The Scottish Government

« Previous | Contents | Next »

Listen

EXTERNAL-TO-VEHICLE DRIVER DISTRACTION

CHAPTER FOUR STUDIES OF DRIVER DISTRACTION AND WHAT THEY FOUND

4.1 To reiterate, therefore, there is a large body of scientific data relating to attention which is of relevance to driver distraction. Now, however, we will look at specific road accident databases which contain information relating to distraction, to see whether there is empirical data supporting the hypothesis that driver distraction is, in fact, a major problem in terms of road safety.

KINDS OF DATA: U.S. DATABASES

4.2 There are two U.S. databases which are of interest here: FARS (Fatality Analysis Reporting System), and NASS (National Automotive Sampling System). NASS can be further subdivided into NASS/GES (General Estimates System), and NASS/CDS (Crashworthiness Data System). The vast majority of studies relating to driver behaviour use data from one or both of these databases.

FARS

4.3 FARS is the oldest and most prestigious of these databases (collection of data began in 1975), and uses data from Police Accident Reports (PAR) to create an electronic database of contributory factors to road accidents. Since FARS deals only with fatalities (deaths must occur within 30 days of the accident), it avoids many of the problems discussed above as to gauging the 'seriousness' of an accident. Its major deficiency as regards studying external-to-vehicle driver distractions is that it does not collect data on Non-Technological Distractions. Moreover, FARS uses police reports from many different states, and it is not clear if data-collection methodologies are identical (see above) (Tessmer, 2000). Wang et al also point out that 'PAR based data are generally superficial and not designed to provide a scientific determination of crash causation' (Wang et al 1996: 3).

NASS

4.4 NASS (created in 1979), on the other hand, is a proactive database. That is, accidents (of any sort) are selected randomly in order to produce a representative sample of accidents in any given year. Then trained accident investigators obtain data (via interviews, and 'on the spot' investigation), and a database of accident causes is created (NHTSA 2001).

NASS CDS and GES

4.5 The NASS CDS data deal only with accidents serious enough for one or more cars to be towed away from the scene of the crash. 'The CDS (Crashworthiness Data System) data collection process includes review of the PAR, vehicle and accident site investigation, reconstruction of component trajectories, interviews with drivers and other persons, and review of medical records' (Wang et al, 1996: 3). In the GES (General Estimates System) a similar process takes place directly on PARs, with the results being coded, and then problem areas identified (U.S. Department of Transportation, 1999).

THE INDIANA STUDY

4.6 The largest study yet carried out on general traffic safety is the Tri-Level Study of the Causes of Traffic Accidents. This is sometimes referred to as the Indiana study as it took place under the auspices of the University of Indiana.

4.7 The Tri-Level study is so-called because of the three levels of investigation used. These were: level A, in which basic demographic and driver data were obtained from standard databases and populations surveys (for example, data relating to age and gender was obtained from a general study of licensed drivers, and vehicle registration data was obtained from the Indiana Bureau of Motor Vehicles).

4.8 In Level B teams of accident investigators were assembled, such that by 1974 they were able to reach any accident in the target area 24 hours a day seven days a week. When an accident was reported, an investigation team was dispatched who were able to take photographs, inspect vehicles and interview drivers. Later they were able to produce a report as to what 'caused' the accident.

4.9 Level C was a more intensive post hoc investigation in which drivers were interviewed by experienced psychologists or sociologists, accident scenarios were reconstructed and experiments carried out to test hypotheses.

4.10 The 'on-site' team (responsible for the level B investigations) found that human factors were identified as being the 'definite cause' of 64.3% of accidents. These figures rose when discussing probabilities so that the level B team were prepared to state that 90.3% of accidents were probably or definitely caused by human factors. The level C investigations were slightly higher (92.6% probably or definitely caused by human factors).

4.11 It should be noted that Indiana is a predominantly rural state, and that it may not be possible to extrapolate these results to more urban areas.

