The air bag has proven effective in reducing fatalities in frontal crashes with estimated decreases ranging from 11% to 30% depending on the size of the vehicle [IIHS-1995, Kahane-1996]. At the same time, some air bag designs have caused fatalities when front-seat passengers have been in close proximity to the deploying air bag [Kleinberger-1997]. The objective of this study was to develop an accurate and repeatable out-of-position test fixture to study the deployment of air bags into out-of-position occupants. Tests were performed with a 5th percentile female Hybrid III dummy and studied air bag loading on the thorax using draft ISO-2 out-of-position (OOP) occupant positioning. Two different interpretations of the ISO-2 positioning were used in this study. The first, termed Nominal ISO-2, placed the chin on the steering wheel with the spine parallel to the steering wheel. The second, termed Chest On Module, relaxed the chin positioning criterion while maintaining the spine parallel to the steering wheel. In this position, the chest was forced against the air bag module with a 100 N force. All tests were performed in one of four nominal positions with respect to the steering wheel plane. The reference position had the center of the sternum on the center of the module. Variations from the reference position examined included the occupant sternum displaced 2 cm and 4 cm vertically in the steering wheel plane and displaced 4 cm left. Tests were performed using a production ‘depowered’ air bag from the current automobile fleet. To minimize the effect of external conditions, tests were performed in a climate-controlled room.Dummy positioning on the test fixture was found to be repeatable to within 0.3 cm on all axes. This variation was within the dimensional similarity of the two available 5th percentile female Hybrid III dummies. A large variation in occupant response was found with a very small change in effective distance from the sternum to the air bag module. Nearly 50% variation in peak chest center-of-gravity resultant acceleration was found when moving from the sternum pressed on the air bag module to the sternum effectively being 2 cm from the module. In addition, large variations in occupant response were found with vertical and horizontal displacements of the occupant with respect to the air bag module center. Also, a qualitative change in air bag deployment was found on changing the horizontal position by 4 cm to the left. These variations have significant implications for expected response from in-vehicle out-of-position dummy tests.