A Review of Human Factors Studies on Cellular Telephone Use While Driving

 

Chapter 5 Table of Contents

5.1 Introduction

5.2 Research Methodologies

5.3 Simulator and Closed Course Studies of Cellular Telephone Use

5.4 On-Road Studies of Cellular Telephone Use While Driving

5.5 Epidemiology Studies

5.6 Summary and Conclusions

 

5.1 Introduction   A number of research investigations have examined driving behavior and performance during cellular telephone use. This chapter provides a review of those studies that have been published and are available in English or were accessible for translation. It organizes them chronologically according to the experimental method used: simulator and test track studies on the one hand, and on-the-road studies on the other hand. In addition, epidemiological and observational investigations are also presented and discussed. The summaries of each study highlight key aspects and findings of the research. Appendix C contains additional details and critiques for each of the studies mentioned here. See Parkes (1993) and Petica (1993) for other literature reviews on the subject. Before proceeding with this review, some additional comment is in order. Cellular telephone use while driving can be characterized by the tasks that make up such use. These tasks include the following: Accessing the Cellular Telephone This may involve removing a handset from a "pod" installed in the vehicle, reaching into a pocket or briefcase to retrieve the phone, or initiating a "hands-free" connection (e.g., answering a call, placing a call). Dialing This may include accessing a directory or stored number and keying in one or more digits. Voice Communications Usually relates to dialogue, listening, and talking. Associated Tasks Additional actions the user might carry out in association with Cellular Telephone use (e.g., taking notes, referencing a calendar or a map). The first task has typically been considered as trivial, especially if the location of the handset is well learned or the unit is of the hands-free variety. However, in a recent Japanese study (see Chapter 3), it was found that 42% of cellular telephone related crashes occurred in response to a call, and involved being startled or distracted by the ring, dropping the phone, or turning to pick up the phone. While the relevance of these data to crashes in the United States is unclear, the results suggest that receiving a call may have more significance than is readily apparent. The dialing and voice communications tasks have not typically been considered trivial among researchers and these have been the focus of most of the published research in this area.

An Investigation of the Safety Implications of Wireless Communications in Vehicles

Chapter 5: A Review of Human Factors Studies on Cellular Telephone Use While Driving

The benefits of on-road studies are that they provide the driver with real driving task demands and priorities. However, they are far from perfect (Smiley, 1995). There are limits as to what can be done experimentally while on the road. The test participants are usually screened to have good driving records. They are usually driving under conditions where there is no real sense of urgency to get from one place to another as quickly as possible. Perhaps most importantly, the test participants are keenly aware that they are driving an unfamiliar vehicle, with a stranger in the passenger seat, and everything that is going on is being recorded. Would drivers likely behave differently if alone, in their own vehicles, running late, without their behavior and performance being captured? Probably yes. Would they likely behave in a riskier fashion? Probably yes. The real question is how much riskier would drivers act with regard to cellular telephone use. To capture such data may require "black box" technology installed in a volunteer's own vehicle to randomly sample behavior over a long period of time. However, none of the studies to be described here made use of such a method. Thus, on-the-road studies represent the most realistic, though still imperfect means of studying cellular telephone use while driving.

An Investigation of the Safety Implications of Wireless Communications in Vehicles

Chapter 5: A Review of Human Factors Studies on Cellular Telephone Use While Driving

5.3 Simulator and Closed Course Studies of Cellular Telephone Use   The earliest published study on mobile telephone use and its impact on drivers was that of Brown, Tickner, and Simmonds (1969). They point out that mobile phone use may involve two sources of interference. The first source is the manual-visual demand of dialing. The second source is the attentional demand of the communications task. Brown et al. (1969) focussed on the latter only by simulating a hands-free phone application. A sample of 24 male subjects drove a car on a 1.5-mile closed course without traffic to collect measures on judgments of gap size (possible vs. impossible to clear), number of gaps actually cleared successfully, total course travel time (interpreted as speed), and control inputs (steering and foot controls and associated lateral and longitudinal accelerations). The telephone communications task was a paced grammatical reasoning task in which the driver heard a short sentence followed by the letters "A" and "B" where each sentence claimed to describe the order of the letter pair that followed. The driver decided whether the sentence was true or false and responded accordingly. Examples are provided below:
Driver (Correct) Incoming Phone Message: Response   "A follows B ... BA" "True" "B precedes A ... AB" "False"
Results indicated that gap judgments were significantly degraded during the communications task and travel speed was reduced. Additionally, concurrent driving was associated with longer decision times for the grammatical reasoning task and more errors relative to performing the grammatical reasoning task alone while the car was parked. Unfortunately, travel speed was a global measure based on circuit completion time. Thus, it is impossible to tell if drivers drove more slowly throughout the telephoning task or took the additional time driving around incorrectly judged "impossible" gaps. The "intelligence test" nature of the dialogue materials were highly demanding, probably more so than normal cellular telephone conversations. Finally, there were no other vehicles on the closed course and no serious consequences to making a gap judgment error, factors that might have prompted the test participants to work harder at answering the logic questions than on the driving task. This study demonstrated that driver judgments about gaps could be disrupted by concurrent dialogue of a demanding nature. Kames (1978) also made use of an instrumented vehicle on a closed course to examine the effects of three types of dials (rotary dial, push-button dial, and push button dial-in-handset) on driving performance and behavior while concurrently dialing. Eighteen (18) test participants drove a 4.4 mile course on a deserted airfield and each worked with six different versions of dials over six different sessions. At predetermined locations, an experimenter asked the test participants to dial a number. Measures taken concurrently included lane position, range of speed, reaction time to a subsidiary task, steering wheel movement rate, range and duration of head movements, and dialing completion times. In general, results indicated that rotary dials were the slowest with which to work but that other varieties of dial designs and locations (dash-mount, visor area-mount) had relatively minor impacts on driver lane position variability and apparently no significant effects on other measures of driving performance. Drivers nonetheless reported being uncomfortable about dialing while driving.