Findings

4.12 The findings are given in Table 4.1, below.

Table 4.1

4.13 The numbers of cases with 'External Distraction' as a possible or probable cause were not a sufficiently large category to be included in the table.

4.14 Perhaps more suggestive is another finding that analysed the frequently occurring causal factors associated with property damage. (Frequently occurring defined as factors occurring 16 or more times in on-sight investigation). In Table 4.2 the 'Human Direct Causes' of this variable are shown.

Table 4.2

4.15 Unfortunately the 'External Distraction' Category is not broken down anywhere in the report (Treat et al 1979 a and b).

AAA STUDY

4.16 The AAA foundation for Traffic Safety funded the University of North Carolina to interrogate the NASS-CDS database in order to attempt to identify how many accidents were caused by driver distraction (Stutts et al, 2001). The report also contained narrative searches (qualitative searches) of the database of CDS and North Carolina data. 1995 was the year that 'driver inattention' was included as a code in the NASS-CDS taxonomy. The AAA study amalgamates data from 1995-1999 to create an average.

4.17 The report defined driver distraction as a situation 'where a driver is delayed in the recognition of information needed to safely accomplish the driving task because some event, activity, object, or person within or outside the vehicle compelled or tended to induce the driver's shifting attention away from the driving task' (Stutts et al, 2001: 6). They specifically exclude 'looked but did not see' incidents, and Task Unrelated Thought (TUT). The form had a section for 'Driver Attention Status', which consisted of five subcategories: 'Attentive, Distracted, Looked but Didn't See, Sleepy or Fell Asleep, Unknown/No Driver'. Only if the driver was noted as being 'Distracted' was a further series of options made available to describe why the driver was distracted. Here there were thirteen separate headings.

4.18 It will be seen that category 1, and categories 3-13 represent 'inside distracters', with only category 2 representing 'outside distracters'.

Problems with the Study

4.19 Before moving on to the findings of the study, we should first look at some of its limitations, as explained by the AAA themselves.

Missing Data

4.20 In spite of 'intensive investigation' the 'driver attention' status was marked as 'unknown' for 36% of the accidents presented. Moreover, even when 'driver distraction' was marked, 34% of the 'causes' were listed as being 12 or 13 above: that is, 'other' or 'unknown'.

Sample Size

4.21 Because of small numbers these figures should be extrapolated to the national level with extreme caution. For example, the estimates as to mobile phone distraction are based on only 42 cases.

Exposure time

4.22 It is not noted in the CDS database how long the distracting activities lasted. Therefore it is hard to quantify to what extent they qualify as a risk factor. For example, if drivers using mobile phones were travelling cross-country and had been on their phones all day, then we would have to conclude that mobile phones are not much of a risk. On the other hand, if the drivers started to make calls and then immediately crashed, we would have to conclude it was much more of a risk. It might be added that the AAA study applies only to the United States. It is not clear to what extent these findings apply to other countries (University of North Carolina, 2001).

Findings of the AAA Study

4.23 Between 1995 and 1999, in 'serious' accidents available on the CDS database, 8.3% of the drivers were described as being distracted at the time of the accident. This could be broken down into internal-to-vehicle distraction and external-to-vehicle distraction. External-to-vehicle distraction accounted for 29.4% of distracters. Internal-to-vehicle distraction accounted for 36.4% of distracters. There were also 25.6% distracters counted as 'other' and 8.6% counted as 'unknown'.

Further Breakdown of Statistics

4.24 Unfortunately the 'external-to-vehicle distracters' were not further broken down in this database. However, in addition to the coded variables on the CDS data file, it was also possible to analyse the written narratives that describe the accidents in free text. The years 1997 and 1998 were made available for analysis in this way. The results were (presumably) classified (the report does not make this clear). It should be noted that only 84-88% of the (raw number of) distraction cases contained any data at all, and of these, 43% contained insufficient data to clarify the nature of the distraction. Therefore the stricture above about small numbers should be particularly borne in mind regarding these data. Only the 'Outside Person, Object or Event' category has been further subdivided, as this is the specific object of this study (See Table 4.3).

