You are on page 1of 8

Applied Ergonomics 42 (2011) 619e626

Contents lists available at ScienceDirect

Applied Ergonomics
journal homepage: www.elsevier.com/locate/apergo

Effects of advertising billboards during simulated driving


Jessica Edquist*, Tim Horberry 1, Simon Hosking 2, Ian Johnston
Accident Research Centre, Monash University, Building 70, Victoria 3800, Australia

a r t i c l e i n f o a b s t r a c t

Article history: There is currently a great deal of interest in the problem of driver distraction. Most research focuses on
Received 30 October 2009 distractions from inside the vehicle, but drivers can also be distracted by objects outside the vehicle.
Accepted 19 August 2010 Major roads are increasingly becoming sites for advertising billboards, and there is little research on the
potential effects of this advertising on driving performance. The driving simulator experiment presented
Keywords: here examines the effects of billboards on drivers, including older and inexperienced drivers who may be
Distraction
more vulnerable to distractions. The presence of billboards changed drivers patterns of visual attention,
Roadside advertising
increased the amount of time needed for drivers to respond to road signs, and increased the number of
Driving simulation
Lane change test
errors in this driving task.
Response time 2010 Elsevier Ltd and The Ergonomics Society. All rights reserved.
Eye movements

1. Introduction billboards might impair driving performance, controlled laboratory


experiments are necessary.
1.1. Background A small number of these sorts of experiments have been per-
formed: Johnston and Cole (1976) found that centrally presented
Advertising billboards are highly conspicuous due to their size, non-driving-related signs did not affect a tracking task, but did
colouration, and location usually near major roads. The outdoor affect peripheral detection. Luoma (1986) found that participants
advertising industry promotes billboards as attention-grabbing, for were less likely to correctly recall road signs in slides of road scenes
example Out-of-home advertising provides visual impact that that contained billboards. Clark and Davies (2008) used a PC-based
commands complete attention, and offers total cut-through simulated driving task and found that non-driving-related signs
(Eyecorp, 2004). It might therefore be expected that they would delayed responses to simultaneously presented road signs.
distract attention from the driving task. Several studies of eye movements while driving (Beijer et al.,
Roadside advertising has received attention with other potential 2004; Crundall et al., 2006; Lee et al., 2007; Smiley et al., 2004)
driver distractions, as evidenced by the symposium at the recent provide evidence that billboards may capture drivers visual
International Conference on Driver Distraction. There is not space attention and hold it for relatively long periods. Long glances away
in the present paper to cover the entirety of the relevant literature from the road increase crash risk, as demonstrated in the 100-car
and international policies around roadside advertising, but it is naturalistic driving study (Klauer et al., 2006).
summarised by Horberry et al. (2009) at that conference. Wickens (2008) suggests that task interference will occur when
Reviews of crash rates associated with billboards seem to show two stimuli or tasks share the same input modality, processing
that billboards are correlated with higher crash rates, particularly in requirements, and response type. Roadside advertising attracts
certain situations such as at intersections, when driver workload is visual attention and cognitive central processing (as the observer
very high or very low (Cairney and Gunatillake, 2000; Farbry et al., determines whether the advertisement is relevant to their current
2001; Wallace, 2003). However correlation does not necessarily goals), so billboards would be expected to cause task interference
imply causation. To ascertain how (and in what circumstances) with driving tasks that require visual xation and central
processing.

* Correspondence author. Tel.: 61 3 9905 5815; fax: 61 3 9905 4363. 1.2. The present experiment
E-mail address: Jessica.Edquist@muarc.monash.edu.au (J. Edquist).
1
Tim Horberry is now at Minerals Industry Safety and Health Centre, University
of Queensland, Australia.
It therefore seems likely that billboards can distract drivers such
2
Simon Hosking is now at Defence Science and Technology Organisation, that their response to road signs and potentially other road users is
Australia. delayed or absent. The present experiment examined a range of

0003-6870/$ e see front matter 2010 Elsevier Ltd and The Ergonomics Society. All rights reserved.
doi:10.1016/j.apergo.2010.08.013
620 J. Edquist et al. / Applied Ergonomics 42 (2011) 619e626

