« Previous | Contents | Next »
Listen
EXTERNAL-TO-VEHICLE DRIVER DISTRACTION
CHAPTER FIVE INTERNAL TO VEHICLE DISTRACTION: MOBILE PHONES
5.1 Since 1990 there has been a great increase in studies on the possible safety effects of mobile phones. It is likely, in fact, that there is a relationship between the increase in studies on this subject and the comparative decrease in studies on external-to-vehicle distraction (for example, billboards). This subject is not, strictly speaking, relevant to the study of external-to-vehicle distraction. Nevertheless, a brief discussion of mobile phone use and its impact on accident rates will help to demonstrate the difficulties of demonstrating causal links between such complex phenomena.
KINDS OF STUDIES
5.2 As noted before, FARS and NASS are the major databases concerned with car accidents in the US. Therefore it is important to note that only Minnesota and Oklahoma included data relevant to mobile phone use on their Police Accident Reports. Given that causal elements in fatal crashes are generally considered to be more reliable in terms of assessing numbers (given that fatal accidents are rarely not reported) the low numbers should be noted. For example, from the FARS database, only 2 incidents where mobile phone use played a part in a fatal accident were reported in 1994, and only 1 in 1995. However, having said that, it does appear, at first glance that 'the NASS and FARS files, and anecdotal observations of driver performance, […]' demonstrate 'an apparent link between cellular telephone use and driver inattention'. (Goodman et al, 1997: Chapter 3: 13). However, proving that this is the case, and then making the leap to associating driver inattention (caused specifically by mobile use) to crash rates is more difficult.
5.3 A more complex analysis of the raw data demonstrates that there does seem to be a correlation between mobile phone prevalence and mobile phone related accidents. However, if the data are analysed such that the number of accidents is divided by the number of mobile phones in use in that year (in other words, the accidents are expressed as a percentage of mobile phones in use) then mobile phone related accidents are seen to be decreasing (Goodman et al, 1997: ch. 4: 11). This is compatible with the Neisser theories discussed earlier, in which multi-tasking is a skill. When mobiles were introduced they were a safety hazard because people had not learned to drive and talk at the same time. Therefore, accident rates rose. However, after a while, this effect peaked and then, as drivers learned this skill, declined.
5.4 Of course, there may be other reasons for this statistical pattern, which again indicate the difficulties in this kind of study. It may be that the increasing prevalence of hand held phones (as opposed to in-car phones) facilitates drivers hiding their phone in the event of an accident and then denying that they were talking on the phone for insurance or legal reasons. Alternatively, phones might be becoming easier to use, and hence safer (Goodman et al, 1997 Chapter .4:14).
LAB STUDIES
5.5 The classic demonstration of the deleterious effect of mobile phone use on performance was Brown, Tickner and Simmonds (1969). In this experiment, subjects carried out a task (basic logical questions) relayed to them by in-car telephone while driving. Results suggested that driving ability was reduced during task performance. The fundamental problem with this experiment, of course, was ecological validity (that is, how 'realistic' the experiment was). Specifically, there were no serious consequences to making a driver error. Therefore it may have been that subjects spent more time thinking about the telephone task than they would have done 'in real life'. If this is the case, then this is a case of 'absorption' rather than distraction (Tellegen, and Atkinson, 1974). That is, drivers became too absorbed in the task, and devoted insufficient concentration to the driving task. Given that absorption, as previously discussed, is related to 'highway hypnosis' and the 'moth effect', this might be better conceptualised as drivers finding a task (talking on the 'phone) as being (comparatively) more interesting than the more monotonous driving task, rather than as an example of 'cognitive overload'. Of course, most telephone conversations are rather easier to follow than the complex logical questions used in this task.
5.6 Drory, (1985) carried out an experiment in which truck drivers drove a simulator and were asked to perform a very easy verbal task via hands-free phone during a 7 hour driving course. This actually increased driver performance, supporting a point made in section 3.44. In monotonous driving conditions, additional stimulus in the form of information can help maintain arousal and keep drivers alert (of course, if the task was too 'interesting' this might have led to a decrease in performance due to absorption, as above).
EPIDEMIOLOGICAL STUDIES
5.7 The major advantage of epidemiological studies is of course that they contain 'real world' data. The major disadvantage is that correlation does not imply causation. That is, even if the statistical tests used are beyond reproach, and the data is trustworthy, then even perfect correlations do not prove causal relations. This point has been made before, but is worth stressing.
5.8 The classic study in this regard is Redelmeier, D, Tibshirani, R. (1997) (published in the highly prestigious New England Journal of Medicine) which has often been claimed to have provided proof of the causal connection between mobile phones and accidents.
5.9 The authors studied 699 Toronto drivers who had had crashes and who owned mobile phones. Each person's mobile 'phone records were studied through their billing records. This was then cross-referenced with the time of their accident. Their conclusion helped to make the study famous: that the relative risk of having a crash whilst using a cellular telephone was estimated to be the same as that of driving while at the legal limit of blood alcohol.
5.10 To reiterate, this is by far the best and the most comprehensive of the epidemiological studies. And since this is an epidemiological study it must be pointed out, yet again, that this kind of study cannot demonstrate that mobile phone use causes car accidents. However, even with this caveat, there are grounds for caution.
5.11 Firstly, and most importantly, the authors made a distinction between accident times that were 'exact', and accident times that were 'inexact'. 'Exact' times were accident times which could be substantiated from at least three separate sources: personal recollection, police records, and telephone listings to emergency services. Inexact times where fewer than three of these data were available. 33% of the crash times given were exact, and the remaining two thirds were inexact. Therefore only a third of the time data used for the correlations were known to be accurate. Moreover it should be remembered that personal recollection of accident time (even when taken on the spot) might be inaccurate. How many drivers who have had an accident immediately check their watches to find out the exact time at which it happened? And who checked whether the driver's watch was fast or slow?
5.12 Police records might seem to provide a check on this phenomenon, until it is realised that in many cases the police estimate the time of the accident by asking the driver. Using telephone records to emergency services as an 'objective' check on these two estimates presupposes that all drivers immediately (or at least within a minute or so) phone the emergency services. Surely, however, many drivers would be briefly unconscious immediately after the accident, or, if not, would check that the occupant of the other vehicle was uninjured or needed first aid, before calling for help.
5.13 If this is the case, then cross-tabulation of mobile phone records and accident times becomes extremely problematic. Crudely put, how do researchers know that drivers were actually on the phone at the time of the accident? This problem is made worse by the fact that the authors analysed their data on five minute 'time intervals'. However the average call length in this study was 2.3 minutes. Therefore, even if a call was made in the five-minute interval before an accident, this does not prove that the driver was on the phone at the time of the crash. In any case, as the authors themselves point out, they did not take 'culpability' into account. That is, the police or the courts did not necessarily decide that the driver using the mobile 'phone actually caused the accident. Mobile phone use may just have been a coincidence, with the accident being caused by another driver (who was probably not using their mobile 'phone at the time of the accident).
5.14 Therefore, even this study, which seemed to prove a correlation (NOT a causal relation) between mobile 'phone use and accident rates must be treated with extreme caution. It should be noted that there has been far more work done on mobile phones and accidents than on external-to-vehicle distraction and accidents. The fact that, as of yet, there is no proof that mobile phone use causes accidents (although there is much circumstantial evidence) indicates that the relationship between external-to-vehicle distraction and accidents is likely to be even more tenuous.
« Previous | Contents | Next »