While shame is essential for adaptive functioning, experiencing shame more often or intensely than others is strongly associated with psychopathology. To date, no measure of the behavioral expression of shame exists, despite the great potential for use in research and clinical settings. The present study aimed to assess the Shame Code, a new behavioral coding system of the expression of shame. Participants included 149 youth between the ages of 12 and 17 (50 % female, M = 14.5). Shame was elicited with a spontaneous speech task. Participants’ overall Shame Code scores were correlated only with a state measure of shame, however, structural equation modeling results showed that Shame Code variables combined differentially to assess state and trait shame scores. A two-factor model was the best fit to the data. The first factor, Fidget, consisted of Hiding, Fidget (positively loaded), Nervous Positive, and Stillness (negatively loaded). The second factor, Freeze, was comprised of Stillness, Facial Tension, and Silence (positively loaded). The Fidget factor was associated with higher Trait Shame and the Freeze factor was associated with higher State Shame but lower Trait Shame. Therefore, the Shame Code not only effectively captured the behavioral manifestations of shame, but Shame Code variables also differentially predicted state and trait shame.