performance measures using a high-delity driving simulator, as 1.3. Study overview and summary of hypotheses
this method provides a balance between the control of a laboratory
study and the realism of an on road study. The aim of the study was to examine visual behaviour and
One driving task previously used to investigate distraction is the responses to road signs in the presence and absence of billboards. It
ISO standardised Lane Change Test (Burns et al., 2005). This was expected that billboards would impair responses to road signs,
involves changing lanes in response to road signs at regular inter- with changeable billboards having more impact than static billboards.
vals. Burns et al. (2005) found that when drivers were distracted by Billboards were expected to change visual behaviour by drawing
a visual-manual task, they performed worse at the Lane Change attention towards roadsides, at the expense of the road ahead. Novice
Test. The task of responding to road signs requires visual xation drivers and older drivers were expected to be more affected by the
and cognitive processing, which as discussed above, are vulnerable billboards than the comparison group. Lead vehicles were expected to
to interference from roadside advertising. It is therefore expected increase the proportion of time xating on the road ahead; this could
that distraction due to billboards would impair drivers ability to interact with the effect of billboards on lane changes.
correctly follow road signs and change lanes quickly. The task used As the billboards used in this study were very simple (the logo
in the present experiment was not the standard Lane Change Test, and tagline of a company only) they are potentially less conspicuous
as rather than speed being set at a predetermined limit, it was and less likely to retain attention than billboards used in the real
desirable for drivers to be able to choose their own speed in world (which tend to contain a picture and more text). To ensure
response to the road environment. that we could determine the effects of looking at billboards even if
The present experiment also examined visual behaviour. While the simple simulated billboards were not sufciently attention-
the research of Beijer et al., as well as Lee et al. (2007), demon- grabbing, half of the participants in each age/experience group were
strates that drivers xate on billboards, it is not possible for on road instructed to report any billboards they saw. These participants were
studies to analyse how billboards change gaze patterns as the expected to show greater effects of billboard presence.
environment cannot be sufciently controlled to attribute differ-
ences to billboards alone. This problem does not exist in a simu- 2. Material and methods
lated environment. The proportion of time xating on the road
ahead is expected to be high, as the present experiment involved no 2.1. Participants
hazards approaching from the sides (so there was no reason for
drivers to xate away from the road ahead). It was hypothesised 48 participants (30 males and 18 females) from three age/
that billboard presence would draw xations away from the road experience groups took part. The novice driver group consisted of
ahead, towards roadsides where billboards were situated. sixteen drivers aged 18e25 who had gained their licence to drive
In addition to static billboards, the present experiment included unsupervised less than one year ago. The sample was restricted to
billboards that changed what they were displaying. These billboards rst-year drivers as the rst year of unsupervised driving has been
are becoming more common and there is little experimental research reported to be the most dangerous (Mayhew et al., 2003). The older
about their effects. The visual attention literature suggests that the driver group comprised sixteen drivers aged over 65 years who
motion onset cue provided by the changing display will involuntarily held full drivers licences. The comparison driver group consisted of
attract attention (Abrams and Christ, 2003). Therefore it was expec- sixteen drivers aged 25e55 who held full licences. Table 1 shows
ted that these changeable billboards should attract eye movements further details of participant characteristics. Participants were
and affect responses to road signs more than static billboards. recruited by notices in the universitys staff bulletin, on the student
We also wished to investigate whether the effect of billboards job website, and at nearby Senior Citizens clubs and Universities of
changed in the presence of trafc, so one of the drives included lead the Third Age. All received a small cash payment as compensation
vehicles. Lead vehicles may increase the effect of the billboards due for their time and travel costs.
to the increased workload of maintaining an acceptable headway
distance. Strayer et al. (2003) found that adding vehicles increased
2.2. Apparatus and stimuli
perceptual load and the effect of a secondary task. However it is
also possible that lead vehicles may attenuate the effect of bill-
The experiment was performed in Monash University Accident
boards, as drivers will concentrate on the vehicle ahead in order to
Research Centres driving simulator (see Fig. 1). This consists of
maintain the specied headway. Lead vehicles have been found to
a 2003 Holden Calais sedan mounted on a motion platform, with
increase the proportion of time xating on the (vehicles and) road
three projection screens at the front. From the drivers viewpoint
ahead (Crundall et al., 2004), and this effect is expected in the
the three screens in front of the car provide a eld of view sub-
present experiment.
tending angles of approximately 180 horizontally and 40 verti-
It was anticipated that age and inexperience might affect the
cally. The display has a resolution of 1280  768 pixels for the front
extent of any billboard effects, so young novice drivers and older
panel and 640  480 pixels for the side panels. The projectors
drivers were compared with drivers who were neither old, young
update the image at a rate of 30 Hz. A Crystal River Audio Reality
nor inexperienced (comparison drivers). Novice drivers were
Accoustetron II audio system generates appropriate sounds. Data is
expected to be particularly distracted by billboards, as their mental
collected from the brake pedal, accelerator pedal and steering
model of how to nd relevant information is less well developed;
wheel at a rate of 30 Hz. Participants head and eye movements
novices have been found to xate more signs than experienced
were tracked using faceLab head and eye tracking hardware and
drivers, and more irrelevant signs the rst time they drive a route
(Sprenger et al., 1997). Older drivers have been found to report fewer
Table 1
roadside signs (Wood, 2002); if this reects differences in percep-
Participant characteristics for driving simulator experiment.
tion rather than reporting, they are more likely to respond to road
signs later or not at all. Older drivers have been found to make fewer Group Proportion Mean age Mean driving (Self-estimated)
of males experience Mean distance driven
xations than younger drivers overall, but make more xations to
peripheral areas, perhaps because they are unable to monitor Novice 7/16 19.4 years .6 years 10,400 km/year
Comparison 11/16 35.1 years 15.4 years 14,700 km/year
peripheral events with covert attention (Mapstone, 2001). This may Older 12/16 73.6 years 53.0 years 15,400 km/year
leave them more vulnerable to distraction from billboards.
J. Edquist et al. / Applied Ergonomics 42 (2011) 619e626 621

