Transportation Research Part F 7 (2004) 77–94

www.elsevier.com/locate/trf

Speed regulation by in-car active accelerator pedal

Effects on driver behaviour
almdahl *, Andr

Magnus Hj€ as V

arhelyi
Department of Technology and Society, Lund University, John Ericssons vag 1, 221 00 Lund, Sweden
Received 17 February 2003; received in revised form 13 February 2004; accepted 18 February 2004

Abstract

The long-term effects of driving with an active accelerator pedal on driver behaviour were studied by
using an in-car observation method over the period beginning 2000 until 2001. The system produced a
counterforce in the accelerator pedal when the speed limit was reached, but could be overridden by pressing
the accelerator pedal harder. Twenty-eight drivers were studied when driving without the system and then
when driving with the system after they had used it in their own cars for at least six months. The results
showed that their behaviour towards other road users improved, they had a yielding behaviour correct to a
higher degree and were more likely to give pedestrians the right of way at zebra crossings when driving with
the active accelerator pedal. It was also found that the time gap to the vehicle in front increased slightly
with the system. There were also signs of negative behavioural modifications in the form of drivers for-
getting to adapt their speed to the speed limit or the prevailing traffic situation when they were not sup-
ported by the system and in low speed areas; these effects, however, were not statistically significant.
Together with studies showing improved speed behaviour, the results of this study augur well for great
safety effects of the system.
Ó 2004 Elsevier Ltd. All rights reserved.

Keywords: Active accelerator pedal; In-car observations; Driver behaviour; Instrumented vehicle

1. Introduction

The strong relationship between the speed level and accidents has been shown in several studies;
small speed level changes result in significant changes in the number of accidents (see e.g. Elvik,

*
Corresponding author. Tel.: +46-46-222-48-24; fax: +46-46-12-32-72.
E-mail address: magnus.hjalmdahl@tft.lth.se (M. Hj€
almdahl).

1369-8478/$ - see front matter Ó 2004 Elsevier Ltd. All rights reserved.
doi:10.1016/j.trf.2004.02.002
78 M. Hj€almdahl, A. V
arhelyi / Transportation Research Part F 7 (2004) 77–94

€
Vaa, & Ostvik, 1989; Finch, Kompfner, Lockwood, & Maycock, 1994; Nilsson, 1982; Salusj€arvi,
1981). An additional finding is that lower speed variance is correlated with fewer accidents (see
e.g. Finch et al., 1994; O’Cinn
eide & Murphy, 1994; Salusj€arvi, 1981). These observations indicate
that lower and more even speeds effectively reduce the number of accidents and mitigate the
outcome of collisions. However, compliance with speed limits is low, for example, a survey of
actual speed levels in Europe (Drask
oczy & Mocs
ari, 1997) showed that speeding is a common
phenomenon, especially widespread on urban roads and motorways. So far, speed management
has typically been concentrated on the road and its environment (engineering), and on the driver
(e.g. enforcement). Engineering measures such as traffic calming have proved to be effective in
reducing speed at isolated sites, but the effects are localised in time and space (for an overview see
Comte, V
arhelyi, & Santos, 1997). Visible enforcement at specified sites, directed towards certain
groups of drivers and administered regularly at times when many of the passing drivers are
commuters, usually results in an immediate speed reduction. Still, the extent of speed reduction in
time and space is very small (see e.g. Hauer, Ahlin, & Bowser, 1982; Østvik & Elvik, 1990; Teed,
Lund, & Knoblauch, 1993).
Recent technological advances in information and communications technology offer a much
greater flexibility and a broader possibility of influencing speed. Equipment in the road envi-
ronment or the vehicle itself may provide the driver with information on prevailing speed limits or
speed limits in critical conditions or warn him of speed errors or register inappropriate speed or
prevent the driver from exceeding the prevailing speed limit.
The concept of speed adaptation via in-car devices has been researched since the eighties. The
most promising system, i.e. tactile feedback via the accelerator pedal, has been most researched.
Earlier field trials (Almqvist, Hyd
en, & Risser, 1991; Almqvist & Nyg ard, 1997; Brookhuis & de
Waard, 1999; Carsten & Fowkes, 2000; Persson, Towliat, Almqvist, Risser, & Magdeburg, 1993;
Saad & Malaterre, 1982; Schulman, 1985; V
arhelyi & M€akinen, 2001) and simulator experiments
(Comte, 1996; 1998a; 1998b; Godthelp & Schumann, 1991) have demonstrated the positive effects
of such systems. These trials showed that the ‘‘intelligent accelerator pedal’’ had positive effects on
speed level, speed variance and car-following behaviour on urban roads. However, one study
reported a slight tendency to negative behavioural effects in the form of increased approach speeds
and turning speeds at intersections (Persson et al., 1993), while another reported smoother ap-
proach speeds and unchanged turning speeds at intersections (V
arhelyi & M€akinen, 2001).
Behaviour in interactions with other road users deteriorated according to Persson et al. (1993),
improved according to Almqvist and Nyg ard (1997) and it was unchanged according to V
arhelyi
and M€ akinen (2001). V
arhelyi and M€ akinen (2001) reported shorter time gaps in car-following
situations on rural roads. However, in these studies the test drivers did not have a chance to get
used to the system in everyday driving, except in one study (Almqvist & Nyg ard, 1997), where
they used the system for two months before being observed. Hence, no long-term adaptation to
the system could be observed. It can be assumed that this fact is the reason for some of the
conflicting results in these earlier studies, which raised the need for a more comprehensive
investigation. From these earlier trials it could be concluded that negative behavioural adapta-
tions and the long-term effects on driver behaviour in real traffic needed more research.
When it comes to the phenomenon of ‘‘behavioural adaptation’’, an OECD scientific expert
group was set up in 1990 (OECD, 1990) to examine the empirical evidence for its ‘‘existence’’.
Behavioural adaptation was then defined as ‘‘those behaviours which may occur following
 M. Hj€almdahl, A. V
arhelyi / Transportation Research Part F 7 (2004) 77–94 79

the introduction of changes to the road-vehicle-user system and which were not intended by the
initiators of the change’’. On the basis of a review of a large number of empirical studies, the
expert group concluded that ‘‘. . . behavioural adaptation to road safety programmes does occur
although not consistently’’. It was stated that the magnitude and direction of its effect could not be
precisely stated. The reviewed studies suggested that behavioural adaptation generally did not
eliminate safety gains from measures, but tended to reduce the size of the expected effects.
Behavioural adaptation can, in practice, appear in many different driving manoeuvres: in change
of speed, change of following distance, way and frequency of overtaking, way and frequency of
lane changing, late braking, change of level of attention, etc. (Drask

