The performance of pedestrian surrogates including a mathematical model is assessed on the basis of a comparison between a sample of real accidents and the results of simulated collisions. For this purpose a representative standard test program was derived from the characteristics of real accidents. A substantial increase in average injury severity is observed in reality between impact speeds of 25 and 35 km/h. This finding is reflected in the simulated collisions insofar, as the measured or calculated loadings of the surrogate in general exceed the known tolerance limits only if the impact speed is in excess of 30 km/h. The lack of surrogate motion prior to impact is shown to cause the largest differences in comparison to real accident circumstances and their outcome. Moreover, the details of the measured accelerations during contact exhibit a large variability and their significance with regard to injury mechanisms remains to be established.