ArticlePDF Available

An analysis of 'looked but failed to see' accidents involving parked police vehicles

Taylor & Francis
Ergonomics
Authors:

Abstract and Figures

Drivers who collide with a vehicle that is parked on the hard shoulder of a motorway or dual-carriageway sometimes claim not to have seen it before the collision. Previous research into vehicle conspicuity has taken such 'looked but failed to see' claims at face value, and concentrated on attempting to remedy the problem by making vehicles more conspicuous in sensory terms. However, the present study describes investigations into accidents of this kind which have involved stationary police cars, vehicles which are objectively highly conspicuous. Two laboratory studies showed that experienced drivers viewing a film of dual-carriageway driving were slower to respond to a parked police car as a 'hazard' if it was parked directly in the direction of travel than if it was parked at an angle; this effect was more pronounced when the driver's attention was distracted with a secondary reasoning task. Taken together with the accident reports, these results suggest that 'looked but failed to see' accidents may arise not because the parked vehicle is difficult to see, but for more cognitive reasons, such as vigilance failure, or possession by the driver of a 'false hypothesis' about the road conditions ahead. An emergency vehicle parked in the direction of travel, with only its blue lights flashing, may encourage drivers to believe that the vehicle is moving rather than stationary. Parking at an angle in the road, and avoiding the use of blue lights alone while parked, are two steps that drivers of parked emergency vehicles should consider taking in order to alert approaching drivers to the fact that a stationary vehicle is ahead.
Content may be subject to copyright.
An analysis of `looked but failed to see’ accidents involving
parked police vehicles
MARTIN LANGHAM{*, GRAHAM HOL E{, JACQUELINE EDWARDS{and COLIN O’NEIL{
{School of Cognitive and Computing Sciences, University of Sussex, Falmer,
Brighton BN1 9QH, UK
{Sussex Police Tra c Division, Police Headq uarters, Mallin g H ouse, Lewes, East
Sussex, UK
Keywords: Vehicle conspicuity; Police cars; Collisions; Hazards.
Drivers who collide with a vehicle that is parked on the hard shoulder of a
motorway or dual-carriageway sometimes claim not to have seen it before the
collision. Previous research into vehicle conspicuity has taken such `looked but
failed to see’ claims at face value, and concentrated on attempting to remedy the
problem by making vehicles more conspicuous in sensory terms. However, the
present study describes investigations into accidents of this kind which have
involved stationary police cars, vehicles which are objectively highly conspicuous.
Two laboratory studies showed that experienced drivers viewing a ®lm of dual-
carriageway driving were slower to respond to a parked police car as a `hazard’ if
it was parked directly in the direction of travel than if it was parked at an angle;
this eVect was more pronounced when the driver’s attention was distracted with a
secondary reasoning task. Taken together with the accident reports, these results
suggest that `looked but failed to see’ accidents may arise not because the parked
vehicle is di cult to see, but for more cognitive reasons, such as vigilance failure,
or possession by the driver of a `false hypothesis’ about the road conditions
ahead. An emergency vehicle parked in the direction of travel, with only its blue
lights ¯ashing, may encourage drivers to believe that the vehicle is moving rather
than stationary. Parking at an angle in the road, and avoiding the use of blue
lights alone while parked, are two steps that drivers of parked emergency vehicles
should consider taking in order to alert approaching drivers to the fact that a
stationary vehicle is ahead.
1. Introduction
Perceptual factors are often claimed to play an important role in many road tra c
accidents (review in Hills 1980). Following an accident, one or both of the drivers
often claim not to have seen the other vehicle until it was too late to avert a collision
Ðthe so-called `looked but failed to see’ (LBFS) error (Staughton and Storie 1977).
Cairney and Catchpole (199 5) estimated that 69 ± 80% of all intersection accidents
result from a failure by one driver to detect another’s presence until it was too late to
avoid a collision. Rumar (1990) has identi®ed two important causes of LBFS errors:
sensory limitations (the stimulus may fail to be detected because it falls below or near
perceptual thresholds) and cognitive factors (e.g. failures of attention or
inappropriate expectations about what is likely to happen next).
*Author for correspondence; e-mail: Martinl@cogs.susx.ac.uk
ERGONOMICS, 2002, VOL. 45, NO. 3, 167 ± 185
Ergonom ics ISSN 0 014-0139 print/ISSN 1366-5847 online #2002 Taylor & Francis Ltd
http://www.tandf.co.uk/journals
DOI: 10.1080 /001401301101 15363
There has been a great deal of research into late detection errors within the context
of drivers’ problems in detecting motorcyclists and cyclists (Olson 1989). In this
context, most research has tended to focus on sensory limitations as the root cause of
the problem, rather than cognitive failures: motorcyclists and cyclists have been
considered to be di cult to detect due to their small size compared to other vehicles
(although see Olson 1989, Wulf et al. 1989, Hole and Tyrrell 1995, for alternative
explanations). Researchers have tended to take at face value drivers’ claims to
accident investigators that they failed to see a motorcyclist or cyclist. Correspond-
ingly the solution to the problem has been thought to be to make the motorcyclist or
cyclist as physically conspicuous as possible (for example, by use of daytime
headlights and bright clothing, e.g. JanoVand Cassel 1971, Fulton et al. 1980).
However late detection accidents are not con®ned to accidents involving
motorcyclists and cyclists (Olson 1989, Rumar 1990): `looking and not seeing’
appears to have been a factor in accidents involving other types of vehicle as well,
some of which one might expect to be easily detected, for example railway engines
(Leibowitz 1985) and buses (Draskoczy 1989). Cercarelli et al. (1992) produce
evidence from accident database studies to suggest that many crashes between cars
can be interpreted as LBFS errors. The explanation of collisions purely in terms of
sensory conspicuity problems does not therefore seem likely to provide a complete
explanation of this type of accident (Olson 1989; Wulf et al. 1989).
Previous studies of vehicle conspicuity have often investigated situations where a
lack of sensory conspicuity (small size, together with positioning within the
peripheral visual ®eld as opposed to foveal vision) and failure to detect have
coincided, such as in the case of motorcyclists and cyclists: in these situations, it is
di cult to rule out sensory limitations as a contributory factor in accident causation.
In the case of collisions between cars and motorcycles or bicycles, one has to allow
for the possibility that in some cases the behaviour of the motorcyclist or cyclist may
have contributed to the accident. There is also the possibility that the oVending
driver is using the claim that they failed to detect the other vehicle as an excuse,
rather than admitting that they had deliberately violated the other vehicle’s right of
way (e.g. in an attempt to force their way out at a junction rather than wait for a
suitable gap to appear in the oncoming tra c).
However, what if one found a situation where an objectively conspicuous and
relatively large object, appearing within the driver’s central visual ®eld, and which no
sane person would want to collide with, was nevertheless not detected? Such failures
would argue against an explanation of conspicuity failure purely in terms of sensory
factors: the explanation must lie somewhere else. Such a situation does exist:
sometimes objectively highly conspicuous vehicles are collided with by drivers who
claim that they failed to see them. These include police vehicles and other emergency
vehicles parked on the hard shoulder of a dual carria geway or motorway.
In the UK, a `motorway’ is usually a six-lane road (three lanes in either direction)
with a speed-limit of 113 km/h (70 miles per hour), from which certain classes of
road-user (e.g. cyclists and learner drivers) are excluded. A `dual-carriageway’ is
usually a four-lane road (two lanes in either direction) with a 113 km/h speed limit.
Both types of road usually have a `hard shoulder’ on which vehicles can stop only if
they have broken down. There are numerous exceptions to these rules, since some
dual-carriageways have six lanes, and some motorways have only four. The
important points for the present discussion are that (a) both types of road are
designed for high speed (so that the road generally lacks sharp bends or turns); (b) all
168 M. Langham et al.
tra c on a carriageway is moving in the same direction; and (c) stationary and very
slow-moving vehicles are relatively rare, with stationary vehicles almost always
con®ned to the hard shoulder (where one exists).
The initial motivation for the present study came from our local police force, who
reported an increased number of accidents involving collisions with stationary police
vehicles on motorways. There was concern that this increase was related to a change
in parking practices: before 1996, UK police vehicles were parked across the
carriageway in the lane that contained the hazard, so that the side of the vehicle was
visible to oncoming drivers (hereafter referred to as `echelon’ parking). Since 1996,
guidelines issued by the Association of Chief Police O cers (ACPO 1996) have
required police vehicles to park so that the rear of the police vehicle is visible to the
oncoming tra c (i.e. the vehicle’s long axis is parallel to the lane’s line markings, a
practice hereafter referred to as `in line’). This change was intended to increase the
police vehicle’s conspicuity by making the car’s roof-mounted lighting strip more
visible to the tra c than if the car were parked at an angle.
