Causality between drivers' successive eye fixations.

Abstract

T o govern a continuous, adequate information input in driving the visual modality must function as a self-regulating system. A proper visual search strategy, as manifested by the discrete fixations of the eye, can only be expected if the perceptual system is both interiorly, as well as exteriorly, oriented. This means that the next fixation of the eye, i.e., information input, must be determined by the subject's present cognitive schemata in relation to the task-oriented relative importance of different environmental elements of the visual display. T o rest rhis assumption, the eye movements, when passing an unexpected building site, of five drivers whose ages were between 22 and 32 yr. were analyzed by means of a systems theoretical approach (see Hirsig, in press). In this driving situation, three essential conditions were fulfilled to guarantee the correspondence between information input and points of fixations. These were a great load of information, avoidance of anticipation, and well-defined distinctions between the road elements. If it is possible to determine a dynamic time-discrete process model which exactly describes the pattern of eye fixations, then a causal relationship between successive fixations is indicated. The model used consists essentially of two kinds of variables which are !he pattern of eye fixation within each interval of say 0.5 sec., judged from the last known information observed, and the task-oriented importance of defined road elements ahead (summarized as environmental vector), which the driver reaches within the next few time intervals. Only the most important finding can be described here. For each driver a dynamic process model could be determined individually. This described the observed sequence of eye fixations quite accurately. The accuracy of these individual models, i.e.. the prediction error, varied among the five drivers from 0.0% to a maximum of 2.6%. The essential differences between the individual models concern the number of the environmental vectors, e.g., the length of driving path ahead in relation to travel velocity, which had to be considered to allow a proper mathematical description of the pattern of eye fixations. The number of these discrete intervals varied among the subjects on n scale of one and four, corresponding to an assumed relevant preview time of 0.5 to 2.0 sec. Because the driving route was a short distance, it was not possible to validate the models established on an independent set of data. Nevertheless, this finding supporred a causal relationship between successive fixations of the eye. The finding suggesred that the present information from fixated targets and the relative importance of the road elements ahead are an integral part of a feed-forward program governing the movements of the driver's eye toward the next target of fixation. This conclusion confirms the general notion that the visual system is both interiorly and exteriorly oriented. The information the driver seeks depends, therefore. on the information he has already picked u p in relation to the road characteristics ahead when he is under the pressure of relevant information to be processed.