oczy, 1994). In hypothesising
and testing behavioural adaptation it is important to take into consideration the fact that it is an
effect that does not appear immediately when the driving context is changed, but usually appears
only after a familiarization period (ibid).
Mechanisms that can lead to behavioural adaptation with regard to a system like an intelli-
gent accelerator pedal are ‘‘delegation of responsibility’’, ‘‘behaviour diffusion’’, and ‘‘compen-
satory behaviour’’. Carsten (1993) describes ‘‘delegation of responsibility’’ as follows: ‘‘Studies
had shown that in situations people consider uncontrollable, they want to know who is
‘responsible’ for certain events. If other, generally more powerful, people assume responsibility,
it is not unusual to delegate responsibility to them. This delegation of responsibility can lead to
behaviour which is potentially more risky, e.g. in emergency situations where those at risk
should make their own decisions’’. In the case of vehicle-based systems for driver support the
driver might delegate the responsibility to the system. A driver supported by an intelligent
accelerator pedal is able to devote more attention to the other driving tasks. On the other hand
he might become over-reliant on the system. For example, the driver might consider that the
system will always know what the speed limit is and will always issue a warning at inappropriate
speeds. ‘‘Behaviour diffusion’’ might occur in situations where drivers are not supported by the
system, e.g. when driving outside the areas covered by the system, driving non-equipped vehicles
or when the system fails (Carsten, 1994). In these cases drivers who become totally reliant on the
system might have difficulty in following the changes in the actual speed limits. The notion of
‘‘compensatory behaviour’’ has its origin in the ‘‘risk compensation’’ theory of Wilde (1994) (the
notion that road users will use up some of the margin afforded by safety improvements by, for
example, driving faster) and the ‘‘risk homeostasis’’ theory of Wilde (1994) (the notion that road
users seek to keep their risk constant).
This study aimed at observing how driver behaviour was affected after long-term use of an
Active Accelerator Pedal (AAP), and was carried out within the framework of the Swedish Na-
tional Road Administration’s large-scale trial with Intelligent Speed Adaptation (1999–2002)
where 284 vehicles were equipped with the AAP in the city of Lund. The system was activated
automatically when the test vehicles were within the test area and could not be turned off. The
system was not activated outside the test area.
The AAP consists of a mechanical unit connected to the accelerator pedal, a digital map with
all the speed limits within the city boundaries and a GPS-navigator. When the driver tries to
exceed the speed limit a counterforce is activated in the accelerator pedal. However, if necessary,
the driver can exceed the speed limit by pressing the accelerator with approximately five times
more pressure than normally required.
80 M. Hj€almdahl, A. V
arhelyi / Transportation Research Part F 7 (2004) 77–94

1.1. Hypotheses

The hypotheses to be tested in this study have been derived from earlier studies on the effects
of the Active Accelerator Pedal on driver behaviour (Almqvist & Nyg ard, 1997; Persson et al.,
1993; V
arhelyi & M€akinen, 2001) and on theoretical considerations of possible effects of driver
assistance systems on driver behaviour. Drivers’ change in speed behaviour is, of course, of great
interest when equipment such as the AAP is being studied. The present study, however, is fo-
cused on how the AAP affects driver behaviour apart from speed. Nevertheless, in the discussion
chapter the AAP’s effect on speed behaviour, presented in more detail in Hj€almdahl, Almqvist,
and V
arhelyi (2002) and V
arhelyi, Hj€almdahl, Hyd
en, and Drask
oczy (in press), will also be
discussed.

1.1.1. Hypothesis 1
Drivers do not lower their speed in low speed situations to the same extent when driving with an
AAP as when driving without an AAP.
This hypothesis has its origin in the phenomenon of ‘‘compensatory behaviour’’ discussed
above. Speed adaptation is a significant safety indicator. Based on in-car observation studies
Risser, Teske, Vaughan, and Brandst€ atter (1982) analysed the correlation between the frequency
of errors of different types and the number of observed conflicts for individual drivers. They found
significant correlations between, on the one hand, the frequency of error types ‘‘speed badly
adapted to traffic situation’’ and ‘‘late or absent deceleration at intersections or on approaches’’
and, on the other, the number of observed conflicts. Carlqvist and Persson (1988), in an in-depth
study of accidents in the city of V€
axj€
o, identified the critical behaviour that led to these accidents.
They found that too high speed in relation to the environmental and interactional demands ac-
counted for 21% of the accidents.

1.1.2. Hypothesis 2
Behaviour towards other road users will improve (the drivers will be more willing to give
priority to other vehicles and pedestrians).
This hypothesis originates from the empirical findings that drivers entering an interactive
situation at lower speeds are more willing to give priority. Hyd
en and V
arhelyi (2002), in an
evaluation study of the effects of small roundabouts as speed reducing measures, concluded that
car drivers showed more consideration to vulnerable road users and pedestrians got priority
twice as often after the introduction of the roundabouts. They also showed that the number of
expected injury accidents for pedestrians decreased by 80%. Towliat (2001) showed that the
introduction of speed cushions as speed reducing measures improved drivers’ give-way behav-
iour towards pedestrians, and at the same time decreased the number of serious conflicts be-
tween these groups. Yielding behaviour was shown to be an important safety indicator.
Carlqvist and Persson (1988), in their in-depth study of accidents, found that erroneous yielding
behaviour accounted for 26% of the accidents. Carsten, Tight, Southwell, and Plows (1989)
found that failure to yield was one of the main driver failures leading to urban traffic accidents.
Drivers’ behaviour towards vulnerable road users is naturally an important safety indicator, but
it is also an indicator of the formers’ situation awareness, communication skills and, in many
countries, their law abidance.
 M. Hj€almdahl, A. V
arhelyi / Transportation Research Part F 7 (2004) 77–94 81

1.1.3. Hypothesis 3
The law abidance of drivers in AAP-cars can change in a positive or negative direction (fewer/
more stops at stop signs, larger/smaller proportion of driving against red).
This hypothesis has its origin in the notion of ‘‘behavioural adaptation’’ discussed above.
Behaviour at traffic lights was shown to be an important safety indicator as well. Risser et al.
(1982) found significant correlations between the error type ‘‘unlawful behaviour at traffic lights’’
and the number of observed conflicts for individual drivers.

