Experimental car-pedestrian collisions were performed with a modified PART 572 dummy and cadavers; they involved some reconstructions of real accidents. These collisions brought to light the differences between the kinematics and the impact responses when dummy and human subject are compared under identical and realistic test conditions to simulate a pedestrian struck sideways.These differences are mainly due to the overall relative stiffness of the PART 572 dummy when compared to cadavers. Same-type collisions were therefore carried out again with other dummies which were designed so as to simulate human response in lateral impact better; thus they were also assumed to display better kinematics as pedestrians. APROD and ONSER dummies were used; when compared to PART 572, their flexibility and deformation capabilities are greater, in particular as regards their thoraxes and shoulders.The corresponding runs are analyzed; results in terms of comparisons between their kinematics and that of PART 572 dummies are obtained. In addition, some runs were performed with human subjects under strictly identical conditions in order to obtain references for the ranking of these dummies' responses. Data obtained are discussed. Findings of this research should be used as a starting point for the specification of a correct pedestrian dummy, i.e. with humanlike responses when struck by a vehicle.