Fig. 1. The Advanced Driving Simulator at MUARC. Fig. 2. Example of the simulated environment showing billboard and lane change sign.

software. This system has a resolution of three degrees of visual billboard was present on the opposite side of the road. Similarly,
angle. Data collection from faceLab and the vehicle was linked to four signs had changeable billboards opposite to them, while four
the display updates. were control sites with no billboards. The remaining 25 signs were
The simulated scenarios consisted of three-lane divided arterial not analysed; they were used to direct the driver back into the
roads through commercial and industrial environments, with centre lane, or from the left to the right lane (so that the driver
occasional intersections controlled by signal lights. This environ- could not predict which lane they would be directed into next).
ment was chosen because, of the simulated environments available, Billboards used were logos of companies with a large Australian
it was closest to the type of road where advertising billboards are advertising presence (see Fig. 2 for an example). These were chosen
most often located, and these higher speed, higher demand envi- so that they would be clearly recognisable in the simulator to most
ronments are where billboards are likely to have the greatest participants (More complex real-world billboards were piloted,
effects. In one drive there was no trafc travelling in the same however they were difcult to read due to resolution in the
direction as the participant; the other drive involved three lead simulator). Billboards measured 8  5 m (substantially larger than
vehicles, one in each lane, driving at 70 km/h (this speed limit was the lane change signs; this is a common size format for billboards
consistent with the simulated road environment). Each drive was next to roads) and popped up when the participant was 140 m
approximately 9 km long. Participants occasionally had to stop at away from the billboard location. Static billboards displayed the
intersections, but the main task was to change lanes in response to same logo for the whole 140 m; changeable billboards switched
roadside lane change signs at regular intervals. This task was from one logo to another when the participant was 85 m away from
broadly based on the ISO standard Lane Change Test as described by the billboard. The simulator was only capable of performing one
Burns et al. (2005), but drivers controlled their own speed; there- change to the changeable billboards, so the change was timed to
fore it was not possible to calculate deviations from the optimal maximise the chance of capturing attention (it occurred approxi-
lane change path as used in the standard Lane Change Test. Instead, mately 3 s after the billboard rst appeared, when drivers might be
the dependent variables were errors (number of times drivers expected to have returned their attention to the road; the second
missed a sign or entered the wrong lane) and time to change lanes; logo appeared larger and clearer as the driver was closer, and was
worse performance would involve more errors and longer lane visible for a further 5 s).
changes. The lane change signs measured 2  1 m and were spaced
approximately every 150 m; they did not become visible until the 2.3. Procedure
participant was 140 m from them. Each sign contained two crosses
and one arrow indicating the lane into which the participant should At the start of the experimental session, participants completed
change. Fig. 2 shows an example of a lane change sign. the demographic questionnaire. Participants then drove a short
Signs became visible when the driver was 140 m away so that sequence with the experimenter seated in the passenger seat, to
there were never two signs visible at the same time. Time to change familiarise themselves with the controls of the vehicle. The
lanes was measured from about 50 m after the appearance of the experimenter then returned to the control room, from where there
lane change sign (when the sign was 90 m away), so that the driver is two-way communication with the participant in the simulator
could nish responding to the previous sign (These distances were vehicle. The next drive was performed solo, and allowed the
chosen as piloting showed that most drivers did not start their lane participant to practice the car following and lane change tasks.
change until they were 90 m from the sign, and completed the lane Finally, the two experimental drives were performed. The order of
change shortly after they passed the sign). Signs were spaced drives was counterbalanced within age/experience and instruction
approximately every 150 m so that the driver was almost contin- groups. In the no trafc drive, participants were instructed to stay
ually switching lanes. If the driver did not enter the correct lane as close as possible to the speed limit of 70 km/h; in the drive with
within the distance 90 m before passing the sign to 50 m after it, no lead vehicles, they were instructed to maintain a following
time data was recorded; these occurrences were analysed as errors. distance of approximately 2 s behind the lead vehicles. In both
Each experimental drive contained a total of 37 lane change drives, participants were instructed to change lanes as soon as they
signs. At two signs instructing the participant to change from the could perceive the arrow on the next lane change sign. Half of the
centre lane into the left lane, and at two signs instructing the participants in each age/experience group were given only these
participant to change from the centre to the right lane, a static normal instructions. The other half were also asked to report what
622 J. Edquist et al. / Applied Ergonomics 42 (2011) 619e626

