A data correction method to eliminate the effect of local tissue-accelerometer vibration from surface measurements of vibration over the spine has been developed and compared with previous direct measurements. A single degree-of-freedom linear model for the local tissue-accelerometer system in the vertical and the fore-and-aft axes is assumed. The natural frequency and the damping ratio of the local system are estimated so as to form a correction frequency function, using spectral analysis of the free vibration response of the local system caused by transient displacements of the accelerometer attached to the body surface. Accelerometers were attached to the skin over the spinous process of the vertebra L3 and on the abdominal wall. For four different masses and each site, correction frequency functions were computed. Seated subjects were then exposed to vertical random vibration (0.5-35 Hz) and acceleration transfer functions from the seat to each accelerometer were calculated. Different transfer functions were obtained with different additional masses but the differences were eliminated by the correction method so as to indicate the transfer functions to the spine and the viscera. For vertical responses, the correction method was effective at frequencies below the estimated natural frequencies of the local system. Fore-and-aft response over the spine did not require correction at frequencies below 35 Hz.