With the aid of the Institute of Tra c Accident Investigators, we mounted a
nationwide campaign to solicit information from police forces about any accidents
that they had experienced which involved a driver colliding with a stationary police
car. Detailed descriptions of 47 collisions were obtained, from twelve UK police
forces. After excluding reports for which an alternative explanation might exist (such
as driver impairment due to fatigue or alcohol, poor weather, or the presence of a
physical obstruction to the approaching driver’s view) 29 reports remained of
daytime accidents in which a stationary police car, ®tted with conspicuity
enhancements (re¯ective markings and ¯ashing lights) had been crashed into by a
driver who claimed either not to have seen the vehicle at all, or at least not in time to
avoid an accident.
A total of 59% (17 of 29) of the accidents occurred when the police vehicle was
de®nitely parked `in-line’; in the case of the other reports, the orientation of the
vehicle was not always explicitly recorded. All but one of the accidents occurred
when there was only one police vehicle parked, due to it being either the ®rst or last
vehicle at the scene. The early deployment of warning signs and tra c cones did not
guarantee detection of the parked vehicle.
Some 62% (18 of 29) of the accidents occurred within 15 km of the oVender’s
home address, i.e. on roads that were presumably relatively familiar to the drivers
concerned. (This ®nding has to be treated with caution, given the small size of the
sample and the fact that familiarity is inevitably confounded with exposure to risk:
drivers might be more likely to have an accident close to home because they spend
more time driving in that area).
The oVending drivers were over the age of 25 years except in one case, which
tentatively suggests that that this type of accident can involve drivers of all ages and
hence levels of experience.
Evidence from these reports suggests that in this type of accident the driver was
not detecting the hazard too late, but was failing to detect it altogether. Some 39% of
the reports (11 of 29) contain no evidence that the driver braked at all before the
collision. A total of 70% (20 of 29) of the oVending drivers’ statements contained the
phrase `I did not see it’.
Interpretation of accident statistics is fraught with di culties, especially when the
sample is small and concerns events as uncommon as this particular type of accident.
However, the data do enable us to make a number of observations. First, while this
169
`Looked but failed to see’ accidents
is not a common type of accident, it does occur with an appreciable frequency. It was
suggested to us that many other cases had occurred, apart from the ones o cially
reported to us. Police forces evidently ®nd this a sensitive issue to discuss, even when
con®dentiality is assured. If these claims are to be believed, it suggests that there may
be as many as 150 such cases in the UK per year amongst the police alone. When one
includes similar collisions with breakdown vehicles and the general public, this type
of LBFS error may be more common than might be suggested by the statistics
reported here.
Second, we have been able to isolate a core sample of reports that appear to
demonstrate clearly that, despite being conspicuous in sensory terms, police vehicles
were being hit by drivers who claimed not to have seen them. We have good reports
concerning drivers who have, in broad daylight and with an unrestricted view of the
road ahead, nevertheless failed to detect a parked vehicle covered in re¯ective stripes
and using ¯ashing blue and red lights.
The survey data, however, provide few clues to the cause of this detection failure.
In order to try to identify contributory factors, we undertook two laboratory
experiments in which we attempted to simulate the essential aspects of the drivers’
situation before the accident took place.
2. Experiment 1: hazard detection as a function of vehicle parking orientation and
drivers’ experience
2.1. Introduction
The accident data suggested to us that one factor in these LBFS accidents might be
that drivers collide with police cars that are parked `in-line’ because they are relying
on certain expectancies about the road ahead. Experience may have taught the driver
that the `in-line’ parked vehicle displaying blue lights is a moving vehicle rather than
stationary. The accident data are inconclusive about whether or not `echelon’-parked
cars are more readily detectable as being parked than are cars parked `in-line’.
However, if this kind of accident is due to drivers having inappropriate expectations
about their surroundings, then it seems likely that this would be the case: compared
to cars parked `in-line’, the `echelon’-parked car’s orientation should present drivers
with a clearer signal that they are approaching something which is out of the
ordinary and which is unlikely to be a moving vehicle. If so, this eVect should be
aVected by the driver’s level of experience: inexperienced drivers may not have
developed such strong expectancies about the road environment and hence should be
more likely to recognize that a police car parked `in-line’ is stationary.
This laboratory experiment therefore investigated the ability of drivers to detect a
parked police car in a video of a dual carriageway ®lmed from the driver’s viewpoint,
as a function of the parked car’s orientation (in-line versus echelon) and the level of
the driver’s experience (inexperienced versus experienced).
Participants were asked to report any hazards they saw on the video. Towards
the end of the video, a stationary police car came into view, parked either `in-line’ or
`echelon’. If detection of the parked vehicle is governed solely by sensory (physical
conspicuity) factors, then there might be diVerences in detectability produced by the
two diVerent modes of parking, but there should be no diVerence in performance
between experienced and inexperienced drivers. However, if drivers’ expectancies are
a factor in this kind of accident, there should be an eVect of experience on detection
performance: to an experienced driver, a car parked `in-line’ may look no diVerent to
any other car that is travelling in the same direction as the driver, and consequently
170 M. Langham et al.
experienced drivers should take longer to detect a stationary police car when it is
parked `in-line’ than when it is parked at an angle (`echelon’ fashion). In the absence
of such well-developed expectancies, inexperienced drivers should detect either
orientation equally well.
2.2. Method
2.2.1. Design: The experiment used an independent measures design. Experienced
or inexperienced drivers viewed either a police vehicle parked `in-line’ or parked
`echelon’ in one video clip amongst a series of clips that were otherwise identical for
all participants. Participants therefore participated in one of four conditions:
(1) `experienced in-line’;
(2) `experienced echelon’;
(3) `inexperienced in-line’; and
(4) `inexperienced echelon’.
Participants’ reaction times were recorded, together with their decision about
whether or not each clip contained a `hazard’.
2.2.2. Participants: A total of 59 participants (20 males and 39 females) were
recruited by campus advertisement at the University of Sussex. Mean age was 27
years (range 17 ± 49 years). `Experienced’ drivers were de®ned as drivers with more
than 2 years’ driving experience. `Inexperienced’ drivers were either learning to drive
or had passed their test within the previous month. The experienced drivers had a
mean of 13 years’ driving experience.
2.2.3. Preparation of stimulus materials: Video-recordings were made of police
cars parked in the left-hand lane of the west-bound carriageway of the A27(M),
west of Chichester, West Sussex. This road is a dual-carriageway with motorway
status: on each side of the road there are two lanes plus a hard shoulder, and
usage is restricted to certain classes of road-user (e.g. learner drivers, cyclists,
moped riders and invalid carriages are excluded). The site where the stimulus
materials were collected had a gentle leftwards bend in the road, so that there
was a well-de®ned point at which an approaching driver could ®rst see the parked
police vehicle. This site had been the location for two LBFS-like accidents in the
past.
The video-®lming of the parked cars was from the viewpoint of an approaching
motorist. Filming was through the front windscreen, by means of the in-car `Provid’
video-camera system ®tted to Sussex police cars. Recordings were made during
September 1997 in ®ne weather between 06:00 and 09:15 h. A marked police car was
parked in the left-hand lane of the dual carriageway in either an `in-line’ position or
an `echelon’ position (®gure 1). In the `in-line’ condition the vehicle was parked 1 m
from the white line which designates the beginning of the hard shoulder. In the
`echelon’ condition the vehicle was parked facing the central reservation, at an angle
of 458from the near-side of the road, with the rear left wheel on the white line of the
hard shoulder. In both cases the front of the parked car faced the direction of tra c
¯ow. The car was a white Volvo T5 Estate with the array of conspicuity
enhancements which are becoming the UK standard for police vehicles used on
high-speed roads: the so-called `Battenberg’ striping pattern of large blue and yellow
171
`Looked but failed to see’ accidents
squares on the vehicle’s sides plus diagonal red and yellow retro-re¯ective stripes on
the tailgate, together with ¯ashing blue and red lights on the roof-mounted lighting-
bar.
Echelon
(45
8)
Figure 1. Plan of the experimental site used in Experiments 1 and 2, showing the
approximate location of the parked police car on the west-bound carriageway of the A27.
The plan is not to scale: in particular, the curvature of the road is exaggerated to
emphasize how west-bound drivers were unable to see the parked vehicle before a
reasonably well-de®ned point on the road.
172 M. Langham et al.
Two police video vehicles preceded a rolling road block, produced by a marked
police car which reduced tra c ¯ow towards the parked police car. Tra c was
stopped on the carriageway 4 km before the site of the parked car. When the road
past the parked police car had become clear of tra c, the ®rst police video vehicle
drove towards the parked car, travelling in the left-hand lane of the carriageway.
When 77 m away from the parked car, the moving car moved progressively into the
right-hand lane. After the ®rst video vehicle passed the parked car, the latter changed
its parking orientation before being passed by the second video vehicle. After both
video vehicles had passed the scene, the following police car escorted tra c past the
parked car.
2.2.3.1. Contents of video clips: Participants were shown six video clips, each of
which lasted 2 min. Five clips were the same for all participants: only the ®nal clip
contained the police vehicle, parked in either an `in-line’ or `echelon’ position
depending on the experimental condition. In both cases this came into view 90 s after
the beginning of the clip.
