ERGONONUCS, 1993, VOL. 36, NO.

6, 627-644

Decision-making

style, driving style, and self-reported involvement in

road traffic accidents

D. J. FRENCH,
R. J. WEST,J. ELANDER
and J. M. WILDING
Psychology Department, Royal Holloway and Bedford New College, University of London,
Egham, Surrey TW20 OEX, UK
Keywords: Driving; Accident involvement; Survey; Questionnaire; Decision-making; Driving

style.
In an exploratory postal survey of 711 drivers stratified by age, sex, annual
mileage, and accident involvement, decision-making style was measured using a
Decision-Making Questionnaire (DMQ) and driving style was assessed using a
Driving Style Questionnaire (DSQ). Responses to 21 items of the DMQ formed
seven independent and internally coherent dimensions according to a principal
components (PC) analysis. These were labelled: control, thoroughness,
instinctiveness, social resistance, hesitancy, perfectionism, and idealism. PC
analysis also revealed that responses to 15 items of the DSQ formed six
independent dimensions of driving style. These were labelled: speed, calmness,
social resistance, focus, planning, and deviance. Multiple regression analysis
indicated that drivers of 60 years and under who scored lower on thoroughness
were at greater risk of a traffic accident and that this relationship was mediated by
faster driving. This relationship was independent of age, sex, annual mileage, and
all other factors measured. In the drivers over 60 years, lower thoroughness,
greater hesitancy, and faster driving were independently associated with higher
accident rates independent of all other factors measured. The results provide
preliminary support for the view that people import aspects of their general
decision-making style into the driving situation, and that in so doing they put
themselves at differential risk of having a road traffic accident.
1.

Introduction

Previous research has in general failed to reveal an association between psychomotor

ability and road traffic accident rates (Goldstein 1961). Factors which have emerged
as predictors of accident involvement (taking account of annual mileage) include age,
experience, ability to detect hazards quickly and tendency towards risk taking
(Brown and Groeger 1988, Jonah 1986, Mayhew et al. 1981, Quimby et al. 1986). It
may be that differential accident involvement has more to do with the way that
people make judgements and decisions than merely ability to control the car, e.g., the
decision to overtake, change lanes, accept a gap of a particular size when parking, etc.
Thus it seems plausible that decision-making would be a useful focus for research
efforts aimed at understanding differential accident involvement.
There are two main ways of approaching the study of decision-making. One is to
examine the beliefs and values that enter into the decision process (e.g., Edwards
1954, Fishbein and Ajzen 1975). This has been the subject of much research in
psychology. The other is to examine the style of decision-making, i.e., the way that
individuals habitually approach decision problems and use information. Little
research has been directed at this latter issue. The present research programme was
designed to investigate possible relationships between decision-making style and
road traffic accident liability. A major feature of this approach has been to collect
information about general aspects of decision-making and relate these to accident
()()14-0139/93

$1O'()()

1993 Taylor

&

Francis

Ltd.

628

D. J. French et al.

liability by looking at the way that they are reflected in driving style (e.g., choice of
speed).
Style of decision-making involves those aspects of the decision process in which
individuals may be presumed to adopt a common mode of operation across a wide
range of decision domains. Thus for the concept to apply there must be crosssituational stability. Examples of decision-making style may be: how far individuals
are prepared to examine the pros and cons of various options before making up their
minds; the level of risk that they will accept in return for an optimum outcome; and
the relative importance of emotions in making a decision which could be reached
analytically.
There are several existing questionnaire measures of decision-making style. One of
these (Buck and Daniels 1984) was developed from Harren's (1979) Model of Career
Decision-Making. Individuals are characterized by scores on one of three scales:
rational, intuitive, or dependent. Rational decision-makers seek relevant information,
look carefully at future consequences, and act deliberately and logically. Those who are
intuitive in their decision-making show little anticipation of future consequences or
systematic information-seeking.
Dependent decision makers do not take
responsibility for their decisions, but take the path of least resistance, having a high
need for social approval. Arroba (1977) devised a classification of decision-making
which was applied subsequently to career decision-making by Hesketh (1982).
Arroba's taxonomy was derived by interviewing subjects regarding recent decisions
that they had made; six styles of decision-making emerged. These were emotional,
intuitive, rational, hesitant, compliant, and no-thought. The six styles were found to
fall along an active-passive continuum, with rational and hesitant being the most
active and compliant the least active. Hesketh found that one year after careers
counselling, subjects who had reported a rational decision-making style were most
likely to have achieved a match between their aspirations and their current situation,
whereas emotional and compliant individuals were least likely to have done so.
Johnson (1978) has proposed a more general theory of decision-making style, but
this too has been investigated and reported only in the context of careers counselling.
It proposes two independent aspects of decision-making style; information gathering
and information analysis. Information may be gathered spontaneously or
systematically and analysed internally or externally. A questionnaire, the Johnson
Decision-Making Inventory (JDMI-Johnson
et al. 1983), has been designed to
assess these dimensions.
Gordon et al. (1986) studied the relationship between Johnson's dimensions and
Harren's styles, which were assessed using the Assessment of Career DecisionMaking Scale (ACDMS-Buck and Daniels 1984). A factor analysis of the scale
scores, three from the ACDMS and four from the JDMI, showed that the first factor
was identified by strong positive loadings for systematic (John son) and rational
(Harren) and an equally high negative loading for intuitive (Harren). The second
factor was characterized by positive loadings for external and spontaneous (Johnson)
and dependent (Harren). The third factor was Johnson's internal.
The link between Johnson's and Harren's inventories suggests that, by self-report
at least, there is an identifiable trait of decision-making style; conceptual similarities
between these and Arroba's (1977) taxonomy support this conclusion. Had Arroba's
styles been included in Gordon et al.'s investigation, one might have expected further
factors to emerge, indicating that neither the JDMI nor the ACDMS are identifying
the full range of decision-making styles.

