In general it was assumed that the effects of the five days of driving and listening to the phone would generate a gradual improvement in both the driving tasks and the phone listening tasks. The critical issue is whether there is a statistical interaction between the learning effects and the phone conditions, and whether that interaction differs for the three different age groups. Such an interaction would be manifested in a diminishing difference in the driving performance as a function of the distracting phone condition from Day 1 (with driving performance being significantly poorer with the distraction than without it) to Day 5.
This general hypothesis – that the interference from the distracting task diminishes over time  is tested below on each of the driving measures: average speed, rootmeansquare (RMS) of speed, average lane position, lane position RMS, average extent of the steering wheel turns, and the subjective workload experienced by the drivers. To test these effects a fourway partially repeated measures analysis of variance (ANOVA) with Age (between subjects), Speed, Distraction, and Day (within subjects) was performed on each of the dependent measures. The general model used for all the analyses was StatSoft’s (2001) General Linear Model, Repeated Measures ANOVA.
Effects of Distraction on Average Speed
A distracting task should make it more difficult for the drivers to maintain the desired speed, so that the requirement to perform the distracting phone task should cause drivers to lower their drivingrelated information processing and drive at a lower speed than required.
The results of the fourway ANOVA on the average speed are reproduced in table 1. As can be seen from this table, all of the main effects and most of the interactions are statistically significant. An examination of the mean speed levels indicated that actual average speed was – as expected  highest for the 65 mph condition and lowest for the 50 mph condition. The interaction between Speed and Day is depicted in figure 1, from which it can be seen that the average speed was fairly constant – and essentially as desired – for the 50 mph and the carfollowing condition. It was higher, and increased over the five days for the 65 mph condition, never quite reaching the required 65 mph. These results indicate that the 65 mph requirement was in fact much more demanding than the other two lower speeds, but that over time drivers managed to approach the desired level.
The differential effects of age on speed are displayed in figure 2, where it can be seen that the effects of age are restricted to the highspeed condition only. At all required speeds, there are essentially no differences between the two younger groups. However the older drivers’ ability to keep the required speed diminishes as the required speed goes up, so that for the 65 mph requirement, their average speed is approximately 4 mph less than that of the younger drivers. Also apparent from this figure is the fact that even for the younger participants, driving at 65 mph was sufficiently challenging that they could not achieve it immediately, and in fact, there is a very noticeable learning curve even for them.
The effects of the distraction task on the driving of the three age groups are presented in figure 3. It is apparent that the distraction had an impact on speed, but the direction of the effect was unexpected, with speed being the lowest with no distraction, and the highest with the distracting conversation. Of the two distracting tasks, the math operations caused the drivers to drive at lower speeds than when the distracting task was a conversation. Although the interaction indicates that the relative effect of distraction is greater on the younger than the older drivers (based on the formers’ higher speeds) this effect is not a strong one, and the effects of age and distraction do seem almost additive.
Table 1: Analysis of Variance on the Effects of Age, Day, Speed Condition, and Distraction on Average Speed.
Effect  Repeated
Measures Analysis of Variance (avgspeed) Sigmarestricted parameterization Effective hypothesis decomposition 


SS 
Degr.
of Freedom 
MS 
F 
p 

Intercept  4216515

1

4216515

70733.49

0.000000

Age  1066

2

533

8.94

0.001047

Error  1610 
27 
60 

DAY  376 
4 
94 
11.14 
0.000000 
DAY*Age  102 
8 
13 
1.51 
0.162737 
Error  910 
108 
8 

SPEED  28093 
2 
14046 
1430.48 
0.000000 
SPEED*Age  504 
4 
126 
12.82 
0.000000 
Error  530 
54 
10 

DISTRACT  318 
2 
159 
91.20 
0.000000 
DISTRACT*Age  25 
4 
6 
3.52 
0.012627 
Error  94 
54 
2 

DAY*SPEED  653 
8 
82 
11.68 
0.000000 
DAY*SPEED*Age  192 
16 
12 
1.72 
0.045173 
Error  1511 
216 
7 

DAY*DISTRACT  567 
8 
71 
20.31 
0.000000 
DAY*DISTRACT*Age  179 
16 
11 
3.21 
0.000056 
Error  754 
216 
3 

