Effects of Distraction on the Variance of Lane Position

The variance in lane position is an indicator of the driver’s stability in maintaining the car within the lane – independently of its average location within the lane. Thus, it was expected that in the presence of a demanding distracting task, the variance would increase relative to the driving without distractions, and that with practice the variance would decrease.

The results of the analysis of variance on the variance in the lane position are presented in table 4. The only significant main effect was that of age, with the younger drivers being much more stable than the older ones: the youngest drivers had a lane position variance of 0.9 ft, the middle age drivers had 1.4 ft, and the older drivers had 1.6 ft. All the two-way interactions between day, speed, and distraction were also significant, as was the three-way interaction between them. This last interaction is plotted in figure 13.

Table 4. Analysis of Variance on the Effects of Age, Day, Speed Condition, and Distraction on Variance in Lane Position.

It is obvious from this figure that, with few exceptions, the variance was relatively stable around 1.3 ft when the drivers were in the 50 mph and car-following conditions. However, in the most demanding situation of maintaining 65 mph, the variance varied greatly from day to day in the absence of distraction. It remained quite high when the drivers maintained a conversation over the phone; and reflected significant learning process in the most demanding math operations dual task. Thus, when both the driving task and the phone task were most demanding, a significant learning process was evident in the decreasing variance in lane position, to the point when – on the fifth day – it was actually as low as for any other specific combination of day, speed and distraction (0.7 ft).

Figure 13. The Effects of Distraction, Age, and Speed on the Drivers’ Variance in Lane Position

[d]

Effects of Distraction on Steering Variability

Steering variability was measured in terms of the rate and extent of steering wheel corrections. The actual measure was the absolute value of the steering angle deviations in radians/second (rads/s). As with lane position variance, the rationale behind the use of this measure is that the greater the load on the driver, the greater the extent and rate of steering corrections that the driver has to make in order to retain the desired lane position.

The results of the four-way analysis of variance on this measure are summarized in table 5. With the exception of one interaction, all main effects and interactions were highly significant, and in the expected direction. Steering variability decreased with practice (0.67, 0.65, 0.64, 0.60, and 0.60 on Days 1 through 5, respectively), with speed (0.70 at 65 mph, 0.66 in car following, and 0. 54 at 50 mph), and with age (0.70 for the older drivers, 0.57 for the middle age drivers, and 0.63 for the young drivers). The effects of distraction were the only counter-intuitive effects; with steering variability of 0.73 rads/s in the absence of distraction, and 0.58 rads/s when either distraction task was used.

Table 5. Analysis of Variance on the Effects of Age, Day, Speed Condition, and Distraction on Steering Deviations (in radians/second).

Figure 14 reflects the joint effects of speed and practice on the steering wheel deviations. The figure clearly reflects the improvement with practice and the consistently poorer performance at the higher required speed of 65 mph than at the lower speed of 50 mph and the car following mode. Practice and Speed interacted significantly with age, and the three-way interaction is depicted in figure 15. As can be seen from that figure, the most significant learning effect was observed for the older drivers. In fact, the steering deviations of the younger and middle age drivers, which were less than those of the older drivers, improved only slightly over the five days. In contrast, the older drivers started out with very high rate of deviations (in the 65 mph and car-following condition) and improved consistently until, on Day 5, their performance was the same on all conditions, and similar to that of the younger drivers (except for the 50 mph condition, where their performance remained poorer than that of the younger drivers). Thus, as with some of the measures before, the most significant learning is observed in the older drivers when driving under the most demanding conditions.

Figure 14. The Effects of Practice and Required Speed on the Drivers’ Steering Wheel Deviations (in radians/second).

[d]

Figure 15. The Effects of Age, Practice and Required Speed on the Drivers’ Steering Wheel Deviations (in radians/second).