Table 2 Table 4
Between-subjects variables for the simulator experiment, and number of partici- Analysis of variance for time to change lanes (*indicates signicant at a .05).
pants per cell.
Source df F p h2
Probationary Comparison Older Between-subjects
Normal instructions 8 8 8 Instructions 1 .06 .802 .00
*
Report billboards 8 8 8 Age/experience 2 6.17 .004 .23
Instructions  age/exp 2 .20 .817 .01
Error (F MS error) 42 (1.89)
was advertised by any billboards they passed. This instruction was Within-subjects
*
to simulate the effect of being distracted by a particularly Vehicles 1 11.41 .002 .21
conspicuous billboard. After each drive, participants completed the Vehicles  instruct 1 .32 .574 .01
Vehicles  age/exp 2 .18 .838 .01
NASA-RTLX mental workload scale (Hart and Staveland, 1988). To
Vehicles  instruct  age/exp 2 .67 .515 .03
save time, weightings were not elicited, as an unweighted sum has Error(vehicles) (F MS error) 42 (.51)
been found to give an adequate indication of overall workload *
Billboard 2 35.03 .000 .46
*
(Hendy et al., 1993). The full session lasted for approximately 1 h. Billboard  instruct 2 6.87 .002 .14
*
Billboard  age/exp 4 2.85 .029 .12
Billboard  instruct  age/exp 4 2.12 .086 .09
2.4. Design and data analysis Error(billboard) (F MS error) 84 (.32)
*
Vehicles  billboard 2 4.51 .014 .10
The between-subjects factors were age/experience, and Vehicles  billboard  instruct 2 .57 .568 .01
instructions. Table 2 gives the number of participants in each cell. Vehicles  billboard  age/exp 4 .40 .806 .02
Vehicles  billboard  instruct  age/exp 4 1.31 .275 .06
The within-subjects factors were presence/absence of lead Error(vehicles  billboard) (F MS error) 84 (.29)
vehicles and billboard type (Table 3).
The dependent variables examined were time to change lanes,
number of lane change errors and proportion of time xating on the
road ahead, as these together demonstrate the major effects of As workload was anticipated to be a mediating factor for any
roadside advertising on driving performance. Further eye movement effect of the lead vehicles, mean subjective workload ratings for the
and vehicle control measures were examined but are not reported two drives were compared. Participants rated the drive without
here due to lack of space; for the full results see (Edquist, 2008). lead vehicles 44 (out of 100) and the drive with lead vehicles 41.
Time to change lanes and proportion of time xating on the road This difference was signicant, F(1,42) 6.029, p .018; however it
ahead were analysed using separate mixed-model ANOVAs with was not in the anticipated direction. There were no signicant
the within-subject factors being presence/type of billboard and interactions with age/experience or with instructions. However
presence of lead vehicles, and the between subject factors being both these main effects were signicant at p < .01. Novice drivers
type of instructions and age/experience group. Where the full reported higher workloads (mean rating 51) than comparison
ANOVA showed a signicant 3-way interaction, further analyses drivers (mean 34, p .004), while older drivers reported work-
were undertaken; this resulted in a total of 36 comparisons for the loads were not signicantly higher (mean 42, p .2). Drivers who
Time to Change Lanes results and 27 comparisons for the Fixation were asked to report billboards reported higher workload
results. P-values reported are uncorrected; F and df are reported for (mean 48) than drivers given normal instructions (mean 37,
each comparison so that readers can apply their own preferred p .007). As participants rated workload for each entire drive, it
adjustment if desired. was not possible to examine the effects of billboards.
Incorrect lane changes were analysed using Pearsons chi-
square test (as this data is binary, not continuous). Cases when the
3.2. Time to change lanes when directed by a roadside sign
driver changed into the wrong lane, as well as cases where
participants failed to change lanes within the segment of road 90 m
Incorrect or missed lane changes were not included in this
before passing the lane change sign to 50 m after it, were recorded
analysis, however the remaining dataset included 1090 correct lane
as lane change errors.
changes. The overall mean time to change lanes was 6.9 s, which
equates to 125 m at drivers average speed of 66 km/h. As lane
3. Results
change signs were on average 150 m apart, drivers did not have
a lot of leeway to delay lane changes.
3.1. Methodological checks

In both drives, participants reported 89% of billboards correctly,


suggesting that participants obeyed the instructions. All age groups
successfully performed this task: novice drivers reports were 95%
correct, comparison drivers 91% correct, and older drivers 82% correct.

Table 3
Within-subjects variables for the lane change task part of the simulator
experiment.

No trafc Lead vehicles

Lane change Centre lane Centre lane Centre lane Centre lane
sign instructions: / left lane / right lane / left lane / right lane
No billboard 2 signs 2 signs 2 signs 2 signs
Static billboard 2 signs 2 signs 2 signs 2 signs
Changeable 2 signs 2 signs 2 signs 2 signs
billboard Fig. 3. Mean and 95% condence intervals for time to change lanes, by vehicle pres-
ence and billboard type.
J. Edquist et al. / Applied Ergonomics 42 (2011) 619e626 623

Table 5 Table 6
Post-hoc comparisons for interaction effect between vehicles and billboards on time Post-hoc comparisons for interaction effect of age and billboards on time to change
to change lanes. lanes.