SPEED*DISTRACT  1110 
4 
278 
85.69 
0.000000 
SPEED*DISTRACT*Age  30 
8 
4 
1.16 
0.328326 
Error  350 
108 
3 

DAY*SPEED*DISTRACT  862 
16 
54 
16.28 
0.000000 
DAY*SPEED*DISTRACT*Age  253 
32 
8 
2.39 
0.000052 
Error  1429 
432 
3 
Figure 1. The Effects of Day and Required Speed on the Drivers’ Average Speed.
Figure 2. The Effects of Day, Required Speed, and Age
on the Drivers’ Average
Speed.
Figure 3. The Effects of Distraction and Age on the Drivers’ Average
Speed.
The effects of practice in diminishing the distraction effects of the phone task are shown in figure 4, which shows the interaction of Day, Speed, and Distraction on the average speed. The interaction indicates that with the two lower speeds (50 mph and carfollowing) there is no apparent learning in the sense that the speed does not change over time. In these two conditions there is a ceiling effect showing that already on Day 1 all groups of drivers are able to maintain the required speed. In contrast, with the required high speed of 65 mph, there is a learning effect – reflected in an increase in speed over the 5 days, and a diminishing effect of the distraction on the speed: initially the math operations cause a significant reduction in speed relative to the nodistraction and the conversation (which do not differ from each other), but by the end of the 5th day they have no impact on the average speed. This is also indicated by the significant fourway interaction. This interaction showed that in the 50 mph speed condition there were essentially no effects of practice, age or distraction. In the carfollowing mode there were also no effects of practice or age, but the average speed with no distraction was very stable at approximately 55 mph, while with the distracting task it was more variable and significantly higher. Only at the most demanding speed of 65 mph, do we see the effects of learning, distraction, and age. All groups show a learning effect, but the relationship between the learning and the distraction is different for the different age groups, as can be seen in figure 5. The youngest drivers do not show any deleterious effects of the distracting task. The middleaged group initially performs better with nodistraction and conversation than with the distracting math task, but by the fifth session their performance is the same regardless of the presence or absence of distraction. The older drivers show the greatest effect of the distraction initially, but they too are able to combine the two tasks by the fifth session, though their progress is much more variable.
Figure 4. The Effects of Distraction, Practice and Required Speed on the Drivers’ Average Speed.
Fig. 5. The Effects of Distraction, Practice and Age on the Drivers’ Average
Speed when required to maintain a speed of 65 mph.
Effects of Distraction on Speed Variance
A distracting task should make it more difficult for the drivers to maintain the desired speed, and consequently their variance around their mean speed should increase with increasing distraction.
The results of the fourway ANOVA on the speed variance are reproduced in
table 2. As can be seen from the table, all but one of the effects and interactions
were statistically significant.
Table 2. Analysis of Variance on the Effects of Age, Day, Speed Condition, and Distraction on Speed Variance.
Effect  Repeated
Measures Analysis of Variance (varspeed i) Sigmarestricted parameterization Effective hypothesis decomposition 


SS 
Degr.
of Freedom 
MS 
F 
p 

Intercept  35930.48 
1 
35930.48 
744.9535 
0.000000 
Age  1422.92 
2 
711.46 
14.7509 
0.000047 
Error  1302.26 
27 
48.23 

DAY  1011.18 
4 
252.79 
26.9747 
0.000000 
DAY*Age  222.07 
8 
27.76 
2.9620 
0.004908 
Error  1012.12 
108 
9.37 

SPEED  2331.43 
2 
1165.71 
187.5237 
0.000000 
SPEED*Age  149.86 
4 
37.46 
6.0267 
0.000439 
Error  335.68 
54 
6.22 

DISTRACT  215.23 
2 
107.61 
44.8838 
0.000000 
DISTRACT*Age  9.50 
4 
2.38 
0.9910 
0.420404 
Error  129.47 
54 
2.40 

DAY*SPEED  1743.01 
8 
217.88 
51.5541 
0.000000 
DAY*SPEED*Age  223.78 
16 
13.99 
3.3095 
0.000035 
Error  912.85 
216 
4.23 

