Typically, attention toward cues that are threatening to social identity is assessed with self-reports of perceived devaluation (e.g., Branscombe, Schmitt, & Harvey, 1999; Kaiser & Miller, 2001a; Major, Quinton, & Schmader, 2003). However, self-reports of attention to devaluation are problematic. First, they can be biased by self-presentational concerns about the social ramifications of expressing the perception of prejudice (Kaiser & Miller, 2001b) and by motivational concerns to avoid perceiving prejudice (Crosby, 1984; Major et al., 2002). Second, self-reports do not disentangle attention toward cues that are threatening to social identity from interpretation of events as due to social identity, and cues that are threatening to social identity may be attended to outside of conscious awareness. Finally, selfreports of perceived devaluation share substantial method variance and overlap in negative affectivity with individual difference variables typically used as predictors of perceived devaluation (Major et al., 2002). Thus, it is important to examine attention toward cues that are threatening to social identity with measures better suited for capturing attentional processes. In this investigation, we drew upon research on cognitive models of emotion to understand attention toward cues that are threatening to social identity. Research in this tradition often employs dual-task attention paradigms, like the emotional Stroop task, to assess attention to emotionally relevant life concerns (see Williams, Mathews, & MacLeod, 1996, for a review). In the emotional Stroop task, participants view words (some of which represent emotionally relevant life concerns) written in a variety of colors and are instructed to identify each word’s color. If attention is drawn toward the content of emotional words, respondents experience cognitive interference, which causes impaired performance in naming the color of emotional words relative to control words. On emotional Stroop tasks, people with a variety of emotional concerns, including clinical and trait levels of anxiety, posttraumatic stress disorder, and panic disorder, show cognitive interference for words that are semantically related to their current concern or diagnosis (Williams et al., 1996). Furthermore, these effects are particularly pronounced when stimulus words are presented subliminally, presumably because respondents cannot control subliminal attention (MacLeod & Hagan, 1992; Mogg, Bradley, Williams, & Matthews, 1993; Williams et al., 1996). When these research findings are applied toward understanding stigma, it suggests that individuals who chronically expect to face prejudice or who are in situations in which those expectations are salient should show enhanced attention toward cues that are threatening to social identity, whereas individuals who do not chronically anticipate prejudice or who are in identity-safe environments should not show this pattern of attentional bias. Indeed, Davies, Spencer, Quinn, and Gerhardstein (2002; Davies et al., 2005) demonstrated that women were quicker to recognize consciously presented gender-stereotyperelevant words than neutral words after watching commercials that displayed women stereotypically; this difference did not emerge when they watched commercials that displayed women counterstereotypically. Though these results provide evidence that exposure to sexist commercials activates gender stereotypes, Davies et al. did not examine nonconscious attention, nor did they examine attention to words that are threatening to social identity or individual differences in prejudice expectations. Thus, the present investigation provides novel and critical tests of the social-identity-threat perspective.