The principal objective of this study was to assess the degree to which people can time-share driving with a phone task when they are given time to practice both tasks. In general, the results demonstrated the existence of learning effects on most of the driving measures accompanied by reductions in the subjectively measured task load and task difficulty. Also, the results demonstrated that in the driving situations used – a monotonous road with a few turns and with little traffic - the effect of the distracting task – especially on the most demanding driving condition of maintaining a speed of 65 mph, and on the most vulnerable group of old drivers – diminished over time. Table 8 provides a summary of the principal findings relative to the effects of learning, the effects of the distraction, the interaction between the two main independent measures, and the association between these effects and age.
While the study findings do not point in a clear-cut direction, three relatively consistent findings that are critical to the validity of the study were apparent:
There may be several reasons why the results presented above are difficult to present in the form of a simplistic conclusion. These include:
All of these factors probably contributed to the significant amount of variance in the data, and occasionally to statistically significant effects for which we have no plausible interpretation.
The results are also consistent with most of the previous research on the effects of distraction, since they clearly show a highly significant main effect of distraction on several of the driving measures (average speed, speed variance, and steering wheel deviations) and a marginally significant effect on the others (lane position and lane position variance). More important, in general, the effects of the distraction are most noticeable on Day 1. This is important because the results from Day 1 are the most comparable to the results of the previous studies that only employed one day (or one session) on which the effects were tested.
Table 8. Summary of the Effects of Practice and Distraction from Cell Phone Use on Driving Performance and Other Measures
|Performance Measure||Learning/ Practice Effect||Phone Distraction Effect||Practice X Distraction Interaction||Age- Related Interactions|
|Average Speed||Increases for 65 mph requirement||Significant for all age groups but in an inconsistent manner||Yes – at the math task at 65 mph, esp. for the older drivers||Most difficult and greatest learning effects for older drivers|
|Speed Variance||Decreases, esp. for older drivers and at 65 mph||Possibly but only for math operations, and not for emotionally-involving conversation||Not consistent, but less on Day 5 than Day 1||Most difficult and greatest learning at 65 mph, esp. for older drivers|
|Average Lane Position||In car-following, and 65 mph, older and middle age drivers move closer to shoulder||No significant effect||Significant but inconsistent, since effect of distraction is very small||Oldest drivers closest to center of lane. Youngest closest to shoulder|
|Variance in Lane Position||None||None as main effect||Significant, consistent reduction with math task at 65 mph||Variance increased with age|
|Steering Deviations||Decreased||Greater without distraction than with it||Significant for Math operations, but not for emotionally-involving conversation||Greatest reduction for older drivers, at end all ages are the same|
|RT to Peripheral Signals||RT decreases with practice||None||Significant but not systematic for any phone task||RT longer for older drivers|
|Crashes||Insufficient data||Insufficient data||Insufficient data||Insufficient data|
|Math Operations||Errors decreased, especially for young drivers, and at 65 mph||Not relevant||Not relevant||Older drivers initially better than young drivers and have less room to improve|
|Subjective Workload||Decreased on all measures||Not relevant||Not relevant||No age-based differences|
In addition, the inclusion of the more difficult 65 mph requirement and the inclusion of older drivers (both of which were not present in our first study), yielded the learning effects that we sought for most of the measures. This enabled us to test the main hypothesis that with practice, performance on the driving task should improve, and the difference in driving performance with and without distraction should diminish. Consequently, it was in the more demanding situations and for the initially less able drivers (older drivers), that the greatest learning effects were observed, whereby the effects of the interfering task on the driving task diminished over time.
The results of this study also demonstrated another phenomenon that is consistent
with the active nature of the driving task: that drivers can, and sometimes
do, adjust their behavior to compensate for information overload. In the present
study this was apparent in the driving speed, especially in the high-speed
requirement. In the 50 mph and in the car-following condition, the demands
were not that high, so drivers achieved the desired speed already on the first
session and maintained it throughout the five days. In contrast, in the 65
mph condition, there was a typical logarithmic learning curve, and even on
the last day the drivers still drove below the desired 65 mph.