Decision-making and RTAs

629

All of the above attempts to characterize decisions and decision makers have
potential contributions to make to a comprehensive account of individual
differences in decision-making style. However, there is no comprehensive decisionmaking questionnaire available.
This paper reports the development of a questionnaire measuring general
decision-making style and its relationship with driving behaviour, also measured by
questionnaire, and involvement in road traffic accidents. The ultimate aim was to
trace a path from certain aspects of decision-making style to differential likelihood of
accident involvement.

2. Methods
2.1. Materials
In order to develop a questionnaire on decision-making style, a number of ways in
which the decision process might differ were generated. Some of these corresponded
to dimensions used in questionnaires already in existence. Thus Johnson's
internal/external distinction and the compliant and dependent dimensions were
represented by items about preference for consultation and advice and susceptibility
to social pressure. The emotional and logical versus intuitive categories were
reflected in questions about the role of feelings in decision-making.
Questions were also added asking about locus of control (Rotter 1966). In
addition, examination of normative theories of decision-making (Simon 1957, Janis
and Mann 1977, Edwards 1954) led to consideration of depth of search through
consequences of courses of action, breadth of search through different options,
degree of planning ahead, use of principled versus pragmatic solutions, use of
satisficing versus optimizing decision rules, risk acceptance, degree of certainty about
a decision, and level of commitment to it.
Several versions of a questionnaire containing items relating to these dimensions
were tested and items deleted or modified according to whether subjects appeared to
be able and willing to answer them consistently and informatively.
This resulted in a 30-item Decision-Making Questionnaire (DMQ). DMQ items
were phrased as questions asking about frequency of a given type of behaviour and
subjects were instructed to tick one of six boxes that indicated that they behaved in
this way: never or very infrequently, infrequently, quite infrequently, quite
frequently, frequently, very frequently or always.
A questionnaire about driving style was also developed. The choice of items for
inclusion was based on behaviours that had previously been shown, or were
suspected, lo be related to accident involvement or risky driving behaviour. These
were speed (Wasielewski 1984), headway (distance to the car in front-Evans and
Wasielewski 1983), seat belt use (Evans et al. 1982), gap acceptance (size of gap in the
flow of traffic before attempting to pull out-Bottom and Ashworth 1978), and traffic
light violations (Koneci et al. 1976). In addition, the questionnaire included items
about behaviours thought to be directly related to decision-making style. For
example, specific questions about feeling in control when driving, some of which
were based on Montag and Comrey's (1987) scale of driving internality and
externality which proved to have a reliable factor structure, were included. Questions
were also asked about reactions to advice when driving, route planning, and risktaking on the road. Responses were on the same six-point scale of frequency as the
decision-making questionnaire.

630

D. J. French et al.

The DMQ and driving style questionnaires were put together in a package which
also included questions about biographical details, miles driven annually and
accident involvement during the previous year. Data on accident involvement for up
to three years prior to this was available from a questionnaire previously completed
by the same subjects (see below).
2.2. Subjects and procedures
The questionnaire package was sent to 980 drivers in all parts of the UK. The sample
was a subset of 30 000 drivers who had previously been randomly selected from those
licensed at the Driver and Vehicle Licensing Agency (DVLA) Swansea to take part in
a questionnaire survey conducted by National Opinion Polls for the Transport and
Road Research Laboratory. In the survey, they gave details of their driving habits
and accident records for the past three years, as well as personal details, and at the
end indicated whether they would be willing to help with further research. The data
gathered during this exercise provided information necessary for stratification ofthe
sample selected for this study by age, sex, annual mileage and accident involvement.
Thus our sample was not representative of the population at large, but it enabled us
to take account of the relationship of exposure, age, and sex with accident rates.
Table 1 shows the stratification of the sample.
Males24
15
25
16
60+
25-59
50
No
Low
accid.
50
50 1.=of1-10765
2:
2:1
1Table
accid.
low
mileage
Stratification
the sample.
miles p.a.
Age
18-24
Females-Iow
high
mileage
mileage>
10765
miles
= 1-4848
miles p.a.
p.a.miles p.a.
high
mileage>4848

Completed questionnaires were returned by 711 drivers, 73% of those who were
sent one. There was a tendency for a slightly higher response rate among older
subjects (chi squared = 714 with 2 degrees of freedom, p<005), but otherwise there
were no differences in response rates between different cells of the stratification. In
the analyses described below, degrees of freedom may be reduced in certain cases
because of missing data.
2.3. Statistical analyses
The analysis was carried out in six stages:
(1) assessment of the internal coherence and derivation of subscales for the
Decision-Making Questionnaire (DMQ);
(2) investigation of relationships between DMQ scores and age and sex;

