Model estimation

Some non-forward focal points, such as left forward and right mirror, were rarely glanced. To satisfy the sample size requirement, grouping similar focal points is necessary. In the renewal cycle generation, the focal points were treated separately to calculate the duration and identify the repetition. However, in the process of model estimation, the three focal points on the left (or right) side were included into one single model dealing with the path shifting from the left (or right) side of a vehicle. In total, 1461 complete renewal cycles without missing value were generated, of which 45.45%, 41.96% and 12.59% were classified as starting a new renewal cycle, repeated renewal cycle and direct vision transition to other non-forward focal point, respectively. Estimation result derived in this study will be divided into four sections for the path of transiting vision from the four non-forward focal points.

(1) Path from left side

Table 4-8 shows the estimated logit models for vision shifting from the left side of vehicles. The results in the first layer shows that the drivers were more likely to transit vision back to the front and, then, begin another renewal cycle after glancing at the left side. The drivers would transit vision to the identical focal point on the left side after the forward glance, particularly under conditions of high speed, distraction and LOS C. It shows a compensatory behavior of which drivers constantly check the frontal side for securing the safety against leading traffic. Driving in such conditions of heavy task load urges drivers to pay less attention to safety irrelevant areas and repeatedly transit vision between left and frontal side more frequently. The type of vision transition that the drivers undertook least frequently is transiting vision from the left side directly to another non-forward focal point, especially in LOS D or E. Meanwhile, the probability of directly transiting vision to other non-forward focal points decreases with the increase of on- and off-road glance durations.

Table 4-8 Estimated logit models for the path from the left side

a: Set as the base alternative in model estimation

b: Represent a path from any one of the focal points on the left side to another one on the identical side.

**: Significant at the level of 0.05

*: Significant at the level of 0.1

The estimated logit model in layer 2 shows the choice of non-forward focal points when starting a new renewal cycle. The most frequent path observed in these samples is transiting vision from left side, to forward direction and to rearview mirror.

Path between these two non-forward focal points indicates a usual manner that drivers adopted to gather information. Additionally, instead of directly transiting vision from one side of vehicle to another side, a substantial path from left to right side via a glance at forward area occurred when drivers approached an intersection. Despite the reduced effort stemmed from connecting two renewal cycles, transiting vision across vehicle is still not a comfortable way for allocating attention and observing surroundings. Such a type of vision transition occurred less frequently than transiting vision to other non-forward focal points, particularly under high speed and heavy traffic (LOS C to E).

Moreover, the duration that drivers glance on-road in the current renewal cycle decreases the probability of the significant paths of shifting attention from left side to rearview mirror and to right side. Meanwhile, the probability of shifting vision to in-vehicle distraction increased. Long glance on the frontal side may represent a relatively stable driving status. Less evident paths would occur after glancing at the left side. The attention allocation may be close to a random pattern rather than following certain paths.

(2) Path from right side

Table 4-9 is the estimated logit model for paths from the right side of vehicles. Comparing with other two types of vision transition, starting a new

renewal cycle share highest portion of renewal cycles choices after glancing at the right side. Relatively, drivers less likely transit vision directly to other non-forward focal points immediately after glancing at the right side, particularly in the conditions of high speed, rain, LOS C and long glance off-road. These results stated the drivers’ cautiousness behavior of inhibiting possible long off-road glances in such restricted conditions. Moreover, probability of repeatedly transiting vision between right and forward side is lower than that of starting a new renewal cycle. Only when driving in the conditions of restricted speed choices (such as LOS D/E), probability of repeated renewal cycle after a glance to right side will increase.

Table 4-9 Estimated logit models for the path from the right side

Layer 1 Layer 2

Left side Right side^aRearview mirror

a: Set as the base alternative in model estimation

b: Represent a path from any one of the focal points on the right side to another one on the identical side.

**: Significant at the level of 0.05

*: Significant at the level of 0.1

Similar with the result of paths from the left side, the probability of transiting vision from right side to left side via a forward glance was found significantly higher than the base alternative. Drivers decreased the probability of such a path when driving in high speed and under LOS B to E, possibly owing to the heightened task complexity in these conditions. Moreover, the path of transiting vision to forward and rearview mirror sequentially after looking at the right side occurred. In line with the path from the left side, probability of transiting vision to rearview mirror decreases when approaching intersections;

while the path of transiting vision to other three non-forward focal points. As for the distraction, the path from right side to in-vehicle distraction, comparing with the ones to other non-forward focal points, was not significant. Yet, when

suffering distractions, probability of transiting vision from right side to in-vehicle distraction increases.

(3) Path from rearview mirror

Table 4-10 shows the estimated logit model for the path from rearview mirror. Unlike the result of left and right side, constant of repeated renewal cycle is insignificant and close to that of choosing a new renewal cycle. The off-road duration provides positive effect on choosing repeated renewal cycles. This result suggests that drivers rely deeply on the rearview mirror and consider it as a focal point that drivers must keep checking. Moreover, comparing with other types of renewal cycle, drivers would increase the probability of the repeated ones with the increasing speed and under rain conditions. These conditions imply a scenario with higher expectancy and lower conspicuity on the leading area. Thus, repeatedly looking at rearview mirror can be seen as a compensatory behavior enabling drivers to constantly check the rear side in harmful situations.

Furthermore, the probability of transiting directly from rearview mirror to other non-forward focal point is exceedingly low. It may support the notion of looking at rearview mirror being the important mean of maintaining situational awareness. Driver would rather transit vision back to frontal side (for a new renewal cycle or a repeated one) than allocate attention to other areas.

Table 4-10 Estimated logit models for the path from rearview mirror

Layer 1 Layer 2

Left side Right side^aIn-vehicle distraction

a: Set as the base alternative in model estimation

**: Significant at the level of 0.05

*: Significant at the level of 0.1

Though the result obtained in layer 2, the path of transiting vision from rearview mirror is relatively simple. The only path found through the sample data is transiting vision from the rearview mirror sequentially to forward side and

to left side. The path of transiting vision from rearview mirror to right side or to glancing at in-vehicle distraction, especially in the condition of suffering distraction and high speed. The repetition becomes more evident with the increase of duration glancing off-road and the decrease of duration glancing on-road. In other words, drivers would spent long time on processing the distracted information; meanwhile, they glance back to the frontal side only shortly for checking the conditions ahead. In this condition, unlike the results obtained in other paths, the majority of attention resource seems to be allocated on the distraction rather than the driving tasks.

Table 4-11 Estimated logit model for path from in-vehicle distraction

Layer 1 Layer 2

Left side Right side^aRearview mirror

a: Set as the base alternative in model estimation

**: Significant at the level of 0.05

*: Significant at the level of 0.1

After finishing the interaction with in-vehicle distractions, results in layer 2 shows that the drivers usually transit vision to the left side or rearview mirror rather than the right side. Glancing at these two focal points would help drivers retrieve their awareness of traffic in surrounding areas, which they may not be able to comprehensively update during the distraction. The result also shows significant effect of duration glancing off-road on the probability of choosing rearview mirror. Comparing to the one of left and right side, drivers’ peripheral vision can still partially cover these two areas. Thus, the longer time that drivers

spent glancing off-road, the more likely drivers would transit vision to the rearview mirror.