Deception can produce cognitive dissonance (Festinger, 1962). We explored whether it was possible to leverage classic research on attributions of arousal to improve deception detection. In Study 1, targets voluntarily lied or told the truth about a transgression under one of three conditions: the room's lighting reportedly induced arousal (arousal attribution), induced relaxation (relaxation attribution), or was not mentioned (control). Study 2 featured similar attribution conditions; however, targets were explicitly instructed to lie or tell the truth about a witnessed event (i.e. behavioral choice was eliminated). As predicted, observers' deception detection improved in the arousal and relaxation attribution conditions (vs. control) only when targets' behaviors were volitional. This work provides preliminary evidence that misattributions of arousal might affect deception under high-choice conditions. In turn, attribution manipulation is a promising theory-based approach to improving deception detection that merits further investigation.