Comparison df F p h2 Comparison df F p h2
Lead vehicles e effect of billboards 2, 84 22.38 <.001 .35 Novice drivers e effect of billboards 2, 28 6.21 <.01 .31
- static vs. control 1, 42 33.09 <.001 .44 - static vs. control 1, 14 12.43 <.01 .47
- changeable vs. static 1, 42 5.261 <.05 .11 - changeable vs. static 1, 14 3.323 >.05 .19
No vehicles e effect of billboards 2, 84 18.54 <.001 .31 Comparison drivers e effect of billboards 2, 28 34.22 <.001 .71
- static vs. control 1, 42 22.05 <.001 .34 - static vs. control 1, 14 45.25 <.001 .76
- changeable vs. static 1, 42 3.816 >.05 .08 - changeable vs. static 1, 14 .002 >.1 .00
Older drivers e effect of billboards 2, 28 8.29 .001 .37
- static vs. control 1, 14 7.77 <.05 .36
- changeable vs. static 1, 14 .854 >.1 .06
A mixed-model ANOVA including the within-subject factors of
billboard type and presence of lead vehicles, and the between-
subjects factors of instructions and age/experience, found several
signicant effects and interactions. Table 4 summarises the results 3.3. Errors when changing lanes (including changes to the wrong
of the ANOVA (effects signicant at a .5 are highlighted). lane, and failure to change lanes)
There was a signicant effect of billboard type on time to change
lanes. This effect was consistent despite two-way interactions Errors were consistent with the results for time to change lanes.
between billboard type and the other variables: in all cases, lane Out of 1152 lane changes, there were 62 errors, of which 50
changes when passing sites with billboards took longer than lane occurred in the presence of billboards (24 at static and 26 at
changes when passing control sites. The effect was similar for both changeable billboards; these categories were collapsed to enable
static and changeable billboards. calculation of odds ratios). Overall, drivers were signicantly more
Lane changes were on average faster in the presence of lead likely to make errors at sites with billboards: c2(1, N 1152) 5.76,
vehicles, however there was a signicant interaction between p < .05, odds ratio 2.16. This effect was stronger during the drive
vehicle presence and billboard type. As can be seen in Fig. 3, the with lead vehicles (OR 5.50 with vehicles, 1.11 without). Older
interaction between lead vehicles and billboard was due to the drivers (who made 63% of the errors across all conditions) showed
longest lane changes being at static billboard sites in the drive with a stronger effect than the comparison group (OR older 2.47,
lead vehicles and at changeable billboard sites in the drive without comparison 1.39). Novice drivers also showed a strong effect of
lead vehicles (F, p and h2 values for post-hoc comparisons are billboards (OR 3.57), however low error numbers for this group
reported in Table 5). meant that the effect was not signicant. Finally, the effect of bill-
Older drivers were slowest to change lanes overall, followed by boards was stronger for the group instructed to report billboards
novice drivers. There was a signicant interaction between bill- (OR 11.61; OR normal instructions 1.29) (Higher interactions
board type and age/experience group, as can be seen in Fig. 4. were not analysed as individual cell sizes would be insufcient).
Further analyses revealed that lane changes were signicantly
faster at control sites than sites with static billboards, and sites with 3.4. Proportion of time xating on road ahead (including lead
changeable billboards did not differ signicantly from sites with vehicles)
static billboards. The interaction is presumably due to the slightly
smaller effect of the changeable billboards in the novice driver Overall, drivers spent 55.9% of time on the approach to a lane
group (See Table 6 for F, p and h2 values for these analyses). change sign xating the road ahead. A mixed-model ANOVA
Billboard type also interacted signicantly with instructions (see including the within-subject factors of billboard type and presence
Fig. 5). The effect of billboards was larger for the group instructed to of lead vehicles, and the between-subjects factors of instructions
report billboards, although it was still highly signicant for the and age/experience, found a main effect of billboard type, and an
group given normal instructions. Again, changeable and static interaction between billboard type, instructions and age/experi-
billboards had similar effects (For F, p and h2 values for comparisons ence group. Table 8 summarises the results of the ANOVA (effects
see Table 7). signicant at a .5 are highlighted).
The effect of billboards was to reduce time xating the road ahead
for almost all subgroups of participants. To examine the signicant

Fig. 4. Mean and 95% condence intervals for time to change lanes, by age and bill- Fig. 5. Mean and 95% condence interval for time to change lanes, by instructions and
board type. billboard type.
624 J. Edquist et al. / Applied Ergonomics 42 (2011) 619e626

Table 7
Post-hoc comparisons for interaction effect of instructions and billboards on time to
change lanes.

