tion for a moving target is a more complicated task than the same judgment for a stationary target. Using the Gibsonian model for TTC judgment (e.g., Lee, 1976), in the case of a moving target, the optic flow originating from the target is inconsistent with that originating from the background scene. The estimation of relative motion and the subsequent calculation of required braking point based on this information will therefore be more difficult. It was noticeable that the variability in subjects' brake points was significantly higher for slow relative speeds (i.e., faster targets) than for fast speeds of approach (slower or stationary targets). This inconsistency can be taken to support the previous hypothesis. If moving target situations are more complex, the results could be due to the calculation of relative motion simply taking longer, leading to later braking decisions.
There are several implications of the findings of the first study to the design of collision avoidance headway displays. First, given that only the abstract visual warning achieved better performance results than the nondisplay-mediated control, and that in general it was liked by subjects, it is recommended that this type of visual warning display should be employed. The display should also provide continuously available feedback as the one used in this study. It should be noted that the visual displays used in the experiment were presented in head-up mode. The assumption was that if one display was more effective than the others, then it was certain to be apparent in this superior presentation mode. Of course, there is no guarantee that the previous findings can be extrapolated to other modes of presentation (i.e., midhead and head-down displays).
Second, in collision avoidance systems dedicated to a particular type of hazard (e.g., a headway warning system), the use of a discrete nonspeech auditory warning combined with the visual display is recommended. This conclusion is based on the grounds that subjects clearly preferred the nonspeech warnings. Also, whereas this type of display was not shown to result in braking responses that were significantly earlier than the nondisplay-mediated control, it incurred fewer collisions. Third, where an integrated collision avoidance system is to be designed the use of speech warnings is advisable. Bertone ( 1982) noted that speech warnings are more informative than simple auditory tones, and they not only alert the user to the problem but also provide more cues as to its nature. The fact that language is highly overlearned means that speech is likely to be more effective in conditions of high workload or stress, where the meaning of coded auditory tones may be forgotten ( Edman, 1982). As stated previously, a number of experimental factors contributed to the unpopularity of the