631

Decision-making and RTAs

(3) assessment of the internal coherence and derivation of subscales for the
Driving Style Questionnaire (DSQ);
(4) investigation ofrelationships between DSQ scores and age and sex;
(5) investigation of relationship between decision-making and driving style and
accident rates;
(6) modelling of the role of decision-making style and driving style in the
causation of accidents using multiple regression.
Some of the variables used in the analyses did not conform to a normal distribution
therefore differences and correlations were analysed using both parametric and nonparametric methods, with near identical results. For the purposes of presentation
parametric statistics are reported throughout. In tests of significance, two-tailed tests
were used unless otherwise indicated.
3. Results
3.1. Internal coherence and derivation of subscale scores for the DMQ
A principal components analysis of the DMQ yielded seven components with
eigenvalues greater than one. Together they accounted for 543% of the variance.
These seven independent dimensions were interpreted by examining the contents of
the variables loading highly upon them. Table 2 shows the item loadings on each of
the components after varimax rotation. A criterion of 04 as used as the minimum
loading for an item to be incorporated in a dimension. Nine of the 30 items did not
meet this criterion for any dimension and were excluded from further analysis. The
Table 2. Items loading on the dimensions of decision-making style.
Factor

Loadings

Items

I. Thoroughness
(179% of variance)

079
-070
065
064
076
075
072
-058

Do you work out all the pros and cons?*

Do you decide without considering all the implications?*
Do you plan well ahead?
Is your decision-making deliberate and logical?*
Do you remain calm?*
Do you enjoy making decisions?
Do you feel in control of things?
Do you avoid making decisions if you can?*
Do you find it difficult to think c1early?*
Do you favour first one option then another?*
Do you change your mind about things?
Do you take the safe option if there is one?
Do you avoid taking advice over decisions?
Do you like to consult with others?
Do you make up your own mind about things?*
Do you carry on looking for something better?*
Do you settle for an option that will just about do?*
Are practicalities more important than principles?*
Are your decisions governed by your ideals?*
Do you rely on 'gut feeling' when making decisions?
Do you stick by your decisions come what may?

2. Control
(95% of variance)

-045
3. Hesitancy
(79% of variance)
4. Social resistance
(65% of variance)
5. Perfectionism
(58% of variance)
6. Idealism
(52% of variance)
7. Instinctiveness
(50% of variance)
*Item abbreviated.

070
057
055
070
-057
055
070
069
080
-069
067
061

D. J. French et al.

632
FIG 1a CONTROL

30

SUBSCALE

FIG 1b THOROUGHNESS

SCORE

SUBSCALE

25

SEM

28

23

26

21

24

SCORE

SEM

19

//*22

"

,+----+-/

//"

/-..j</

15

-+

MALES

FEMALES

SCORE

-+

MALES

FEMALES

SUBSCALE

SCORE

15~----~~~~~~---~~--~,

SE.M.

SEM

13

13

"

11

FIG 1d SOCIAL RESISTANCE

15 -,~~-~--~~--~--~---~-

.+

AGE CATEGORY

FIG 1c HESITANCY
SUBSCALE

'

AGE CATEGORY
-

/+

-+//

17

20

//

~--+

......--1-<-

'~/

-- --~+

11

/+---+
~
9

f-

AGE CATEGORY
-

MALES

-+

FEMALES

"'+'/

""..--

AGE CATEGORY
-

MALES

-+

FEMALES

remaining 21 items were used to produce scores for each of the seven dimensions of
decision-making style by reversing the scoring for items with negative loadings and
summing the ratings for items contributing to each dimension.
As a check on the stability of the factor solution, we carried out separate principal
components analyses on two randomly-drawn subsampIes, each comprising half the
overall sample. The results indicated the same pattern ofloadings in both subsamples

633

Decision-making and RTAs

FIG 11 IDEALISM

FIG 1e PERFECTIONISM
SUBSCALE

SUBSCALE

SCORE

15~------

SCORE

15 ~,------~-----------~

S.EN

13

13

11

11

S.EM

~--~5

5'~-~--~--~-~--~--~-~
3

-+

MALES

AGE CATEGORY

AGE CATEGORY
-

/:'><

FEMALES

MALES

-+

FEMALES

FIG 19 INSTINCTIVENESS
SUBSCALE

15~~-----

SCORE

SEM

13

11

//+-

, ---+----+---+//~

~.

AGE CATEGORY
-

MALES

-+. FEMALES

Figure 1. Decision-Making Questionnaire

(DMQ) subsca1e scores broken down
by age group and sex. Age group
1=17-19;
2=20-29;
3=30-39;
4=40-49; 5=50-59; 6=60 and over.

and these were the same as were found in the analysis that had been carried out on the
full sample. The distributions of scores on each of the seven subscales were examined.
They were distributed approximately normally with a good spread of values.
3.2. Relationships between DMQ scores and age and sex
Each dimension of decision-making style was correlated with age using Pearson's R.

634

SD
014
-091
mean
0172'79**
76
015
016
222
11-3
79
015062
016
182
-108
94
177
value
for
t112
018
017
-076
67
019
019020
013
349***
234
-014***
sex
015
difference
mean
95
020***
Females
scores
66
72
017***
79
015***
008
Males
scores
009**
D. J. French
et al. with age and sex.
(n=366)
(n=343)
(n=709)
Relationships
of
decision-making
dimensions
Correlation

*p<005

**p<OOl

***p<OOO1.

Table 4. Items loading on the dimensions of driving style.