Comparison df F P h2
Normal instructions e effect of billboards 2, 42 5.47 <.01 .21
- static vs. control 1, 14 8.34 <.01 .28
- changeable vs. static 1, 14 .103 >.1 .01
Report billboards e effect of billboards 2, 28 37.03 <.001 .64
- static vs. control 1, 14 41.44 <.001 .66
- changeable vs. static 1, 14 .269 >.1 .01

3-way interaction, analyses were conducted separately for each


instruction group. The simple interaction of billboard type and age Fig. 6. Report billboards group: Mean and 95% condence intervals for proportion of
was only signicant for the group instructed to report billboards, F(4, time xating on road ahead.
42) 4.88, p < .01, h2 .32. Results for this group are illustrated in
Fig. 6. When this group was divided by age, only the comparison Table 9
drivers did not show a signicant effect of billboard type. In both Post-hoc comparisons for interaction effect of age and billboards on proportion of
time xating the road ahead, report billboards group.
novice and older driver groups, contrasts showed that both static
and changeable billboards were associated with reduced time spent Comparison df F p h2
xating on the road ahead, although the older drivers showed Novice drivers e effect of billboards 2, 14 28.06 <.001 .80
somewhat less of an effect for the changeable billboards (Table 9 - static vs. control 1, 7 41.33 <.001 .86
- changeable vs. static 1, 7 2.628 >.1 .27
shows the F, p and h2 values for these analyses).
Comparison drivers e effect of billboards 2, 14 1.52 >.1 .18
For the group given normal instructions, the interaction Older drivers e effect of billboards 2, 14 15.97 <.001 .70
between billboard type and age was not signicant; nor was the - static vs. control 1, 7 31.04 <.001 .82
main effect of age. The main effect of billboard was signicant: both - changeable vs. static 1, 7 9.71 <.05 .58
static and changeable billboards were associated with less time
xating on the road ahead, as shown in Fig. 7 and Table 10.

4. Discussion The present study suggests that individual factors can affect the
distraction caused by billboards. Scanning patterns for novice and
As predicted by multiple resource theory (Wickens, 2008), the older drivers were both more affected by billboards than scanning
presence of billboards in the present experiment distracted eye for comparison drivers. Older drivers made more lane change
movements from the road ahead and delayed responses to road errors overall, but particularly when billboards were present. This is
signs. These results are similar to results found by other studies on consistent with previous research suggesting that older drivers
distraction by auditory and visual in-vehicle tasks (Burns et al., have greater difculty responding in complex road environments
2005; Engstrom et al., 2005; Parkes et al., 2007) and suggest that (e.g. Caird et al., 2005). As the population ages, it will become more
billboards, like in-vehicle forms of distraction, should be consid- important to design roads that are safe for all drivers, not just the
ered as potentially harmful. These results are practically as well as average driver.
statistically signicant; responses to road signs were delayed by The effect on response to road signs was not so marked for
half to 1 s in the presence of billboards, in which time a car trav- novice drivers. The novices took longer to change lanes in situations
elling at 70 km/h will have moved almost 20 m. without distracting billboards, and this combined with the high
variability in this group may account for the nonsignicant effect of
changeable billboards (as there is little theoretical basis for the idea
Table 8
Analysis of variance for proportion of time xating on the road ahead (*indicates that changeable billboards are less distracting than static bill-
signicant at a .05). boards). The signicance of the differences between lane change
errors at billboard and control sites was restricted by the low
Source df F p h2
number of errors overall, however the trend was clear: billboards
Between-subjects
Instructions 1 .056 .815 .00
greatly increased the number of errors in responding to road signs
Age/experience 2 1.906 .161 .08 for this group. With a young novice driver group it is not possible to
Instructions  age/exp 2 .212 .810 .01 determine whether these effects are due to age (immaturity), or
Error (F MS error) 42 (.075) inexperience; however the results are consistent with previous
Within-subjects research showing that novice drivers have less well developed
Vehicles 1 .07 .791 .00 mental models of where to look for driving-related information
Vehicles  instruct 1 .21 .651 .00 (Underwood et al., 2002).
Vehicles  age/exp 2 .76 .473 .04
Vehicles  instruct  age/exp 2 2.40 .103 .10
Individual factors include not only consistent differences such as
Error(vehicles) (F MS error) 42 (.02) a drivers age and experience, but also temporary factors such as the
*
Billboard 2 43.21 .000 .51 drivers current set of goals. The group who were asked to report
Billboard  instruct 2 .05 .953 .00 billboards spent less time looking at the road ahead, and presum-
Billboard  age/exp 4 1.45 .225 .07
*
ably more time looking at billboards than the group who received
Billboard  instruct  age/exp 4 3.00 .023 .13
Error(billboard) (F MS error) 84 (.01) normal instructions. This was particularly pronounced in the
Vehicles  billboard 2 1.70 .188 .04 novice and older driver groups. This resulted in a greater delay in
Vehicles  billboard  instruct 2 2.15 .122 .05 the time taken to change lanes when billboards were present, and
Vehicles  billboard  age/exp 4 1.39 .244 .06 an increase in the number of signs which were not followed
Vehicles  billboard  instruct  age/exp 4 .44 .780 .02
Error(vehicles  billboard) (F MS error) 84 (.00)
correctly. This has implications for drivers who are less focussed on
the task of driving (perhaps due to fatigue, or personality factors).
J. Edquist et al. / Applied Ergonomics 42 (2011) 619e626 625

Table 10
Post-hoc comparisons for proportion of time xating on the road ahead for normal
instructions group.

