The analysis of eye glance behavior also supported the findings that some of the experimental messages were perhaps too complex. Rockwell ( 1988) commented that poor display design is usually reflected in more glances not longer glances, although illegible displays encourage longer glance durations. In the present study, both glance frequency and glance duration increased with screen complexity, suggesting that both the design and legibility of messages could be improved. Although Zwahlen et al.'s ( 1988) cut-off point of 2-sec durations is fairly arbitrary, effort must be spent in reducing glances as far as possible. In particular, the longest glances, of up to 3.8 sec in the present study, must be eradicated through improved message design. Some recommendations for improvements to all aspects of the system are presented next.
Although the experiment was carried out with a visual only system, the findings could be extended to auditory and auditory-plus-visual in-vehicle displays. Gatling ( 1976) found information recall performance in a real- road trial to be comparable across display modalities, but Labiale ( 1990) observed that longer messages were memorized better when presented visually. An optimal solution might therefore be a system using both modalities; the auditory for short summary messages, and the visual giving further details of these messages. In this case, drivers would only have to take their eyes off the road when particular message details were needed (e.g., for route diversion or for confirmation of the auditory message).
The results of the present experiment support the idea that only the simplest four element messages were satisfactory in terms of safety, usefulness, and driver acceptability. Messages with 9 elements were frequently forgotten and resulted in worrying eyes-off-the-road time. Even messages with seven elements seemed to be at the upper limits of drivers' capabilities. Further research is necessary to investigate whether certain types of message element can be more successfully processed than others, and which elements are not considered essential (see Graham, Mitchell, & Ashby, 1995). An initial recommendation to reduce complexity is that the road numbers of junction destinations be removed.
Increasing the display time of messages on screen or providing a "repeat last message" facility would alleviate some of the pacing that drivers experienced with the system. This would particularly benefit older drivers whose speed of cognitive processing is slower. In addition, a short break between screens in a message pair would allow drivers to reorient themselves to the road environment, reduce pacing and avoid excessive glance frequencies over short periods.