1.1.4. Hypothesis 4
The time gap to the vehicle in front will increase.
This hypothesis also has its origin in the notion of ‘‘behavioural adaptation’’ discussed above.
Car-following behaviour is another important safety indicator. Risser et al. (1982) found signif-
icant correlation between the error type ‘‘short distance to the preceding car’’ and the number of
observed conflicts for individual drivers. In a similar study, Chaloupka, Risser, Zuzan, and
Lukascheck (1985) concluded that ‘‘keeping too short a distance’’ was among the most serious of
errors. M€akinen and Kulmala (1987) studied accidents and headway on urban roads and found
that rear-end collisions were the second most frequent accident type on the road type with a mean
headway of four seconds. Moreover; Risser (1997) reported a relationship between police reported
accidents and too short headway to the vehicle in front. On the other hand Nilson (1993), who
reviewed the literature regarding this issue, concluded that a direct connection between traffic
safety and short headways could not be established.

1.1.5. Hypothesis 5
The drivers get used to the system ‘‘taking control’’ and thereby delegate responsibility for
certain driving tasks.
This hypothesis has its origin in the phenomenon of ‘‘delegation of responsibility’’ and
‘‘behaviour diffusion’’ discussed above.

2. Method

2.1. Selection of test drivers

From the total population of the 284 test drivers participating in the large-scale trial in Lund 28
were selected for the study. These selected individuals were assigned to groups with regard to
gender, age and initial attitude to the AAP (all drivers in the large-scale trial answered a short
questionnaire with questions about, among others, their attitude towards the AAP-system). The
drivers in the large-scale trial were randomly selected from the national motor vehicle registry, the
objective being to have an even mix of gender, age and positive/negative attitude to the AAP.
However, since the test drivers had to agree to the installation of the system voluntarily, there was
a slight bias towards drivers with a positive attitude to the AAP. There was also a bias in vehicle
ownership towards middle-aged men, which was reflected in the test driver sample in this study as
well. Table 1 shows the test driver distribution.
82 M. Hj€almdahl, A. V
arhelyi / Transportation Research Part F 7 (2004) 77–94

Table 1
The distribution of test drivers with regard to age, gender and initial attitude towards the AAP
18–24 25–44 45–64 65+
Positive Negative Positive Negative Positive Negative Positive Negative
Men 1 1 3 2 6 2 5 0
Women 0 0 3 1 3 1 0 0

2.2. Observation method

The selected test drivers were observed by means of an in-car observation method, originally
developed by Risser (1985) and designed to observe learning drivers. The method, however, also
proved to be useful for studying driver behaviour in real traffic. The observations are carried out
by two observers, riding along in the car with the driver, where one of the observers (called the
coding observer) studies standardised variables such as speed behaviour, yielding behaviour, lane
changes and interaction with other road users. The other observer carries out ‘‘free observa-
tions’’ such as conflicts, communication and special events that are hard to predict, let alone to
standardise. The method was validated by Risser (1985) when he showed that there was a
correlation between observed risky behaviour and accidents. Other validation work was done by
Hj€almdahl and V
arhelyi (2004) who showed that drivers’ speed levels with observers in the car
did not differ from their speed levels when driving their own cars. They also demonstrated that it
was possible to train observers to perform the observations objectively and reliably. The method
has been used with good results in several observational studies before this (see for instance
Almqvist & Nyg ard, 1997; Comte, 2001 and Risser & Lehner, 1997). In the present study, an
instrumented vehicle was used in addition to the observers to increase the quality of standardised
variables, e.g. speed, and to make it possible to measure and register time gaps to the vehicle in
front.

2.3. Study design

The test drivers were observed twice. The first test drive was carried out without the AAP being
activated and before they had had any experience of the AAP. The second test drive was carried
out when the test drivers had had at least six months of experience of driving with the AAP in
their own cars. During this second test drive the AAP was activated within the test area but not
outside, see Table 2.
Whether to use a control group or not was discussed in the start-up phase of the study. The
advantage of using a control group is that effects caused by the studied system can be separated
from possible effects due to the drivers’ ‘‘learning’’ the test route or becoming accustomed to the
observers and the test situation. The disadvantage is of course that it takes extra resources. It was
argued that, for a control group to be useful, it would have to be about the same size as the test
group, but that was not feasible with the resources allocated to this study. It was further argued
that the chance of learning effects was rather small since there was a period of at least six months
between the two studies.
 M. Hj€almdahl, A. V
arhelyi / Transportation Research Part F 7 (2004) 77–94 83

Table 2
The different street types included in the study
Street type Speed limit Driving AAP activated Number of Description
(kph) speed (kph) automatically observed
sections
Arterial road 50
50 Yes 6 Mainly car traffic, signal-controlled
junctions and crossings
Main street 50
50 Yes 5 Cyclists and roadside parking to
some extent
Main street with 30–50
30 Yes 2 A large proportion of pedestrians and
mixed traffic cyclists, interactions mid section
Central street 30
30 Yes 1 Cars and vulnerable road users on
equal terms
Rural through 30–50 30–60 No 5 Rural road through small towns, few
road interactions
Rural road 70–90 60–100 No 6 Rural road
Motorway 110
110 No 1 Motorway

2.4. Test route

A specific route of 33 km, which took approximately 45 min to drive, was designated for the
drives. In addition, each driver drove 15 min before the observations started to get used to the
instrumented vehicle and the test situation. The test route consisted of varying driving conditions
including all the legal speed limits in Sweden, both inside and outside Lund. It was divided into
smaller parts with the same characteristics. In total there were 26 different observed sections,
categorized into 7 different street types, see Table 2.

2.5. Apparatus

The instrumented vehicle used in this study was a 1999 Toyota Corolla with instrumentation
designed and developed by VTT, Technical Research Centre, Finland. The car was equipped with
three cameras, one facing forward, one facing backward and one directed at the driver’s face. The
cameras were mainly used by the observers to discuss and clarify uncertain situations after the test
drives, for instance whether a pedestrian actually intended to cross the road when the vehicle was
approaching or not. The vehicle was also equipped with logging facilities and the variables
measured and used in this study were: time, distance to vehicle in front and speed. Data was
stored with a 5 Hz frequency.

2.6. Observed variables

2.6.1. Speed adaptation in low speed situations

Speed adaptation in this study was analysed based on speed with regard to the situation. Speeds
with regard to speed limits and average speeds were analysed elsewhere (Hj€almdahl et al., 2002
and V
arhelyi et al., in press). The variables that describe speed adaptation are the coded variables
‘‘speed adaptation at obstacles’’ and ‘‘speed with regard to the situation’’. Examples of obstacles
84 M. Hj€almdahl, A. V
arhelyi / Transportation Research Part F 7 (2004) 77–94

are road works, parked cars etc., i.e., obstacles that require some kind of evasive action from the
driver, like braking or swerving. Speed with regard to the situation implies that there is no direct
obstacle that has to be passed, but that the speed should be adapted anyway, for instance if there
are children in the vicinity of the road or if a bus is stopping to let passengers off.