[d]

Distraction had a significant effect on steering wheel deviations both as a main effect and in its interactions with all of the other variables. However, the effects were either opposite than expected (more steering wheel deviations without distraction than with it), or complex. Figure 16 shows the 3-way interaction of distraction, practice and age on steering wheel deviations. No consistent pattern is apparent in the absence of distraction or when the distraction is limited to a conversation. However, the requirement to perform math operations yields a consistent pattern in which all drivers exhibit learning. More relevant to the central hypothesis of this study, is the fact that initially the older drivers’ performance is significantly poorer than that of the two younger groups, but over time, with practice, all three groups converge so that on Day 5 their performance is essentially the same. The effect of practice on the older drivers is almost an inverse image of the effect of practice on the younger drivers. Since neither one shows a consistent trend, we have no explanation for this.

Figure 16. The Effects of Age, Practice and Distraction on the Drivers’ Steering Wheel Deviations (in radians/second).

[d]

A detailed examination of the significant four-way interaction showed that the source of the consistent effect of the math operations was, as might have been expected, in the 65 mph condition. This is demonstrated in figures 17a and 17b, from which it can be seen that in this most demanding combination of speed and distraction there is a very consistent and large learning process in which drivers of all ages initially start out with deviations of 1.1-1.4 rads/s, and end up with nearly zero rads/s. Furthermore, initially there is a significant age effect where the older drivers are the poorest performers and the middle-aged ones are the best, and eventually all groups perform the same – with close to perfect performance.

Figure 17a. The Effects of Age, Practice and Speed Condition on the Drivers’ Steering Wheel Deviations (in rads/s), when Distracted by Requirement to Perform Math Operations.

[d]

Figure 17b. The Effects of Age, Practice and Distraction on the Drivers’ Steering Wheel Deviations (in rads/s), when Required to Maintain a Speed of 65 mph.

[d]

Effects of Distraction on Crashes

The number of crashes – including collisions and driving off the road – was very small: a total of 45 crashes for the 450 combinations of age X day X speed X distraction. With only 10% of the segments having crashes, it was impossible to analyze that data in a factorial design. Still, using Chi Square analysis it was possible to analyze the main effects. There were no consistent and significant effects of Day (13, 14, 8, 2, and 8 crashes for Day 1 through 5), for age (11, 14, and 20 crashes for young, middle-aged, and older drivers, and for Distraction (13, 16, 16 crashes for no distraction, math operations, and conversation). The only significant effect was obtained for speed, where the number of crashes was 26 for 65 mph, 14 for car following, and 5 for 50 mph [Chi Square (2 df) = 7.99, p=.01]. This result also demonstrates the greater difficulty of driving at 65 mph than car-following at a lower speed or driving at 50 mph.

Effects of Distraction on Reaction Time to Peripheral Signals

Reaction times (RT) to the peripheral signals were analyzed to a more limited extent because the signals appeared randomly during each session. Since there were 12 signals in each session and nine specific combinations of speed and distraction conditions, in some of the conditions no signals appeared at all. Consequently, because of these missing data, only main effects and two-way interactions could be analyzed.

The ANOVAs yielded a main effect of age [F(2,26)=20.75, p<.0001], indicating longer reaction times for the older drivers (2.38s) than for the middle age (1.59s) or younger drivers (1.26s). Reaction times were also longer at the 65 mph speed requirement (1.90s) than at the lower speeds (1.74s for the car-following, and 1.59s for the 50 mph) [F(2,52)=5.37, p=.008]. The effect of practice was also significant [F(4,64)=4.87, p=.002], with reaction time generally decreasing over the five days (1.89s, 1.65s, 1.47s, 1.51s, and 1.32s on Days 1-5). Thus, as expected, reaction time increased with increasing difficulty of the driving task, increased with increasing driver age, and decreased with practice, indicating that whatever effects were observed in the driving measures they were not offset by a speed-accuracy effect in which drivers compensated for the poor driving by paying more attention to the peripheral signal reaction time task.