All of the video clips showed stretches of dual carriageway, ®lmed from the
driver’s perspective when travelling in the near-side lane. The ®rst, third and sixth
clips in the sequence contained `hazards’ (appendix 1). All clips showed the
participant’s vehicle travelling at 113 km/h (70 mph, the legal speed limit in the UK),
except the fourth, which showed the participant travelling at speeds between 55 and
135 km/h. These speeds were the minimum and maximum speeds of tra c recorded
on this stretch of road before testing commenced. The speed of approach was
displayed in the lower section of the video screen and was visible throughout the
experiment. Participants were made aware of this before the experiment began.
2.2.4. Apparatus: The video was projected onto a white screen situated in front of
the participant by a Bell and Howell LCD10e projector (Lincolnwood, Illinois) and a
Panasonic SVHS video player (Osaka, Japan). Participants viewed the video in a
darkened room, seated 1.1 m in front of the screen. The resultant image was 318
horizontal and 228vertical.
Participants indicated when a `hazard’ was present by pressing one of two
buttons on a button box. A BBC microcomputer recorded each button-press,
together with the time at which it occurred. All response-times were recorded with
reference to time-pulses that were inserted onto the videotape at the beginning of
each video-clip: every time a pulse occurred on the videotape, the computer’s timing
routine was reset to zero and the computer began timing again until it either
encountered another pulse or the participant pressed a button.
For both versions of the ®nal clip (containing a parked police car), the timing
pulse occurred at the same ®xed point, immediately before the parked vehicle ®rst
came into view. A period of 6 s then elapsed before the car from which ®lming took
place began to move from the left lane to the right-hand lane, in order to avert a
collision with the parked vehicle. Some 3.2 s later, the moving car passed to the right
of the parked car.
2.2.5. Procedure: Each participant was tested individually. Before beginning the
experiment, they read a short instruction sheet and completed details about their
accident involvement and driving experience (these data will not be reported here,
for reasons of brevity).
173
`Looked but failed to see’ accidents
Participants were instructed to look for possible hazards, with the de®nition of
what constituted a `hazard’ left for them to decide. Participants held a two-button
response-box and pressed the left button if they believed a hazard was in the left lane
or the right button if they believed a hazard was in the right lane. They were asked to
respond as soon as they detected a hazard. It was explained to them that the video
might contain many hazards, some or none at all.
Participants were allowed a short time to practice. The videotape took 12 min to
view, but to minimize the possibility of an increase in vigilance near the end of the
task, participants were not told how long the experiment would last.
2.3. Results
Data were collected from 60 participants. One participant’s data were
subsequently discarded because this participant experienced di culty in using
the button box at the same time as viewing the video.Two measures were taken
for each participant: time (in seconds) to respond to the police vehicle as a hazard
in the ®nal video clip (measured from the timing pulse on the videotape, which
occurred just before the parked car became visible); and the total number of
`hazard’ responses for all clips.
2.3.1. Time taken to detect the parked police car: The reaction time data were
analysed using a two-way ANOVA with independent measures on both variables.
The independent variables were the orientation of the vehicle (`echelon’ or `in-line’)
and the driver’s experience-level (`experienced’ or `inexperienced’). Main eVects were
found for vehicle orientation, F(1 ,56) = 10.49, p50.01, and driver experience level,
F(1,56) = 6.67, p50.05. There was a signi®cant interaction between the driver
experience level and vehicle orientation, F(1,56) = 6.39, p50.05. This results from
the fact that vehicle orientation signi®cantly aVected reaction-times only in the case
of experienced drivers, who responded to `echelon’-parked vehicles faster than `in-
line’ parked vehicles, t(28) = 4.22, p50.001. Table 1 shows the mean time to respond
by participants in each of the four experimental conditions.
All participants responded to the parked police car as a `hazard’, regardless of
whether it was parked `in-line’ or `echelon’. However, in evaluating these data, recall
that approximately 9 s elapsed between the parked car ®rst coming into view, and it
being passed by the car from which ®lming took place. If these results can be
generalized to real life, then most participants, even in the `echelon/experienced’
condition, would have had little time (6 s or less) in which to take action to avoid a
collision with the parked car.
Table 1. Mean time to resp ond to the statio naryp olice car in the ® nal video clip (ex periment 1)
Condition Mean time to detect (s) SD
Echelon/Experienced 3.12 0.57
In-line/Experienced 4.76 +1.64 0.99
Echelon/Inexperienced 4.56 +1.44 0.42
In-line/Inexperienced 4.50 +1.38 0.46
The ®gures preceded by + signs shows the diVerence between each condition’s mean
detection time and the shortest mean detection time (for the echelon/experienced condition).
174 M. Langham et al.
2.3.2. Overall number of `hazard’ responses: Experienced drivers considered that
the video clips contained fewer hazards than did the inexperienced drivers: the
mean number button-presses for `hazards’ was 5 for experienced drivers
(SD = 2), compared to 16 (SD = 17) for inexperienced drivers. There may be
a number of reasons for this diVerence, but one interpretation is that the
experienced driver appears to consider very little of a motorway environment to
be hazardous.
2.4. Discussion
During a hazard detection task, experienced drivers were faster to respond to an
`echelon ’ parked vehicle than they were to respond to one that was parked `in-
line’. For experienced drivers, but not for inexperienced drivers, the orientation of
the vehicle appears to have facilitated recognition that it was stationary and
hence a `hazard’. Experienced drivers were 1.64 s faster to react to the `echelon’-
parked car. This represents a considerable diVerence, bearing in mind that at the
UK speed-limit of 113 km/h on this type of road, a driver will cover
approximately 31 m/s. None of the participants failed to respond to the parked
car altogether, but it must be kept in mind that these participants were in an alert
state in which they were actively looking for hazards for a comparatively short
period of time.
3. Experiment 2: hazard perception as a function of vehicle parking orientation and
drivers’ level of attention
3.1. Introduction
This experiment attempted to more closely simulate the relative lack of
attentiveness that may be characteristic of prolonged highway driving under
normal conditions. The previous experiment was repeated, but with the
addition of a secondary task that was intended to distract participants from
attending exclusively to the hazard detection task and to encourage them to
adopt a state of inattention that is perhaps more similar to that experienced
during driving.
The task used was an adapted version of Baddeley et al.’s (1985) Working
Memory Span Test, as used by Alm and Nilsson (1994) (appendix 2). Participants
were presented with eight blocks of ®ve sentences, some of which were
meaningful (e.g. `slippers are sold in pairs’) while the rest were nonsensical (e.g.
`archbishops are made in factories’). They had to decide whether or not each
sentence made sense, and also had to recall the last word of each sentence within
a block, when prompted to do so. In terms of the demands placed on
comprehension and memory, this task simulates some of the essentials of mobile
telephone conversations while drivingÐthe purpose for which it was used by Alm
and Nilsson (1994).
3.2. Method
3.2.1. Design: The experiment used an independent measures design. Participants
viewed either a police vehicle parked `in-line’ or parked `echelon’ in one video clip
amongst a series of clips that were otherwise identical for all participants. Half of
the participants performed this task under divided-attention conditions, while the
other half were able to devote their undivided attention to the hazard-perception
175
`Looked but failed to see’ accidents
task. Each participant therefore participated in one of the following four
conditions:
(a) `undivided-attention echelon’;
(b) `divided-attention echelon’;
(c) `undivided-attention in-line’; and
(d) `undivided-attention in-line’.
Participants’ reaction times were recorded, together with their decision about
whether or not each clip contained a `hazard’.
3.2.2. Participants: 81 participants were recruited, principally staVand students at
the University of Sussex. There were 40 males and 41 females. Mean age was 39 years
(range 25 ± 61 years). All participants held a full driving licence, with a mean of 17.8
years experience (rang e 5 ± 43 ye ars). Pa rticipants were ra ndomly allocated to one of
the four experimental conditions. There were 21 participants in the `undivided-
attention echelon’ condition; 20 in the `divided-attention echelon’ condition; 19 in
the `undivided-attention in-line’ condition; and 21 in the `undivided-attention in-line’
condition.
3.2.3. Preparation of stimulus materials: Participants saw ®ve of the six video clips
used in Experiment 1: the ®rst clip was discarded in order to make the divided-
attention task last as long as the video-clip sequence.
The ®rst four clips were the same for all participants. The ®nal clip contained the
hazard of a marked police vehicle, parked in either an `in-line’ or `echelon’ position,
depending on the experimental condition.
3.2.4. Apparatus: All details of video-clip display and measurement of participant
responses were the same as for Experiment 1. The divided attention task was
presented auditorily via an Hitachi TRK-640E cassette recorder (Tokyo, Japan).
3.2.5. Procedure: Each participant was tested individually. Before beginning
the experiment, they read a short instruction sheet and completed details about
their accident involvement and driving experience. Participants were instructed
to look for possible hazards, with the de®nition of what constituted a `hazard’
left to them to decide. As in Experiment 1, participants held a two-button
response-box and pressed the left button if they believed a hazard was present
in the left lane or the right button if they considered there was a hazard in
the right lane.
