This paper analyzes acceleration measured at different locations on the spine of a post-mortem human surrogate (PMHS) subjected to restraint loading in a frontal impact. The study was performed to assess the use of acceleration measured at the first thoracic vertebra (T1) in the development of thoracic injury criteria for use with anthropomorphic test dummies. This study has an experimental component involving sled tests with a PMHS and with a Hybrid III 50th percentile male dummy, and a computational component involving the MADYMO human body model and the MADYMO dummy model. Mass-scaled PMHS T1 acceleration is found to reflect changes in test conditions (airbag-only, belt-with-airbag, driver-side and passenger-side) differently than dummy chest cg acceleration. PMHS T8/9 acceleration-time history is found to reflect transient restraint loading better than T1 acceleration-time history, but PMHS T8/9 maximum acceleration, even scaled for differences in PMHS and dummy mass, is found to be significantly greater than dummy chest cg acceleration for some test conditions. Spinal curvature and differences between human and dummy spinal flexibility are shown to be significant factors in the interpretation of PMHS-based acceleration measurements, particularly for use as a proxy for dummy chest cg acceleration.