Laboratory testing measured the response of a 1984 Chevrolet S-10 Blazer seatbelt buckle to impact on the back of the buckle. The peak acceleration, pulse duration and webbing tension were recorded to map the unique circumstances necessary to inertially unlatch the buckle. The conditions necessary to inertially unlatch the buckle in the laboratory were compared with the measured buckle responses in fifteen sled tests and six rollover crash tests using anthropomorphic dummies. All of the crash tested buckles remained latched and all had dynamic responses well below those required to produce inertial unlatching. Dummy hip areas were measured to be significantly stiffer than humans. Buckle accelerations measured in the “parlor trick” of intentionally striking the hip with a buckle are not representative of crash conditions.