Accident Analysis and Prevention 42 (2010) 196–202

Contents lists available at ScienceDirect

Accident Analysis and Prevention

journal homepage: [Link]/locate/aap

The visual search patterns and hazard responses of experienced and

inexperienced motorcycle riders!
Simon G. Hosking ∗ , Charles C. Liu, Megan Bayly 1
Monash University Accident Research Centre, Clayton, VIC 3900, Australia

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

Article history: Hazard perception is a critical skill for road users. In this study, an open-loop motorcycle simulator was
Received 29 January 2009 used to examine the effects of motorcycle riding and car driving experience on hazard perception and
Received in revised form 1 July 2009 visual scanning patterns. Three groups of participants were tested: experienced motorcycle riders who
Accepted 29 July 2009
were experienced drivers (EM-ED), inexperienced riders/experienced drivers (IM-ED), and inexperienced
riders/inexperienced drivers (IM-ID). Participants were asked to search for hazards in simulated scenarios,
Keywords:
and click a response button when a hazard was identified. The results revealed a significant monotonic
Hazard perception
decrease in hazard response times as experience increased from IM-ID to IM-ED to EM-ED. Compared to
Motorcycles
Riding experience
the IM-ID group, both the EM-ED and IM-ED groups exhibited more flexible visual scanning patterns that
Eye movements were sensitive to the presence of hazards. These results point to the potential benefit of training hazard
Simulation perception and visual scanning in motorcycle riders, as has been successfully demonstrated in previous
studies with car drivers.
© 2009 Elsevier Ltd. All rights reserved.

1. Introduction One of the factors underlying the faster hazard response times of
experienced drivers may be improved deployment of overt atten-
Hazard perception has been defined as a driver’s “ability to read tion (e.g., see Underwood, 2007; Underwood et al., 2002, 2005).
the road and anticipate forthcoming events” (McKenna et al., 2006; That is, differences in the visual search patterns of experienced and
p. 2). Hazard perception skill has been shown to be associated novice drivers may influence their ability to detect and discrimi-
with accident involvement (e.g., Quimby et al., 1986; Horswill and nate hazards such that drivers with a larger spread of visual search
McKenna, 2004, for a review), and the over-representation of inex- may more quickly detect objects approaching from the peripheral
perienced drivers in crash statistics may be partly explained by field. A number of studies have demonstrated that experienced
findings that inexperienced drivers are slower to detect hazards drivers visually search the road environment in fundamentally dif-
than experienced car drivers (e.g., Groeger and Chapman, 1996; ferent ways to those of inexperienced car drivers (e.g., Chapman
McKenna and Crick, 1994; Quimby et al., 1986; Wallis and Horswill, and Underwood, 1998; Crundall et al., 2003; Mourant and Rockwell,
2007). Early hazard detection affords additional time for respond- 1972). In what Shinar (2008) described as one of the most influential
ing appropriately to hazards and is regarded as an important studies of drivers’ eye movements, Mourant and Rockwell (1972)
driving skill. Therefore, recent efforts have been directed towards measured the visual search patterns of novice and experienced car
improving hazard perception using tools that train novice drivers drivers and found that novice drivers scanned a smaller area on
to identify hazards (e.g., McKenna et al., 2006), or to direct their the horizontal plane, and looked closer to the front of the vehicle,
visual attention to hazardous regions of the road environment (e.g., than experienced drivers. Mourant and Rockwell concluded that the
Pollatsek et al., 2006). more constrained visual search patterns of novice drivers reduced
their capability to detect hazards. In a similar study, Chapman and
Underwood (1998) compared the reaction times and visual search
! This project was undertaken for VicRoads and funded from the Victorian Motor-
patterns for detecting hazards for a group of novice drivers and
cycle Safety Levy.
a group of experienced drivers. In contrast to the results found
∗ Corresponding author. Present address: Defence Science and Technology Organ- by Mourant and Rockwell (1972), Chapman and Underwood found
isation (DSTO), 570 Lorimer Street, Fishermans Bend, VIC 3700, Australia. that novices did not differ from experienced drivers in terms of the
E-mail addresses: [Link]@[Link] (S.G. Hosking), area of visual search on the horizontal plane, but showed a larger
[Link]@[Link] (C.C. Liu), [Link]@[Link]
spread of search in the vertical plane. Furthermore, novices fixated
(M. Bayly).
1
Present address: Centre for Behavioural Research in Cancer, The Cancer Council further ahead of the vehicle than experienced drivers. Chapman
Victoria, Drummond Street, Carlton, VIC 3053, Australia. and Underwood concluded that novices may exaggerate instruc-