Factor
1. Speed
(183% of variance)
2. Calmness
(125% of variance)
3. Social resistance
(94% of variance)
4. Focus
(87% of variance)
5. Planning
(73% of variance)
6. Deviance
(69% of variance)

Loading

Item

081
079
078
081
-069
068
-082
081
070
069
061
-082
079
084
060

Do you break the motorway speed limit?*

Do you drive fast?
Do you exceed the speed limit in built up areas?
Do you become flustered?*
Do you remain ca1m?*
Do you respond to pressure from other drivers?*
Are you happy to receive advice from peop1e?*
Do you dislike people giving advice?*
Do you drive cautiously?
Do you find it easy to ignore distractions?*
Do you ignore passengers?*
How often do you set out without looking at a map?*
Do you plan long journeys in advance?*
Do you overtake on the inside?*
Do you ever drive through a red traffic 1ight?*

*Item abbreviated.

Sex differences were examined using t-tests. The results are given in table 3. Figure I
shows the mean scores on each dimension in various age categories for the males and
females.
Control increased with age, the greatest increase being between the under 20s and
over 20s. Males scored significantly higher than females overall, the difference
disappearing at 60 years (figure Ia). There was a steady increase in thoroughness
between late teens and 30s with no discernible change thereafter (figure Ib). There
was no sex difference. Hesitancy showed a small negative correlation with age. The
group means showed that there was a decline up to the age of 30 with no change
thereafter (figure le). Social resistance showed a very small positive association with
age and there was no sex difference (figure Id). Perfectionism showed a negative
correlation with age brought about by a fall after the age of fifty (figure le). Idealism
showed a significant positive correlation with age, the main difference being between
those under and over fifty (figure If). Instinctiveness showed a positive correlation
with age caused by a rise after the age of 50, and a sex difference, with females being
more instinctive than males (figure 19).

635

Decision-making and RTAs

FIG 2b CALMNESS

FIG 2a SPEED

15

SUBSCALE

SUBSCALE

SCORE

20

SEM

13

18

11

16

+-,

,,,
,

"+----+

'I

SCORE

------------------~

SE.M.

14

-v--

+----+..,

",+
__

-+

MALES

FEMALES

S.E.M

13

18

11

16

_'_ __

_'__ _

__1

-+. FEMALES

MALES

SUBSCALE SCORE
20,--------------------,

SCORE

15

FIG 2d FOCUS

FIG 20 PLANNING
SUBSCALE

__L __

AGE CATEGORY

AGE CATEGORY
-

--,+--

_-I"

12

10 ~I--~--~-~

-+---+-

/ /,

SEM.

-+----+

"'

14

+-_---1"

*---+-- - //

+---- //

~-+

12

10

MALES

-+. FEMALES

AGE CATEGORY

AGE CATEGORY
-

MALES

-+

FEMALES

3.3. Internal coherence and derivation of subscale scores for the DSQ
A principal components analysis of the driving style questions revealed six
components accounting for 394% of the variance. The six dimensions were
interpreted as speed (made up of items about driving fast and exceeding the speed
limit), calmness (items about staying calm in dangerous situations and when there is
little time to think), planning (consulting a map and planning places to stop and rest

D. J. French et al.

636

FIG 21 DEVIANCE

FIG 2e SOCIAL RESISTANCE

15

SUBSCALE

SUBSCALE

SCORE
r-------------------,

10

SCORE
r-------------------~
I
S.EM

S.E.M.

13

11

......

, --

+-~---+.-

"-

-+-

~-+-- -r-- _ _..J......_

,,

-r- __ _..I

--

" ---+
I

--'2

o~' -~-~-~-~~-~-~3

MALES

-+

FEMALES

AGE CATEGORY

AGE CATEGORY
-

__

MALES

-+

FEMALES

Figure 2. Driving Style Questionnaire (DSQ) subscale scores broken down by age group and
sex. Age group 1= 17-19; 2=20-29; 3=30-39; 4=40-49; 5=50-59; 6=60 and over.

before setting out), focus (driving cautiously and ignoring distractions), social
resistance (disliking being given advice about driving), and deviance (jumping the
lights and overtaking on the inside). Table 4 shows the dimensions, the percentage of
variance that they accounted for and the item loadings. As with the DMQ, subscale
scores were calculated for driving style dime.,,:sionsby reversing the coding of ratings
with negative loadings and then adding up the items with loadings greater than 04.
Also, as with the DMQ, separate principal components analyses were carried out on
two randomly-selected subsamples each comprising half of the full sample. The
resulting loadings were similar in both subsamples and both were similar to the
results of the analysis on the full sample, indicating a stable structure.
3.4. Relationships

between DSQ scores and age and sex

The relationship between each driving style dimension and age and sex was tested
using Pearson's R correlations to investigate changes with age and t-tests to look at
the sex differences. In the case of deviance and planning scores, we also examined
relationships in which these variables had been recoded into dichotomies to avoid
the problem of undue influence of extreme scores. The results were nearly identical to
those using the untransformed variables so, for the sake of presentational
consistency, we report the simple Pearson correlations here. The mean-scores for
each age group of males and females on each dimension were examined in order to
establish the nature of the relationships. Figure 2 shows the results graphically. Table
5 shows the correlations with age, and the sex differences in the driving style
dimensions.
Speed showed a steady decrease with age in both sexes and females scored lower

