Experimental simulations of car pedestrian collisions were conducted with production cars impacting a pedestrian dummy derived from PART 572 ; the tests were performed under well standardized conditions to try to limit experimental dispersions.The results corresponding to one series of tests are presented. The comparison of kinematical behaviour of this dummy and human subjects impacted by an other way under identical realistic test conditions displayed differences in kinematics and response at impact between these two kinds of surrogates. These differences can be explained largely by the greater stiffness of the PART 572 dummy, as compared to human subjects.For this reason, it was determined to duplicate the tests performed with the PART 572 dummy with a new series of tests with one APROD dummy, modified in a pedestrian version. This dummy is more flexible, especially at the thorax and shoulder levels than the PART 572 one and it may be supposed that these improvements for side impact would give also a better response, in pedestrian accidents situation. So, ten tests are conducted with the APROD dummy in as identical as possibletest conditions ; the kinematics between this dummy and the PART 572 one at impact against the vehicle are compared and the measurements values related to protection criteria are analyzed with reference to the results obtained previously with the PART 572 dummy.