Zwahlen, Adams, and Schwartz (1988) conducted two experiments in automobiles on a closed course (an unused airport runway) to investigate the impact of phone dialing on driving. Drivers used a standard (i.e., not a cellular) push-button phone (to simulate a cellular telephone) and manually entered an 11-digit number. The experiment varied phone location (high vs. low on the dash), and whether the driver could look at the road while dialing (allowed vs. not allowed). Also drivers did not correct dialing errors.   An assessment of lane keeping for straight road driving indicated that manual dialing increased lane standard deviations to potentially dangerous levels. Not being able to look at the roadway ahead while dialing was most disruptive; low-mounted phone position was also disruptive to lane keeping, though much less so than the look/no-look manipulation. When averaged across two different vehicles (a compact passenger car and a station wagon), the standard deviation for lane position was 15.4 inches for a 675 foot run. Zwahlen et al. predict that for a 12 foot wide lane this would lead to lane exceedences under ideal conditions (e.g., daylight, dry, straightaway) 11.9 percent of the time at 40 mph.   The manipulation of looking behavior was an attempt to emulate "worst case" driving behavior. This, plus the absence of other traffic or serious consequences for poor lanekeeping may account for the magnitude of effects. Somewhat reassuringly, when the telephone was mounted in the low position, drivers who were permitted to look at the road while dialing did so on 47 out of 50 runs. This compares to looking on only 37 out of 50 runs when the telephone was mounted in the high position, presumably because this position afforded some monitoring of the path via peripheral vision. This suggests that drivers can be sensitive to at least some of the performance- degrading features of telephones in vehicles and attempt to compensate for the degradation

.

Boase, Hannigan, and Porter (1988) investigated the effects of talking while using a hands-free phone (again a simulated device) while the driver was engaged in a laboratory computer game of "squash" that the authors claimed (without rigorous justification) to involve some of the same task demand characteristics as driving. Without a rigorous means to tie the computer game to driving demands, the results are judged not directly applicable to driving and so will not be discussed here. The main telephoning variable was type of dialogue. Informational dialogue (ID) was mostly simple question-and-answer dialogues like making appointments, checking dates, and so on; it was mimicked by asking subjects simple questions like their favorite foods or educational experiences. The negotiation dialogue (ND), uncovered in focus groups with businessmen, involved negotiation and deal-making; this was simulated by having subjects engage in a dialogue such as to return faulty merchandise to a store or to modify an airline itinerary altered by the air carrier. This represents an interesting attempt to use dialogues that might arise naturally in mobile or cellular telephone use. Alm and Nilsson (1990) conducted a motion-base driving simulator study of the effects of hands-free mobile phone conversation on driver performance as measured by reaction time, lane position, speed level, and Task Load Index (TLX) subjective workload assessments for easy (80 km two-lane rather straight road) and hard (80 km two-lane very curvy road) driving tasks. The "conversation" was actually the Baddeley Working Memory Span Test. A number of 3-to-5-word sentences were presented over the phone of the form "X does Y" and the subject had to answer "YES" if it was reasonable or "NO" if not. For example, one sentence might be, "The train bought a newspaper", to which the correct response is "NO"; another sentence might be, "The boy brushed his teeth" to which the correct response is "YES". After 5 sentences, the subject was to recall the last word in each sentence. This telephone task was paced. Results for 40 test participants were complex but generally indicated that this telephone conversation increased driver brake reaction time and resulted in a reduction in travel speed when the driving task was easy (i.e., mostly straight road segments). It degraded lane position and this was most pronounced when the driving task was hard (i.e., mostly curvy road segments). Finally, subjective workload was always rated higher with telephone conversation. The nature of the conversational materials is such that they are likely to be more cognitively demanding than normal cellular telephone conversations. The increase in brake reaction time to a visual stimulus presented in the simulator scene during easy routes but no increase during hard (curvy) routes implies that test participants were somewhat sensitive to the primary driving task demands and attempted to manage their attention to the communications task accordingly. The speed reduction found may represent an attempt by the test participants to reduce the primary driving task demand by slowing things down. However, going substantially slower than the prevailing travel speed is also associated with traffic mishaps, as noted earlier. Thus, this speed reduction, rather than being a safety positive outcome, may actually represent an increase in crash hazard exposure.

Nilsson and Alm (1991) extended the previous study to an older test participant sample by carrying out the same experiment for "easy" road segments, only this time with 20 test participants (equal numbers of males and females) between the ages of 60 and 71 years. The analysis used the previous study data set in combination with the newly acquired data to assess the impact of the voice communications task on elderly test participants. Results indicated that, compared to younger drivers, the elderly drivers had longer average brake reaction times, showed greater lanekeeping variability during the conversation task, and drove faster than younger drivers while using the telephone. This last finding, the authors note, may be an artifact of the limited perceptual information about speed provided by the driving simulator. Nilsson (1993) (see also Alm and Nilsson, 1995) then extended the driving simulator investigations to a car following situation. Forty (40) test participants, 20 participants below 60 years of age and 20 participants aged 60 years or older, performed the voice communications task described earlier while simultaneously driving the simulator so as to catch up to and then follow another car. When the lead vehicle brake lights came on, the test participant was to apply the brakes as quickly as possible. When the lead vehicle's right turn signal was turned on, the test participant was to turn on the simulator left turn signal. The speed level was constrained in the simulator so that it was not possible to overtake the lead vehicle. When compared to driving without the voice communications task, drivers had longer brake reaction times, and headway distance decreased. Older drivers had longer brake reaction times than younger drivers, all else being equal, and older drivers on average allowed greater following distances than did younger drivers.

Younger drivers on average approached the lead vehicle at an 11.5 km/hr higher closing rate while engaged in the voice communications task than did older drivers. No effects on lanekeeping performance were found. All test participants reported greater subjective workload during the voice communications task than without the voice communications task. Alm and Nilsson (1995) point out that the failure to increase headway to accommodate increased brake reaction times might be interpreted to mean that test participants were unaware of the impact cellular telephone tasks were having on their ability to react quickly. On the other hand it might also be interpreted to mean that the test participants already believed they had sufficient headway during the telephone task to compensate for any decreased reaction time. Again, because of the extreme nature of "intelligence test" voice communications materials, the relevance of this study to normal cellular telephone communications is unclear. The driving simulator environment, with its lack of real-world consequences, might also have had subtle effects on test participants so, relative to real world driving, they paid more attention to the intelligence test questions than to the driving task. Finally, because of the limited duration of the testing, it remains to be seen if drivers could learn to modify their driving over time to adapt to cellular telephone use or if a non-paced voice communications task might lead to altogether different outcomes.   McKnight and McKnight (1993) (see also McKnight and McKnight, 1991) conducted a study of cellular telephone use on driver attention using an open-loop simulator that consisted of videotaped road scenes. Because of the simulator's nature, driver inputs had no effect on what was presented on the simulator screen. The simulator testing included 47 situations to which a driver might ordinarily respond. These situations included vehicles stopping, turning, and entering a motor way; road changes such as lane drops, narrow bridge, and so forth; pedestrians or animals entering the travel lane; route changes; traffic control signals like stop signs and light changes; etc. As the subject "drove" along the video scene, accelerator pedal use, brake onset, and steering and turn signal use were recorded. The dialing component of the telephone task was to place a call to the subject's home number manually. The conversation component of the telephoning task involved both simple conversations (e.g., gathering demographic information, chit-chat on what the subject did the previous weekend), and complex conversations (i.e., math problems of the form 2+ 3 + 4 + 1/2 + 3 + 4 + 6 = ? or short term memory problems that required the subject to listen to a list of 5 or 6 digits and then answer whether certain digits were in that list).