Decision-making and RTAs

637

than males at all ages (figure 2a). Calmness showed a positive correlation with age
which was brought about by a rise in the 50s and over. Females reported themselves
to be less calm than did males (figure 2b). There was no sex difference in scores on
planning and the positive correlation with age was largely accounted for by an
increase in scores between the 20s and 30s with little change thereafter (figure 2c).
Scores on focus showed a small positive correlation with age, a more or less linear
effect, whilst females reported being more focused than males (figure 2d). Social
resistance in a driving context showed no sex difference. The negative correlation
with age was significant but small (figure 2e). Self reports of deviant behaviour
showed a significant decrease with age (figure 2t). There was no sex difference.

IH

SD
27
1409
37
39
28
31
29
384
73
-24
051
*102
91
Table
5.
Female
scores
1-3
27
63131
mean
mean
136
-04
72
016***
-004n.s.
91
027***
Male
n.s.
28
147
017*** style dimensions with age and sex.
***
-0'12**
-035***
-018***
(n=366)
(n=343)
(n=709)
Sex
difference
(t) scores
Relationship
of
driving
Correlation

*p<005

**p<O'Ol

***p<OOOl.

3.5. Relationships between DMQ, DSQ, and accident liability

Relationships between the DMQ subscale scores and dimensions of driving style
were investigated using Pearson's R correlations. Table 6 shows that thoroughness
correlated significantly with all six dimensions of driving style, control correlated
with all but speed and deviance, and hesitancy correlated with calmness and
planning. Social resistance in the DMQ correlated with social resistance and
planning in the DSQ. Perfectionism in the DMQ correlated with calmness in the
DSQ and instinctiveness in the DMQ with focus in the DSQ.
To reduce the skew in the accident rate distribution, accident rates were recoded
as 0, I or
1. The results of Pearson correlations using this recoded variable were
very similar to those of non-parametric
correlations.
Before assessing the extent to which the psychological variables were associated
with accident liability, we wished to investigate the stability of the accident liability

>

Calmness
Soc.
Focus
res. -000
001
003
002
007
005
003
001
001
-010**
-011
-002
-003
Deviance
0'12**
-001
009*
004
-007
021***
045***
001
-015***
017***
041
004
0003
06
**
***
-005
-001
-025***
-001
-002
-017***
-000
-0'11**
006
Table -005
6.
021***
0'13***
023***
-015***
**p<OOl
***p<OOOl.
Correlations
between dimensions of decision-making and driving style.
Planning
Speed

638

D. J. French et al.

measure. The annualized accident rate for the three years prior to our study (the total
number of accidents divided by the reporting period in years), as determined by the
NOP survey which preceded ours (see section 2), was correlated with the rate during
the following year as assessed by our postal questionnaire. This yielded an R of O 305
(p<000l). This gives an estimate of the extent to which accident rates are a reliable
indication of underlying accident liability and sets an approximate upper limit of
around 9% on the percentage of variance in our recorded accident rates that we can
predict with psychological measures.
The possibility that the correlation for the annualized rates between the two
periods over which accidents were recorded was artificially inflated by covariance in
the annual mileage was considered but annual mileage was not, in fact, related to the
annualized accident rates for the first three years because of the stratification of our
sample by mileage and accident involvement (R=OOl).
Table 7 includes correlations between DMQ subscale scores and accident rates.
Two dimensions of decision-making style correlated to a small but significant degree
with annualized accident rates calculated over all four years; they were thoroughness
and instinctiveness.
Table 7. Correlations between decision-making style and driving style dimensions and
accidents per year, recoded 0, 1, > 1.
-005
-011
-003
-006
-002
006
-001
-006
011*
-002
-006
022***
001sample
-007
000
-004
-006
011
002
R:
008*
over
60s
003
008*
004
-018*
-009*
-012**
017**
-010**
022***
-009**
-001
002
-002
004
-001
0'18*
(n=
130)
(n=575)
R:
whole
**p<OOl
***p<OOO1.
(n=
575)

R: 60s and under

nd under

Table 7 also shows the correlations obtained between the driving style dimensions
and accident rates. Speed was the most important predictor of accident rates but
planning and deviance also correlated significantly.
3.6. Modelling causes of accident liability
In order to ensure that all possible confounding between predictor variables was

eliminated, a forced entry multiple regression was carried out in which all those
variables that had shown significant bivariate correlations with accident rates were
entered as possible predictors.

Decision-making

639

and RTAs

Fig 3a Drivers aged 60 and under

~
-.12 p<.02

Thorough I '"

--+

-.24 p<.001 . ~
~

.16/

.15
p<.

001

-.20 p<.001

c:J
Fig 3b Drivers aged over 60

G
S..
r-:::I---. E

':;:P<OI

esltant

.25 p<.OI

Speed

20p<.05

.18
~P<.05

Figure 3. Models of accident liability in the postal survey sample. Figure 3a shows the model
for drivers under 60 years. Figure 3b shows the model for drivers over 60 years. Each path
is accompanied by the partial standardized regression coefficient and associated
significance level.