Participants in the two divided-attention conditions performed the Working
Memory Span test at the same time as they performed the hazard-perception task.
The Working Memory Span test was pre-recorded onto audiocassette tape. Forty
sentences were presented auditorily to the participant, while the videotape was being
shown. After each sentence was presented (e.g. `archbishops are made in factories’),
there was a 4-s silence in which the participant had to respond `true’ or `false’. After
each set of ®ve sentences, a tone signalled the start of a 20-s period of silence, during
which the participant was expected to recall the last word from each of the preceding
5 sentences. Each presentation of 5 sentences (including response breaks) took
approximately 60 s. The participant’s responses were recorded by the experimenter.
176 M. Langham et al.
3.3. Results
One measure was taken for each participant: the time (in seconds) to respond to the
parked police vehicle in the ®nal video clip. A total of 76 participants pressed a
button in response to the parked police car: ®ve failed to respond to it altogether.
3.3.1. Time taken to respond to the parked police car: The reaction time data for the
76 participants who made a response were analysed using a two-way ANOVA with
independent measures on both variables. The independent variables were the
orientation of the vehicle (`echelon’/or `in-line’) and the attention condition
(`undivided attention’ or `divided attention’). Main eVects were found for vehicle
orientation (F(1,72 ) = 13.3 3, p50.001) and attention (F(1,72) = 15.96, p50.001).
There was no signi®cant interaction between orientation and attention
(F(1,72) = 0.006, n.s.).
Table 2 shows the mean time to respond by participants in each of the four
experimental conditions. Inspection of these means, together with the results of the
ANOVA, suggests that there were two independent in¯uences on participants’
reaction times. First, performing the logical reasoning task markedly increased
reaction times, regardless of the parked police car’s orientation. Second, the police
car’s orientation aVected reaction times, with participants slower to respond when
the car was positioned in-line than when it was parked in the echelon position
(replicating the main ®nding of Experiment 1).
3.3.2. Frequency of hazard detection failures in the ®nal video-clip: Of the ®ve
participants who failed to respond to the parked police car as a `hazard’, four were in
the in-line/divided attention condition, and one was in the in-line/undivided
attention condition. No participants in the two echelon conditions failed to respond
to the presence of the police car.
3.3.3. Performance on the divided attention task, in relation to hazard-detection
performance: For each participant, a Working Memory Span test-score was
computed: this consisted of the number of correct judgements on the logical
reasoning task, added to the number of words correctly recalled. To assess whether
there was a trade-oVbetween performance on the divided attention task and
performance on the hazard-detection task, the reasoning task measure was
correlated with the participant’s reaction time to respond to the parked police car.
A signi®cant positive relationship was found between the two measures (Pearson’s
Table 2. Mean time to respond to the stationary police car in the ®nal video clip
(experiment 2).
Condition Mean time to detect (s) SD
Echelon/Undivided Attention 3.72 0.53
Echelon/Divided Attention 4.44 +0.72 0.77
In-line/Undivided Attention 4.38 +0.66 0.94
In-line/Divided Attention 5.13 +1.41 0.94
The ®gures preceded by the + signs show the diVerence between each conditions mean
detection time and the shortest mean detection time (for the echelon/undivided attention
condition)
177
`Looked but failed to see’ accidents
r= 0.67, 75 df, p50.01): participants who scored highly on the logical reasoning
task tended to take longer to respond to the parked police car.
3.4. Discussion
There was a highly statistically signi®cant eVect of attention-state on response-times:
participants were signi®cantly slower in the divided attention conditions than in the
undivided attention conditions. There was also a highly statistically signi®cant eVect
of orientation of vehicle: participants were slowest to identify the police car as a
`hazard’ when it was parked `in-line’ than when it was parked at an angle. This
experiment replicated the advantage of `echelon’ parking over `in-line’ parking that
was found for experienced drivers in Experiment 1, although the diVerence in
response-time was somewhat less than in the ®rst study.
No participants in either of the `echelon’ parking conditions failed to respond to
the police car; however, four participants in the in-line/divided attention condition,
and one participant in the in-line/undivided attention condition, failed to identify it
as a `hazard’. These numbers are too small to perform any statistics on, but are
suggestiveÐespecially given that the experimental session was short and participants
were alert, conditions which would tend to militate against detection failure.
4. General discussion
Any individual accident is likely to be the outcome of a number of contributory
factors, and collisions with parked vehicles may occur for diverse reasons. Let us
consider various explanations for this kind of accident in turn.
It seems highly unlikely that drivers were unable to detect the parked vehicles due
to sensory limitations. Our accident reports were limited to collisions that had
occurred during daylight conditions, when there was good unrestricted visibility. A
parked police car presents a large, bright stimulus, with ¯ashing multi-coloured
lights and retro-re¯ ective striping that are designed to enhance conspicuity and
attract attention: it would be well above physical detection thresholds long enough
for an approaching driver to see it, and take avoiding actionÐand yet at least some
of the accident reports suggest that the driver failed to swerve or brake until just
before impact. Moreover, the vehicle would have been directly ahead before the
collision occurred, and hence located for some time in central as opposed to
peripheral vision. While some collisions may occur because the driver failed to see a
stationary vehicle due to sensory limitations (e.g. because of fog or because the view
of the parked vehicle was obscured in some way), there remains a core sample of
accidents that cannot be explained in these terms.
The results of our two experiments are also di cult to explain purely in terms of
sensory conspicuity. First, the orientation of the parked police vehicle aVected how
quickly participants responded to it as a `hazard’, even though the two orientations
of parked car were (objectively) similarly visible on the video. Second, experience
aVected response times: in Experiment 1, experienced drivers identi®ed a parked
police car as a `hazard’ faster when it was `echelon’-parked than when it was parked
`in line’. This was not the case for inexperienced drivers. Neither of these ®ndings can
be accounted for by theories that account for conspicuity solely on the basis of the
physical properties of the objects to be detected.
There are other explanations for these kinds of collisions which can also be ruled
out: for example, while it might bene®t a driver to attempt to force a motorcyclist or
cyclist to give way to them at an intersection at which the latter has right of way,
178 M. Langham et al.
there is nothing to be gained by colliding with a stationary vehicle! Also, since the
vehicle that was collided with is stationary, all of the responsibility for the accident
falls on the driver of the moving vehicle; to some extent this simpli®es interpretation
of why the accident occurred, since one can rule out the possibility that the other
party involved in the collision had contributed to the accident occurring.
The accident data that we collected are more consistent with the hypothesis that
these LBFS accidents arise for `cognitive’ reasons. There are a number of possible
explanations, all of which have some plausibility. First, there may have been a failure
in vigilance on the part of the oVending driver. Since Mackworth’s studies in the
1950s (Mackworth 1957), it has been appreciated that a pronounced deterioration in
vigilance performance may occur after as little as 20 min (see Cabon et al. 1993,
Matthews et al. 1993, and Edkins and Pollock 1997 for discussions of vigilance
failure in long-haul pilots and train drivers).
Motorway or dual-carriageway driving conditions may be ideal for promoting a
decline in vigilance. Although motorway driving involves high speeds, it is relatively
undemanding and unarousing most of the time. Due to advances in vehicle and
highway design, together with the fact that everyone around the driver is travelling in
the same direction and at similar speeds, little sensation of speed is obtained. Driving
becomes largely a matter of monitoring one’s lane position, maintaining a safe
distance from the vehicle in front, and changing lanes to overtake slower-moving
vehicles if necessary. Other vehicles are generally behaving quite predictably, and
little attention has to be paid to them, other than cursorily registering their presence.
Also, drivers very rarely encounter stationary vehicles on such roads (at least, not
immediately in front of them), and so have no reason to look out for them. In terms
of vigilance theory, the event which is to be detected occurs with a very low
frequency amongst many other, fairly similar events. This militates against its
successful detection (Lewis 1973).
Another factor that may contribute to these accidents is driver fatigue. In our
accident data, there were some indications that collisions occurred most often in the
middle of the day, with a peak around 14:00 h. One has to be wary of drawing
conclusions from such a small sample, but if this is a valid observation, it is
consistent with Mavje and Horne’s (1994) ®nding that there is a propensity for
sleepiness in the early afternoon, between midday and 16:00 h. Moreover, they claim
that if a task lacks interestÐas is the case with motorway drivingÐthen the eVects of
this post lunch `dip’ in arousal are more noticeable.
Although it is possible to distinguish between fatigue and vigilance decrement on
theoretical grounds, in practice their eVects are likely to be di cult to disentangle: a
tired driver is unlikely to be highly vigilant. An explanation entirely in terms of
physical fatigue would not explain the results of our two laboratory studies, where
participants were alert, performing for a relatively short period of time and (one
hopes!) reasonably awake. As with explanations in terms of sensory conspicuity, a
simple `fatigue’ explanation has di culty in accounting for the eVects of vehicle
orientation and driver experience in these experiments.