2.6.2. Behaviour towards other road users

The variables describing behaviour towards other road users were yielding behaviour and
interaction with unprotected road users, (in priority situations and in interactions). The yielding
behaviour to other vehicles was classified as ‘‘OK’’, ‘‘hesitant’’ or ‘‘dangerous/short gap’’. For a
situation to be registered as a yielding situation there had to be another vehicle present.
Approaching an intersection or a roundabout without any other vehicle present was not classified
as a yielding situation. Behaviour towards pedestrians was observed in interactions at pedestrian
crossings. Swedish law gives the pedestrian the right of way at pedestrian crossings.

2.6.3. The drivers’ law abidance

Law abidance of the drivers was studied by the free observer since traffic violations are hard to
standardise.

2.6.4. Time gap

The distance to the vehicle in front, presented as time gap (i.e., the distance in meters from the
front of the test vehicle to the rear of the vehicle in front divided by the speed measured in m/s)
was measured by a lidar (laser-radar) mounted in the front of the instrumented vehicle. It was
registered five times per second. The distance was measured in meters with 0.1 m accuracy and the
maximum measurable distance was 127 m. Apart from registering distance to the vehicle in front,
the lidar could pick up reflections from parked cars, road signs etc. Video-recordings were used to
separate this data from the relevant data. Car-following situations were analysed according to the
definition below.
For a situation to be classified as a car-following one, several criteria have to be established: (1)
The time gap between the vehicles should be 4.5 s or less. A commonly used figure for mea-
surements with pneumatic tubes is 5 s between the vehicles’ front axels (see e.g. Nilsson, 2001).
With the assumption that a normal vehicle is approximately 4.5 m long, the two measures will be
the same at 32.4 kph. (2) The difference in speeds between the vehicles should be less than ±4 kph.
This is a commonly used figure to filter out overtaking situations when measuring with pneumatic
tubes (see e.g. Nilsson, 2001), but it can also be considered a reasonable limit for the vehicles to
interact with each other. The difference in speed was calculated using the speed of the test vehicle
and the change in distance. There was some instability in the measured distance that occurred due
to road and vehicle geometry, hence an average over 7 measurements, i.e. 1.4 s was used to in-
crease stability in the data. (3) The car-following situation should last for at least five seconds.
This was to filter out situations where the vehicles were together at a traffic light and then chose
different routes at the next junction without really interacting with each other. (4) The speed
should be at least 10 kph to filter out traffic queues. (5) The final condition is that the difference in
speed should not be one-sided; that is, if the test driver drives away from a traffic light following
a vehicle, which chooses a driving speed up to 4 kph faster and then after a while exceeds the 4.5 s
boundary; alternatively, the test driver catches up with another vehicle and then overtakes it.
 M. Hj€almdahl, A. V
arhelyi / Transportation Research Part F 7 (2004) 77–94 85

If these five criteria are met the registered time gaps are not affected by the fact that the drivers
might be restricted in speed when driving with the AAP, and it can thereby be excluded that any
differences between the two test-situations are due to system effects.

2.6.5. Delegation of responsibility

The phenomenon ‘‘delegation of responsibility’’ is represented by events that can be interpreted
as the driver, relying on the fact that the system takes care of tasks other than giving support to
keep the speed limit, delegates to the system some of the tasks other than the system was designed
for. With the AAP, tasks that the drivers might delegate to the system are those of keeping the
speed limit and keeping track of the current speed limit. The events that could be an indication of
delegation of responsibility were noted by the free observer.

2.7. Analyses

The analyses of the observations aimed at finding patterns and correlations that would suggest
behavioural changes. The standardised variables were analysed quantitatively and tested for
statistical significance by Chi2 -test for those on the nominal scale and by t-test for those on the
quota scale. The free observations were organised into groups to match a certain phenomenon or
hypothesis and were analysed qualitatively. The number of free observations in each group was
often too small to test for statistical significance; instead the analysis of free observations aimed at
explaining findings from the standardised variables. In some cases they also contributed to gen-
erating new hypotheses for further research.
In the analysis of time gaps each driver’s car following was compared individually with and
without the AAP for the different road sections described in Table 2. Consequently, only those
drivers who had a car-following situation on the same street types with and without the AAP were
included in the analysis. The time gaps were tested for statistical significance for each driver and
street type by independent samples t-test and the overall difference was tested by sign test.

3. Results

The analysis of the observed data was made for each of the different stretches described in Table
2. The number of observations differed between the different stretches. For instance, the number
of interactions with unprotected road users was significantly higher on central streets than on
arterial roads where most of the pedestrian crossings are signal controlled. To make the results
more transparent and to get sufficient data for statistical analysis the stretches were grouped into;
‘‘within test area’’ and ‘‘outside test area’’. Within the test area the drivers were supported by the
AAP-function in the after situation, whilst this was not the case outside the test area. The results
are presented according to the hypotheses they correspond to.

3.1. Speed adaptation in low speed situations

The events belonging to the category ‘‘speed adaptation at obstacles’’ were registered as ‘‘OK’’,
‘‘late/hard braking’’ and ‘‘badly adapted speed’’ and there were 58 registered events without AAP
86 M. Hj€almdahl, A. V
arhelyi / Transportation Research Part F 7 (2004) 77–94

Table 3
The test drivers’ speed adaptation at obstacles without and with the AAP inside the test area
Without AAP With AAP
n % n %
OK 54 93.1 42 91.3
Late/hard braking 4 6.9 2 4.3
Badly adapted speed 0 0.0 2 4.3
Sum 58 100 46 100

and 46 with AAP within the test area. The registrations are presented in Table 3. The percentage
of cases where speed adaptation was ‘‘OK’’ decreased from 93.1% to 91.3% when driving with the
AAP, but the difference is statistically not significant ðp ¼ 0:244Þ. The number of registered events
of speed adaptation with regard to the traffic situation was too low to draw any conclusions from;
there were only 10 registered events evenly distributed with and without the AAP.

3.2. Behaviour towards other road users

In total, there were 94 yielding situations without and 94 situations with the AAP within the
test area. The percentage of correct yielding behaviour increased from 88% to 96% when driving
with the AAP. The difference is statistically significant when the ‘‘OK’’ registrations are compared
with the registrations of erroneous yielding behaviour: ‘‘hesitant’’ and ‘‘dangerous/short gap’’,
however, only on the p < 0:10 level. Outside the test area there were 47 yielding situations without
and 46 with the AAP. The percentage of correct yielding behaviour increased from 94% without
the AAP, to 96% with the AAP, but that increase is not statistically significant ðp ¼ 0:851Þ.
When analysing the recorded data it turned out that the number of recorded events for
pedestrians outside the test area was too low to draw any conclusions, so a comparison could only
be carried out for pedestrians within the test area. The number of observations of interactions
with pedestrians within the test area was 118 without the AAP and 115 with the AAP and the
classification and distribution can be seen in Table 4.