0001-4575/$ – see front matter © 2009 Elsevier Ltd. All rights reserved.
doi:10.1016/[Link].2009.07.023
 S.G. Hosking et al. / Accident Analysis and Prevention 42 (2010) 196–202 197

tions to look as far ahead as possible, to a degree that is greater ios at slower speeds. However, the use of a closed-loop training
than necessary. In contrast, experienced drivers may have opti- simulator that consisted of hazards that were not configurable
mised their visual search strategies over the course of their driving meant that response times and eye movements to hazards could not
experience. be measured. Since riders could adopt one of many response strate-
The studies of drivers’ visual search patterns reviewed above gies in order to avoid the hazards (e.g., slowing down or swerving
provide seemingly contradictory results, perhaps due to the differ- and maintaining speed), response times could not be defined by
ent methodologies that they employed. Nonetheless, an important any particular response.
and consistent finding from such studies is that experienced drivers As in Liu et al. (2009), the purpose of the present experiment
use more flexible visual search strategies than those of inexpe- was to investigate the influence of experience on the hazard per-
rienced drivers that depend on both the type of road and the ception skills of motorcycle riders. However, in this experiment, a
prevailing traffic conditions (e.g., Falkmer and Gregersen, 2005; more controlled experimental setting was used, such that hazards
Mourant and Rockwell, 1972). For example, the spread of visual were presented at the same points in time, and at the same speeds,
search on the horizontal and vertical planes for experienced drivers for all participants, and where hazards could been seen to approach
have been shown to increase in the presence of hazards and haz- from within a larger field of view. A group of experienced riders
ardous driving contexts, whereas inexperienced drivers do not and two groups of inexperienced motorcycle riders with different
demonstrate such flexibility (e.g., Chapman and Underwood, 1998; levels of driving experience rode an open-loop motorcycle simula-
Falkmer and Gregersen, 2001). A recent review of the role of expe- tor and were asked to search for and respond to hazards. The visual
rience in hazard perception (Underwood, 2007) also found that search patterns and response times of each of the three groups were
response times to hazards vary depending on whether the road compared.
environment is rural, suburban, or urban. Furthermore, mean eye It was predicted that experienced motorcycle riders would
fixation locations and variance in eye fixations consistently vary as a respond to hazards faster and display different visual search pat-
function of an interaction between driving experience, the presence terns from inexperienced motorcycle riders. Furthermore, if the
of hazards, and the road environment. hazard perception skills learned in the driving domain are trans-
ferable to motorcycle riding, then inexperienced motorcycle riders
1.1. Hazard perception for motorcycle riding who were experienced car drivers were expected to display faster
response times and different visual search strategies compared to
Visual scanning patterns for motorcycle riding may differ from inexperienced motorcycle riders who were inexperienced at car
those for driving a car or other vehicle for a number of reasons. The driving.
axis of the rider’s head and eye-height relative to the ground plane
may differ to that of car drivers, and could contribute to different
visual search patterns. Motorcycle riders must also be able to detect 2. Method
a larger set of hazards than car drivers. For example, road-surface-
based hazards (e.g., potholes and debris), while of only low-priority 2.1. Participants
to car driver safety, can severely impact a rider’s ability to remain
upright. Finally, motorcycle riders are more vulnerable than car A total of 37 motorcycle riders were recruited from Monash
drivers to injuries sustained in crashes, and it is possible that this University students and staff, motorcycle user groups, and
vulnerability also leads to visual search strategies that are more motorcycle-training providers. All participants were paid $AUD 50
responsive to changing road conditions and hazardous events. to take part. Participants were allocated into three separate groups:
There is some evidence that motorcycle riders display better (i) Experienced motorcycle riders that were experienced car drivers
hazard perception skills when driving a car than car drivers. For (EM-ED); (ii) Inexperienced motorcycle riders that were experi-
example, Horswill and Helman (2003) found that motorcycle riders enced car drivers (IM-ED); and (iii) Inexperienced motorcycle riders
had faster response times to hazards than car drivers. Motorcy- that were inexperienced car drivers (IM-ID). Table 1 presents a sum-
cle riders and car drivers have also been shown to differ in their mary of demographic information, and shows the average number
visual search patterns. Tofield and Wann (2001) found that motor- or years or months experience at car driving and motorcycle riding
cycle riders who were driving a car looked further ahead than car for each of the three groups.
drivers who had no motorcycle riding experience. They argued that
the motorcycle riders applied hazard perception skills that they
had learned specifically from motorcycle riding. Nagayama et al.
(1980) showed that riders look more at the road-surface immedi- Table 1
Summary of demographics of participants.
ately ahead of the motorcycle to detect any surface irregularities or
other road-surface-based hazards. They found that the proportion EM-ED IM-ED IM-ID
of road-surface scanned while riding a motorcycle was much larger Licence type
than while driving a car. In addition, the distribution of gaze fix- Motorcycle Full Learner permit Learner permit
ations for car drivers was more frequently located at or above the Car Full Full Probationary
horizon, whereas the gaze fixations of motorcycle riders were more Gender (N)
frequently located on the road surface. Underwood and Chapman Male 13 14 7
(1998) found that, compared to car drivers, motorcycle riders had Female 1 0 2
shorter reaction times to hazards, fewer fixations on hazards, and Age (years)
were faster to classify an object as a hazard. Range 28–69 21–48 18–25
If the demands of motorcycle riding do lead to the development Mean 41 31 20