DAY*DISTRACT  1330.85 
8 
166.36 
59.5925 
0.000000 
DAY*DISTRACT*Age  214.60 
16 
13.41 
4.8048 
0.000000 
Error  602.98 
216 
2.79 

SPEED*DISTRACT  611.41 
4 
152.85 
39.3604 
0.000000 
SPEED*DISTRACT*Age  105.41 
8 
13.14 
3.3841 
0.001673 
Error  419.41 
108 
3.88 

DAY*SPEED*DISTRACT  1856.23 
16 
116.01 
45.0814 
0.000000 
DAY*SPEED*DISTRACT*Age  611.46 
32 
19.11 
7.4251 
0.000000 
Error  1111.73 
432 
2.57 
Looking first at the effects of age and practice we can see from figure 6, that older drivers had more difficulty at the task, manifesting much greater variance than the two younger groups, which did not differ from each other. Some learning is apparent for all groups, though the younger two groups’ variance nearly levels off after the 2nd day, while the older drivers continue to improve on all days. On their last day all groups perform at the same level, suggesting that additional practice would not have yielded significant further improvements for any of the groups. The main effect of practice, across all conditions and driver groups was almost linear, with the speed variance decreasing from 6.5, to 5.5, to 5.3, 4.6, and to 3.9, from Day 1 to Day 5, respectively.
Figure 6. The Effects of Practice and Age on the Drivers’ Speed Variance.
The effects of the required speed and the interactions of speed with age and
practice on the speed variance are demonstrated in figure 7. Whereas the difficulties
that older people had in maintaining average speed were mostly in the high
65 mph condition, their inability to drive smoothly at the required speed – with
minimal variance around their average speed – was apparent and significant
at all three speed conditions. Older drivers had a much greater speed variance
than the younger two groups and showed much more of a learning effect than
them. For all groups the variance was highest at the most demanding 65 mph
condition, and the learning effects were greatest at the most demanding 65
mph condition. Thus, in the easier carfollowing and lowerspeed conditions,
all three groups eventually converged on the 5th day, but in the more difficult
65 mph condition, the older drivers retained the higher variance throughout.
Figure 7. The Effects of Speed, Practice and Age on the Drivers’ Speed Variance
The main effect of distraction on speed variance was significant with an average
variance of 5.2 mph for the nodistraction condition, 4.6 mph for the conversation
distraction, and 5.6 mph for the math operations distraction. Thus, the speed
variance was significantly higher with distraction than in the control condition
only when the phone task required math operations. The requirement to converse,
inexplicably, actually lowered the variance relative to the control condition.
The absence of a significant interaction with age indicated that this pattern
was identical for all age groups. The significant interaction of Distraction
and Day is displayed in figure 8. While the distraction clearly shows the learning
effect, the effects of the distraction and its interaction with learning are
not very consistent. Nonetheless, the figure does reflect the fact that speed
variance disparities between the nodistraction and the math distraction conditions
are much larger on Day 2 than they are on Days 3 and 4. However, on Day 1,
the variance is actually smaller with the distracting conversation (but not
with the math operations) than without it.
Figure 8. The Effects of Distraction and Practice on the Drivers’ Speed Variance
Further examination of the fourway interaction between the variables, revealed
that in the car following and 50 mph condition, the older drivers showed
a learning effect while the younger drivers had essentially the same variance
on all days – the level equivalent to the one eventually reached by
the older drivers on Day 5. However, in the most demanding 65 mph condition,
displayed in figure 9, the effects of practice – though not very consistent
 were significant for the younger and older drivers, and less so for the
middleaged drivers. For all drivers, speed variance in the presence of the
demanding math operations decreased over the first four days, but then (inexplicably)
increased on Day 5 for the younger and older drivers.
Figure 9. The Effects of Distraction, Practice and Age on the Drivers’ Speed Variance when Required to Maintain a Speed of 65 mph.
Effects of Distraction on the Average Lane Position
In the driving simulator, lane position is measured relative to the center of the road. Since the lane width was 8.33 feet, a driver who perfectly centers his/her car in the middle of the lane would have an average lane position of 4.16 feet. Drivers who keep their car closer to the shoulder than to the median would have an average lane position greater than 4.16 ft, and those who keep it closer to the median than to the shoulder would have an average lane position less than 4.16 ft. A summary of the results of the analysis of variance on the average lane position is presented in table 3.
Table 3. Analysis of Variance on the Effects of Age, Day, Speed Condition, and Distraction on Average Lane Position.
Effect  Repeated Measures Analysis of
Variance (avglanepo) Sigmarestricted parameterization Effective hypothesis decomposition 