Comparison df F p h2
Interaction between billboard and age >.1 .04
Main effect of age >.1 .07
Main effect of billboards 2, 42 16.94 <.001 .45
- static vs. control 1, 21 39.07 <.001 .65
- changeable vs. static 1, 21 .037 >.1 .00

Simulator experiments are limited in their ability to emulate the


visual complexity of real road scenes. However, while the absolute
values for simulator measures may not precisely match values
Fig. 7. Normal instructions group: Mean and 95% condence intervals for proportion
obtained on road, relative differences between conditions still hold,
of time xating on road ahead. and simulator measures have been found to be valid predictors of
accident risk (Rudin-Brown and Lenn, 2010).
Lead vehicles did not have the expected effects of drawing While much research remains to be done (particularly on
xations towards the road ahead and increasing workload. This changeable and moving billboards), the present experiment has
could be because the vehicles did not change speed; thus the driver demonstrated that even simple billboards can affect drivers
only needed to match speeds once and then maintain approxi- attentional allocation and performance on a high-delity simulated
mately the same speed, essentially the same task as the drive driving task. This adds to the evidence that billboards also have
without vehicles, where they were instructed to drive at 70 km/h. these potentially dangerous effects when driving on road. This
Nonetheless, the increase in lane change errors at sites with bill- research has important implications for evidence-based policies on
boards was larger in the drive with lead vehicles. The effect on time roadside advertising around the world.
to change lanes was not so straightforward, and would need to be
investigated further with better control over mediating variables. Acknowledgements
The present study attempted to investigate the effects of
changeable billboards by including billboards that changed as Jessica Edquist was supported by an Australian Postgraduate
drivers approached. Unfortunately technology limitations meant Award (Industry) sponsored by the Department of Main Roads,
that only one image change could be programmed. The changeable Queensland; and by a Postgraduate Publications Award from
billboards used thus did not show the expected consistently greater Monash Research Graduate School.
effects than the static billboards. Future research should examine Programming of the driving simulator was performed by Nebojsa
the effect of billboards that change more frequently than once Tomasevic. Data parsing was performed by Ashley Verdoorn.
during the period for which they are visible, as change frequency
for changeable billboards is currently under debate in many juris- References
dictions. A further concern is the issue of full motion billboards in
which the image is constantly changing; these have the potential to Abrams, R.A., Christ, S.E., 2003. Motion onset captures attention. Psychological
Science 14 (5), 427e432.
be even more distracting than other types of billboards. Beijer, D., Smiley, A., Eizenman, M., 2004. Observed driver glance behaviour at
It should be noted that the billboards used in this study were very roadside advertising signs. Transportation Research Record 1899, 96e103.
simple, comprising only the logo and tagline of a company. Bill- Burns, P.C., Trbovich, P.L., McCurdie, T., Harbluk, J.L., 2005. Measuring distraction:
task duration and the Lane-Change Test. In: Human Factors and Ergonomics
boards used in the real world (with more complicated pictures and
Society 49th Annual Meeting.
text) are likely to hold drivers attention for longer. This increases the Caird, J.K., Edwards, C.J., Creaser, J.I., Horrey, W.J., 2005. Older driver failures of
probability of adverse effects on driving safety; Horrey et al. (2006) attention at intersections: using change blindness methods to assess turn
decision accuracy. Human Factors 47 (2), 235e249.
found that the length of glances away from the road ahead (in their
Cairney, P., Gunatillake, T., 2000. Roadside Advertising Signs e A Review of the
paper, to an in-vehicle display) correlated with response time to Literature and Recommendations for Policy. ARRB Transport Research for RACV.
a hazard, and Klauer et al. (2006) found that glances at objects away Clark, O.J., Davies, S.P., 2008. Ads on the road: a study into the effects of perceptual
from the road ahead totalling longer than 2 s increase crash risk. load and expertise on reaction time to road signs. In: British Psychology Society
Annual Conference.
This study also involved a fairly low workload driving environ- Crundall, D., Shenton, C., Underwood, G., 2004. Eye movements during intentional
ment, in which drivers were able to devote their attention to the car following. Perception 33, 975e986.
forward roadway 56% of the time, compared to 47% of the time in Crundall, D., van Loon, E., Underwood, G., 2006. Attraction and distraction of attention
with roadside advertisements. Accident Analysis & Prevention 38 (4), 627e832.
the baseline epochs of the 100-car naturalistic driving study Edquist, J., 2008. The Effects of Visual Clutter on Driving Performance (PhD thesis).
(calculated from gures in Klauer et al., 2006). This may have Monash University. http://arrow.monash.edu.au/hdl/1959.1/64951.
lessened the effects of the billboards in distracting attention from Engstrom, J., Johansson, E., Ostlund, J., 2005. Effects of visual and cognitive load in
real and simulated motorway driving. Transportation Research Part F: Trafc
the forward roadway. Given their frequent occurrence in complex Psychology and Behaviour 8 (2), 97e120.
road environments, future research could investigate billboards in Eyecorp, 2004. Out-of-home advertising. http://www.eyecorp.com.au/content.asp?
combination with high workloads, perhaps through a more siteid200 Retrieved 15 April 2004.
Farbry, J., Wochinger, K., Shafer, T., Owens, N., Nedzesky, A., 2001. Research Review
demanding driving task, the occurrence of unexpected hazards, or of Potential Safety Effects of Electronic Billboards on Driver Attention and
the addition of a secondary loading task. Distraction. Federal Highway Administration, Washington, DC.
The results of this study correspond well to the results of Hart, S.G., Staveland, L.E., 1988. Development of NASA-TLX (Task Load Index):
results of empirical and theoretical research. In: Hancock, P.A., Meshkati, N.
another recent driving simulator experiment on billboards. Young
(Eds.), Human Mental Workload. Elsevier Science, Amsterdam, pp. 139e183.
et al. (2009) found that the presence of roadside advertisements Hendy, K.C., Hamilton, K.M., Landry, L.N., 1993. Measuring subjective workload:
increased the number of separate eye glances across the scene, and when is one scale better than many? Human Factors 35 (4), 579e601.
impaired lateral control of the vehicle. They also found that drivers Horberry, T., Regan, M. Edquist, J., 2009. Driver distraction from roadside adver-
tising: the clash of road safety evidence, highway authority guidelines and
subjective ratings of mental workload were higher in the condition commercial advertising pressure. In: First International Conference on Driver
with advertisements. Distraction and Inattention, Sweden, September 2009.
626 J. Edquist et al. / Applied Ergonomics 42 (2011) 619e626

