The majority of research on the contemporary radical-right has aimed its attention at political and electoral politics while at the same time overlooking the more informal grass-root components (Rydgren, 2018) and the individual motivations that draw people toward the extreme right (Parker, 2018). This paper quantitatively explores evaluations and stance expressions in the caller section of The Savage Nation, a popular ultra-conservative radio show. We randomly selected 106 conversations from 37 shows, aired in August and September 2018, and from March to June 2020. We extracted every occurrence of a stance object and its evaluation(s), whether expressed by the host or a caller. We manually extracted and annotated 873 tokens (Krippendorff, 2004) for the type of evaluation produced using the Appraisal framework (Martin & White, 2005), the alignment (Du Bois, 2007;  Jaffe, 2012; Kiesling, 2015; Morgan et al., 2013; Stivers, 2008) of stance takers with one another, and the evidentials (Bednarek, 2006; Morgan et al., 2013) they provided to back their claims. We then used a Random Forest, a machine learning technique for classification, regression, and survival analysis, to show that both the callers and the hosts (i) have departed from the "old racism" model to instead target opposing ideologies as proxy; (ii) construe the American society in a dichotomous frame that distinguishes between the "pure" and the "corrupt"; (iii) in a paranoid style ripe of over-suspicion and over-aggressiveness.