of faster response times to hazards and superior visual search pat- Riding experience
terns, then it would seem likely that the hazard perception skills of Mean 14.4 years 4 months 7 months
Range 5.5–30 years 1–13 months 2–18 months
motorcycle riders vary as they gain riding experience. A recent study
by Liu et al. (2009) found that relative to inexperienced or novice Driving experience
riders, experienced motorcycle riders crashed less often, received Mean 22.8 years 12.5 years 27 months
Range 10–48 years 4–29 years 8–48 months
better performance evaluations, and approached hazardous scenar-
198 S.G. Hosking et al. / Accident Analysis and Prevention 42 (2010) 196–202

crossing lane markers into the path of the motorcycle rider, fallen
tree branches on the road, cars or trucks merging/turning/crossing
in front of the rider, and pedestrians walking on the side of the road
with cars or trucks approaching from the other direction.
The speed and heading of the motorcycle were completely under
computer control. The motorcycle travelled at the signed speed
limit, and while the hazards were simulated to be a threat to the rid-
ers’ safety (as confirmed in a pilot study that measured the response
rate of 10 experienced drivers), the motorcycle did not strike any
hazards nor did it swerve to avoid them. Each of the scenarios
simulated dry weather conditions with bright ambient lighting.
Textural effects were used to simulate the visual characteristics of
road-surface hazards.

2.3. Procedure

Fig. 1. The experimental set-up used in the hazard perception experiment. Note, Participants first completed a questionnaire regarding their age,
however, that participants did not wear a helmet during the experiment (see text
riding and driving experience, and completed a Monash Univer-
for details).
sity Ethics Committee consent form. Participants were asked to
maintain a comfortable riding posture with both hands on the
2.2. Apparatus and scenarios motorcycle handlebars. They then completed a 5-min practice ride
in the simulator to adapt to the simulator environment and become
Simulated motorcycle riding scenarios and hazards were gener- familiar with the experimental task. The practice ride was a differ-
ated by a Silicon Graphics Onyx computer and projected by four ent scenario to those used in the experimental drives.
BarcoGraphics 808 High Performance Graphic Projectors onto a Following the practice ride, the faceLab and scene camera sys-
display screen that subtended a visual angle of 180◦ horizontally tem were calibrated. Participants then completed the rural and
and 40◦ vertically. The scenarios were displayed with a refresh CBD/residential scenarios in a block-randomised order. Participants
rate of 30 Hz and a resolution of 1280 × 768 pixels (front panel) were asked to position themselves as they would normally when
and 640 × 480 pixels (front-side panels). A Crystal River Engineer- riding, observe the road environment, and press a response button
ing Audio Reality Accoustetron II audio system produced accurate (located on the right side of the handlebar) when they detected
localised sound, such as sound from other vehicles. a hazard or hazardous event. Participants were told that a hazard
All participants viewed the scenarios while seated on a Honda was a feature of the road or environment that has an immediate
NSR 150 motorcycle, which was fixed in a vertical position to allow or potential threat to their personal safety, and were asked to use
the riders to have both feet on the foot pedals and emulate a real- this definition of a hazard when responding in the experiment. Par-
istic riding posture (see Fig. 1). The motorcycle was not connected ticipants were also asked to make only a single response to each
to a motion platform and the simulation was an open-loop sys- hazard or each hazardous event.
tem that did not provide interaction between the motorcycle and Immediately after a scenario had been completed, the hazards
the simulation displays. The faceLab TM (version 4.1) eye-tracking that were identified by the participant were replayed and paused
system was used to monitor the eye movements of riders. FaceLab at the time corresponding to each button press. Participants were
cameras were mounted on an adjustable mounting plate that was then asked to name the hazard that they had responded to, and
fixed to the handlebars of the motorcycle. Data from the faceLab their responses were recorded. At the end of the experiment, par-
system were recorded in real time at 60 Hz. A faceLab Scene Cam- ticipants were debriefed about the purpose of the experiment, and
era mounted on a tripod behind the motorcycle riders’ viewpoint paid for their participation. Each participant took approximately 2 h
was used to record the scenarios from the riders’ viewpoint, and to complete the experiment.
integrate and overlay the gaze direction data from faceLab to pro-
vide real-time information about where riders were looking. Pilot
studies found that faceLab could not accurately record the gaze 3. Results and discussion
positions of riders when wearing a helmet because the system was
perturbed by reflections from the helmet. Therefore, participants Preliminary analyses revealed that two of the participants in the
did not have a helmet fitted in the present experiment. In an unpub- IM-ED group did not follow experimental instructions (they clicked
lished control experiment, we found that although riding without a repeatedly for any new object in the simulator environment). An
helmet increased riders’ head movements, eye gaze locations were additional four of the participants (1× IM-ED; 1× EMED; 2× IMID)
not affected. did not complete the experiment (due to simulator discomfort)
The experiment used two simulated riding scenarios—one sce- resulting in missing data. The data from these six participants
nario set in a rural environment, and a second scenario set in a were subsequently omitted from the analyses. Analyses of the
suburban residential area and a central business district (CBD; i.e., unbalanced design (unequal sample sizes) were approached using
inner city). The rural scenario was a 10-min ride on a one-lane dual weighted means and SS Type III in the SPSS (version 13) statistical
carriageway in an 80–100 km speed zone, and the CBD/residential package (see Tabachnick and Fidell, 2001). No other assumptions of
scenario was a 5-min ride on a dual-lane dual carriageway in a analysis of variance (ANOVA) were violated.
60 km zone. Both scenarios included hazards that were identified
in a focus group with experienced motorcycle riders as critical 3.1. Analyses of hazard response and fixation times
to motorcycle hazard perception. Hazards in the CBD/residential
scenario included cars turning in front of the motorcycle rider, Mixed design two-way ANOVAs were performed on: (i) the aver-
pedestrians walking out from behind parked vehicles, potholes, cars age latencies between when a hazard was first visible and when
hook-turning in front of the rider, and cars merging in front of the participants pressed the response button (response time), and (ii)
rider. In the rural scenario, the hazards included oncoming trucks the average amount of time between when a hazard was first visible
 S.G. Hosking et al. / Accident Analysis and Prevention 42 (2010) 196–202 199