Results indicated that complex phone conversations led to the greatest increases in missed events and time to respond, followed by a radio tuning task, with simple phone conversation having the least effect. Placing a call (e.g., dialing home) did not degrade performance any more than simple conversation but delayed responses about the same as complex conversations. The relative increase in chances of a highway traffic situation going unnoticed ranged from about 20 percent for placing a call to 29 percent for complex conversations.

Older drivers (i.e., 50 - 80 yrs) were most adversely affected by telephoning in their ability to detect driving situations. However, time to respond was only affected by age when placing calls. This study is laudable for being the largest conducted up to that time (150 test participants of both genders and spanning the ages 17 to 80 years). The dialing task of dialing one's own number (probably highly over learned) is likely to beeasier than dialing a less familiar key sequence. On the other hand, the paced "intelligence test" nature of the intense conversational materials may be so difficult as to be unrepresentative and generally irrelevant to the interference potential of most calls placed or received while driving. Furthermore, the high percentages of missed detections would, if encountered in the real world, lead to a veritable epidemic of crashes, yet no such epidemic has yet been reported through either formal or informal means. This suggests caution in interpreting the results as absolutely applying to real world driving. Serafin, Wen, Paelke, and Green (1993) have reported on a car simulator study of mobile phone usability in terms of features such as manual vs. voice dialing, instrument panel vs. head-up displays (HUD), length of phone number (7 vs. 11 digits) and number familiarity (unfamiliar vs. previously memorized). The communications tasks included loose ends (how many unconnected ends are there in a capital letter), listing (name as many items in a category as possible in a fixed time period, e.g., "a type of furniture") , talking (answer the question, "What did you do last weekend?"), and listening (i.e., listen to a hypothetical situation and answer multiple choice questions about it). Each task lasted about 30 seconds and all test participants were given the same materials. The driving simulator run simulated a night drive on a single lane, slightly curvy road. The dependent driving measure was standard deviation of lane position. In terms of the driving performance of 12 test participants, lane standard deviation was greater with manual dialing, voice input resulted in less lane position deviation for all drivers, and dialing times were faster for older drivers dialing unfamiliar numbers. The effects of the communications tasks on lane variation were not reported, presumably because none were found. In addition to the manual dialing-induced lane keeping disruption, this study also found that age influenced both driving performance and dialing times. Pachiaudi and Chapon (1994) reported results of research into cellular telephone use on a car simulator conducted in France. The nature of the car phone communications task was not described but the simulator scenario was a simple route for which drivers were to try to maintain constant speed (either 90 or 130 kph). Of the 17 subjects in the study, only two showed no change in travel speed while telephoning. For nine subjects, telephoning caused them to modify their travel speed (slow down) while for the other subjects, speed control was momentarily lost and this led to increased speed variability or increase in speed without attempts to correct for this. Clearly, such disruption with vehicle control would be of concern on the highway. Less clear is the extent to which drivers would allow such disruptions to occur on the highway.

An Investigation of the Safety Implications of Wireless Communications in Vehicles

Chapter 5: A Review of Human Factors Studies on Cellular Telephone Use While Driving

5.4 On-Road Studies of Cellular Telephone Use While Driving   A number of on-road studies have been conducted that bear upon questions of the effects of cellular telephone use while driving. Hayes, Kurokawa, and Wierwille (1989) studied driving performance while engaged in various instrument panel tasks (including entering a 7-digit number into a telephone keypad) by means of an instrumented vehicle driven on public roads. Twenty-four test participants (12 males and 12 females grouped into three groups by ages: 18-25, 26-48, and 49-72) each drove four 15-minute runs along a preselected route. During each run, the test participant completed various in-vehicle tasks. Of particular interest was the manual dialing task. Results indicated that the dialing task took less time to complete than a radio tuning task and demanded fewer glances, regardless of age. This suggests that 7-digit input, using a keypad and mounting location like that reported, is no more time consuming or visually demanding than radio tuning. No vehicle performance measures were reported.

Brookhuis, De Vries, and De Waard (1991) examined the impact of telephone use on driver performance in an instrumented passenger car on the road measured every work day for 3 weeks. Twelve (12) subjects drove while and while not operating a mobile telephone under three driving conditions: light traffic, heavy traffic on an outer belt following a confederate lead vehicle, and driving in city traffic. The dialing component was either manual or hands-free, and the telephone conversation consisted of a 3-minute test, presented over the phone, of a paced serial addition task that was a fairly hard combination of a memory test and an addition test. Results indicated a variety of interesting patterns. There was a decrease in lane standard deviation while in telephone conversation, particularly while driving on a quiet motor way. Steering wheel amplitudes were substantially higher with manual dialing. There was a decrease in the number of rearview mirror checks during conversation, but only on the quiet motor way. There was also a statistically significant increase in brake reaction time while telephoning of about 600 milliseconds (ms) to adapt to a slowing lead vehicle and a non-significant increase of about 130 ms to brake for a stopping lead vehicle (i.e., lead vehicle brake lights suddenly come on). However, drivers did not decrease their average travel speed (termed speed coherence in the paper), and this compounded with the greater reaction times could lead to an increase in rear-end collision hazards. An alternative interpretation, however, is that drivers believed they nonetheless had sufficient headway distances to lead vehicles so as to make slowing down unnecessary. In general, the results of this study show that cognitively intensive cellular telephone communications tasks undertaken while driving may increase driver reaction time to objects and events, and may decrease situational awareness such as that updated by means of mirror sampling. Drivers appeared in this study to modulate their performance in instances where they thought it safe to do so (e.g., decreased mirror sampling while driving, but only on quiet motor ways, where presumably such a decrease would be more acceptable). Fairclough, Ashby, Ross, and Parkes (1991) compared cellular telephone use with speaking to a passenger while the driver drove on an open roadway. Driving behavior was measured in terms of route completion time, eye movement behavior, heart rate, and subjective workload assessments. Both speaking conditions resulted in higher subjective workload assessments and longer route completion times as a consequence of reducing travel speed about 10 percent in the mixed urban and rural route. No differences in driver visual allocations were noted. Heart rate was significantly higher while using a cellular telephone than while speaking to a passenger directly or driving with no speaking. The nature of the communications task was one in which drivers negotiated a predetermined topic, e.g., booking a summer holiday or negotiating a partial exchange deal of a car until they reached a conclusion that was satisfying to them. This study is interesting in that it suggests that cellular telephone conversations and conversations with a passenger need not be substantially different in terms of their effects on the driver. The speed reduction was relatively small and may have no substantial safety implications. It is not clear that similar results would necessarily arise given emotionally laden conversations, e.g., negotiations with a belligerent caller and potential losses at stake.

