A statistical theory is used to quantify the amount of information transmitted from a transducer (i.e., accelerometer) to the air bag controller during a vehicle crash. The amount of information relevant to the assessment of the crash severity is evaluated. This quantification procedure helps determine the effectiveness of different testing conditions for the calibration of sensor algorithms. The amount of information in an acceleration signal is interpreted as a measure of the ability to separate signals based on parameters that are used to assess the severity of an impact. Applications to a linear spring-mass model and to actual crash signals from a development vehicle are presented. In particular, the comparison of rigid barrier (RB) and offset deformable barrier (ODB) testing modes is analyzed. Also, the performance of front-mounted and passenger compartment accelerometers are compared.