and when each participant first fixated on the hazard (first fixa-
tion time). The repeated measures factor was the type of scenario
(two levels; rural environment, and CBD/residential environment)
and the between-subjects factor was riding and driving experience
(three levels: EM-ED, IM-ED, and IM-ID).
There was a significant main effect of riding and driving expe-
rience on mean response times F(2, 29) = 3.61, p < .05, !2 = .20
(see Fig. 2). As can be seen in Fig. 2, response times decreased
monotonically as experience progressed from IM-ID to IM-ED to
EM-ED. Linear contrasts confirmed that this ordinal relationship
between riding and driving experience and response times was
significant (p < .05). There was also a significant main effect of
scenario type F(1, 29) = 455.86, p < .001, !2 = .94. Mean response
times were longer in the rural scenario (15.97 s, se = 0.71) than in the
CBD/residential scenario (8.11 s, se = 0.77). The two-way interaction
between scenario type and riding and driving experience was not
significant.
The analysis of mean first fixation time data found that there
was no significant main effect of riding and driving experience,
nor was there a significant interaction between riding and driv- Fig. 2. Mean response time (in s) for the experienced motorcycle rider/experienced
ing experience and scenario type. However, irrespective of riding driver (EM-ED) group, the inexperienced motorcycle rider/experienced driver (IM-
and driving experience, participants took significantly longer to fix- ED) group, and the inexperienced motorcycle rider/inexperienced driver (IM-ID)
group. Error bars represent 1 standard error of the mean.
ate on hazards in the rural scenario (mean=4.0 s, se = 0.30) than in
CBD/residential scenario (mean=1.46 s, se = 0.16) F(1, 29) = 55.96,
p < .001, !2 = .66. indicative of participants having a larger spread of search in either
Wallis and Horswill (2007) proposed that the earlier detec- the horizontal or vertical planes.
tion of hazards by experienced drivers could be due to either: (i) It was also mentioned earlier that the eye fixation locations vary
differences in their ability to visually detect and discriminate haz- as a function of the interaction between driving experience and the
ards, or (ii) differences in the threshold, or criterion, for judging complexity of the traffic situation; that is, experienced drivers tend
that an object or event is hazardous. Our results support the latter to have less variance in their eye fixation locations in more haz-
explanation because while all participants fixated on the hazards ardous situations (e.g., Chapman and Underwood, 1998; Chapman
at approximately the same time, experienced motorcycle riders et al., 2002). In the present study, the degree of danger to riders
responded to hazards faster than inexperienced riders. This finding varied as they traversed roads that changed from having no haz-
is consistent with the response times of inexperienced and novice ards to having one or more hazards. The presence of a hazard was
car drivers (e.g., Underwood and Chapman, 1998), and suggests that defined as the time extending from when a hazard first appeared on
one aspect of skilled motorcycle hazard perception is the ability to the simulation display to the time that it was no longer visible (i.e.,
quickly recognise which objects in the environment are potential the hazard had been passed). Viewing plane data were extracted
threats. The longer average response times and average times to first from the two-dimensional gaze angle data recorded during each
fixation on hazards that occurred in the rural scenarios relative to scenario.
the CBD/residential scenarios were also consistent with other stud- Participants’ eye movement data were analysed for three
ies showing differences in hazard perception that are dependent on repeated measures factors (hazard presence [two levels: hazard
traffic context (e.g., see Underwood et al., 2002). It seems plausible present, and hazard not present]; viewing plane [two levels: hori-
that in the absence of road-side clutter, hazards in rural scenario zontal, and vertical], and scenario type [two levels: CBD/residential,
are visible from greater distances and therefore riders have more and rural]) and one between-subjects factor (riding and driving
time available to fixate and respond to them. experience [three levels: EM-ED, IM-ED, and IM-ID]). Mean eye fix-
ation locations (") were defined as the sum of single gaze fixation
3.2. Analyses of gaze locations angles (x) recorded at 60 Hz divided by the total number of eye fix-
ations (n) in both the hazard present and hazard absent sections of
As mentioned earlier, previous research has shown that both each driving scenario. Variance of eye fixation locations was calcu-
mean eye fixation locations (the mean point in the visual field that lated using the standard
! deviation (SD) of gaze fixation angles using
participants look overall) and the variance of eye fixation locations the formula SD = ( (x − ")2 )/n.
(the spread of visual search in a scene) vary as a function of the A four-way mixed model ANOVA on mean eye fixation loca-
interaction between driving experience and viewing plane (e.g., tions using the factors described above did not find any significant
Chapman and Underwood, 1998; Mourant and Rockwell, 1972). main effects of riding and driving experience, nor did riding and
These viewing planes are classified as either horizontal (i.e., gaze driving experience significantly interact with any other of the fac-
rotation about the y-axis in a Cartesian coordinate system) or ver- tors. However, there were significant main effects of scenario type
tical (i.e., gaze rotation about the x-axis), where gaze angle is the F(1, 28) = 42.61, p < .001, !2 = .60, and viewing plane F(1, 28) =
angle at the observer’s eye in the right triangle defined by the eye, 92.51, p < .001, !2 = .77. There were also significant two-way
the center of field of vision, and the fixation point. Larger positive interactions between hazard presence and viewing plane F(1, 28) =
or negative values of mean fixations on the horizontal plane are 4.73, p < .05, !2 = .14, and between riding scenario and viewing
indicative of participants looking further to the right or left of the plane F(1, 28) = 10.17, p < .01, !2 = .27. A three-way interaction
center of the field of view; larger positive values on the vertical between hazard presence, viewing plane, and riding scenario was
plane map onto participants looking further ahead in the field of marginally non-significant F(1, 28) = 4.14, p = .052, !2 = .13 (see
view, and larger negative values indicate participants looking closer Table 2). Table 2 shows that in the vertical plane, there was no sig-
to the point of view (e.g., an area on the ground plane closer to the nificant difference between participants’ mean fixation locations
front wheel of the motorcycle). Larger variances in eye fixations are in the CBD/residential and rural scenarios, and that mean fixation
200 S.G. Hosking et al. / Accident Analysis and Prevention 42 (2010) 196–202

