This paper presents a quantitative comparison technique for evaluating similarities between sets of data resulting from a common random process. The normalized integral square error (NISE) criterion is derived using correlation techniques and is applied to deceleration-time histories obtained from six vehicle crash tests in an attempt to relate the vehicle response to the crash event. Intercomparison of the six sets of data using NISE suggests that differences in amplitude and phase shift may account for much of the discrepancy between sets of test data. That is, the magnitude and the time of occurrence of deceleration peaks determine to a large extent the value of the comparator, NISE.These discrepancies in phase shift and amplitude are examined further with the breakdown of NISE into components which isolate errors due to phase shift and amplitude. We find that differences in phase shift account for most of the error detected by the NISE criterion.Two sets of data resulting from computer simulations of vehicle crashes are also analyzed to determine if NISE is usable as a computer model validation tool. Finally, a possible graphical method for simplifying the calculation of NISE is presented.