Anthropometric test devices (dummies) provide a means for evaluating vehicle/occupant restraint systems for compliance with safety standards. Inasmuch as all tests cannot be performed with the same dummy, it is important that all individuals of the dummy population exhibit as nearly the same response as possible when tested under identical conditions. Moreover, a particular dummy should also exhibit highly repeatable performance when subjected to the same impact environment. This paper describes efforts to measure (1) performance difference attributable to test-to-test variability of a particular dummy, (2) dummy-to-dummy variability of two or more dummies of the same manufacture, and (3) variability from one make of dummy to another. In recognition of the fact that repeatability depends on the restraint environment, comparions were conducted under a number of restraint conditions.

Two makes of dummies were tested:​ Sierra and Alderson. Two dummies of each make were employed, and each was subjected to three replicate tests. For each run, head and chest accelerations and femur loads were recorded as functions of time. Two aspects of the time histories were isolated as being of interest in the repeatability comparions:​ (1) general shape of the response pulse, and (2) high-frequency perturbations about this general shape. The time history records were digitized for computer analysis and appropriate numerical measures extracted for statistical analysis. This analysis consited of partitioning the total variance of the observations into test-to-test and dummy-to-dummy components.

With regard to qualification of dummies for use in compliance testing, it was concluded that dummies should be qualified no on the basis of response repeatabilit alone but rather with due regard to their ability to discriminate between severity levels of the environment to which they are exposed Thus, sensitivity and repeatabilit must bth be considered in judging the efficay of a dummy for safety compliance testing. Lack of repeatability of dummy response when subjected to supposedly identical conditions can arise because of either (a) difference in the response characteristics of the dummies or (b) differences in the test environment. Moreove, various quantities can be viewed as measures of repeatability. Some of these quantities evidence less variability than others because they, in effect, emphasize difference aspects of the dummy response. Repeatability of response also depends on the nature of the test environment. In some instances, minor variations in the test environment may induce sizable changes in dummy response;​ in other instances, dummy response may be relatvely insensitive to such variations. Finally, it was concluded that for statistically valid conclusions a larger number of tests may be required than is often realized.