Parkes (1991) used a set of intelligence tests for conversations conducted over a cellular telephone or with a passenger concurrent with driving. The route involved a mixture of suburban and rural English roads. Results showed that test participants scored significantly lower on the intelligence test items when using the cellular telephone than in the other experimental conditions. Part of this apparent discrepancy between this study outcome and that reported by Fairclough et al. (1991) is that videotape analysis revealed the experimenter-as-passenger naturally made allowances for traffic movements when administering the test. Such allowances could not be made in the cellular telephone call to a remote office. One suggestion made by Parkes is that an "intelligent answer phone" be developed that can divert, record, and interrupt messages appropriately based on driving circumstances. Parkes (1993) summarized an on-the-road study in "low complexity driving" using a three-lane motor way with moderate traffic flow. Subjects drove two 20- minute journeys, one in silence, and one involving four conversations. These communications involved mental arithmetic and memory tasks. Driving behavior was measured in terms of accelerator depression, steering wheel reversals, and travel speed. Subjective workload assessments were collected by means of the Modified Cooper-Harper Scale and the TLX. Finally, global observations by a ride-along experimenter included recordings of the number of lane changes, overtaking behavior, and proportion of time spent in each of the three lanes. Results indicated no evidence of change in driving behavior during phone conversations. Speed choices, lane occupancy, and accelerator depressions were consistent across all experimental conditions. Subjective workload assessments did reveal an increase in perceived workload.

The duration of conversations was limited to 2 minutes only, however, and Parkes notes that longer calls might involve a greater demand on driver resources. Furthermore, the driver input and speed measures may simply have not been sensitive to any disruptive effects of the conversations, though perhaps lane keeping measures would have been (had they been taken). Green, Hoekstra, and Williams (1993) (see also Serafin, Wen, Paelke, and Green, 1993) conducted a small scale on-road study of cellular telephone use and route guidance system use while driving. Eight (8) test participants drove a 1991 Honda Accord station wagon over a 19-turn 35-minute route that included sections of residential neighborhoods, city streets, and expressways. During the trip, the driver was asked to dial a familiar telephone number and to participate in a simulated cellular telephone conversation. The simulated conversations included: a) a 30-second description (e.g., three options for dining out) to which they had to respond with one choice, b) talking (e.g., describe what they did last weekend), and c) a listing task (e.g., list all the fruits you can in the time allotted). Differences were found in the standard deviation of steering wheel angle (dialing was more difficult than conversations, which were not significantly different from each other but greater than with driving alone). Differences were also found in standard deviation of throttle position, with throttle position varying most during the talking task and least in dialing. The safety implications of such effects are unclear. No other significant effects were found. This may be an artifact of the small sample size used in the study. Alternatively, it may signify that drivers appropriately prioritized the driving task and generally maintained adequate control over the vehicle. Tijerina, Kiger, Rockwell, and Tornow (1995a) (see also Tijerina, Kiger, Rockwell, and Tornow, 1995b) carried out an on-road study of heavy vehicle driver workload while engaged in both dialing and voice communications tasks. Sixteen (16) male professional truck drivers between 32 and 60 years of age, accompanied by a ride-along experimenter, drove a conventional tractor with a 53-foot single trailer loaded to a gross weight of approximately 76,000 pounds on non-revenue runs. The routes taken included undivided and divided rural and urban highways, conditions of light (non-rush hour) and heavy (rush hour) traffic density, and both daytime and night/dusk driving conditions. Driver visual allocation (number of glances, duration of glance and time between glances to various locations), lanekeeping, speed maintenance, and driver steering and pedal inputs were measured. Dialing included 7-digit, 10-digit, and auto-dial (RCL 1) dialing; radio tuning was included as a baseline manual task. Two types of voice communications materials were also included. "Easy" communications included paced, question-and-answer dialogue where questions related to driver demographics (e.g., "What is your date of birth?"). "Moderate" dialogues consisted of paced question-and-answer dialogue consisting of questions that required some mental arithmetic (e.g., "How much longer can you drive today before you reach your hours of service limit?"). Dialogues were paced to last approximately 1 minute. Results indicated that 7-digit and 10-digit manual dialing took even more glances on average than radio tuning, and took significantly greater total glance time away from the road than did radio tuning. Steering holds and accelerator holds showed patterns indicative of greater driver inattention during 7-digit and 10-digit manual dialing and radio tuning than that associated with the auto-dial task.

Speed variability was not practically significant and lane variability did not differ substantially among the dialing and radio tuning tasks. However, lane exceedences were found in 27 percent of all manual task trials, compared to only 14 percent of trials where the driver only read a 1-line message from a CRT display. This suggests that cellular telephone dialing and manual radio tuning, though not significantly different, can be disruptive of lanekeeping performance, even among professional truck drivers. Note that the greater girth of the semi-tractor-trailer also makes lane exceedences more likely than in a passenger car. Results obtained during the voice communications tasks indicated that there was no concurrent degradation in lanekeeping or speed maintenance measures during the conversations. However, there was a reduction in mirror sampling during dialogues (approximately 6 percent of time spent mirror sampling, regardless of dialogue type) when compared to driving without dialogues (approximately 12 percent of time spent mirror sampling). This suggests that even a non-visual task like dialogue can affect driver situational awareness such as that maintained by mirror sampling.