The nature of the distracting task is very critical to any evaluation of its effects on driving. This is because theoretically at least, the more demanding the cognitive aspect of the subsidiary task, the more it should interfere with the driving task. For this reason in order to evaluate the impact of a phone conversation on the driving task and driving safety, it is very important that the distracting task resemble in its demands the demands placed on the driver by a real phone conversation. Unfortunately many of the studies on the effects of cell phone use on driving used tasks such as “shadowing” (Strayer and Johnston, 2001), math operations (Harbluk, Noy, and Eizenman, 2002; McKnight and McKnight, 1993; Parkes and Hooijmeijer, 2001), memory tasks (Alm and Nilsson, 1995), and reasoning operations (Haigney, Taylor, and Westerman, 2000; Parkes and Hooijmeijer, 2001; Strayer and Johnston, 2001) to simulate the phone conversation. Whether or not these tasks are more demanding, less demanding, equally demanding, or task cognitive functions that are different from most phone conversations is often unknown.
In the few studies where simulated conversations were compared to other tasks, the conversation was either significantly less distracting than mathematical operations (e.g. McKnight and McKnight, 1993), or not measurably distracting at all (Briem and Hedman, 1995). One difficulty of simulating a true phone conversation is that it cannot be standardized across subjects and even kept constant in its cognitive demands across time within subjects. This is because every conversation is unique in its effects on the different participants, and drivers always have the option of pacing their responses to accommodate the driving requirements. Briem and Hedman (1995) who failed to find driving impairment due to conversation on the phone concluded that “the use of a hands-free telephone does not in itself increase the risk of an incident… while simple manipulation-free conversation may, actually, contribute to a decrease in the deviation from the correct path.” The problem is further exacerbated when distraction tasks are labeled as “conversations” even though they are not. For example, McKnight and McKnight (1993) called their math operations task “intense conversation” as distinct from a “casual conversation” that consisted of asking the drivers what they “did for a living, what they did with their free time, etc.” Then when math operations were determined to be significantly more distracting than the “casual conversation”, the authors concluded that “intense conversations provided the greatest overall degree of performance decrement,” when in fact no intense conversations were conducted at all. Similarly, Strayer and Johnston’s (2001) “conversing” task was a complex shadowing task where the participants heard words, and in response to each word they had to generate a different word beginning with the last letter of the word they just heard. Yet, the title of Strayer and Johnston’s paper referred to “conversing on a cellular phone.” Further distortion of the findings was made when this terminology was adopted verbatim by the press.
As with the phone conversation, so often the driving task used in experimental studies is at best a very partial simulation of actual driving. For example, Irwin, Fitzgerald, and Berg (2000), measured foot reaction time to a light in the presence of different phone tasks – presumably conversations with varying amounts of cognitive or emotional stress – and titled their study “Effect of the intensity of wireless telephone conversations on reaction time in a braking response.” In fact, their study had no elements of driving at all (not even braking) other than a reaction time task. Strayer, and Johnston (2001), used a pursuit tracking task and titled their study as “Driven to Distraction: Dual-Task Studies of Simulated Driving and Conversing on a Cellular Phone.” Interestingly, in the first of two studies that they conducted they complicated their “driving” task by having the driver also respond whenever the target “flashed” red or green. However in the second study, in which the detrimental effect of the “conversation” was noted, they omitted the color change because it added “substantial noise in the tracking data.” Thus when the option to add some of the complexities of driving was considered, it was rejected in favor of using a simpler task.
The design of the present study was based on a relatively high-fidelity driving simulator (though still fixed-base with a single 40o field of view screen), and compared the effects of distraction from math operations with the effects of distraction from a presumably demanding conversation that challenged the drivers’ belief systems and self perceptions. As in the previous studies quoted above, the conversation – even though it was emotionally engaging – turned out to be much less demanding than the math operations. Perhaps one reason for the difference is that the conversation – at least in part – can be paced by the driver while the math operations (as most of the other distracting tasks used in other studies) are experimenter-paced and thus can be much more demanding vis-à-vis the time-sharing requirements.
Finally, it is of interest to note that the car-following task was less affected
by the phone task, than the requirement to keep a constant and high speed.
It implies that as long as the driver is not required to keep a fixed distance-headway
or time-headway, simply driving behind another car that changes its speed from
time to time can be safely accomplished while performing a secondary task such
as a phone conversation. However, in our study, the lead car never stopped
abruptly and so the driver’s readiness to respond to a sudden change
was never tested.