Table 2
Standard deviation, mean, and RMS of gaze angle (in ◦ ) as a function of riding scenario, hazard presence, and viewing plane.

Plane Eye gaze CBD/residential Rural

No hazard Hazard No hazard Hazard

Horizontal SD 5.22 (0.31) 4.97 (0.32) 4.42 (0.39) 4.32 (0.38)

Mean −11.14 (1.33) −11.31 (1.33) −9.70 (1.17) −8.94 (1.24)
RMS 13.12 (1.03) 11.57 (0.80) 12.72 (1.09) 10.77 (0.80)

Vertical SD 13.83 (0.59) 12.44 (0.50) 8.31 (0.49) 8.33 (0.44)

Mean −0.15 (0.64) −0.22 (0.62) 4.03 (0.45) 3.73 (0.58)
RMS 14.21 (0.63) 9.57 (0.46) 11.75 (0.46) 7.96 (0.40)

Note: Numbers in brackets are standard errors.

locations were not affected by the presence of a hazard. However, fixations. It has been suggested that the location of gaze most pre-
in the horizontal plane, there were significant differences in mean ferred by experienced and novice drivers is straight ahead, in line
fixation locations between the CBD/residential and rural scenarios with the direction the vehicle is travelling (Mourant and Rockwell,
in both the no hazard (p < .001) and hazard (p < .001) conditions. 1972; Underwood et al., 2003) at a point in space that maps onto
A separate four-way mixed model ANOVA using the same fac- the focus of optic expansion (FOE). The FOE is an optic variable that
tors described above on the standard deviation of eye fixation data has been shown to be a critical stimulus for the guidance of driving
also did not find a significant main effect of riding and driving behaviours such as judgements of heading (direction) and time-
experience. However, there was a significant three-way interaction to-collision, and the control of steering (e.g., Gibson, 1950; Lappe
between viewing plane, hazard presence, and riding and driving et al., 1999; Lee, 1980). However, in hazardous driving situations,
experience F(2, 28) = 5.12, p = .01, !2 = .28 (see Table 3). Table 3 previous research has shown that experienced drivers spread their
shows that while fixation variations were larger in the horizontal fixations further to the left and right from the FOE than inexpe-
plane than in the vertical plane, the spread of fixations were sig- rienced drivers (Falkmer and Gregersen, 2005), inspecting other
nificantly different for the IM-ED group in the no hazard condition road-users and potential hazards. Hence, an additional analysis
compared to the IM-ID group in the hazard condition (p < .05). was conducted on the absolute deviation of gaze rotations from
The analysis of SD of eye fixation data also found significant the origin (centre) of the visual field (corresponding to the FOE)
main effects of scenario type F(1, 28) = 65.18, p < .001, !2 = .70, using a" root mean square (RMS) statistical measure, defined as
the presence of a hazard F(1, 28) = 52.93, p < .001, !2 = .65, and RMS = (x12 + x22 + · · · + xn2 )/n.
viewing plane F(1, 28) = 198.2, p < .001, !2 = .88, and significant Mean RMS data for each participant were analysed for both the
two-way interactions between scenario type and viewing plane rural and CBD/residential scenarios, and for both the vertical and
F(1, 28) = 112.39, p < .001, !2 = .80, hazard presence and view- horizontal planes. Larger RMS values indicate that the mean abso-
ing plane F(1, 28) = 42.22, p < .001, !2 = .60, and scenario type lute deviation of eye gaze from the origin was larger. A four-way
and hazard presence F(1, 28) = 65.47, p < .001, !2 = .70. mixed model ANOVA, using the same factors as described above,
The three-way interaction between scenario type, viewing on mean RMS gaze data found a significant three-way interaction
plane, and hazard presence on SD of eye fixation data was also between viewing plane, hazard presence, and riding and driving
significant F(1, 28) = 47.65, p < .001, !2 = .63. As can be seen in experience F(2, 28) = 4.57, p = .02, !2 = .25 (see Table 3). As can
Table 2, the variance in visual fixation angles on the vertical plane be seen in Table 3, the differences in mean RMS values between
was not significantly different between the CBD/residential sce- hazard and no hazard conditions on the vertical and horizontal
nario and the rural scenario, nor was it significantly affected by planes varied as a function of the riding and driving experience
the presence of a hazard. However, on the horizontal plane, the of participants. The mean RMS values in the horizontal plane were
variance in visual fixation angles was significantly larger for the significantly reduced in the hazard conditions relative to the no
CBD/residential scenario than that of the rural scenario in both the hazard conditions; however, this reduction only occurred for the
hazard (p < .001) and no hazard (p < .001) conditions. The vari- EM-ED (p < .001), and IM-ED (p < .001) groups—the IM-ID group
ance in visual fixation angles was also reduced when a hazard was did not reduce their mean RMS when a hazard was present. Table 3
present in the CBD/residential scenario (p < .05), but not in the shows that the mean RMS in the vertical plane of the EM-ED group
rural scenario. was greater than that of the IM-ED and IM-ID groups. However,
In contrast to the eye movement data reported by Chapman and these differences did not reach statistical significance, and did not
Underwood (1998) and Mourant and Rockwell (1972), our analysis vary significantly as a function of the presence of a hazard.
did not find any meaningful differences between groups in terms In summary, the EM-ED and IM-ED groups reduced the distance
of their mean eye fixation locations, nor their mean variance in eye of their eye fixations from the FOE when a hazard was present,
whereas the IM-ID riders did not. These results are consistent with
Table 3 previous research showing that experienced drivers have a more
Standard deviation and RMS of gaze angle (in ◦ ) as a function of riding experience, flexible search strategy than inexperienced drivers, and that expe-
hazard presence, and viewing plane. rienced drivers reduce their spread of search in the horizontal plane
Group Eye gaze Horizontal Vertical in the presence of hazards (e.g., Chapman and Underwood, 1998).
There was also a trend for mean RMS of eye fixations to be larger on
No hazard Hazard No hazard Hazard
the vertical plane for the experienced riders relative to the inexpe-
EM-ED SD 11.49 (0.71) 10.64 (0.62) 4.92 (0.48) 4.70 (0.48) rienced riders, which could indicate that experienced riders looked
RMS 12.09 (0.69) 9.80 (0.57) 13.92 (1.26) 13.52 (1.32)
further ahead. However, these differences did not reach statistical
IM-ED SD 11.62 (0.77) 10.72 (0.67) 4.25 (0.52) 4.12 (0.52) significance.
RMS 12.32 (0.74) 9.51 (0.63) 11.46 (1.38) 10.89 (1.43)
The analysis of the RMS data also found significant main effects
IM-ID SD 10.09 (0.97) 9.79 (0.84) 5.29 (0.65) 5.12 (0.66) of scenario type F(1, 28) = 50.80, p < .001, !2 = .65, and presence
RMS 11.29 (0.92) 10.20 (0.80) 11.63 (1.78) 10.83 (1.78) of a hazard F(1, 28) = 51.41, p < .001, !2 = .65, and significant
Note: Numbers in brackets are standard errors. two-way interactions between scenario type and viewing plane
 S.G. Hosking et al. / Accident Analysis and Prevention 42 (2010) 196–202 201