Interestingly, the effect of distraction was not significant either as a main effect, but was significant in its interaction with day, or practice [F(8,128)=2.77, p=.0007]. The interaction was due to the fact that on Days 1 – 3, RT to the signals in the absence of any distraction was greater than with the distraction, and on Days 4-5, RT was faster without a distracting task than with it.

Effects of Driving on Performance of Math Operations as a Distraction Task

The best estimate of the effects of distraction on driving can be made when performance on the distraction task remains the same throughout all conditions. To that end participants were asked to be as accurate as possible on the math operations task under all driving conditions. Therefore, it is important to look not only at the effects of the distracting task on the driving, but also on the effects of the driving task on performance of the distracting task, with the hope of finding no significant effects. Nonetheless, performance on the math task also improved with practice. There were significant effects of Practice, age, and speed, and significant interactions of practice with both driver age and the required speed condition. These ANOVA effects are summarized in table 6.

Table 6. Analysis of Variance on the Effects of Age, Day, and Speed Condition on Performance in the Distracting Math Operations Task.

The effects of practice and age are illustrated in figure 18. As can be seen from this figure, in addition to the main effect of practice, drivers of all ages display a learning effect, but the learning gradient is much greater for the young novice drivers than for the older and middle-aged drivers (which, in turn, do not differ significantly from each other). The differences between the young drivers and the drivers belonging to the older groups persist only for the first two days, after which they are not statistically significantly different from each other. A similar pattern is observable for the interaction between practice and speed, which is reproduced in figure 19. The gradient of learning is greatest for the most demanding 65 mph condition, which initially generates the most errors. However, by the third day the number of errors is essentially the same at all speed conditions, though the improvement in all speed conditions continues into Day 5. Taken together, these results suggest, that if anything, the learning effects, the greater difficulty experienced with the 65 mph, and the greater difficulty initially experienced by the young drivers as they were observed on the driving tasks are all probably an underestimate. No tradeoffs were observed between the driving tasks and the math operations task that would complicate the understanding of the results of the driving tasks, except for the fact that the continued reduction in errors on the math operations task beyond Day 3 is accompanied by an increase in the speed variance (figure 8), especially for the younger and older drivers on the difficult 65 mph task (figure 9).

Figure 18. The Effects of Age and Practice on the Number of Errors in the Distracting Math Operations Task.

[d]

Figure 19. The Effects of Speed Condition and Practice on the Number of Errors in the Distracting Math Operations Task.

[d]

Effects of Practice on the Subjective Estimates of Fatigue and Task Difficulty.

As expected, the subjective evaluation of the workload and related measures decreased significantly over time. The participants rated their workload on a scale of 1 (very little) to 9 (very much) on six different dimensions. The results of the Analysis of Variance are presented in table 7. There was a significant main effect of a decrement in the effort experienced over the five days, and separate ANOVAs on each measure showed that the practice effect was significant for each of the six measures of workload. There was also a significant Days X Question interaction indicating that the rate of decrease was not the same for all measures. The rate of decrease in effort on each of the subjective measures is depicted in figure 20. The rate of decrease in workload was less for the question about the “physical demands” of the driving task than for all other measures, at least in part because on this measure even the initial rating was not very high.

Table 7. Analysis of Variance on the Effects of Age, Day, Question on the Subjective Workload Assessment.

Figure 20. The Effects of Practice and Subjecive Workload Component on Subjective Workload Estimates*.

[d]

* Q4 is not included is this analysis since it was qualitatively different: asking the driver to evaluate his/her success in the task

The drivers’ response to the question, “How successful were you at the driving task today?” showed no significant effect of day or age, but a marginally significant interaction of these two variables, F(8, 108) = 1.93, p=.06. This interaction is presented in figure 21, from which it can be seen that while the drivers belonging to the youngest and oldest groups felt that their driving improved over time, the middle-aged drivers, if anything, felt that their driving actually deteriorated (however, the only significant contrast was between the youngest and middle-aged drivers on Day 5).

Figure 21. The Effects of Practice and Age on Drivers’ Estimates of How Well They Performed the Driving Task.

[d]