The purpose of this study is to evaluate the underreporting of violence and aggression on the Staff Observation Aggression Scale-Revised (SOAS-R) when compared to a simpler assessment: the Aggression Observation Short Form (AOS). During a period of one year, two open and two closed wards gathered data on both the SOAS-R and the AOS for all of their patients. The 22-item SOAS-R is to be filled out after each violent episode. The 3-item AOS is to be filled out during each shift and should also record the absence of violence. The SOAS-R registered 703 incidents and the AOS registered 1,281 incidents. The agreement between the SOAS-R and the AOS was good (kappa = 0.65, 95% CI = 0.62-0.67). Among the 1,281 AOS episodes, 51% were also registered on the SOAS-R. For the 176 AOS episodes with harm, 42% were also registered on the SOAS-R. We found 44% missing registrations on the AOS, primarily for open wards and for patients with short admission lengths. Standard instruments such as the SOAS-R underreport aggressive episodes by 45% or more. Underreporting can be reduced by introducing shorter instruments, but it cannot be completely eliminated.