Kantowitz, Hanowski, and Tijerina (1996) (see also Hanowski, Kantowitz, and Tijerina, 1995) presented the results of a part-task simulator study that used the same tasks and conversational materials as the Tijerina et al. (1995a) study. Fourteen (14) male commercial truck drivers each drove eight simulator sessions on an STI fixed-base driving simulator mounted in a heavy truck test buck and with equations of motion for the visual scene appropriate to the heavy vehicle. Con current with driving, drivers completed the same types of dialing and voice communications tasks as were just described. There were many differences between the on-road and simulator study results. With regard to manual tasks, for example, mean lane position was closest to lane center for the autodial task and farthest for the radio tuning task in the simulator. Yet on-road data indicated no reliable differences in effects of manual task on mean lane position, lane variability or lane exceedences. In the simulator study, lane deviations during the manual tasks were smaller in high traffic density than in low, yet no such effects arose on the road. Similarly, during the voice communications tasks, no degradation of lanekeeping or speed maintenance were found on the road. In the simulator, however, this was not wholly the case. For example, the simulator study found lane exceedences were greater during dialogues of either type than when driving only, yet no significant differences were found between dialogues and baseline driving on the road. Despite the similarity of tasks, materials, and procedures, numerous differences existed between the simulator study and the on-road study. These differences range from differences in the heavy vehicle cab layout versus that of the test buck, to very different road scene characteristics. This pattern of differences suggests that heavy vehicle drivers in the simulator adopted a more lax attitude toward the driving task, (lanekeeping) perhaps because there is no safety risk associated with degraded lanekeeping. On the road, the drivers maintained more or less consistent lanekeeping and speed control throughout all phases of the testing, thus providing evidence that they accorded appropriate priorities to the driving task and the cellular telephone tasks. Briem and Hedman (1995) studied hands-free mobile telephone use by 20 test participants (half of them male and half female) while they engaged in a pursuit tracking task in a laboratory setting. Test participants "drove" for 20 minutes while engaged in three secondary tasks. A "radio" manipulation task required turning on, tuning and listening, and turning off a car radio. "Easy" telephone conversations were 2 minute dialogues about current events. "Difficult" telephone conversations were 2-minute working memory tests similar to those used by Alm and Nilsson (1990). Velocity and acceleration dynamics were applied to the tracking task to simulate firm road and slippery road conditions which test participants tracked half of the time. Obstacle avoidance was also included in the tracking task. Measured responses were road position (measured as pixels off of the curved line), small and large position deviations, speed, steering wheel movements, and collisions. The pattern of results is complicated but the authors summarize the key findings as follows. On the slippery road condition, radio manipulation led to the greatest deterioration of tracking performance. Male drivers exhibited better control while driving under difficult conditions. The authors concluded that simple conversation with a hands-free phone does not impair performance but that difficult conversations may, particularly under conditions that put heavy demands on the driver's attention and skill. While the results may be intuitively appealing, the limited fidelity of the simulated driving task and the artificial nature of the difficult conversation suggest a need for validation of the findings under more realistic conditions.

An Investigation of the Safety Implications of Wireless Communications in Vehicles

Chapter 5: A Review of Human Factors Studies on Cellular Telephone Use While Driving

5.5 Epidemiological Studies   The advantage of an epidemiological approach to research is that there are fewer experimental artifacts to contaminate the data. Drivers, for example, are assumed to be driving under normal circumstances and are not aware they will be part of a study until after the study period is over.

The main disadvantage of epidemiological studies is that such studies cannot establish causal relationships, because many factors that may influence the results are not controlled or even measured. They nevertheless can uncover interesting associations that merit further experimental investigation. Two recent epidemiological studies of cellular phone use and traffic safety are reviewed below. Violanti & Marshall (1996) used a case-control design and logistic regression techniques to examine the association of cellular telephone use in motor vehicles and traffic crash risk. The amount of time per month spent talking on the cellular telephone along with eighteen other driver inattention factors were examined. Data was obtained from 100 randomly selected drivers involved in crashes within the last two years and compared to that of a control group of 100 randomly selected drivers not involved in crashes within the last ten years. Groups were matched on the basis of geographic residence. To assess risk, data concerning the frequency of attention diverting driver behaviors and other factors that might influence the association between cellular telephone use and crashes were collected using a mail survey. The authors reported that talking more than 50 minutes per month on the cellular telephone while driving was associated with a 5.59-fold increase in crash risk. Those users involved in crashes tended to be younger, with less driving experience and more previous crashes than non-crash involved subjects. Crash involved subjects also talked longer and appeared to be engaged in more intense business calls than non-crash involved drivers. The authors conclude that talking on the cellular telephone was associated with an increased risk of a crash. Violanti and Marshall (1996) discuss a number of limitations to their research. In their subject sample, there were only 7 cellular phone users among the 60 study participants who had a reportable crash and 7 cellular phone users among the control group of 77 study participants who had not had a crash in at least 10 years. While the researchers report that the power of their statistical testing to reject a false null hypothesis of "no difference" between the case and control group was over 90 percent, they acknowledge that a larger sample of cellular telephone users are needed for validation of their findings. Response bias is mentioned but discounted by virtue of the moderate response rates in both groups (from initial samples of 100 persons in each group) and the finding that non-responders were demographically similar to the responders. More problematic was the lack of direct evidence that the persons were using a cellular telephone at the time of the crash. The researchers did not ask about this directly for fear of an inappropriate or misleading response. Finally, while they did include 18 reported driver in attention behaviors (e.g., talking with others, lighting cigarette/cigars, drinking beverages while driving), Violanti and Marshall indicate that many other factors that influence driver attention have gone unmeasured. The researchers emphasize that their findings suggest a statistical association and not a causal relationship between cellular telephone use and crashes. In a later study (to be published - see Appendix E, Violanti, 1997), Violanti analyzes the Oklahoma crash data for the period between 1992 and 1995. For a discussion of the Oklahoma data set see Chapter 3, Section 3.4. The most recent epidemiological study on the relationship between cellular telephone use and traffic safety is that of Redelmeier and Tibshirani (1997). Because of its unique approach and the unusually high degree of media attention paid to this study, it will be examined in some detail. The editorial prepared by Maclure and Mittleman (1997) on this study will also be referenced. Redelmeier and Tibshirani studied 699 Toronto drivers who had cellular telephones and who were involved in motor vehicle collisions resulting in substantial property damage but no personal injury. Each person's cellular telephone calls on the day of the collision and during the previous week were analyzed through detailed billing records. The time of each collision was estimated from each subject's statement, police records, and telephone listings made to emergency services. These collision times were classified as "exact" if information from all three sources was available and consistent or when one source was missing but the remaining two sources were available and consistent. Otherwise, a given collision time was classified as "inexact" and the earliest time given by the different information sources was used.

