Affective Intelligence (AI) theory proposes to answer a fundamental question about democracy: how it succeeds even though most citizens pay little attention to politics. AI contends that, when circumstances generate sufficient anxiety, citizens make informed and thoughtful political decisions. In Ladd and Lenz (2008), we showed that two simpler depictions of anxiety's role can explain the vote interactions that apparently support AI. Here, we again replicate Marcus, Neuman and MacKuen's (2000)'s voting model, which they contend supports AI, and again show that it is vulnerable to these alternative explanations, regardless of how candidate choice is measured. We also briefly review the broader literature and discuss Brader's (2005, 2006) important experimental results. Although the literature undoubtedly supports other aspects of AI, few studies directly test AI's voting claims, which were the focus of our reassessment. In our view, the only study that does so while ruling out the two alternatives is our analysis of the 1980 ANES Major Panel (Ladd & Lenz, 2008), which finds no support for AI, but ample support for the alternatives. None of the responses to Ladd and Lenz (2008) addresses these findings. Overall, evidence that anxiety helps solve the problem of voter competence remains sparse and vulnerable to alternative explanations.