SS 
Degr. of Freedom 
MS 
F 
p 

Intercept  31185.91 
1 
31185.91 
3640.894 
0.000000 
Age  126.28 
2 
63.14 
7.371 
0.002790 
Error  231.27 
27 
8.57 

DAY  4.65 
4 
1.16 
1.328 
0.264114 
DAY*Age  5.51 
8 
0.69 
0.787 
0.614875 
Error  94.57 
108 
0.88 

SPEED  1.08 
2 
0.54 
1.422 
0.250210 
SPEED*Age  1.70 
4 
0.43 
1.115 
0.359220 
Error  20.60 
54 
0.38 

DISTRACT  1.01 
2 
0.51 
2.417 
0.098782 
DISTRACT*Age  0.65 
4 
0.16 
0.772 
0.548522 
Error  11.34 
54 
0.21 

DAY*SPEED  21.57 
8 
2.70 
8.868 
0.000000 
DAY*SPEED*Age  9.51 
16 
0.59 
1.955 
0.017288 
Error  65.68 
216 
0.30 

DAY*DISTRACT  13.87 
8 
1.73 
8.901 
0.000000 
DAY*DISTRACT*Age  3.36 
16 
0.21 
1.079 
0.376151 
Error  42.07 
216 
0.19 

SPEED*DISTRACT  8.93 
4 
2.23 
12.437 
0.000000 
SPEED*DISTRACT*Age  2.89 
8 
0.36 
2.010 
0.051798 
Error  19.39 
108 
0.18 

DAY*SPEED*DISTRACT  28.15 
16 
1.76 
10.960 
0.000000 
DAY*SPEED*DISTRACT*Age  6.13 
32 
0.19 
1.193 
0.220459 
Error  69.36 
432 
0.16 
There was a significant main effect of age on lane position, with the oldest
drivers being closest to the center of the lane, at an average position of
4.5 ft, the middle age drivers maintaining a position farther to the shoulder
at 4.7 ft, and the youngest staying the farthest to the right at an average
of 5.2 ft. There was no significant main effect of practice on the average
lane position, though there was an interaction between day and speed, and between
day, speed, and age. The last threeway interaction is depicted in figure 10.
It is difficult to discern a consistent pattern in the twoway interaction
(not shown), and it appears that the interaction was due mostly to the difference
in the lane position between the 50 mph speed condition and the other two conditions
on Day 2, and the absence of such a difference (or any significant differences)
on the other days. The threeway interaction demonstrates that the youngest
drivers were clearly different from the other two groups: at all speeds they
drove the farthest away from the centerline, and showed essentially no change
over the five days of practice at the task. The two older driver groups were
similar to each other in their distance away from the center of the lane when
driving in the car following mode, but differed from each other in the fixed
speed mode, with the older drivers driving closest to the center of the road.
Furthermore, there was a slight – but albeit noisy – practice effect
with the two older groups: getting farther away from the centerline with practice
in the 65 mph and carfollowing conditions.
Figure 10. The Effects of Practice Speed and Age on the Drivers’ Average Lane Position
The effect of distraction was only marginally significant, but its interaction with practice was highly significant, as was its interaction with speed. The threeway interaction of day, speed and distraction is depicted in figure 11. The practicerelated pattern is quite noisy, but in general there is a practice effect in the 65 mph and carfollowing condition but not in the easier 50 mph condition.
Figure 11. The Effects of Distraction, Practice, and Speed on the Drivers’ Average Lane Position
[d]
There was also a marginally significant interaction between the driver age, required speed and distraction, which is depicted in figure 12. The consistent effects of age are very large, but no consistent patterns appear with respect to the effects of speed or distraction. Furthermore, at the most demanding speed of 65 mph, there are no statistically significant differences between the nodistraction and the most difficult distraction of math operations, for any of the age groups.
Figure 12. The Effects of Distraction, Age, and Speed on the Drivers’ Average Lane Position
 
 
 
 