F(1, 28) = 11.54, p < .001, !2 = .29, and presence of hazard and perienced riders has important implications for hazard perception
viewing plane F(1, 28) = 33.03, p < .001, !2 = .54. There was also training. Some success in training visual scanning and anticipa-
a significant three-way interaction between scenario type, view- tory behaviour has been found with novice drivers (McKenna et
ing plane, and hazard presence F(1, 28) = 7.394, p = .01, !2 = al., 2006; Pollatsek et al., 2006). However, Crundall et al. (2002)
.21 (see Table 2). Table 2 shows that the mean RMS on the have argued that eye movement training in new drivers may be
vertical viewing plane were not significantly different for the counter-productive, if such programs are asking novice drivers to
rural and CBD/residential scenarios and did not vary as a func- perform visual scans that exceed their attentional resources. Fur-
tion of the presence of a hazard. However, mean RMS of gaze ther research is needed to determine whether such training could
angles in the horizontal plane were significantly larger for the also be applied successfully to novice motorcycle riders. It should
CBD/residential scenario in both the no hazard (p < .001) and haz- also be noted that while eye movements are an important indicator
ard (p < .001) conditions. The reduction in mean RMS values that of visual search patterns, the visual system is capable of directing
occurred in the hazard conditions was also significantly larger in attention to areas of the visual field that are outside of foveal vision
the CBD/residential scenario (p < .001) than in the rural scenario (i.e., the area of the visual field that is in focus), and may also use
(p < .001). These data provides converging evidence with other peripheral vision to detect hazards.
research showing that variances in eye fixations between subur- In summary, we have shown that experienced motorcycle rid-
ban and rural scenarios occurs primarily in the horizontal plane ers respond to hazards earlier than inexperienced riders, and have
(e.g., Chapman et al., 2002). more flexible visual search patterns. Furthermore, there appears to
be some degree of hazard perception skill learned through prior
driving experience that can be transferred to the motorcycle riding
4. Summary and conclusions domain. Skilled hazard perception, as demonstrated by experienced
motorcycle riders, involves faster recognition of hazardous objects,
This study investigated the response times to hazards and visual and flexibility in visually scanning that is dependent on road-types
scanning behaviours of three different groups of motorcycle rid- and road-contexts.
ers: (i) experienced motorcycle riders that were experienced car
drivers (EM-ED), (ii) inexperienced motorcycle riders that were
References
experienced car drivers (IM-ED), and (iii) inexperienced motorcycle
riders that were inexperienced car drivers (IM-ID). It was predicted Chapman, P.R., Underwood, G., 1998. Visual search of driving situations: danger and
that the EM-ED group would respond to hazards faster and dis- experience. Perception 27, 951–964.
Chapman, P.R., Underwood, G., Roberts, K., 2002. Visual patterns in trained and
play more flexible visual search patterns across different road and
untrained novice drivers. Transportation Research F: Psychology and Behaviour
traffic contexts relative to the IM-ED and IM-ID groups. It was also 5, 157–167.
predicted that the greater amount of driving experience of the IM- Crundall, D., Chapman, P., Phelps, N., Underwood, G., 2003. Eye movements and
ED group would transfer to better hazard perception skills when hazard perception in police pursuit and emergency response driving. Journal of
Experimental Psychology: Applied 9, 163–174.
riding a motorcycle relative to that of the IM-ID group. Crundall, D., Underwood, G., Chapman, P., 2002. Attending to the peripheral world
The results showed that experienced motorcycle riders were while driving. Applied Cognitive Psychology 16, 459–475.
faster to respond to hazards than inexperienced riders, and that Falkmer, T., Gregersen, N.P., 2001. Fixation patterns of learner drivers with and with-
out cerebral palsy (CP) when driving in real traffic environments. Transportation
such faster response times may be due to experienced riders having Research: Psychology and Behaviour F 4, 171–185.
a visual search pattern that is more flexible than that of inexpe- Falkmer, T., Gregersen, N.P., 2005. A comparison of eye movement behavior of inex-