Table 4
The distribution of registrations regarding the test drivers’ yielding behaviour when interacting with pedestrians at
pedestrian crossings
Interaction with pedestrians at Without AAP With AAP
pedestrian crossings
n % n %
The driver yields early 64 54 78 68
The driver yields late 7 6 5 4
The pedestrian forces the driver to stop 5 4 2 2
The pedestrian waits at roadside 29 25 25 22
The driver forces the pedestrian to stop 13 11 5 4
The driver puts the pedestrian in danger 0 0 0 0
Total 118 100 115 100
 M. Hj€almdahl, A. V
arhelyi / Transportation Research Part F 7 (2004) 77–94 87

The difference in distribution of the different types of interactions is not statistically significant
when driving without and with the AAP, but, if the registrations of correct interaction behaviour
(‘‘the driver yields early’’) are compared with those for erroneous interaction behaviour (all of the
above except ‘‘the driver yields early’’), there is a statistically significant ðp < 0:05Þ increase in
correct yielding behaviour.

3.3. The drivers’ law abidance

The majority of the traffic violations occurred within the test area. There were 14 events reg-
istered without the AAP and 5 with the AAP (note that speeding was not included in traffic
violations in this study). The occurrence of traffic violations can be considered to be Poisson
distributed and under this assumption the difference is statistically significant ðp < 0:05Þ. It should
be mentioned, however, that two types of traffic violations dominated; consciously driving against
yellow lights and stopping in an area designated for cyclists before signalised intersections. The
severity of the traffic violations can be considered as low and there might even have been a
learning effect when driving along the route the second time, even though there was a period of at
least six months between the test drives. A conservative interpretation is therefore that the number
of traffic violations did not increase.

3.4. Time gap

In total there were 212 h of driving in a car-following situation, evenly distributed between with
and without AAP and approximately 70% registered within the test area. The number of drivers
that had car-following situations with and without the AAP varied from 16 for the arterial roads,
10 for the main streets and 8 for the main streets with mixed traffic down to less than 5 for the rest
of the street types. Due to the low number it was only on the arterial roads that any statistically
significant difference could be found ðp < 0:05Þ, see Table 5.
The average, unweighted, time gap for the drivers presented in Table 5 is 1.72 s without the
AAP and 1.89 s with the AAP. The mean speed for the drivers in Table 5 during the car-following
situation on the arterial roads was 37.5 kph without the AAP and 38.7 kph with the AAP. The
distribution of time gaps for these drivers is shown in Fig. 1.
An analysis was also carried out to compare time gaps in different speed intervals. This was to
see whether the differences were dependent on street types or on speed level, and to make a
comparison between the results of this trial in Lund and the results of an earlier trial con-
ducted within the framework of the EU-project MASTER (V
arhelyi & M€akinen, 2001), see
Table 6.
As can be seen in Table 6 the results of the two trials are in line with each other even though the
relative lengths of time gaps are different. This is most likely due to different measurement
methods and a slightly different definition of a car-following situation. For none of the trials could
any difference be found in the lower speed interval, while there was a difference in the interval
30 < V < 50 kph. A comparison, where speeds above 50 kph were included (above 50 kph means
that the driver is speeding, which occurred to a higher degree without the AAP), gave an even
larger difference indicating that the AAP’s effect on time gaps increases when the speed increases.
The same effect could be found on mean speeds where the AAP had a high effect on arterial roads
88 M. Hj€almdahl, A. V
arhelyi / Transportation Research Part F 7 (2004) 77–94

Table 5
The time gap for the different drivers without and with the AAP on arterial streets
Driver Time gap (s) Difference Sign. level
a a
n Without AAP n With AAP
52 832 1.357 399 1.693 +0.336 0.000
117 375 1.86 143 2.508 +0.648 0.000
1293 299 1.911 308 2.169 +0.258 0.008
1514 216 1.12 754 1.201 +0.081 0.216
1837 62 1.71 91 1.573 )0.137 0.005
1838 128 1.183 291 2.123 +0.940 0.000
1866 599 1.948 172 1.504 )0.444 0.000
1895 344 1.66 395 2.053 +0.393 0.000
2105 1123 2.043 448 1.645 )0.398 0.000
2150 56 1.972 249 1.866 )0.106 0.000
2277 247 1.798 77 1.818 +0.020 0.069
2382 85 1.596 515 1.8 +0.204 0.008
3289 171 1.266 474 1.75 +0.484 0.000
3592 805 1.677 1100 1.852 +0.175 0.105
4083 104 1.936 710 2.106 +0.170 0.000
5094 190 2.4 127 2.545 +0.145 0.000
a
n ¼ the number of registrations from the instrumented vehicle which was sampled at 5 Hz.

20.00
Without AAP
With AAP

15.00

10.00
%

5.00

0.00
0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8 2 2.2 2.4 2.6 2.8 3
Time gap (sec)

Fig. 1. The distribution of time gaps without and with the AAP for arterial roads.

but a very small effect in central streets in which the speeds were already low without the AAP
(Hj€
almdahl et al., 2002).
 M. Hj€almdahl, A. V
arhelyi / Transportation Research Part F 7 (2004) 77–94 89

Table 6
Time gaps in different speed intervals without and with the AAP for this trial and the European project MASTER with
and without the Speed Limiter (SL)
0 < v < 30 30 < v < 50
Lund trial MASTER Lund trial MASTER
Without With AAP Without With SL Without With AAP Without With SL
AAP SL AAP SL
n 4917 4877 87 94 6445 8151 280 318
Avg. time 2.03 2.04 1.44 1.49 1.83 1.87 1.27 1.36
gap
Std. dev. 0.49 0.48 0.55 0.63 0.52 0.52 0.50 0.55
Skewness )0.03 0.10 1.13 0.97 0.31 0.09 0.79 0.82
Kurtosis )0.81 )0.83 1.46 0.16 )0.82 )0.72 0.39 0.23
Sign. level 0.583 0.839 0.000 0.094
n : In the Lund trial n denotes the number of registrations with the lidar which means that n=5 is the number of
measured seconds. In MASTER n denotes the number of observed situations.

3.5. Delegation of responsibility

The observed events having to do with ‘‘delegation of responsibility’’ are listed in Table 7.
As seen in Table 7 most of the events that can be attributed to ‘‘delegation of responsibility’’
occurred outside the test area where the AAP was not active. The numbers are small but the
difference is apparent ðp < 0:01Þ assuming that the events are Poisson distributed. Thus, it can be
stated that a certain degree of delegation of responsibility occurred.