One contributory factor in this type of accident may be that drivers detect the
stationary vehicle in front of them, but misinterpret what they see. The driver who
claims not to have seen a police car before colliding with it may have detected it but
believed that it was moving. A stationary car on a motorway or dual-carriageway
looks very similar to one that is moving in the same direction as oneself: there are no
obvious cues that it is diVerent to any of the other vehicles travelling on that road.
179
`Looked but failed to see’ accidents
The ¯ashing blue lights on a police car may actually contribute to this false
impression, since police cars with ¯ashing lights are normally seen moving rapidly on
their way to an emergency (Shinar and Stiebel 1986). Therefore a driver may well
adopt an erroneous hypothesis about the parked police car on the road ahead, based
on their previous experiences of seeing police vehicles. By the time radial expansion
of its retinal image occurs to any signi®cant degree, it is probably too late for the
driver to take evasive action. This is compounded by the fact that drivers’ reaction
times may be as long as 1.5 s when response is required to an unexpected hazard, and
a lane-changing manoeuvre may take 8 s or more to complete (review in Olson
1996).
Davis (1958) used the concept of false hypotheses in an attempt to explain why
train drivers sometimes failed to stop at a red light (see also Borowsky and Wall
(1983) and Hurst and Hurst (1982) for an explanation of pilot error in terms of false
hypotheses). Davis suggested that a false hypothesis is particularly likely to be
adopted when the operator’s expectancy is very high because of repeated exposures
to the situation, and when attention is `elsewhere’ (i.e. the operator is distracted by
another task). These conditions are likely to apply in the case of motorway and dual-
carriageway driving. Once an operator forms a hypothesis about a given situation, it
appears to be resistant to revision, despite information to the contrary. A driver who
sees a police vehicle displaying lights might construct an initial hypothesis that it is
moving, and retain this despite the subsequent development of con¯icting cues such
as looming, until it is too late to avert a collision. This explanation would account
for the results of Experiments 1 and 2. Because it is at an angle in the road, the
`echelon’-parked police car not only provides an unusual stimulus to oncoming
drivers, but one that is at odds with the hypothesis that the vehicle is moving. In the
case of Experiment 1, this orientation cue appears to have been an aid only to
experienced drivers who have extensive knowledge of what position a moving vehicle
ought to have on the road ahead.
False hypotheses or vigilance decrements are possibly most likely to occur when
the driver believes that he is `as good as home’ (Davis 1958). This may be re¯ected in
our obtained accident data, given that many of these accidents occurred close to the
oVending driver’s home (although there are problems in interpreting these data, as
mentioned in the Introduction).
One of our reviewers suggested that drivers’ perceptual sampling strategies, and
their distribution of attention, might have a role to play in LBFS accidents. Drivers
may base at least some of their behaviour on periodic `samples’ of the view ahead,
rather than monitoring their surroundings continuously. The best evidence for this
comes from Godthelp’s (1985) measurements of lane-changing performance in
drivers, under varying conditions of visual feedback. He found that even when an
occluding visor prevented drivers from seeing the road ahead for 3 s, eVects on lane-
positioning were small. We do not know of any studies that have demonstrated a
comparable sampling strategy in relation to other aspects of driving, such as
interpretation of the surroundings. However, there are now numerous studies on
`change blindness’ (Rensink et al. 1996), which demonstrate that individuals
maintain remarkably impoverished representations of their surroundings, and may
fail to notice highly marked changes to objects. This is well-documented when the
changes occur to objects to which the individual is not attending, but can occur even
when the objects are the focus of attention (Levin and Simon 1997). As one of our
reviewers suggested, it is possible that a driver might detect a parked police vehicle,
180 M. Langham et al.
decide from the available cues (lighting and road position) that it is a moving vehicle,
and not `sample’ it again until it is too late to avert a collision. It is clear to see how
there might be an interaction between the adoption of a false hypothesis, faulty
perceptual sampling of this kind, and change blindness.
Finally, the results of Experiment 2 are consistent with a growing body of data
suggesting that driving performance may be impaired when the driver’s attention is
distracted, for example by the use of mobile telephones (Redelmeier and Tibshirani
1997). Using the same Working Memory Span task as we used in Experiment 2, Alm
and Nilsson (1994) found that distracted drivers had lengthened reaction times and
an increased subjective mental workload. Some studies (Briem and Hedman 1995)
have found that the eVects on performance of simulated mobile telephone use are
greatest when the driving task is di cult. The present study suggests that distraction
may have consequences even when the driving task is perceived as relatively
undemanding, such as when driving on a motorway or dual-carriagewayÐprecisely
the kinds of conditions under which most drivers would consider using a mobile
telephone. As Summala (personal communication, 2001) has pointed out, the
increases in response times produced by a secondary task in our experiment are
similar not only to those reported from the laboratory studies of Alm and Nilsson
(1994), but also comparable to those found in real-life studies of the eVects of mobile
telephone use on drivers’ responses to braking by the car in front of them (Brookhuis
et al. 1991, Lamble et al. 1999).
So far, we have assumed that our experimental manipulations were a valid
simulation of the conditions preceding a collision. However, laboratory studies
cannot adequately simulate all of the properties of the real-world conditions under
which this type of accident takes place. An obvious problem is that video very poorly
reproduces the physical characteristics of the parked vehicle and its surroundings
(ambient lighting, vehicle lighting, ¯uorescence and re¯ectivity of materials).
Conditions are very far from those found in the real world, and video does not
allow for the evaluation of the ¯uorescent properties of the materials ®tted to police
vehicles.
Second, objections might be raised to the extent to which our task was a valid
simulation of `driving’, given that participants had an essentially passive role, with
no control over the vehicle’s movements. It might be thought that it would have been
preferable to run our experiments in a driving simulator. However, driving consists
of many diVerent processes, including tactical decision-making, interpretation of
perceptual input, and vehicle control. Most driving simulators emphasize vehicle
control at the expense of perceptual realism. This may be satisfactory for
investigating many aspects of driving, but it was felt that in the current context,
we were justi®ed in ignoring the motor control elements of driving, and focusing on
the perceptual components of the task. Although our methods simulate only a small
part of the driving task, they may be simulating the more important part of driving,
at least as far as LBFS accident causation is concerned: since vehicle control is
largely automatized in the experienced driver, it could be argued that removing this
component of the driving task in our experiment is unlikely to make much of a
diVerence to the cognitive `overheads’ of the driver’s task.
Ideally, a complete simulation of the driving task would include all aspects of
driving, but this is unfeasible at present: even the best driving simulators do not
provide su ciently realistic representations of the outside world, and even if they
did, they would fail to provide the threat to personal safety that real-life driving
181
`Looked but failed to see’ accidents
provides. No laboratory experiment, no matter how eVective the simulation, can
adequately simulate the dangers of real-life driving: participants know that they are
in an experiment, and that their mistakes have no important consequences for
themselves or other road-users. The only way to test our hypotheses conclusively
would be to run ®eld experiments, with the vehicle under the participant’s control,
but there are obvious ethical objections to this; one should not underestimate the
potential dangers that were involved in collecting our video footage, given that the
road was open to the general public at the time.
A further problem concerns how our participants construed their task. It was left
to them to de®ne what might constitute a `hazard’ that needed a response. Although
our main reason for this procedure was to enable us to present participants with an
unexpected hazard (the parked car), we can also justify it on the grounds of
ecological validity: in real life, each driver decides for themselves what represents a
hazard, whether or not a response is required, and how urgently it must be executed.
However, by leaving the participant’s task so ill-de®ned, we may potentially have
introduced some ambiguity into the interpretation of our results: participants may
have responded slowly or not at all to the parked car, not because they did not see it
for what it was, but because they did not consider it to be a `hazard’.
In practice, given that the parked car occupied an entire lane, we believe that it is
unlikely that participants did not include this stimulus in their personal de®nitions of
`hazards to be responded to’, especially given the implicit demand characteristics of
the experiment. Recall that we had asked participants to make diVerent responses
depending on whether a hazard was located in the left or right lane of the
carriageway, a request which might be expected to give them a strong hint about how
they should respond when confronted with a lane that was eVectively closed to
approaching vehicles. Also, this would not explain the diVerences in participants’
responses in the `in-line’ and `echelon’ parking conditions, since in both cases, the
parked car clearly blocked its lane to oncoming tra c, and required a similar
urgency of response by the view er.
With all of these limitations in mind, the results of this study have a number of
implications. On a theoretical level, the accident data clearly demonstrate that high
levels of conspicuity (in sensory terms) do not guarantee detection of a vehicle, a
conclusion supported by the results of our two experiments. They also suggest that
cognitive factors, such as drivers’ expectations, may play an important role in
causing this kind of `looked but failed to see’ accident. Precisely which cognitive
factors are involvedÐfatigue, false hypotheses, inattention or a combination of all of
theseÐremains to be determined by future studies.
