Recent epidemiological and biomechanical studies have suggested that whiplash injury is related to a vehicle's average acceleration rather than its speed change during a rear-end collision. To further explore this phenomenon, the effect of various kinematic properties of the collision pulse on seven proposed whiplash injury criteria was quantified. A BioRID II rear-impact dummy was seated on a programmable sled and exposed six times to each of 15 different collision pulses. Five properties of the collision pulse were varied: peak acceleration (1.3-4.4 g), speed change (3-11 km/h), duration (52-180 ms), displacement (2-26 cm) and shape (square, sine and triangular). Linear and angular accelerations and displacements of the head, and linear accelerations of the T1 and pelvis were measured in the sagittal-plane. Upper neck loads in the sagittal-plane were also measured. Variations within the proposed injury criteria between the different pulses were compared using analyses of variance. Six criteria--peak upper neck shear force, peak upper neck moment, peak retraction, the neck injury criterion (NIC) and two normalized neck injury criteria (Nij and Nkm)--exhibited graded responses that were most sensitive to the average acceleration of the collision pulse. Peak extension angle between the head and T1 decreased with both increasing speed change and peak acceleration, and was, therefore, deemed unsuitable as a whiplash injury criterion for the BioRID dummy. Of the seven criteria, Nij and Nkm were best able to distinguish between the 15 pulses. If the six graded injury criteria are related to the risk of whiplash injury, then the results of this study indicate that the risk of whiplash injury can be reduced by bumper and seat designs that prolong the collision pulse and thereby reduce the average vehicle and occupant accelerations for a given speed change.