Table 7
Events indicating delegation of responsibility
Description Number of registrations
Without AAP With AAP
Within test Outside test Within test Outside test
area area area area
Positive – – – –
Negative Drives at 70 kph on 90 kph road until – 1 – –
a vehicle approaches from behind
Accelerates then up to 100 kph
Drives 50 kph on a 30 kph road – – – 3
Drives 70 kph on a 50 kph road – – – 3
Drives 50 kph on a 70 kph road – – – 2
Accelerates up to 90 kph on a 70 kph – – – 1
road and then slows down

Total 0 1 0 9
90 M. Hj€almdahl, A. V
arhelyi / Transportation Research Part F 7 (2004) 77–94

4. Discussion

The observations in this study aimed at establishing whether the AAP leads to any changes in
driving behaviour and driving style, and especially to see if there are any negative behavioural
changes which can reduce the safety effect that the AAP should have had due to lower and more
even speeds (Almqvist et al., 1997; Brookhuis & de Waard, 1999; Carsten & Fowkes, 2000;
Hj€almdahl et al., 2002; V
arhelyi & M€akinen, 2001).
The numbers of events recorded with and without the AAP are similar in size, which indicates
stability in the measurement method and the test conditions. Together with the large number of
observations for some of the variables, this provides for good possibilities for analyses. However,
for some variables there were not enough observations to state with certainty that there is an
effect. Still, it does raise the need for further research.
Hypothesis 1, ‘‘Drivers do not lower their speed in low speed situations to the same extent when
driving with AAP as when driving without AAP’’ can neither be verified nor rejected due to too
few observations. There were no clear signs of compensatory behaviour when the drivers drove
with the AAP. The number of ‘‘badly adapted speed at obstacles’’ and ‘‘late/hard braking’’ in-
creased within the test area but the difference is small and not statistically significant. Outside the
test area there were too few situations to trace any effects. Studies of approach speeds and turning
speeds based on the logged data in the large-scale trial in Lund (V
arhelyi et al., in press) could
reject the hypothesis that drivers do not lower their speed to the same extent as they do without
the system. The findings from this study, however, cannot rule out that such an effect exists.
Hypothesis 2, ‘‘Behaviour towards other road users will improve (the drivers will be more
willing to give priority to other vehicles and pedestrians)’’ can be verified. The percentage of
correct yielding behaviour increased from 88% to 96% within the test area when driving with the
AAP. Thus, there was a strong tendency towards improved yielding behaviour. Interactions with
pedestrians improved statistically significantly, and correct yielding behaviour increased from 54%
to 68% when driving with the AAP ðp < 0:05Þ. These results are in line with the results from
Almqvist and Nyg ard (1997) who found that the behaviour towards other road users improved
after the test subjects had used an in-car speed limiter for two months. V
arhelyi and M€akinen
(2001) did not find any difference in behaviour towards other road users and Persson et al. (1993)
found that behaviour actually deteriorated when driving with a speed limiter. Those studies,
however, did not look at long-term behavioural modifications. Instead, these trials were carried
out as a comparison between two consecutive test drives. The present results reinforce earlier
empirical findings of Hyd
en and V
arhelyi (2002) and Towliat (2001) who showed that drivers
entering in an interactive situation with lower speeds are more willing to give priority to pedes-
trians at pedestrian crossings. V
arhelyi (1998), observing the behaviour of drivers approaching a
pedestrian crossing, found that in encounters, three out of four drivers maintained the same speed
or accelerated (even over the speed limit of 50 kph) to signal that they do not intend to give way to
the pedestrian at the crossing. The AAP, reminding the driver of the speed limit, obviously
prevents the driver from such signalling. Subjective measures through questionnaires have also
shown that the drivers believe they are more aware of pedestrian presence and they yield to a
higher degree (Falk, Hj€ almdahl, Risser, & V
arhelyi, 2002).
Hypothesis 3, ‘‘The law abidance of drivers in AAP-cars can change in a positive or negative
direction (fewer/more stops at stop signs, larger/smaller proportion of driving against red)’’ can
 M. Hj€almdahl, A. V
arhelyi / Transportation Research Part F 7 (2004) 77–94 91

neither be verified nor rejected. Due to the small amount of observations and to the low severity of
the traffic violations it is too daring to say that the law abidance has improved. A conservative
interpretation of the results is that the law abidance has not worsened due to the AAP.
Hypothesis 4, ‘‘The time gap to the vehicle in front will increase’’ can partly be verified. The
time gap to the preceding vehicle increased on arterial roads. The analysis shows that the effect of
the AAP increases when speed increases. This result can be compared with the results of other
studies in the field with and without a speed limiter; Persson et al. (1993) concluded that distance
to the vehicle in front increased and V
arhelyi and M€akinen (2001) found that distance to the
vehicle in front increased within urban areas but decreased on rural roads.
Hypothesis 5, ‘‘The drivers get used to the system ‘taking control’ and thereby delegate
responsibility for certain driving tasks’’ can be verified. There were some signs of delegation of
responsibility in the form of the drivers not changing their speed when entering a new speed limit
zone where the system was not active (i.e. outside the test area). Without the AAP this was only
registered once while it was observed nine times when the test drivers had been driving with the
AAP in their own cars for at least six months. There is only one variable in this study that
indicates delegation of responsibility and the number of observations is relatively small. However,
the difference between the two test drives is so apparent that it is highly probable that ‘‘delegation
of responsibility’’ exists to a certain degree.
In-vehicle speed adaptation systems, such as the AAP, have proved to have a large safety
potential due to reduced speeds, but earlier studies had also shown that there were some
behavioural modifications when driving with such systems. Examples of this from earlier studies
are: deteriorated interaction with other road users, changed headway and law abidance. The
current study differs from the earlier studies in that in this study it was possible to observe
behavioural modifications after using the system for a longer period of time. The drivers in this
study had been driving with the AAP in their own cars for at least six months and in some cases
for up to one year before their behaviour in the ‘‘after’’ situation was observed.
Analysis of speed data logged in the drivers’ own vehicles during everyday driving showed that
the test drivers’ compliance with the speed limits improved considerably when driving with the
AAP (see Table 8) and the safety effect due to reduced mean speeds was estimated to range from a

Table 8
Changes in mean speeds at mid-block sections and the expected safety effect on different street types (from Hj€
almdahl
et al., 2002)
Street type/speed No of measured Mean speed at mid-block Change (kph) Expected
limit (kph) stretches (unweighted) for all stretches (kph) decrease in fatal
Without AAP With AAP accidents (%)