Horrey, W.J., Wickens, C.D., Consalus, K.P., 2006. Modeling drivers visual attention Rudin-Brown, C.M., Lenn, M.G., 2010. Driving simulation to support road safety
allocation while interacting with in-vehicle technologies. Journal of Experi- policy: Understanding crash risks to better inform speed setting guidelines. In:
mental Psychology: Applied 12 (2), 67e78. Australasian Road Safety Research, Policing and Education Conference, Can-
Johnston, A.W., Cole, B.L., 1976. Investigations of distraction by irrelevant infor- berra, September 2010.
mation. Australian Road Research 6 (3), 3e23. Smiley, A., Smahel, T., Eizenman, M., 2004. Impact of video advertising on driver
Klauer, S.G., Dingus, T.A., Neale, V.L., Sudweeks, J.D., Ramsey, D.J., 2006. The Impact of xation patterns. Transportation Research Record 1899, 76e83.
Driver Inattention on Near-Crash/Crash Risk: An Analysis Using the 100-Car Sprenger, A., Schneider, W., Derkum, H., 1997. Trafc signs, visibility and recogni-
Naturalistic Driving Study Data. Virginia Tech Transportation Institute, Blacksburg, tion. In: Vision in Vehicles VII Conference.
Virginia. Strayer, D.L., Drews, F.A., Johnston, W.A., 2003. Cell phone-induced failures of visual
Lee, S.E., McElheny, M.J., Gibbons, R., 2007. Driving Performance and Digital Bill- attention during simulated driving. Journal of Experimental Psychology:
boards Final Report. Virginia Tech Transportation Institute e Centre for Auto- Applied 9 (1), 23e32.
motive Safety Research (Prepared for Foundation for Outdoor Advertising Underwood, G., Chapman, P., Bowden, K., Crundall, D., 2002. Visual search while
Research and Education). driving: skill and awareness during inspection of the scene. Transportation
Luoma, J., 1986. The Acquisition of Visual Information by the Driver: Interaction of Research Part F: Trafc Psychology and Behaviour 5 (2), 87e97.
Relevant and Irrelevant Information. Liikenneturva e Central Organization for Wallace, B., 2003. External-to-vehicle Driver Distraction. Scottish Executive Social
Trafc Safety, Research Department, Helsinki, Finland. Research, Edinburgh.
Mapstone, M., 2001. Dynamic Allocation of attention in aging and Alzheimer's Wickens, C.D., 2008. Multiple resources and mental workload. Human Factors 50
disease. Archives of Neurology 58, 1443e1447. (3), 449e455.
Mayhew, D.R., Simpson, H.M., Pak, A., 2003. Changes in collision rates among novice Wood, J.M., 2002. Age and visual impairment decrease driving performance as
drivers during the rst months of driving. Accident Analysis & Prevention 35 measured on a closed-road circuit. Human Factors 44 (3), 482 (413).
(5), 683e691. Young, M.S., Mahfoud, J.M., Stanton, N.A., Salmon, P.M., Jenkins, D.P., Walker, G.H.,
Parkes, A.M., Luke, T., Burns, P.C., & Lansdown, T., 2007. Conversations in cars: the 2009. Conicts of interest: the implications of roadside advertising for driver
relative hazards of mobile phones (Report No. 664): Transport Research attention. Transportation Research Part F: Trafc Psychology and Behaviour 12
Laboratory. (5), 381e388.

You might also like