Interracial interactions produce anxiety among dominant group members, due to a fear of behaving prejudicially (e.g., Plant, 2004). In turn, anxiety promotes self-monitoring strategies during interracial interactions, reflecting attempts to avoid expressions of prejudice (e.g., Monteith, 1993). In the present research, we investigated the effects of jury racial composition (all White vs. racially mixed) on mock jury deliberations, expecting that interracial juries will trigger social anxiety, self-monitoring, and in turn, diminished communication with fellow jurors (i.e., word count). Testing these hypotheses, mock jurors’ online text responses ostensibly explaining their verdict preference to either a racially diverse or all-White jury were subjected to text analysis via Linguistic Inquiry Word Count (LIWC) software. Supporting hypotheses, mock jurors spoke significantly less (used fewer words), yet simultaneously spent more time developing their response in racially mixed juries than all-White juries. The effect of jury racial composition on word count was serially mediated by fewer social words, and in turn, greater use of first-person singular pronouns in racially diverse juries, likely reflecting increased social anxiety and self-monitoring efforts in interracial contexts.