Arterial road/70 8 76.0 71.1 )4.9 23

Arterial road/50 10 55.3 50.3 )5.0 32
Arterial road/50 19 52.8 49.1 )3.7 25
Main street/50 12 45.2 43.2 )2.0 17
Main street, 12 38.1 37.1 )1.0 10
mixed traffic/50
Central street/30 8 28.7 27.0 )1.7 22
*
Statistically significant difference according to t-test and sign test ðp < 0:05Þ.
92 M. Hj€almdahl, A. V
arhelyi / Transportation Research Part F 7 (2004) 77–94

10% reduction in the number of fatal accidents on main streets with mixed traffic up to a 32%
reduction on arterial roads with a 50 kph speed limit (Hj€almdahl et al., 2002). These estimations
were made by using the ‘‘power model’’ (Nilsson, 2000) which only considers the change in mean
speed and presupposes that no other changes have taken place. Hj€almdahl et al. (2002) also
showed that the variance in speed decreased and these changes in speed behaviour indicate a great
traffic safety potential.
The present study has shown that the AAP brings other effects too, apart from changes in the
speed level and speed variance, such as improved behaviour towards other road users. The drivers
showed a more correct yielding behaviour at intersections and yielded early for pedestrians to a
higher extent when driving with the AAP. This indicates that the system does not change
behaviour immediately but driving with it for a longer period of time does. This finding reinforces
earlier empirical findings on the mechanisms of behavioural adaptation, articulated by Drask
oczy
(1994), saying that behavioural adaptation does not appear immediately when the driving context
is changed, but usually appears only after a time.
Another positive effect is that the time gaps increased on arterial roads when driving with the
AAP. On other street types no effect could be found due to the low number of car-following
events. The increase in time gap is quite small but it means that the drivers on average will have an
extra 0.17 s disposable for reaction time or 1.79 m extra to stop when driving at 38 kph, which was
the mean speed during the car-following situations.
There were also signs of behavioural modifications that could have a negative effect on safety.
The most apparent phenomenon was the fact that, after having driven with the AAP for a longer
period, some of the test drivers forgot to change their driving speed when entering a new speed
limit zone in areas where the AAP was not active. Such a result is to be expected when the AAP is
only active in certain areas. The number of observations of this phenomenon was quite low but
there were clear differences between driving without the AAP and driving with it. This is a very
important aspect, since during an introductory phase there will be areas not covered by the system
and there will be cars not equipped with the system. Further research is needed on how drivers
react, after getting used to the system, when driving in areas or cars where the AAP is not in
operation.
In addition to these observed behavioural effects, subjective measurements of the test drivers’
workload and experience with the system via interviews (see Falk et al., 2002) revealed that they
felt a slightly higher time pressure and that they were holding up others in traffic when driving
with the AAP. In-depth interviews revealed that these negative effects were most apparent on
roads with a mean speed above the speed limit. Such effects can be expected when the number
of equipped vehicles in traffic is low. On the other hand, some positive indications of
decreased driver workload were found too, such as the test drivers stating that they looked less
frequently at the speedometer and that they needed to use the accelerator and the brakes less
frequently.
An overall conclusion regarding the safety effect of the AAP must be that; even though there
are signs of behavioural modifications that could diminish the positive effects on traffic safety,
these diminishing effects are by far exceeded by the additional safety potential due to lowered
speed, in combination with the positive change in car-following behaviour and improved inter-
action with other road users.
 M. Hj€almdahl, A. V
arhelyi / Transportation Research Part F 7 (2004) 77–94 93

References

Almqvist, S., Hyd
en, C., & Risser, R. (1991). Hastighetsbegr€ ansare i bil. Effekter pa f€
orares beteende och interaktion
[Speed limiter in car. Effects on driver behaviour and interactions]. Department of Traffic Planning and Engineering,
Lund University, Lund, Sweden [in Swedish].
Almqvist, S., & Nyg ard, M. (1997). Dynamic speed adaptation. Field trials with automatic speed adaptation in an urban
area. Bulletin 154, Department of Traffic Planning and Engineering, Lund University, Lund, Sweden.
Brookhuis, K., & de Waard, D. (1999). Limiting speed, towards an intelligent speed adapter (ISA). Transportation
Research Part F, 2(2), 81–90.
Carlqvist, T., & Persson, H. (1988). Polisrapporterade olyckor, Projekt V€ axj€o [Accidents reported to the police, Project
V€axj€o]. Bulletin 3, Department of Traffic Planning and Engineering, Lund University, Lund, Sweden [in Swedish].
Carsten, O. M. J. (1993). Framework for prospective traffic safety analysis. Deliverable 6 in the DRIVE II Project V2002,
HOPES.
Carsten, O. M. J. (1994). Guidelines on safety evaluation of in-vehicle information systems. In M. Drask
oczy (Ed.),
Guidelines on safety evaluation. DRIVE Project V2002, HOPES. Bulletin 118, Lund University, Lund, Sweden.
Carsten, O. M. J., & Fowkes, M. (2000). External vehicle speed control. Executive summary of project results. University
of Leeds, UK.
Carsten, O. M. J., Tight, M. R., Southwell, M. T., & Plows, B. (1989). Urban accidents: why do they happen? AA
Foundation for Road Safety Research, Institute for Transport Studies, University of Leeds, UK.
Chaloupka, Ch., Risser, R., Zuzan, W. D., & Lukascheck, H. (1985). Unfall- und Fahrverhaltensanalyse ausl€ andischer
€
Kraftfahrer, insbesondere Deutscher, in Osterreich. Teiprojekt 2, Kuratorium f€ ur Verkehrssicherheit, Wien, Austria.
Comte, S. L. (1996). Response to automatic speed control in urban areas: a simulator study. Institute for Transport
Studies, University of Leeds, ITS Working Paper, no. 477, Leeds, UK.
Comte, S. L. (1998a). Simulator study on the effects of ATT and non-ATT systems and treatments on driver speed
behaviour. Working Paper R 3.1.2 in the MASTER project, VTT, Espoo, Finland.
Comte, S. L. (1998b). Evaluation of in-car speed limiters: simulator study. Working Paper R 3.2.1 in the MASTER
project, VTT, Espoo, Finland.
Comte, S. (2001). Intelligent speed adaptation: evaluating the possible effects of an innovative speed management system on
driver behaviour and road safety. Doctoral thesis, Institute for Transport Studies, University of Leeds, UK.
Comte, S. L., V
arhelyi, A., & Santos, J. (1997). The effects of ATT and non-ATT systems and treatments on driver speed
behaviour. Working Paper R 3.1.1 in the MASTER project, VTT, Espoo, Finland.
Drask
oczy, M. (1994). Guidelines on safety evaluation. DRIVE Project V2002, HOPES. Bulletin 118, Department of
Traffic Planning and Engineering, Lund University, Lund, Sweden.
Drask
oczy, M., & Mocs
ari, T. (1997). Present speeds and speed management methods in Europe. Deliverable R 2.1.1 in
the MASTER project, VTT, Espoo, Finland.
€
Elvik, R., Vaa, T., & Ostvik, E. (1989). Trafikksikkerhetsh andbok [Traffic safety handbook]. Transportøkonomisk
Institutt, Oslo, Norway [in Norvegian].
Falk, E., Hj€almdahl, M., Risser, R., & V
arhelyi, A. (2002). Testf€ orarnas attityd till ISA: resultat fr
an enk€ atun-
ders€okningar [Test drivers’ attitude to ISA: results from questionnaire studies]. Department of Traffic Engineering,
Lund University, Lund [in Swedish].
Finch, D. J., Kompfner, P., Lockwood, C. R., & Maycock, G. (1994). Speed, speed limits and accidents. Project Report
58, Transport Research Laboratory, Crowthorne, UK.
Godthelp, H., & Schumann, J. (1991). The use of an intelligent accelerator as an element of a driver support system. In
24th ISATA International Symposium on Automotive technology and Automation, May 1991, Italy.
Hauer, E., Ahlin, F. J., & Bowser, J. S. (1982). Speed enforcement and speed choice. Accident Analysis and Prevention,
14(4), 267–278.
Hj€
almdahl, M., Almqvist, S., & V
arhelyi, A. (2002). Speed regulation by in-car active accelerator pedal––effects on
speed and speed distribution. IATSS Research, 26(2), 60–67.
Hj€
almdahl, M., & V
arhelyi, A. (2004). Validation of in-car observations, a method for driver assessment.
Transportation Research Part A, 38(2), 127–142.
94 M. Hj€almdahl, A. V
arhelyi / Transportation Research Part F 7 (2004) 77–94