It turned out that, because of the stratification, sex did not correlate significantly
with accident rate. However, in the case of age, there was a relationship and it was
clearly non-linear in our sample. Accident rates decreased with age up to 60 years
(R=-0'15,
p<OOl) and then increased slightly (R=OlO). This required us to
consider those up to 60 and those over 60 separately because when partialling out age
effects, a simple linear regression model would not have been appropriate. This is
why table 7 shows the correlations of DMQ and DSQ scores with accident rates
separately for these age groups. In the younger age group, despite our stratification,
there was a significant relationship between annual mileage and accident rates
(R=O13, p=O003). There was no relationship in the older age group (R=003).
We carried out conventional forced entry multiple regression analyses for the two

D. J. French et al.

640

Table 8. Correlations of age, sex, mileage, and DMQ subscale scores with speed in 60s and
under and over 60s.
<=60
Thoroughness
Hesitancy
Resistance
Perfectionism
Idealism
Instinctiveness
Age
Sex

Mileage
*p<005

yrs

-028**
-001
001
001
002
005
-027**
-026**
022**

>60 yrs

-004
-006
008
003
-011
025*
-002
015
017

**p<OOl.

age groups separately in which all variables with significant simple correlations with
accident rates were entered together. The predictor variables were speed, deviance,
planning, thoroughness, instinctive ness, age, and annual mileage in the 60s and
under, and speed, thoroughness, hesitancy, and age in the over 60s. In the 60s and
under, both age and speed independently predicted accident rates (see figure 3a). In
the over 60s, speed, thoroughness, and hesitancy all independently predicted
accident rates (see figure 3b). Because of the importance of speed as a predictor of
accident rates, we examined the relationships between DMQ subscale scores and
speed in the 60s and under and the over 60s separately (table 8). In the younger age
group, age, sex, annual mileage, and thoroughness correlated with speed. In the older
age group, only instinctiveness correlated with speed. To examine further the
predictors of speed in the younger age group, we carried out a forced entry multiple
regression analysis in which age, sex, mileage, and thoroughness were entered
together. As shown in figure 3a, all these variables made independent contributions
to predicting speed.
4. Discussion
The 21 items of the DMQ were shown to comprise seven independent dimensions of
decision-making style. The distributions of scores on these subscales were
approximately normal and showed a good spread across subjects. The dimensions
were labelled thoroughness, control, hesitancy, social resistance, perfectionism,
idealism, and instinctiveness.
Fifteen items on the Driving Style Questionnaire were found to comprise six
dimensions labelled speed, calmness, planning, focus, social resistance (driving), and
deviance. Thoroughness correlated significantly with all dimensions of driving style.
Among a large set of significant correlations the two largest were between
throughness in decision-making style and planning in driving style and between
control and calmness in driving style.
Dimensions of both driving style and decision-making style showed significant
correlations with accident rates. These were the driving style dimensions of speed,
planning, and deviance and the decision-making styles of thoroughness and
instinctiveness in those up to 60, and thoroughness and hesitancy in the over 60s.
Multiple regressions showed that in the 60s and under, speed was directly predictive
of accident liability taking account of all other factors, whilst the effect of

Decision-making and RTAs

641

thoroughness appeared to be mediated by speed. In the over 60s, speed,

thoroughness, and hesitancy all played an independent role, and speed was predicted
by instinctiveness.
Part of the purpose of this study was to obtain a preliminary assessment of the
concept of decision-making style. We found that respondents to a self-report
questionnaire about their decision-making style gave answers which revealed quite a
large number of different dimensions. This contrasts with the theories of Harren
(1979), Arroba (1977) and Johnson (1978). It is possible that there may be more than
the seven which emerged from our questionnaire, but on our data there are unlikely
to be fewer. The difference between our results and those of the studies emerging
from the previous theories is probably due to the fact that the DMQ canvassed a
much more diverse set of possible behaviours than did the previous questionnaires.
One of the major dimensions which emerged from the DMQ corresponded to
important aspects of the previous theories: the thoroughness dimension is similar to
the rational styles of Harren and Arroba and the systematic style of Johnson.
However, there can be little doubt that this is only one of many aspects of decisionmaking.
These data represent only the first step of a process of development and
validation of both the concept of decision-making style and its measurement. It is
one thing for self-reports of decision-making behaviour to be explicable in terms of a
set of dimensions such as we have proposed. It is another to show that the behaviour
itself can be thus explained.
According to our drivers' self-reported behaviour they did carry their global
decision-making style into the driving situation. This offers the possibility that
important aspects of driving behaviour can be understood in terms of more generally
applicable traits that drivers may have. At a theoretical level it provides a way of
linking theories of driver behaviour with more general theories of cognition and
motivation. At a practical level it offers the possibility that aspects of driver
behaviour can be indexed in situations, from the respondent's point of view,
unrelated to driving. This may turn out to be useful if assessment of driver
characteristics becomes used in real-life settings which have implications for
individual drivers.
It is only possible to speculate at this stage what might underlie the relationships
between thoroughness and preferred driving speed in drivers of 60 and under. One
obvious possibility is that low thoroughness is a reflection of a more global trait of
impatience. It is not difficult to see how impatience might lead people to drive faster.
One aspect of the Type A behaviour pattern involves a feeling of being under
continual time pressure and this may too be related to impatience. In this regard it is
interesting that one study has examined the relationship between Type A behaviour
and accident liability and found a positive relationship (Perry 1986). To the extent
that extroversion and sensation-seeking also involve impatience, this factor may also
explain association between these variables and accident liability (Pestonjee and
Singh 1980). Further research along these lines will be needed to dissect out the key
variable or variables here.
The direct association of hesitancy and thoroughness with accident rates in the
over 60s must be mediated by some aspects of driving style which we have not
measured (or not measured adequately). It may be that for drivers of this age, ability
to process information rapidly is impaired (Rabbitt 1991) and that if they do not
allow for this by being more deliberate in their decision-making, or if it results in