Of the 699 cases analyzed, 231 (33% of the sample) were judged exact and 468 (67% of the sample) were judged inexact. The authors reported that the risk of a collision was estimated to be between 3.0 and 6.5 times as high within 10 minutes after a cellular telephone call began as when the telephone was not used. Maclure and Mittleman (1997) carried out additional analyses on the same data and confirmed that the risk more than doubled within five minutes after the start of a call. Three additional findings in the Redelmeier and Tibshirani study were: cellular telephone units that allowed hands-free operation offered no safety advantage, thirty-nine percent of the drivers called emergency services after the collisions, suggesting that a cellular telephone may have advantages in collision notification, and the relative risk of having a crash while using a cellular telephone was estimated to be similar to the hazard associated with driving with a blood alcohol level "at the legal limit."   Now consider some of the details. Redelmeier and Tibshirani themselves point out several limitations to their study. They note that causality cannot be inferred from such a study. By way of example, they mention that emotional stress might lead to both increased cellular telephone use and decreased driving ability, so that individual calls may have nothing to do with increased crash risk. They also list four weaknesses in their study.   First, only volunteer drivers participated, perhaps leading to underestimates of risk caused by riskier drivers opting out. Second, they point out that people vary in their driving behavior from day to day, though Redelmeier and Tibshirani consider the findings hard to explain in terms of such variations because of consistent findings between the whole sample and a subset of 72 subjects who remembered (up to a year later) having driven during both the hazard period and the control period. Third, case-crossover analysis does not eliminate all forms of confounding, particularly in regard to temporary conditions, though again, the article's authors believe such factors are unlikely to account for the magnitude of association observed. Finally, they point out that collision involvement did not mean the cellular telephone owner was judged "at fault". This was left unspecified in the article and the authors indicate that perhaps cellular telephone use merely decreases a driver's ability to avoid a collision caused by someone else.   Maclure and Mittleman (1997) point out additional limitations to the study and qualifications to its results. While they applaud the use of the case-crossover design (Maclure was the originator of this approach), they indicate that the use of pilot study subject data to adjust for the "intermittency of driving" (i.e., to account for the fact that some drivers didn't even drive during the control period) was not convincing because of possible unmeasured differences between the pilot subjects and the study subjects.

They have more faith in the analysis of the 72 people who recollected driving during both periods, though they acknowledge that a relative risk result from this group may be an overestimate due to incomplete participation and faulty memory. Maclure and Mittleman indicate that the lack of a safety advantage for hands-free phones may simply be the result of too little statistical power to test for this effect. The risks associated with placing a call, the risk extinction curve over time after a call ends, and the kinds of collisions that are most likely to increase are all in need of future research, they point out. To these caveats and critiques, the present authors add the following. While Redelmeier and Tibshirani distinguish between exact and inexact collision time estimates, no separate analysis of the 231 exact cases is reported. The distinction between exact and inexact, once made in the report, is analyzed in only one instance, reported in a single phrase without comment on p. 455. Determining the exact time of a collision is difficult. Contamination across sources (e.g., driver statement is also used in a police record to indicate crash time) may have occurred. The analysis of the 72 people who remembered up to a year or so later that they were driving in both periods is susceptible to memory errors. By any reckoning, the time of collision is subject to numerous sources of error. Average call length (based on calls placed the week before the collision by this group of subjects) was 2.3 minutes, with 76% lasting 2.0 minutes or less. This suggests a positively skewed distribution with a long right tail, a distribution of mostly short (i.e., less than 2 minute) calls with some calls lasting substantially longer. The importance of this data relates to the fact that the investigators focussed their analysis on 5-minute and 10-minute-long hazard intervals prior to the collision. It is not known if the subject was actually on the cellular telephone at the time of the collision. The study contrasted a time period on the day of the collision with a comparison period on a day preceding the collision. The authors assert that this approach would identify an increase in risk if there were more telephone calls immediately before a collision than would be expected solely by chance. The key measure that was analyzed is termed "relative risk." In other words, relative risk was defined as "the probability of having a collision when using a cellular telephone at any time during a 10-minute interval as compared with the probability of having a collision when not using a cellular telephone at any time during a 10-minute interval." (p. 456). Quantitatively, relative risk is calculated as follows. The example in Table 5-1 is an explanation of the "crude" relative risk assessment given on p.455 of the article, as explained by Redelmeier in a phone interview with one of the present report's authors.

Presumably, the interpretation is that the baseline risk (not observed or estimated) was the same on the crash day and the preceding day. Therefore, by this line of reasoning, the baseline risk was raised by some multiplier equal to the ratio of the observed cellular telephone uses on the crash days and cellular telephone uses on the preceding days. Because of the many variables that can affect crash hazard probabilities but that cannot be equated with the case-crossover study design, the authors point out that a causal relationship between cellular telephone use and crashes cannot be drawn. The implication of causality based on relative risk metrics would require very strong assumptions about the equality of baseline risk for each matched-pair in the study on all accounts except cellular telephone use. Such assumptions may not be plausible unless it can be assured that the situational characteristics (traffic situations, driver states, nature of cellular telephone use, etc.) were the same across the two days. The implausibility of this is reflected in the fact that an adjustment factor of 35% was subsequently applied in their analysis because a subject may not have even been driving during the control period. The comparison of relative risk of a crash associated with cellular telephones and that associated with drivers with blood alcohol levels at the legal limit deserves special mention. While such a comparison may emphasize the potential adverse consequences of using a cellular telephone while driving, it overlooks some important distinctions between the two categories of crashes. First, no causal link has been established between cellular telephone use and crashes in their study. In contrast, the link between driving while intoxicated and crashes is far more clearly established, both in terms of the nature of the influence on driving and the magnitude of the problem. Second, it must be recognized that cellular telephone use is a transient behavior, lasting on the average (in this study) 2.3 minutes, with the majority of calls lasting 2 minutes or less. Intoxicated drivers, however, are impaired throughout a trip and thus exposure is likely to be considerably greater. The comparison given in the article would suggest that cellular telephone use, per unit time, is actually much more hazardous than driving in an intoxicated state. This finding does not accord with what one might reasonably expect. Thus, the comparison in crash hazard exposure between cellular telephone use and driving while intoxicated is specious unless more data than provided in the article are brought forth. With regard to the lack of an apparent safety advantage of hands-free cellular telephones, it should be noted that having such a feature does not mean it was in use at the time of a call. This issue is compounded by the fact that the specific "hands-free" features for a cellular telephone can vary considerably, requiring varying levels of interaction on the part of the driver for both dialing and conversation. Thus, the distinction between the hand-held and hands-free groups in this study are not clear-cut. Finally, apart from the issue of self selection, a threat to the validity of any conclusions suggested by the Redelmeier and Tibshirani study resides in the nature of the study participants themselves. All 699 subjects were cellular telephone owners who had a crash. But three other groups of drivers might be logically identified for comparison: cellular telephone owners who did not have a crash, non-cellular telephone owners who did have a crash, and non-cellular telephone owners who did not have a crash.