Hyd
en, C., & V
arhelyi, A. (2002). The effects on safety, time consumption and environment of large scale use of
roundabouts in an urban area: a case study. Accident Analysis and Prevention, 32(1), 11–23.
M€akinen, T., & Kulmala, R. (1987). Aikav€alien yhteys turvallisuuteen taajamaliikenteess€ a [Safety and headways in
urban traffic]. Valtion Teknillinen Tutkimuskeskus, Esbo, Finland [in Finnish].
Nilson, G. (1993). Tidluckor i landsv€agstrafik: En problemstudie [Time gaps on rural roads: a problem study]. VTI notat
T 139, Link€ oping, Sweden [in Swedish].
Nilsson, A. (2001). Hastighet och tidluckor 2000: resultatrapport [Speed and time gaps 2000: report on results].
V€agverket, Borl€ange, Sweden [in Swedish].
Nilsson, G. (1982). The effect of speed limits on traffic accidents in Sweden. VTI Report 68, Link€ oping, Sweden.
Nilsson, G. (2000). Hastighetsf€or€andringar och trafiks€akerhetseffekter––Potensmodellen [Speed changes and traffic safety
effects the power model]. VTI notat 76-2000, Sweden [in Swedish].
O’Cinn
eide, D., & Murphy, E. (1994). The relationship between geometric road design standards and driver/vehicle
behaviour, level of service and safety. Traffic Research Unit, University of Cork, UK.
OECD (1990). Behavioural adaptations to changes in the road transport system. OECD Road Transport Research, Paris,
France.
Østvik, E., & Elvik, R. (1990). The effects of speed enforcement on individual road user behaviour and accidents. In
Proceedings of the international road safety symposium on enforcement and rewarding strategies and effects,
Copenhagen, Denmark.
Persson, H., Towliat, M., Almqvist, S., Risser, R., & Magdeburg, M. (1993). Hastighetsbegr€ ansare i bil. F€
altstudie av
hastigheter, beteenden, konflikter och f€orarkommentarer vid k€ orning i t€
atort [In-car speed limiters. Studies in the field
of speed, behaviour, conflicts and driver opinions for urban driving]. Department of Traffic Planning and
Engineering, Lund University, Lund, Sweden [in Swedish].
Risser, R. (1985). Behaviour in traffic conflict situations. Accident Analysis and Prevention, 17(2), 179–197.
Risser, R. (1997). Assessing the driver. ‘‘Der Mensch im Verkehr’’ e.V., Cologne.
Risser, R., & Lehner, U. (1997). Evaluation of an ACC (autonomous cruise control) system with the help of behaviour
observation. In Presentation on the 4th world congress on intelligent transport systems, Berlin, Germany.
Risser, R., Teske, W., Vaughan, Ch., & Brandst€atter, Ch. (1982). Verkehrsverhalten in Konfliktsituationen. Kuratorium
f€
ur Verkehrssicherheit, Wien, Austria.
Saad, F., & Malaterre, G. (1982). La regulation de la vitesse: Analyse des aides au controle de la vitesse. ONSER, France.
Salusj€arvi, M. (1981). The speed limit experiments on public roads in Finland. Technical Research Centre of Finland,
Publication 7/1981, Espoo, Finland.
Schulman, R. (1985). The deaccelerator project: research report. The Deacceleration Corporation, Kalamazoo, MI,
1985; Bower, G.H. (1990). Incentive programs for promoting safer driving. Enforcement and rewarding strategies
and effects. In Proceedings of the symposium in Copenhagen, Denmark.
Teed, N., Lund, A. K., & Knoblauch, R. (1993). The duration of speed reductions attributable to radar detectors.
Accident Analysis and Prevention, 25(2), 131–137.
Towliat, M. (2001). Effects of safety measures for pedestrians and cyclists at crossing facilities on arterial roads. Doctoral
dissertation, Department of Traffic Planning and Engineering, Lund University, Lund, Sweden.
V

arhelyi, A. (1998). Drivers’ speed behaviour at a zebra crossing: a case study. Accident Analysis and Prevention, 30(6),
731–743.
V

arhelyi, A., Hj€almdahl, M., Hyd
en, C., & Drask
oczy, M. (in press). Effects of an active accelerator pedal on driver
behaviour and traffic safety after long-term use in urban areas. Accident Analysis and Prevention [corrected proof,
available online].
V

arhelyi, A., & M€akinen, T. (2001). The effects of in-car speed limiters: field studies. Transportation Research Part C,
9(3), 191–211.
Wilde, G. (1994). Target risk. Toronto, Canada: PDE Publications.