Table 4.3

Further Definitions

4.25 This analysis also gives further definitions of specific examples of distracters. For the 'Outside person, Object or Event' category these are as follows.

• 'Outside Traffic/Vehicle: Examples include Vehicle swerved, turned in front of, changed lanes, slowed or stopped, encroached on lane, emergency vehicle, bright vehicle police lights, etc.

• Police: Examples include 'Being chased by, officer directing traffic, thought saw police, other'

• Animal in Roadway: 'Examples, deer, dog, elk, other.

• People/Objects in roadway. Examples: child in road, basketball game, crowd, broken glass, garbage can etc.'

• Other. Examples: 'waved ahead by driver, another person or driver, parachutes in sky, bicycle toll booth, brush obstructing vision, tire blow-out etc.'

4.26 The 'Other' category has been included in table 4.3 above, because the examples make clear that some of its sub-categories refer to external-to-vehicle distraction: for example, 'Other Distraction: Looking Outside Window: Examples, 'in rear view mirror, at road signs, in store window, for gas station, for parking space, for business etc.''

4.27 Also 'Other Distraction: Other' contains the example of 'sun glare' again, arguably, an external distracter. (Note: there is no doubt that sun glare 'distracts attention': (Finlay and Wilkinson, 1984): the question is whether it is genuinely external or not).

4.28 In short here are the totals.

• Percentage of outside distracters excluding Other Distraction: 'Looking Outside Vehicle'. 28.9% (1997).

• Percentage of outside distracters excluding Other Distraction: 'Looking Outside Vehicle' 30.3% (1998).

• Percentage of outside distracters including Other Distraction: 'Looking Outside Vehicle' 31.3% (1997).

• Percentage of outside distracters including Other Distraction: 'Looking Outside Vehicle' 32.5% (1998).

4.29 Given that CDS data External-to-vehicle distraction varied from 19.8% (1998) to 35.4% (1997), the data above falls between the expected parameters, and demonstrates that, very roughly, based on this data, external-to-vehicle distraction was stated as a cause for about a third of all events where 'driver distraction' was noted in the police forms.

THE SUSSMAN STUDY

4.30 Sussman, Bishop, Madnick and Walter (1995) carried out a study of 1982 NASS data (1982 was the first year that 'driver related factors' were noted in the database). The first thing that should be noted about their study is that their definition of driver inattention is 'the attentional state where the driver fails to respond to a critical situation' (Sussman et al, 1995). This is rather different from the more standard definition quoted on paragraph 2.1, above. Since they discovered that in 37% of accidents in the database the driver took no action to avoid the collision, they infer that this 'suggests' that attentional lapses are a major factor in highway accidents. However, they admit that only 8% of accidents listed in the database were defined as being caused specifically by driver inattention. The use of only one year for analysis is another handicap. The paper contains no data on external driver distraction.

WANG KNIPLING AND GOODMAN

4.31 Another important study is the Wang Knipling and Goodman study of 1996. This used the CDS database from one year (1995). It is particularly important in that the CDS database was created to be nationally representative, in a way that more localised databases were not. Moreover, it used a high number of cases (4,536). The main problem with this study is the fact that it only took one year's data: it is unknown how 'average' 1995 was as a year for distraction incidents. Moreover, as the authors themselves point out, the CDS 'was not originally intended to collect crash causation data' (Wang, et al 1996: 11), and, given that CDS uses accident investigation methods, it must be noted that accident investigations are retrospective and that they are always, therefore, to a certain extent conjectural.

