OBJECTIVE: To investigate the factor structure of the Agitated Behavior Scale (ABS) in patients with traumatic brain injury (TBI) in posttraumatic amnesia (PTA). SETTING: Inpatient TBI rehabilitation ward in Victoria, Australia. PARTICIPANTS: A total of 364 patients aged 16 to 92 years meeting diagnostic criteria for TBI and PTA admitted between September 2013 and October 2020. DESIGN: Retrospective cohort study utilizing electronic medical record data. MAIN MEASURES: The ABS and the Westmead Post-Traumatic Amnesia Scale (WPTAS). RESULTS: Exploratory factor analysis uncovered 2 moderately correlated underlying factors (0.52), labeled Restlessness and Aggression/Lability. Two items failed to demonstrate sufficiently large loadings on either factor. Both factors demonstrated adequate reliability (Cronbach α = 0.87 and 0.81 for Restlessness and Aggression/Lability, respectively). Linear regression indicated that higher WPTAS scores were associated with lower levels of Restlessness (β = -.14, P < .001), supporting construct validity. Conversely, WPTAS scores were not significantly associated with Aggression/Lability (β = -.12, P = .08). Subgroup analysis indicated that a history of mood disorder was associated with greater severity of Aggression/Lability (P = .02). Confirmatory factor analysis indicated superior fit of the identified 2-factor solution when compared with previously explored 1-, 2-, 3-, and 4-factor structures. CONCLUSION: This study suggests that the latent structure of the ABS is best explained by a single construct of agitation with 2 discrete facets reflecting Restlessness and Aggression/Lability. These subscales may be used in clinical practice to evaluate the severity of different aspects of agitated behavior, inform treatment decisions, and judge the efficacy of interventions over time. Further research is required to explain low factor loadings demonstrated by 2 items.