On a practical level, the results suggest that drivers of all vehicles that are
stationary on a high-speed road should try to draw attention to the fact that their
vehicle is motionless: parking at an angle is one way to achieve this. However, since
this is not foolproof as a means of avoiding collisions, the safest action civilian
motorists could take is probably to wait in a place of safety, well away from their
vehicle.
Acknowledgements
The authors are deeply indebted to the Tra c Division of Sussex Police, without
whom this research could not have been conducted, and to the Institute of Tra c
Accident Investigators and our anonymous survey respondents for their generous
and enthusiastic help with this research. Thanks also to Heikki Summala and an
182 M. Langham et al.
anonymous reviewer for their constructive comments on an earlier version of this
manuscript.
References
ALM, H. and NILSSON, L. 1994, Changes in driver behaviour as a function of handsfree mobile
phonesÐa simulator study, Accident Analysis and Prevention,26, 441 ± 451 .
ASSOCIATION OF CHIEF POLICE OFFICERS (ACPO) 1996, The Motorway Manual (London: Home
O ce).
BADDELEY , A., LO GIE, R. and NIMMO-SMITH, I. 1985, Components of ¯uent reading, Journal of
Memory and Language,24, 119 ± 131.
BOROWSKY, S. and WALL, R. 1983, Flight experience and naval aircraft misha ps, Aviation Space
and Environmental Medicine,54, 440± 446.
BRIEM, V. and HEDMAN, L. R. 1995, Behavioural eVects of mobile telephone use during
simulated driving, Ergonomics,38, 2536 ± 25 62.
BROOKHUIS, K. A., DE VRIES, G. and DE WA ARD, D. 1991, The eVects of mobile telephoning on
driving performance, Accident Analysis and Prevention,23, 309 ± 316.
CABON, P., COBLENTZ, A., MOLLA RD, R. and FO UILLO T, J. P. 1993, Human vigilance in railway
and long-haul ¯ight operation, Ergonomics,36, 1019 ± 1033.
CAIRNE Y, P. and CATCHPOLE, J. 1995, Patterns of perceptual failure at intersections of arterial
roads and local streets, in A. G. Gale (ed.), Vision in Vehicles,VI (Amsterdam: Elsevier
Science).
CERCA RELLI, L. R., ARNOLD, P. C., RO SMA N, D. L. SLEE T, D. and THORNETT, M. L. 1992, Travel
exposu re and choice of com parison for examining m otorcycle cons picuity by analysis o f
crash data, Accident Analysis and Prevention,24, 363 ± 368 .
DAVIS, R. D. 1958, Human engineering in transport accidents, Ergonomics,2, 24 ± 3 3.
DRASKOCZY, M. 1989 , Bus driv er behav iour at ra il-highway g rade crossing s, in T. R othengatter
and R. De Bruin (eds), Road User Theory and Research (Maastricht, The Netherlands:
Van Gorcum).
EDKINS, G. D. and POLLOCK, C. M. 1997, The in¯uence of sustained attention on railway
accidents, Accident Analysis and Prevention,29, 5 33 ± 539.
FULTON, E. J., KI RBY, C. and STROUD, P. G. 1980, Daytime motorcycle conspicuity,
Supplementary Report 625, Transport and Road Research Laboratory, Crowthorne,
Berks.
GODTHE LP, J. 1985, Precognitive control: open- and closed-loop steering in a lane-change
manoeuvre, Ergonomics,28, 1419 ± 1438.
HILLS, B. L. 1980, Vision, visibility and perception in driving, Perception,3, 434 ± 467.
HOLE, G. and TYRRELL, L. 1995, The in¯uence of perceptual set on the detection of
motorcyclists using daytime headlights, Ergonomics,38, 1326 ± 13 41.
HURST, R. and HURST, L. 1982, Pilot Error: The Human Factor, 2nd ed. (London: Granada).
JANOFF, M. S. and CASSEL, A. 1971, EVect of daytime motorcycle headlight law on motorcycle
accidents, Research Record, N o. 377, 56 ± 63.
LAMBLE, D., KAURANEN, T., LAA KSO, M. and SUMMA LA, H. 1999, Cognitive load and detection
thresholds in car following situations: safety implications for using mobile (cellular)
telephones while driving, Accident Analysis and Prevention,31, 617 ± 62 3.
LEIBOWI TZ, H. W. 1985, Grade crossing accidents and human factors engineering, American
Scientist,73, 55 8 ± 562.
LEVIN, D. T. and SIMO N, D. J. 1997, Failure to detect changes to attended objects in motion
pictures, Psychonomic Bulletin and Review,4, 501 ± 5 06.
LEWIS, M. F. 1973, Frequency of anti-collision observing responses by solo pilots as a function
of tra c density, ATC tra c warnings, and competing behaviour, Aerospace Medicine,
44, 1 048 ± 105 0.
MACKWORTH, N. H. 1957, Some factors aVecting vigilance, Advancement of Science,53, 389 ±
395.
MATTHEWS, G., DAVIES, D. R. and HOLL EY, P. J. 1993, Cognitive predictors of vigilance, Human
Factors,35 (1), 3 ± 24.
183
`Looked but failed to see’ accidents
MAVJE, V. and HORNE, J. A. 1994, Boredom eVects on sleepiness alertness in the early
afternoon vs. early evening and interactions with warm ambient-temperature, British
Journal of Psychology,85, 317 ± 333.
OLSON, P. L. 1989, Motorcycle conspicuity revisited, Human Factors,31 ( 2), 141 ± 146.
OLSON, P. L. 1996, Forensic Aspects of Driver Perception and Response (Tucson, AZ: Lawyers
and Judges).
REDELMEIER, D. A. and TIBSHIRANI, R. J . 1997, Association b etween cellular telephone calls and
motor vehicle collisions, The New England Journal of Medicine,338, 45 3 ± 458.
RENSINK , R. A., O’REGAN , J. K. and CLAR K, J. J. 1996, To see or not to see: the need for
attention to perceive changes in scenes, I nvestigative Op hthalmology and Visual Science,
37, S213.
RUMAR , K. 1990, The basic driver error: late detection, Ergonomics,33, 1281 ± 1 290.
SHINAR , D. and STIEBE L, J. 1986, The eVectiveness of stationary versus moving police vehicles
on compliance with speed limit, Human Factors,28, 3 65 ± 371.
STAUGHTON, G. C. and STORIE, V. J. 1977, Methodology of an in-depth accident investigation,
Survey Report No. 672, Transpor t and R oad Resear ch Laborato ry, Crowthorne, Berks.
SUMMALA , H. 2001, Personal communication.
WULF, G., HANCO CK P. A. and RAHIMI, M. 1989, Motorcycle conspicuityÐan evaluation and
synthesis of in¯uential factors, Journal of Safety Research,20, 153 ± 176.
Appendix 1. Summary of video clips.
Clip 1: this contained the hazard of a slow-moving motorcycle which appeared
30 s after the beginning of the clip.
Clip 2: this showed clear carriageway, with no hazards.
Clip 3: this contained two hazards, in the form of slow-moving vehicles in the
near-side lane. The ®rst vehicle, a truck, occurred 1 min after the start of the clip.
The other, a car, appeared 90 s after the start.
Clip 4: this showed clear carriageway, with no hazards. The vehicle from which
®lming was being performed varied in speed between 55 and 135 km/h.
Clip 5: this showed clear carriageway, with no hazards.
Clip 6: this showed either a police vehicle parked `in-line’ or one parked `echelon’,
coming into view 90 s after the beginning of the clip.
Appendix 2. Working memory span test items.
Slippers are sold in pairs.
The policeman ate the apple.
The train bought a newspaper.
The banker saw his car.
Rivers are crossed by bridges.
The girl sang the water.
The teacher spoke to the student.
The bird swallowed the worm.
Archbishops are made in factories,
The world divides the equator.
The letter spoke to the package.
The cat divorced the milk.
The man saw the woma n.
The removal ®rm took a bed.
The boy brushed his teeth.
The farmer chased the dog.
The freezer was in the ice-cream.
184 M. Langham et al.
The horse is riding the man.
The sailor beat the idea.
The jelly eats the screwdriver.
The plane swims in the sky.
Fish live in water.
The child reads the book.
The waiter served the food.
The man taught the cheese.
The cashier counted the money.
The brick threw the builder.
The soldier fought the battle.
The frog helped the vase.
The cow ate the grass.
The moon orbits the earth.
The cook sailed the kitchen.
The oven heats the food.
The dog wags its tail.
The artist ¯ew the spider.
Elephants travel in cars.
The doctor examines the patient.
Thermometers tell the time.
Wine is bought in carpets.
Potatoes grow on trees.
185
`Looked but failed to see’ accidents
... Modifying the tire color not only costs less than attaching a blinking light but also facilitates self-signaling without any active behavior by the bicyclist. Signal noise from the use of blinking lights on the rear end of a bicycle may result in severe bicycle-vehicle collisions (Langham et al., 2002;Madsen et al., 2013). Finding the optimal level of signaling for the presence of a bicycle can reduce the signal noise and improve the cognitive errors of drivers. ...
