Are women political candidates penalized by voters for displaying emotion during televised debates? At present, the relationship between candidate gender, nonverbal communication, and voter impressions during debates is significantly understudied. We still lack a general understanding of how gender may moderate the influence of agonistic (competitive) and hedonic (bonding) displays by televised debate participants on evaluations of candidates by voters. In this study, we ask to what extent counterstereotypic gender behavior among political elites during televised debates influences real-time voter impressions of political candidates. We conceptualize emotive signaling as occurring through three different channels: nonverbal (facial displays of emotion), vocal (speech pitch), and verbal (sentiment of speech). Using four full-length debate videos from three German federal elections (2017, 2013, 2009), we employ computer vision, machine learning, text analytic, and statistical methods to automatically extract second-by-second measurements of political candidates' facial displays of emotion, vocal pitch, and speech sentiment. After validating our computational approach, we combine second-by-second emotions data with continuous response measures of support for candidates recorded by live audiences, which have been collected by the German Longitudinal Election Study. Controlling for viewer political attitudes and demographic characteristics, we examine whether multimodal emotive displays by female candidates during televised debates influence real-time levels of support from the viewing public.