642

D. J. French et al.

them beginning manoeuvres and changing their mind this could raise their accident
risk. As regards the relationship between instinctiveness and speed in the older
drivers, it may be that a greater feeling ofthe infallibility of one's intuitive judgement
could cause drivers to fail to appreciate the risks they run by driving at higher speed.
Our observation of a link between self-reported driving speed and accident rates
is consistent with the findings of other studies (Engel and Thomsen 1988, Wilson and
Greensmith 1983). Accident rates are a very imprecise measure of accident liability
as demonstrated by the low correlation between accident rates over two successive
periods. In fact the correlation between speed and accident rates was not much less
than the correlation between accident rates measured over two periods. Thus the
importance of speed as a contributory factor to accident liability cannot be
overstated; faster drivers are relatively unsafe drivers. Our self-reported dimension
of speed has been shown in another study to correlate quite well with actual driving
speed. In the study concerned (West et al. 1993), the self-reported speed dimension
of the DSQ correlated 055 with the judgements of drivers' preferred speed made by
two independent in-car observers during a mixed motorway and urban test route. In
that study observed speed also correlated positively with accident rates. Thus, the
relationship between speed and accident liability is consistent and is not an artifact
of our self-report methodology.
The most plausible explanation for the relationship between driving speed and
accident liability is that a direct causal link exists between the two. Minor
misjudgements of distance or timing and unexpected hazards such as patches of ice
on the road are turned into accidents because of excess speed. However, there are
other possibilities which cannot be ruled out. It may be, for example, that faster
drivers engage in particular driving habits which put them at risk. For example, they
may pull out into smaller gaps than other drivers or attend less to the driving task.
Further research is needed to unravel the various possibilities. This may involve
obtaining further data about aspects of driving that are associated with speed, finding
out whether there are certain kinds of accidents to which faster drivers are
particularly susceptible, and assessing whether interventions directed at individual
drivers to reduce their speed reduces their accident rates.
Although deviant driving behaviour correlated significantly with accident rates,
in the multiple regression analysis it was not shown to play an independent role.
Current research indicates that a substantial number of accidents do occur as a direct
result of deviant driving (Reason et al. 1991). However, in our sample they were
relatively uncommon. This may have been due to deviant drivers being less likely to
volunteer for the study. It is also possible that the deviant drivers were unwilling to
report their accidents. Finally, it is possible that there was under-reporting of deviant
driving behaviour. All these factors would act to attenuate the relationship between
self-reported deviant driving and accident rates. The possibility remains, however,
that in the totality of accident statistics, excess speed is a more important factor than
dangerous manoeuvres and that to the extent that traits of fast driving and deviant
driving are separable, fast driving may play a greater role.
Questionnaire methods such as have been used in this study can only be expected
to provide a broad indication of relationships between variables of interest. They
depend on respondents being able to form impressions of their own behaviour and
communicate these using a fixed response format. Thus error of measurement is
likely to limit severely the size of associations. Most of the correlations reported in
this study were low, but reached high levels of statistical significance by virtue ofthe

Decision-making and RTAs

643

large sample size. Thus, while the pattern of associations point the way towards
potentially important relationships and ultimately to a theory of accident causation,
for the practical prediction of accident rates, it will be necessary to develop and refine
the measures used, perhaps supplementing self-report questionnaires with other
forms of behavioural assessment. We have already mentioned the use of direct
observation of dl-iver behaviour. We are also examining ways of using computerized
tasks and presentation of real and hypothetical decisions as a means of assessing
decision-making style.
There is the danger when questionnaire responses are correlated with other
questionnaire responses, that at least part of the associations found reflect
consistency in response biases rather than underlying psychological dimensions of
interest. It is difficult to envisage how response biases could lie behind the particular
pattern of associations we found. The relatively large number of independent
dimensions of decision-making and driving style, and the relationships between
particular dimensions and accident rates would require a much more elaborate
conception of response biases than has been proposed in the literature.
In conclusion, this study has provided preliminary evidence on a new
questionnaire assessing decision-making style indicating that at least seven
independent dimensions can be isolated. One of these dimensions, which we have
labelled thoroughness, correlated significantly with accident rates. In drivers of 60
and under, this relationship appeared to be mediated by a single aspect of driving
style, preferred driving speed. The results suggest several further lines of
investigation which would help to provide a clearer understanding of why some
drivers have more accidents than others.