4.32 The method of analysis is also noteworthy in terms of representativeness. The authors note that 'if any involved driver was coded as exhibiting some form of driver inattention, the whole crash was classified under this category' (Wang et al op cit: 4): thus making a large assumption, that because one driver was classified as being 'inattentive' this therefore proved that inattention was the chief contributory factor to the crash. Table 4.4 below shows the data relating to attention.

Table 4.4

Weighted Driver N=4627000 (7943 unweighted) Weighted Accident N=2619000 (4536). In order for an accident to be classified 'attentive' all involved drivers had to be classified 'attentive'.

4.33 Therefore, Wang et al found that 7.8% of drivers and 13.3% of accidents were associated with driver distraction.

WIERVILLE AND TIJERINA

4.34 In 1996 Wierville and Tijerina conducted an accident base search using search keywords, on the North Carolina accident data base, which contains almost all police reported events in the state. Again, the authors developed their own definition of driver distraction, describing it as problems 'with visual allocation and visual workload': audible and other distracters were therefore not discussed. The search period was on 1989 and the first four months of 1992: as with the Wang et al study it is not clear how representative a sample this was. The 1989 database contained 189,164 records, and the four months of 1992 contained 61,707. The word search located 14,372 possible visual allocation issues from 1989, and 3,247 from 1992. After closer inspection, it was decided that 2,816 were in actual fact visual allocation issues. Of these 1,562 were internal-to-vehicle visual allocation issues, 661 were external-to-vehicle visual allocation issues, and 593 were unknown. Therefore 1.48% of reported accidents in 1989 were noted as involving driver distraction, and 0.34% percent of reported accidents involved external-to-vehicle driver distraction. The 'raw' data for 1991 was not quoted. Unfortunately as the study was investigating mainly internal-to-vehicle distraction, there was little information given about the nature of the external distracters. However, it was noted that 50 events (out of 189, 464) were classed as issues to do with mirrors (e.g. 'observing external vehicle, pedestrian, object, animal etc.), and that a further 41 were 'mirror' incidents where the distracter was unknown.

Problems with Wierville and Tijerina

4.35 Wang et al (1996) point out that the Wierville and Tijerina study has figures for distraction considerably lower than the majority of other studies. They suggest that this is because the North Carolina database includes only data concerning accidents where property damage was over $500. They also suggest that perhaps the North Carolina PAR does not have a field for 'inattention or distraction'. This emphasises a point made earlier: numbers for contributory factors will be different depending on whether drivers are specifically asked whether something might be a factor, or if they are left to volunteer the information themselves.

STEVENS AND MINTON

4.36 Stevens and Minton are of particular importance in this respect in that their study was carried out in the UK. However, it should be noted that the database used covered England and Wales only, and that generalisations to Scotland should be made with care.

Methodology

4.37 Minton and Stevens gathered data from fatal accident reports in England and Wales between 1985 and 1995 and 'coded' them themselves to create a quantitative variable: 'distraction'. As well as this, the degree of certainty as to the extent to which the event really was a 'distraction' event was also noted. For example, if the driver admitted he was distracted, this was 'certain'. If there was strong circumstantial evidence '(for example, the driver died, but was found with a cassette or mobile phone in his hand) this was classed as 'probable'. If there was weak circumstantial evidence this was classed as 'possible'. Unfortunately the published paper only dealt with internal to vehicle distractions but the current database now contains 123 accidents involving external-to-vehicle distraction (Minton, 2003) This constitutes a percentage of 1.06% from a total of 11,529 fatal accidents logged in the database.

Department for Transport

4.38 Brown (2001) wrote a paper reviewing an earlier paper by Sabey and Staughton (1975) reviewing the 'looked but failed to see' accident causation factor in the UK. Sabey and Staughton carried out research on a database of 2130 event reports, which were coded by researchers to create 3704 driver errors from 2130 accidents. These are listed in Table 4.5 below.

Table 4.5

4.39 Therefore distraction accounted for 9.09% of contributory factors listed, and 15.8% of accidents had distraction given as a contributory factor.