None of these three other groups were considered in the analysis. It is possible that the study participants represent members who are in some sense atypical of the driving population or of cellular telephone owners in general. They may be extreme in the nature of their phone use (e.g., greater frequency of calls, longer calls, more intense dialogue), in their driving style (e.g., more aggressive driving with less margin for error), or even in their human abilities (e.g., less capacity to time-share the driving task and telephoning task). Thus, caution is urged in using the Redelmeier and Tibshirani study results alone to infer that cellular telephone use, in general, is hazardous. In summary, Redelmeier and Tibshirani's study represents a unique and suggestive investigation of the relationship between cellular telephone use and highway safety. Increasing the current level of understanding of the nature of this relationship awaits future research that helps untangle the many threads of potentially influential factors present in this study.

An Investigation of the Safety Implications of Wireless Communications in Vehicles

Chapter 5: A Review of Human Factors Studies on Cellular Telephone Use While Driving

5.6 Summary and Conclusions   Experimental Studies   Dialing Task - The simulator and test-track studies described in this review deal with many facets of driver behavior and performance while using cellular telephones. With respect to the dialing task, the studies reviewed suggest the following. When compared to driving alone, cellular telephone manual dialing can be disruptive of vehicle control activities like lanekeeping and speed maintenance (Stein et al., 1987; Zwahlen, et al., 1988; Serafin, et al., 1993). However, this disruption does not always appear, especially in closed-course environments (Kames, 1978). Research does suggest that voice dialing reduces risk and therefore may be viewed as a desirable design goal. Manual dialing is sometimes, but not always, found to be more disruptive than manually tuning a radio (McKnight and McKnight, 1991, 1993; Stein, et al., 1987). Subjective assessments by test participants indicate that they are generally aware of the demanding nature of manually dialing a cellular telephone. Many studies report driver behavior that resembles attempts to compensate for such disruptive effects (e.g., by slowing down the vehicle). Based on the results of the on-road studies of cellular telephone use conducted to date, the following patterns arise. First, on the road, disruptions by manual dialing to lanekeeping or speed maintenance, as compared to manual radio tuning, appear to be small to nonexistent (Hayes, et al., 1989; Green, et al., 1993; Tijerina, et al., 1995).

On the other hand, data indicate that both manual radio tuning and manual dialing can be disruptive to driving (Tijerina et al., 1995a, 1995b) and crash data indicate radio tuning is associated with crash involvement (Wierwille and Tijerina, 1994). The magnitude of visual attention demand while dialing is sometimes less than that associated with manual radio tuning (Hayes, et al., 1989), though at other times dialing may demand greater numbers of glances and total time that the eyes are off the road (Tijerina, et al., 1995a, 1995b). Driver situational awareness (as supported by mirror sampling) appears to be reduced (Brookhuis et al., 1991; Tijerina et al., 1995) though some experimental evidence exists that this reduction occurs only under conditions where drivers judge it to be acceptable, i.e., quiet motor ways (Brookhuis et al., 1991).

Voice Communications - For the voice communications task and its effects on driving, the following can be concluded. On the positive side, voice communications, if sufficiently frequent and simple to perform, appear to enhance driving performance with fatigued drivers (Drory, 1985). Equally important, simple conversations appear to have little impact on lanekeeping and speed maintenance but sometimes affect driver situational awareness (e.g., increased reaction times, reduced mirror sampling). As a rule, however, the simulator and test track studies that make use of cognitively demanding "intelligence test" conversational materials generally show degradations in lanekeeping, speed maintenance, or headway maintenance (Stein, et al., 1987; Alm and Nilsson, 1990; Nilsson and Alm; 1991; Nilsson, 1993; Serafin, et al., 1993). The impact of such voice communications on perceptual and judgment performance and object and event detection is also negative (e.g., Brown, et al., 1969; McKnight and McKnight, 1991, 1993). The relationship between the conversational materials used in these studies and the content of normal cellular telephone communications is unknown. Thus, such results may represent worst- case or atypical voice communications. On the other hand, all simulator and test track studies to date have used conversational materials devoid of emotional content. Emotionally laden communication (e..g, a broker's call to learn that a great deal of money has been lost or a domestic argument) may have a deleterious impact on highway safety that is even greater than that found with cognitively demanding tasks. A better understanding of the nature of actual cellular telephone communications in business and private calls is sorely needed. This characterization would include such factors as call frequency (to both place and receive calls), call duration, call content, and call etiquette. A metric of conversational "difficulty" would also be beneficial, though a fully defensible metric may be as elusive as metrics of reading difficulty have proven to be. Cognitively demanding voice communications appear to also increase driver brake reaction times, again indicating a reduction in situation awareness. In terms of further conversational effects, it appears that cellular telephone conversation need not be any more demanding than conversation with a passenger (Fairclough, et al., 1991) at least in terms of driver visual allocation of attention. On the other hand, cellular telephone conversations can sometimes be more demanding than passenger conversations because the passenger can accommodate the pace of conversation based on the current driving situation (Parkes, 1991). Methodological Considerations - If there is a single common threat to the validity of simulator studies and closed course test track studies, it is the demand characteristics of those environments when compared to real world driving. There is currently no way to determine how closely behavior in the simulator or test track would match behavior exhibited on the roadway other than to compare the two sets of results obtained with identical test materials and protocols. One comparison of on-road study results with those obtained in a part-task simulator using the same dialing and voice communications tasks and materials led to somewhat different results (Hanowski, et al., 1995; Kantowitz, et al., 1996). In general, it appears that in those studies, professional heavy vehicle drivers allowed the driving task to deteriorate more in the simulator than they did on the road. This suggests that the consequences of primary driving task failure on the road provide an incentive to the drivers to maintain consistent performance while driving on public roads. This incentive can be difficult to adequately emulate in the simulator environment. Conclusions - It appears that manual dialing can be disruptive of both vehicle control performance and situational awareness and judgment performance. The incidence and magnitude of disruption while driving on public roads appear to be less than that encountered in driving simulators or on test tracks, but may nonetheless pose a safety concern. Therefore, designs to streamline the visual-manual demand associated with cellular telephone dialing activities appear warranted.