... In these studies, although the drivers looked in the right direction, they could not perceive the signal in time (Hills, 1980;Koustanaï et al., 2008). In the case of a rear light, the flashing beacon signals the presence of a bicycle; however, it is often perceived as light noise that distracts road users and possibly causes cognitive errors (Langham et al., 2002). ...
... On a residential road in Japan with a non-demarcated lane, the frequency of rear-end collisions between vehicles and bicycles was the highest between 2012 and 2016 ((2011) n.d.). Drivers were looking but failing to see bicyclists in time (Langham et al., 2002) owing to a lack of self-signaling. Dynamic objects have a greater potential to attract the attention of road users compared to static objects. ...
Article
Full-text available
Perception time (PT) is a major factor affecting a driver’s ability to detect and recognize a bicycle. Researchers have attempted to enhance the PT and gain insights into why drivers fail to see bicyclists before collisions despite looking, even when they are at a safe driving distance from the bicycle. Previous studies have focused on the detection distance and recognition distance of bicycles within 600 ms. PT is the main factor for avoiding collisions; however, it has been shown that a control bicycle as well as treated bicycles can be detected from greater distances. This study aims to evaluate the early detection and recognition of bicycles owing to the impact of conspicuity treatments, such as white stripes on a red background (WRED), a high-visibility jacket (HVJ), reflective tape (RT), and their combinations, to achieve longer detection and recognition distances under day/night conditions. The detection and recognition distances of WRED tire treatment were compared with those of an HVJ, RT, and their combinations, based on PTs of 250 and 600 ms. The same treatments were applied and compared at the required PT for the safe driving distance of a bicycle. The respondents provided their perceptions based on video surveillance data presented on a computer screen. The detection and recognition distance of WRED treatment combined with an HVJ was significantly greater under all conditions except twilight with car headlights and nighttime with car headlights for a PT of 600 ms. Furthermore, for this combination, the PT was significantly shorter under all conditions except nighttime with car headlights. The effects of gentle self-signaling of a bicycle via the combination of WRED treatment and an HVJ can reduce the PT for detecting a bicycle and increase the detection and recognition distance under all lighting conditions. Passive safety measures based on these results can support drivers, who might otherwise look but fail to see bicyclists in time. In summary, the combination of WRED treatment with an HVJ is strongly recommended to achieve cost-effective self-signaling of a bicycle.
... Modifying the tire color not only costs less than attaching a blinking light but also facilitates selfsignaling without any active behavior by the bicyclist. Signal noise from the use of blinking lights on the rear end of a bicycle may result in severe bicycle-vehicle collisions [22], [23]. Finding the optimal level of signaling for the presence of a bicycle can reduce the signal noise and improve the cognitive errors of drivers. ...
... In these studies, although the drivers looked in the right direction, they could not perceive the signal in time [37], [38]. In the case of a rear light, the ashing beacon signals the presence of a bicycle; however, it is often perceived as light noise that distracts road users and possibly causes cognitive errors [22]. ...
... On a residential road in Japan with a non-demarcated lane, the frequency of rear-end collisions between vehicles and bicycles was the highest between 2012 and 2016 [10]. Drivers were looking but failing to see bicyclists in time [22] owing to a lack of self-signaling. Dynamic objects have a greater potential to attract the attention of road users compared to static objects. ...
Preprint
Full-text available
Perception time (PT) is a major factor affecting a driver’s ability to detect and recognize a bicycle. Researchers have attempted to enhance the PT and gain insights into why drivers fail to see bicyclists before collisions despite looking, even when they are at a safe driving distance from the bicycle. Previous studies have focused on the detection distance and recognition distance of bicycles within 600 milliseconds (ms). PT is the main factor for avoiding collisions; however, it has been shown that a control bicycle as well as treated bicycles can be detected from greater distances. This study aims to evaluate the early detection and recognition of bicycles owing to the impact of conspicuity treatments such as white stripes on a red background (WRED), a high-visibility jacket (HVJ), reflective tape (RT), and their combinations in order to achieve longer detection and recognition distances under day/night conditions. The detection and recognition distances of WRED tire treatment were compared with those of an HVJ, RT, and their combinations, based on PTs of 250 and 600 ms. The same treatments were applied and compared at the required PT for the safe driving distance of a bicycle. The respondents provided their perceptions based on video surveillance data presented on a computer screen. The detection and recognition distance of WRED treatment combined with an HVJ was significantly greater under all conditions except twilight with car headlights and nighttime with car headlights for a PT of 600 ms. Furthermore, for this combination, the PT was significantly shorter under all conditions except nighttime with car headlights. The effects of gentle self-signaling of a bicycle via the combination of WRED treatment and an HVJ can reduce the PT for detecting a bicycle and increase the detection and recognition distance under all lighting conditions. Passive safety measures based on these results can support drivers, who might otherwise look but fail to see bicyclists in time. In summary, the combination of WRED treatment with an HVJ is strongly recommended to achieve cost-effective self-signaling of a bicycle.
... Another reason for missed hazards is when the missed object doesn't fit the driver's attentional expectations. For example, when a cyclist comes from an unexpected direction [50], or when a police car is located in an unexpected location [51]. ...
... Our second aim was to examine gaze behavior, to test whether participants indeed directed their gaze towards the critical stimuli. This was done to demonstrate that the reported invisibility of the stimulus was not due simply to not looking in the general direction of the bus stop (i.e., that participants "looked but failed to see" rather than not gazing in that direction; Langham et al., 2002;White & Caird, 2010). We examined the average gaze duration on all presentations of the intact critical stimulus in each visibility 1 trial by plotting the trial data and performing a Bayesian one-sample t-test against zero. ...
Article
Full-text available
Unconscious processing has been widely examined using diverse and well-controlled methodologies. However, the extent to which these findings are relevant to real-life instances of information processing without awareness is limited. Here, we present a novel inattentional blindness (IB) paradigm in virtual reality (VR). In three experiments, we managed to repeatedly induce IB while participants foveally viewed salient stimuli for prolonged durations. The effectiveness of this paradigm demonstrates the close relationship between top-down attention and subjective experience. Thus, this method provides an ecologically valid setup to examine processing without awareness. Supplementary information The online version contains supplementary material available at 10.3758/s13428-024-02401-8.
... The term LBFTS comes from research on road collisions [15] (Box 1). In 'On-the-spot' interviews, after a collision drivers often insist that, yes, they did look at the location where the pedestrian or motorcyclist was located but, somehow, they just did not 'see' the victim (e.g., [23][24][25]). In medicine, radiologists routinely end up in court [26], sued for malpractice when they fail to report some clinically significant finding that is 'retrospectively visible' [27], meaning that the problem is visible when attention is drawn to it later. ...
Article
Humans routinely miss important information that is ‘right in front of our eyes’, from overlooking typos in a paper to failing to see a cyclist in an intersection. Recent studies on these ‘Looked But Failed To See’ (LBFTS) errors point to a common mechanism underlying these failures, whether the missed item was an unexpected gorilla, the clearly defined target of a visual search, or that simple typo. We argue that normal blindness is the by-product of the limited-capacity prediction engine that is our visual system. The processes that evolved to allow us to move through the world with ease are virtually guaranteed to cause us to miss some significant stimuli, especially in important tasks like driving and medical image perception.
... Authors attributed their findings to the concept of cognitive conspicuity, which posits that experiential knowledge of cycling (i.e., being a driver who cycles) leads to the conspicuity (or salience) of cycling-related stimuli while driving. In this case, a cyclist driver's attention is directed, in a top-down manner, at objects and locations which align with the driver's expectations and experiences, such as other cyclists or VRUs on the road (Hole & Tyrrell, 1995;Langham et al., 2002;Rogé & Vienne, 2015;Wulf et al., 1989). Thus, modulation of visual attentionwhich is based on knowledge and expectationsmay translate to a higher probability of visually scanning VRU-related areas while navigating the roadway. ...
Article
The safety of Vulnerable Road Users (VRUs), such as pedestrians and cyclists, is a serious public health concern, especially at urban intersections. A major reason for vehicle-VRU collisions is driver attentional errors. Prior studies suggest that cross-modal transportation experiences (e.g., being a driver who also cycles) improve visual attention allocation toward VRUs. However, these studies were conducted in simulators or in a laboratory, limiting their generalizability to real world driving. We utilized an instrumented vehicle equipped with eye tracking technology to examine (a) the prevalence of drivers' visual scanning failures toward VRUs at real intersections and (b) whether there is an effect of cycling experience on this prevalence. Twenty-six experienced drivers (13 cyclists and 13 non-cyclists), between the ages of 35 and 54, completed 18 different turns at urban Toronto intersections, for which gaze and video data were utilized to determine drivers' visual scanning failures towards areas where conflicting VRUs could approach. Among the 443 unique turn events, 25% were identified as having a visual scanning failure. Results from a mixed effects logit model showed that the odds of committing visual scanning failures towards VRUs during a turning maneuver at an intersection were 2.01 times greater for drivers without cycling experience compared to drivers with cycling experience. Given that our participants represented a low crash-risk age group, this study suggests that the rate at which VRUs are unattended to may be much higher.