Acknowledgement

The work described in this paper was carried out under a contract placed with Royal
Holloway and Bedford New College by the Transport and Road Research
Laboratory, Crowthorne, Berkshire and the paper is published by permission of the
Director. The views expressed in this paper are not necessarily those of the
Department of Transport.
References
ARRoBA,T. 1977, Styles of decision-making

and their use: an empirical study, British Journal of

Guidance and Counselling, 5, 149-158.
BOTTOM, C. G. and ASHwoRTH, R. 1978, Factors affecting the variability of driver gap
acceptance behaviour, Ergonomics, 31, 721-734.
BROWN, 1. D. and GROEGER, J. A. 1988, Risk perception and decision taking during the
transition between novice and experienced driver status, Ergonomics, 31, 585-597.
BUCK, J. and DANIELS, M. 1984, Assessment of Career Decision-Making Scale (Western
Psychological Press, Los Angeles).
EDWARDS,W. 1954, The theory of decision-making,
Psychological Bulletin, 51, 380-417.
ENGEL, U. and THOMSEN,L. K. 1988, Speeds, speed limits and accidents, Technical Report
3/1988 to the Danish Council of Road Safety Research, Copenhagen.
EVANS,L. and W ASIELEWSKI,
P. 1983, Risky driving related to driver and vehicle characteristics,

Accident Analysis and Prevention, 15, 121-136.

EVANS,L., WASIELEWSKI,
P. and VONBUSEK,C. R. 1982, Compulsory seat belt usage and driver
risk-taking behaviour, Human Factors, 24, 41-48.
FISHBEIN,M. and AJZEN, 1. 1975, Attitude, Intention, and Behaviour: An Introduction to Theory
and Research (Addison-Wesley, Reading, MA).

644

Decision-making and RTAs

GOLDSTEIN,
L. G. 1961, Research on human variables in safe motor vehicle operation: a
correlation summary of predictor variables and criterion measures, The Driver
Behaviour Research project, George Washington University.
GORDON,V., COSCARELLI,
W. C. and SEARS,S. 1986, Comparative assessments of individual
differences in learning and career decision-making, Journal of College Student Personnel,
27, 233-242.
HARREN,V. A. 1979, A model of career decision-making for college students, Journal of
Vocational Behaviour, 14, 119-133.
HESKETH,
B. 1982, Decision-making style and career decision-making behaviours among school
leavers, Journal of Vocational Behaviour, 20, 223-234.
JANIS,1.and MANN,L. 1977, Decision-Making (Free Press, New York).
JONAH,B. A. 1986, Accident risk and risk-taking behaviour among young drivers, Accident
Analysis and Prevention, 18, 255-271.
JOHNSON,
R. 1978, Individual styles of decision-making: a theoretical model for counselling,
Personnel and Guidance Journal, 56, 530-536.
JOHNSON,
R., COSCARELLI,
C. and JOHNSON,
J. 1983, The Decision-Making Inventory (Marathon
Consulting and Press, California).
KONECI,c., EBBESEN,
E. B. and KONECI,D. K. 1976, Decision processes and risk-taking in
traffic: driver response to the onset of yellow light, Journal of Applied Psychology, 6,
359-367.
MAYHEW,
D. R., WARREN,
R. A., SIMPSON,
H. M. and HAAS,G. C. 1981, Young Driver Accidents:
Magnitude and Characteristics of the Problem (Traffic Injury Research Foundation of
Canada, Ottawa).
MONTAG,1.and COMREY,
A. L. 1987, Internality and externality as correlates of involvement in
fatal driving accidents, Journal of Applied Psychology, 72, 339-343.
PERRY,A. R. 1986, Type A behaviour pattern and motor vehicle drivers' behaviour, Perceptual
and Motor Skills, 63, 875-878.
PESTONJEE,
D. M. and SINGH,U. B. 1980, Neuroticism-extraversion as correlates of accident
occurrence, Accident Analysis and Prevention, 12, 201-204.
QmMBY,A. R., MAYCOCK,
G., CARTER,1. D., DIXON,R. and WALL,J. G. 1986, Perceptual
abilities of accident involved drivers, Transport and Road Research Laboratory Report.
RABBIIT,P. 1991, Factors promoting accidents involving elderly pedestrians and drivers, in G.
Grayson and J. Lester (eds) Behavioural Research in Road Safety (TRRL, Crowthorne),
167-183.
REASON,
J. T., MANSTEAD,
A. S. R., STRADLING,
S. G., PARKER,D. and BAXTER,
J. S. 1991, The
social and cognitive determinants of aberrant driving behaviour, TRRL Contractor
Report 253, Crowthorne.
ROITER, J. 1966, Generalized experiences for internal versus external locus of control of
reinforcement, Psychological Monographs: General and Applied, 80, 1-27.
WASIELEWSKI,
P. 1984, Speed as a measure of driver risk: observed speeds versus driver and
vehicle characteristics, Accident Analysis and Prevention, 16, 89-102.
WEST,R., FRENCH,D., ELANDER,J. and KEMP,R. 1993, Direct observation of driving, selfreports of driver behaviour and accident involvement, Ergonomics, 36, 557-568.
WILSON,R. and GREENSMITH,
J. 1983, Multivariate analysis of the relationship between
drivometer variables and drivers' accident, sex, and exposure status, Human Factors, 25,
303-312.
Received 12 October 1990.
Final revison accepted 8 November 1991.