rienced riders. There was also some evidence that prior driving perienced and experienced drivers in real traffic environments. Optometry and
Vision Science 82, 732–739.
experience can lead to some improvements in the hazard percep- Gibson, J.J., 1950. The Perception of the Visual World. Houghton Mifflin.
tion and responding skills of inexperienced riders, although further Groeger, J.A., Chapman, P.R., 1996. Judgement of traffic scenes: the role of danger and
improvements would be required in order for the hazard percep- difficulty. Applied Psychology 10, 349–364.
Horswill, M.S., Helman, S., 2003. A behavioural comparison between motorcyclists
tion skills of these riders to reach the levels attained by experienced
and a matched group of non motorcycling car drivers: factors influencing acci-
riders. Taken together, these results appear to indicate that inex- dent risk. Accident Analysis and Prevention 35, 589–597.
perienced riders, irrespective of driving experience, would benefit Horswill, M.S., McKenna, F.P., 2004. Driver’s hazard perception ability: situation
awareness on the road. In: Banbury, S., Tremblay, S. (Eds.), A Cognitive Approach
from motorcycle-specific hazard perception training.
to Situation Awareness. Ashgate, Aldershot, UK.
It would also appear that at least some of the hazard perception Liu, C.C., Hosking, S.G., Lenne, M.G., 2009. Hazard perception abilities of experi-
skills learned in the driving domain are transferable to motorcycle enced and inexperienced motorcyclists: an interactive simulator experiment.
riding. There was some evidence that inexperienced motorcycle rid- Transportation Research Part F: Psychology and Behaviour 12, 325–334.
Lappe, M., Bremmer, F., van den Berg, A.V., 1999. Perception of self-motion from
ers who were experienced drivers were faster to respond to hazards visual flow. Trends in Cognitive Sciences 3, 329–336.
than the group of inexperienced riders/drivers. Furthermore, the Lee, D.N., 1980. The optic flow field: the foundation of vision. Philosophical Trans-
inexperienced rider/experienced driver group also reduced their actions of the Royal Society of London B, 169–179.
Mayhew, D.R., Simpson, H.M., Pak, A., 2003. Changes in collision rates among novice
spread of search when hazards were present, thus demonstrating drivers during the first months of driving. Accident Analysis and Prevention 35,
more flexible search patterns than the inexperienced rider/driver 683–691.
group. It should be noted that, given the relatively small sample McKenna, F.P., Crick, J.L., 1994. Hazard perception in drivers: a methodology for test-
ing and training. Transport and Road Research Laboratory Report 313. Transport
sizes used in this study, any non-significant differences between Research Laboratory, Crowthorne, Berkshire.
groups should be interpreted with caution because of a possible lack McKenna, F.P., Horswill, M.S., Alexander, J.L., 2006. Does anticipatory training affect
of statistical power in the analyses. Also noteworthy is the poten- drivers’ risk-taking? Journal of Experimental Psychology: Applied 12, 1–10.
Mourant, R.R., Rockwell, T.H., 1972. Strategies of visual search by novice and experi-
tial confound between age and experience in our sample. While
enced drivers. Human Factors 14, 325–335.
this has been noted in many studies of the effects of experience on Nagayama, Y, Morita, T., Miura, T., Watanabem, J., Murakami, N., 1980. Motorcy-
driving performance (e.g., Sagberg and Bjørnskau, 2006), it is gen- clists’ visual scanning pattern in comparison with automobile drivers. Society
of Automotive Engineers (SAE) Technical Paper 790262. Society of Automotive
erally accepted that age and experience have independent effects
Engineers, Warrendale, PA.
on performance (e.g., Mayhew et al., 2003). Pollatsek, A., Fisher, D.L., Pradhan, A., 2006. Identifying and remedying failures of
A critical issue for future research is how visual scanning strate- selective attention in younger drivers. Current Directions in Psychological Sci-
gies may interact with attentional resources (i.e., the capacity of ence 15, 255–259.
Quimby, A.R., Maycock, G., Carter, I.D., Dixon, R., Wall, J.G., 1986. Perceptual Abil-
individuals to deal with multiple competing resources). The issue of ities of Accident Involved Drivers. Transport and Road Research Laboratory,
differences in attentional resources between experienced and inex- Crowthorne, Berkshire.
202 S.G. Hosking et al. / Accident Analysis and Prevention 42 (2010) 196–202

