Monitoring head accelerations as an indicator of possible brain injury may lead to faster identification of injury and treatments. This study investigates the skull-coupling of a tri-axial accelerometer mounted to a back molar and compares it with a tri-axial accelerometer inserted in the boney ear canal. These tri-axial accelerometers were mounted to three post mortem human surrogate (PMHS) skulls, and compared with a rigid, skull-mounted laboratory sensor reference cube. Each specimen was subjected to both a high-g loading from a vertical drop tower and a low frequency cyclic loading from a shaker device. The specimens were subjected to an approximate 150g input acceleration on the drop tower, and up to 10g at a frequency of 9Hz on the shaker device. Each specimen was tested on all three of the anatomical axes on both the drop tower and the cyclic shaker. Both the tooth-mounted accelerometer and the ear-mounted accelerometer were in close agreement with each other, and compared favorably with the rigid reference accelerometers. The coupling of the tooth with the skull did produce an amplification of the resultant acceleration, but maintains the basic biofidelity required to develop a simple transfer function for the sensor data.