4.40 Brown (1984) reanalysed this data and eliminated accidents which occurred at night, or where the driver was under the influence of alcohol, drugs, fatigue or illness, in order to study perceptual errors more closely (Shown in Table 4.6).

Table 4.6

Note: '49%' is not simply the sum of contributions by the individual listed factors because some accidents were associated with more than one perceptual factor.

4.41 Unfortunately, this study was more concerned with Looked but Failed To See Errors (LBFTS) and so the 'distraction' category was not broken down further into 'internal' and 'external'. Nevertheless it is clear that distraction accounts for a large percentage of driver errors leading to accidents in the UK.

THE KOSTYNIUK-EBY STUDY

4.42 One of the few exceptions to the rule that only questionnaire data is used in the study of driver distraction is the Kostyniuk and Eby (1998) study of Rear-End Roadway Crashes, conducted by the University of Michigan for Honda. Kostyniuk and Eby conducted focus groups with Michigan drivers who had had rear-end accidents, and asked them why the accidents had occurred. The main drawback of the study is that, as the authors state, this was an exploratory study, with small numbers (N=16, plus N=10 for separate telephone interviews). Its generalisability is therefore questionable.

4.43 The results of the focus group studies as regards what factors contributed to the crash are shown in Table 4.7.

Table 4.7

4.44 Interestingly, drivers mentioned road design, something not mentioned in the other studies (presumably because there was no entry for it in the form). Actions of other drivers features prominently, but it must be remembered that this study examined rear-end collision only.

4.45 Unfortunately, again, driver distraction was not divided into external-to-vehicle and internal-to-vehicle. Nevertheless, it does show that drivers consider inattention/distraction to be the major contribution to this kind of accident, insofar as drivers feel the accident was 'their own fault'. Further questioning in the focus groups of 'personal error that caused your accident' makes this clearer (Table 4.8).

Table 4.8

N=25

4.46 Finally, two Scottish studies were carried out specifically for this research.

FIFE CONSTABULARY

4.47 This study consisted of 21 records made available with the kind help of Sergeant Andy Edmonston, Traffic Management Sergeant of Fife Police force. It should be noted that there seems to be some confusion in the use of the phrase 'external distraction' in this database in that many of these cases of 'distraction' were (for example) situations in which a pedestrian walked in front of the car. Moreover, many of the examples in the 'animal' category were cases in which an animal moved in front of a moving car and the car had to swerve unexpectedly.

17 of the drivers were male, 4 female.

The average age was 47.38

The categories are as follows (Table 4.9).

Table 4.9

It should be noted that 'sunglare' could also be described as 'internal' distraction.

4.48 It must be stressed that due to the small numbers this list should be treated with caution. Moreover, the definition of 'distraction' seems to be relatively wide.

CENTRAL SCOTLAND POLICE ROAD ACCIDENT DATABASE

4.49 Contact was also made with Records Bureau Supervisor Stuart MacFarlane of the Central Scotland Police force, who kindly agreed to help with this study. Since 1999, as a trial project, police officers investigating accidents must fill in a 'contributory factor' box in the accident report form, which is then stored on an electronic database. As part of a preliminary search, 58 possible accidents were selected as external-to-vehicle distraction, but after a search of the database, it was decided that only 26 accidents genuinely fitted this description. 765 accidents (in total) were reported in 1999, 672 were reported in 2000, 637 were reported in 2001, and 597 were reported in 2002 (up until October: the results for November have not been calculated at time of writing). Therefore, 2,671 accidents were reported in total in this time period. Therefore 0.97% of the accidents in the database were classed as being 'external-to-vehicle distraction'.

4.50 The drivers who claimed to have been distracted (by a factor that was external to the vehicle) were overwhelmingly male (only 15.4% were female). The average age was 37.2.

4.51 The breakdown of the contributory factors is as follows (Table 4.10). (Note: percentages have been rounded up and so will not necessarily add up to 100%).