Sagberg, F., Bjørnskau, T., 2006. Hazard perception and driving experience among Underwood, G., Chapman, P., Bowden, K., Crundall, D., 2002. Visual search while driv-
novice drivers. Accident Analysis and Prevention 38, 407–414. ing: skill and awareness during inspection of the scene. Transportation Research
Shinar, D., 2008. Looks are (almost) everything: Where drivers look to get informa- F: Psychology and Behaviour 5, 87–97.
tion. Human Factors 50, 380–384. Underwood, G., Chapman, P., Brocklehurst, N., Underwood, J., Crundall, D., 2003.
Tabachnick, B.G., Fidell, L.S., 2001. Using Multivariate Statistics, Fourth Edition. Allyn Visual attention while driving: Sequences of eye fixations made by experienced
& Bacon, Needham Heights, MA. and novice drivers. Ergonomics 46, 629–646.
Tofield, M.I., Wann, J.P., 2001. Do motorcyclists make better car drivers? In: Proceed- Underwood, G., Phelps, N., Wright, C., Van Loon, E., Galpin, A., 2005. Eye fixation
ings of Psychological Post-graduate Affairs Group conference, Glasgow, Scotland. scanpaths of younger and older drivers in a hazard perception task. Ophthalmic
Underwood, G., 2007. Visual attention and the transition from novice to advanced and Physiological Optics 25, 346–356.
driver. Ergonomics 50, 1235–1249. Wallis, T.S.A., Horswill, M.S., 2007. Using fuzzy signal detection theory to determine
Underwood, G., Chapman, P., 1998. Eye movements and hazard perception ability. In: why experienced and trained drivers respond faster than novices in a hazard
Grayson, G.B. (Ed.), Behavioral Research in Road Safety VIII. Transport Research perception test. Accident Analysis & Prevention 39, 1177–1185.
Laboratory, Crowthorne, UK.