It is important to point out that the majority of cellular telephones now in use are hand-held (73% of cellular telephones sold in the U. S. in 1995 were hand-held while for Japan the figure was 94%). This fact has potentially important implications for how cellular telephone use might influence driving behavior and performance since the visual allocation and manual tasks are very different for fixed mount vs. hand-held systems. Since the focus of the published research has been on fixed mount systems, care must be exercised in generalizing these results to hand-held. The manual and visual allocation tasks are very different for the fixed mount vs. hand-held architectures, so there may be significant differences in how they influence driving. For example, fixed mount systems may require considerably more glance time for dialing since the driver may have to look further away from the roadway while accessing the phone keypad. In contrast, the hand-held allows the drivers to maintain the phone in a position where the roadway can be more easily monitored, although the hand-held may require two hands to dial, in which case steering control may be compromised. In fact, discussions with hand-held users who dial while driving indicated a variety of strategies to cope with this problem. These include holding the phone and dialing with one hand, or removing both hands from the wheel entirely during dialing while securing the wheel with the knees or wrist/forearm. In addition, hand-held telephones may be stored in glove compartments, briefcases, pockets, etc., and may thus require the driver to reach and/or search for the phone. Finally, hand-held telephones may require that an antenna be extended, and in the case of the "flip-phone," may require additional manipulation. It was noted earlier that 94% of cellular telephones purchased in Japan in 1995 were hand-held and the largest contributor to cellular telephone related crashes in Japan (42%) was associated with responding to a call. Solutions to some of these concerns may be found in the application of hands-free dialing technology. The conclusions to be drawn from assessments of the effects of hands-free voice communications tasks are less clear. On-road studies indicate that if the voice communications activities have any effects at all, they are on driver situational awareness and not on vehicle control performance per se. The simulator studies that show vehicle control disruption may reflect an experimental artifact, i.e., that drivers do not place as high a priority on the driving task in a simulator as they do on the road. The voice communications dialogue materials that have been used in this line of research often involve "intelligence test" type materials that may represent both extreme and atypical cognitive loads when compared to normal cellular telephone communications. On the other hand, all of these studies used voice communications that were free of emotional content as well. Dialogues that involve substantial degrees of conflict, for example, may be even more disruptive than the cognitively challenging materials typically included in human factors testing. There is a great need to better understand the characteristics of cellular telephone communications (frequency, duration, content) that normally arise in the real world in order to better understand how best to represent them in future studies. There also appears to be a need to develop better means to maintain or enhance driver situational awareness while driving. This may be accomplished through intelligent transportation system (ITS) technologies such as the "intelligent answerphone" (Parkes, 1991), driver status monitoring (drowsy driver) or other crash avoidance systems (CAS) that warn the inattentive driver of crash hazards.

Epidemiological Studies Violante and Marshall (1996) and Redelmeier and Tibshirani (1997) represent epidemiological research that has been carried out on cellular telephone use and traffic safety. While useful as an additional research technique that may complement experimental or observational research, it is necessary to recognize certain limitations to the epidemiological method. For this method to be valid, the case and the control have to be similar in every other way that could impact on crash probability. A statistical association does not necessarily imply a causal relationship. For example, other factors may correlate with cellular telephone use, such as driver personality or demographic characteristics, driving style, vehicle characteristics, trip characteristics (purpose, location, time of day, type of roadway), and so forth. It may be that such associated characteristics, and not phone use itself, cause the observed statistical relationship. Furthermore, the epidemiological method addresses the general outcome (crashes), but tells us little about why that outcome occurred. If phone use was affecting driver performance, what aspect of performance, and under what conditions? As Violanti and Marshall were careful to point out, their methods do not even establish whether a cellular telephone was in use at the time of the collision. All that is known is how much the driver tends to use the cellular telephone. If phone use did affect driving, it is not known what aspect of phone use (dialing, talking, etc.) caused the problem, and what aspect of driving performance (lane control, hazard detection, etc.) was degraded and resulted in the collision. There are also other important methodological concerns that must be considered, such as sampling biases. The most recent epidemiological study on the relationship between cellular telephone use and traffic safety is that of Redelmeier and Tibshirani (1997). They studied 699 Toronto drivers who had cellular telephones and who were involved in motor vehicle collisions resulting in substantial property damage but no personal injury. Each person's cellular telephone calls on the day of the collision and during the previous week were analyzed through detailed billing records. The time of each collision was estimated from each subject's statement, police records, and telephone listings made to emergency services. The authors reported that the risk of a collision was estimated to be between 3.0 and 6.5 times as high within 10 minutes after a cellular-phone call began as when the telephone was not used. Maclure and Mittleman (1997) carried out additional analyses on the same data and confirmed that the risk more than doubled within five minutes after the start of a call. Three additional findings in the Redelmeier and Tibshirani study were a) cellular telephone units that allowed hands-free operation offered no safety advantage, b) thirty-nine percent of the drivers called emergency services after the collisions, suggesting that a cellular telephone may have advantages in collision notification, and c) the relative risk of having a crash while using a cellular telephone was estimated to be similar to the hazard associated with driving with a blood alcohol level at the legal limit. This study is suggestive of a relationship between cellular telephone use and crashes that merits further experimental inquiry, but it has several limitations as well. Self-selection of study participants, variability in driving conditions and driving behavior, and no indication that the cellular tele phone users were 'at fault', all limit the definitiveness of the study conclusions. Further, the "relative risk" metric used makes very strong assumptions about the comparability in crash risk between periods where cellular telephone use preceded crash involvement and periods where it did not. The relationship between cellular telephone use and crashes is made more uncertain in light of the fact that the driver with a cellular telephone may not have been using it at the time of the crash, where the time of the crash is itself estimated and subject to various sources of error, and when a substantial number of study participants may not have even been driving during the "control period." While Redelmeier and Tibshirani's study involves a number of shortcomings, it nonetheless represents a unique and suggestive investigation of the relationship between cellular telephone use and highway safety. Increasing the current level of understanding of the nature of this relationship awaits future research that helps untangle the many threads of potentially influential factors present in this study.

An Investigation of the Safety Implications of Wireless Communications in Vehicles

Chapter 5: A Review of Human Factors Studies on Cellular Telephone Use While Driving