Two studies were conducted to examine the internal consistency and validity of the state anxiety subscale of the State-Trait Anxiety Inventory (SAI) in the context of acute exercise. SAI responses typically found in the exercise literature were replicated. Analysis at the item level revealed divergent response patterns, confounding the total SAI score. During moderate and immediately after vigorous exercise, scores on items referring to cognitive antecedents of anxiety decreased, whereas scores on items assessing perceived activation increased. Indices of internal showed exercise-associated decreases. A principal-components analysis of responses immediately postexercise revealed a multidimensional structure, distinguishing “cognitive” and “activation” items. By failing to discern exercise-induced and anxiety-related increases in activation from anxiety-antecedent appraisals, the SAI exhibits compromised internal consistency and validity in the context of acute exercise.