Since 1978, the National Highway Traffic Safety Administration (NHTSA) has been testing the frontal crash protection provided by new cars in the United States. In the New Car Assessment Program (NCAP), vehicles are crashed into a stationary, full width, rigid barrier at 35 mi/h (56 km/h). Occupant protection is measured by comparing accelerations, forces, and deflections experienced by the head, chest, and upper legs of 50th percentile male Hybrid II or III anthropometric dummies restrained in the driver and right front seat passenger positions. The procedures are similar to those specified in Federal Motor Vehicle Safety Standard 208, except that the speed is 5 mi/h faster resulting in a test that requires the car to manage 36 percent more energy. NCAP was created to be a consumer information program, and the higher test speed was chosen to provide consumers with a clear indication of which cars offered significantly greater occupant protection than the minimum levels specified in FMVSS 208.NCAP has been one of NHTSA's most successful motor vehicle safety programs. In the first years of the program, the information provided to consumers was very revealing. For example, in the 1979 model year, head injury criterion (HIC) scores ranged from 500 -- halfway to the injury threshold of 1000 specified in FMVSS 208 -- to more than 4000, or four times the injury threshold. Of the 30 cars tested, 22 had a driver or passenger HIC more than 1000, 13 had a HIC greater than 1500. Despite manufacturers' concerns that the test results are too variable and that the single test represents too few real-world crash situations , analyses of fatality data have confirmed that vehicles with poor NCAP test results also have poor performance in real-world crashes   . Thus, keeping in mind that valid comparisons can only be made among cars of similar size and weight, the early NCAP testing offered consumers valuable information about occupant protection in frontal crashes. Even though the automobile manufacturers continued to object to the program, they could not ignore it, and getting good NCAP scores became a design objective for new models.However, the success of NCAP frontal crash testing has begun to limit the usefulness of its information to consumers. As auto manufacturers responded to consumer interest in the crash test results with improved designs, and especially as air bags have become standard equipment in new cars sold in the United States, the incidence of test scores that exceed injury thresholds has declined greatly. In the 1994 model year, only 10 of the 38 cars and light trucks tested had a driver or passenger HIC greater than 1000, and only three had a HIC greater than 1500. Among the 100 vehicles tested with driver air bags since the 1990 model year, only three had driver HICS greater than 1000, and those were less than 1100. Thus, NCAP tests of new passenger vehicles are increasingly unlikely to produce the driver HICs or other injury scores that approach the levels associated with increased fatality in the field data.At the same time, the limitations of full-frontal, flat barrier crash tests as measures of occupant protection in all types of frontal crashes are increasingly being recognized. Mercedes-Benz, for example, has argued for more than a decade that, although the flat barrier test used in NCAP is a good test of restraint systems, it does not provide an adequate test of the ability of vehicle structure to resist intrusion . Mercedes-Benz and others have suggested that offset barrier crash tests are needed in addition to full frontal tests for assessing frontal crashworthiness  . Achieving low scores in full frontal barrier tests indicates that a vehicle's front end and restraint system can manage the forces associated with the high vehicle decelerations of full frontal crashes. Low scores in offset barrier tests would indicate that the vehicle's front end structures are able to prevent intrusion even when the crash energy is concentrated on only part of the front structure. The potential significance of offset frontal crashes is illustrated by their frequency in real-world crash data   . O'Neill et al., for example, noted that nearly half of the vehicles involved in two-vehicle collisions and about three quarters of those in single-vehicle collisions in the U.S. National Accident Sampling System in 1990-92 incurred direct contact damage to two-thirds or less of their front ends.There has been general agreement that an offset test could result in vehicle designs that improve occupant protection in crashes not currently addressed, or addressed inadequately, by the flat barrier test, although there has been considerable debate on the form and speed of the test. Mercedes-Benz initially adopted a 40 percent overlap test into a rigid barrier at 55 km/h as an internal manufacturing standard . Hobbs has argued, however, that even in the offset configuration, rigid barriers provide too little replication of the structural stresses that vehicles must handle in vehicle-to-vehicle crashes, which constitute a substantial portion of offset frontal crashes  . According to Hobbs, because the fronts of other cars are a mixture of stiff and soft areas, frontal structures must be designed such that the main structural members (e.g., front longitudinal rails) absorb their intended share of crash energy regardless of the surface struck; the front ends should not allow uneven deformation of softer structure to compromise the structural members such that they buckle catastrophically rather than crush uniformly to absorb energy. Rigid barriers do not test this aspect of design because the rigid barrier face directs crash forces into the structural members. Therefore, Hobbs has proposed that offset tests should utilize a deformable element to provide a better assessment of this aspect of design.This argument has been persuasive, and an offset frontal crash test has been tentatively recommended by Working Group 11 (WG-11) of the European Experimental Vehicles Committee (EEVC) for the second phase of new European Economic Community dynamic frontal crash test performance requirements for new vehicles  . Mercedes-Benz has also accepted the argument for deformable barriers in offset tests , and new car comparison testing using the proposed European standard has begun in Australia . Nevertheless, published results from offset tests using deformable barrier faces for validating the representativeness and repeatability of the proposed offset test procedure are still limited. WG-11 participants are currently completing a series of 16 tests with European vehicles to assess the proposed standard.This paper provides additional data from tests at the Insurance Institute for Highway Safety using vehicles from the U.S. fleet. The test series includes car-to-car frontal crashes of recent models of the Oldsmobile Ciera, which is a front-wheel drive unibody midsize American sedan, at nominal speeds of 56 and 64 km/h, with overlaps around 50 percent. These car-to-car tests served as the baseline condition intended to be replicated through offset barrier tests, following Hobbs' argument that vehicle to vehicle crashes exert the greatest demands on front end designs. The ability of various barrier configurations to replicate these crashes was assessed in terms of the frontal deformations that resulted. Subsequent comparisons considered the deformation patterns of stiff underbody structures, occupant compartment intrusion, and velocity change of the occupant compartment.