are being asked to guess what issues other voters think are important; in the latter case, subjects are likely to report issues that are “socially” important to avoid seeming selfish or superficial. Unsurprisingly, such self-reported measures have not been found to explain actual candidate choices by individual voters very well. We introduce a simple process-tracing measure of salience, using mouse-tracking. Experimental participants were asked to rate three hypothetical candidates, using information accessed in a setting where the distribution of attention represents salience in the decision process. Four models were tested: standard city block distance and then the addition of each of the two measures of traditional salience—national and personal—and, finally, the attention distribution measure. Attention distribution improves model fit over the standard distance model and improves classification compared to the traditional salience measures.