Claude Steele's stereotype threat hypothesis has attracted significant attention in recent years. This study tested one of the main tenets of his theory--that stereotype threat serves to increase individual anxiety levels, thus hurting performance--using real-time measures of physiological arousal. Subjects were randomly assigned to either high or low stereotype threat conditions involving a challenging mathematics task while physiological measures of arousal were recorded. Results showed significant physiological reactance (skin conductance, skin temperature, blood pressure) as a function of a stereotype threat manipulation. These findings are consistent with the argument that stereotype threat manipulations either increase or decrease situational-specific anxiety, and hold significant implications for thinking about fair assessment and testing practices in academic settings. (Contains 3 figures and 2 notes.)