Article
Tidak dapat dipungkiri dalam menjalankan aktivitas sehari-hari diperlukan pendukung moda transportasi untuk memudahkan mobilitas. Seiring meningkatnya aktivitas masyarakat Indonesia, maka penggunaan moda transportasi juga meningkat. Adanya peningkatan jumlah kendaraan yang ada di jalan tentunya berpengaruh terhadap permasalahan transportasi, salah satunya adalah peningkatan jumlah kecelakaan lalu lintas. Berdasarkan data peningkatan jumlah kecelakaan dan kerugian materi Badan Pusat Statistik, jumlah kecelakaan yang terjadi meningkat sebesar 1.000.000 dengan kerugian materi sebesar Rp 250.000.000.000,- pada tahun 2004-2014. Penelitian ini bertujuan untuk menganalisis tugas pengemudi dengan memecah operasi-operasi tugas pengemudi menjadi sub-sub operasi sebagai landasan untuk mengetahui sumber gangguan apa saja yang dapat menganggu pengemudi selama mengemudi. Objek penelitian ini merupakan angkutan umum kategori taksi yang merupakan transportasi umum di Indonesia dengan jumlah terbesar, yaitu 27.000 armada pada tahun 2013-2014. Penelitian ini terdiri dari empat langkah, yaitu studi lapangan, document review, wawancara stakeholder dan pembuatan Hierarchical Task Analysis (HTA). Hasil dari penelitian ini berupa diagram dekomposisi yang menampilkan struktur tugas secara spesifik, terbagi dalam bentuk diaram tabular maupun hirarki. Kata kunci : Hierarchical Task Analysis (HTA), Pengemudi Taksi, Taksi Kosti Solo.
Article
En France, depuis le 1er octobre 2008, tous les véhicules en circulation doivent être équipés d'un kit de sécurité comportant un gilet haute-visibilité et un triangle de présignalisation. Le principal objectif de cette mesure est d'améliorer la sécurité des piétons présents près d'un véhicule en panne ou accidenté. L'objectif de la recherche présentée dans cet article est d'examiner l'impact qu'a eu cette mesure sur les accidents de piétons, en particulier les accidents de piétons survenant à proximité d'un véhicule en panne ou accidenté. Les investigations ont porté sur deux départements français et sur la période 2005-2011. La période 2005-2007 constitue la période "avant" et la période 2009-2011, la période « après ». La démarche a consisté à étudier l'évolution entre ces deux périodes de la proportion du scenario type d'accident « Piéton présent près d'un véhicule en panne ou accidenté, [généralement] de nuit, hors agglomération, non vu ou trop tard » décrit par Brenac et coll. (2003) parmi l'ensemble des procès-verbaux d'accidents de piétons survenus hors agglomération. On observe une stagnation de la proportion de ces accidents après l'instauration de la mesure. Cette proportion passe de 27 % en période « avant » à 28 % en période « après ». La différence n'est pas significative (test de Cochran-Mantel-Haenszel ; p = 1.00 ; OR = 0.982 ; 95 % CI [0.439 ; 2.190]). L'analyse des cas suggère que cette absence d'évolution pourrait provenir de la faible utilisation du kit de sécurité par les automobilistes suite à une panne ou un accident. Quelques perspectives pour la prévention sont discutées en fin d'article.
Article
Full-text available
When looking at a scene, observers feel that they see its entire structure in great detail and can immediately notice any changes in it. However, when brief blank fields are placed between alternating displays of an original and a modified scene, a striking failure of perception is induced: identification of changes becomes extremely difficult, even when changes are large and made repeatedly. Identification is much faster when a verbal cue is provided, showing that poor visibility is not the cause of this difficulty. Identification is also faster for objects mentioned in brief verbal descriptions of the scene. These results support the idea that observers never form a complete, detailed representation of their surroundings. In addition, results also indicate that attention is required to perceive change, and that in the absence of localized motion signals it is guided on the basis of high-level interest. To see or not to see: The need for attention to perceive changes in scenes. Available from: https://www.researchgate.net/publication/236170014_To_see_or_not_to_see_The_need_for_attention_to_perceive_changes_in_scenes [accessed Jun 15, 2017].
Article
Full-text available
Our intuition that we richly represent the visual details of our environment is illusory. When viewing a scene, we seem to use detailed representations of object properties and interobject relations to achieve a sense of continuity across views. Yet, several recent studies show that human observers fail to detect changes to objects and object properties when localized retinal information signaling a change is masked or eliminated (e.g., by eye movements). However, these studies changed arbitrarily chosen objects which may have been outside the focus of attention. We draw on previous research showing the importance of spatiotemporal information for tracking objects by creating short motion pictures in which objects in both arbitrary locations and the very center of attention were changed. Adult observers failed to notice changes in both cases, even when the sole actor in a scene transformed into another person across an instantaneous change in camera angle (or “cut”).
Article
The absolute and relative effectiveness of conspicuously marked stationary and moving police units was studied, using a sample of 541 speeding drivers. Driver speeds were monitored prior to police detection, both at the site of the police unit and four km later. It was found that (1) the presence of police was sufficient to slow 95% of the speeding drivers, (2) the magnitude of initial speed reduction was the same in response to stationary and moving police units, (3) the halo effect (the tendency to maintain reduced speed after passing the police) was significantly greater for the drivers exposed to the moving units than to the stationary units, and (4) the greater the initial excess speed, the higher the final speed recorded after the police unit was passed. The results are consistent with the notion that compliance behavior is determined by the concept of perceived risk of apprehension.
Article
There is a widespread belief that motorcycles are more difficult to detect in traffic than are cars and trucks, which has led to much research designed to enhance motorcycle conspicuity. This paper examines the basic concept and finds that it lacks empirical support. Further, a number of other possible explanations could account for the differences one finds when comparing car-motorcycle and car-car collisions. Additional research is required to determine whether conspicuity is a special problem for motorcycles and the extent to which such a problem might contribute to crash involvement.
Article
The paper discusses the part played by three psychological processes, here called ‘ the false hypothesis ’. ‘ pro-occupation ’ and ‘ emergency mechanisms ’, in the causation of transport accidents. Some of the experiments in which those processes have been studied in the laboratory are described. Explanations of the errors responsible for certain train collisions, when the driver had passed one or more signals at danger, and for certain aircraft accidents are suggested by reference to one or other of these processes.
Article
Analyzes the problem of collisions between vehicles and trains in the light of relevant human-factor principles and suggests ways in which knowledge of behavioral, sensory, and perceptual principles can aid in the understanding and amelioration of the problem. Solutions discussed include overpasses, bridges, lights and bells at grade crossings, information campaigns, and multidisciplinary research. (8 ref) (PsycINFO Database Record (c) 2012 APA, all rights reserved)
Article
Motorcycles are overrepresented in fatal motor vehicle accidents: The death rate for motorcycle riders of about 35 per 100,000,000 miles of travel compares with an overall vehicle death rate of 2.57 per 100,000,000 miles. In the attempt to reduce the frequency of automobile-motorcycle collisions, numerous studies have manipulated motorcycle and motorcyclist characteristics to enhance conspicuity. In this paper, we give a review of studies that examined the effectiveness of these measures. Subsequently, we take a critical look at the methods used in these studies to evaluate the effectiveness of conspicuity treatments. Furthermore, we identify factors yet to be considered in the empirical research in this area that may contribute to collisions with motorcycles. These include information-processing failures at the identification and decision stage, as well as more or less permanent factors potentially responsible for different information-processing failures. Transient factors related to the failure to detect motorcycles might include alcohol, fatigue/lack of sleep, inattention, and information overload, whereas more permanent factors might include “cognitive” conspicuity and fi eld dependence.
Article
Two experiments studied individual differences among normal adults in performance on the Nelson-Denny reading test to cast light on the processes involved in reading. Experiment 1 correlated reading comprehension with performance on the Daneman and Carpenter working memory span test, vocabulary, lexical decision with both homophonous and nonhomophonous nonwords, and Posner's letter matching task, based on both physical and name matching. Working memory span proved to be a significant predictor as did lexical decision with nonhomophonous nonwords, letter name matching, and vocabulary. A derived measure indicating degree of phonological coding in lexical decision was only weakly correlated with comprehension while the physical name match difference on the Posner task was uncorrelated with reading performance. A second experiment explored further the working memory span task, comparing it with a nonlinguistic span test devised by Case. Kurland, and Goldberg based on memory and counting. While the verbal working memory measure again correlated significantly with comprehension, the counting measure was much more weakly related. As in Experiment 1, vocabulary also correlated significantly with comprehension. It is concluded that reading comprehension is dependent on a number of separable components including vocabulary, working memory, and a general lexical access process.