Professional Documents
Culture Documents
Crossing Roads Safely An Experimental Study of Age Differences in Gap Selection by Pedstrians
Crossing Roads Safely An Experimental Study of Age Differences in Gap Selection by Pedstrians
Crossing Roads Safely An Experimental Study of Age Differences in Gap Selection by Pedstrians
Abstract
The over-representation of older pedestrians in serious injury and fatal crashes compared to younger adults may be due, in part, to age-related
diminished ability to select gaps in oncoming traffic for safe road-crossing. Two experiments are described that examine age differences in gap
selection decisions in a simulated road-crossing environment. Three groups of participants were tested, younger (30–45 years), young–old
(60–69 years) and old–old (>75 years). The results showed that, for all age groups, gap selection was primarily based on vehicle distance and
less so on time-of-arrival. Despite the apparent ability to process the distance and speed of oncoming traffic when given enough time to do so,
many of the old–old adults appeared to select insufficiently large gaps. These results are discussed in terms of age-related physical, perceptual
and cognitive limitations and the ability to compensate for these limitations. Practical implications for road safety countermeasures are also
highlighted, particularly the provision of safe road environments and development of behavioural and training packages.
© 2005 Elsevier Ltd. All rights reserved.
0001-4575/$ – see front matter © 2005 Elsevier Ltd. All rights reserved.
doi:10.1016/j.aap.2005.04.017
J.A. Oxley et al. / Accident Analysis and Prevention 37 (2005) 962–971 963
choose safe time gaps in traffic in a simulated road-crossing road section, with two approaching vehicles travelling from
task as well as some of the factors involved in such judge- the right-hand side (near-side lane), which is the lane clos-
ments. est to the pedestrian in Australia. There was no traffic in the
far-side lane.
Time gap and speed of the vehicles were manipulated with
2. Experiment one five levels of time gap (1, 4, 7, 10 and 13 s) and three levels
of vehicle speed (40, 60 and 80 k/h) resulting in 15 different
The first experiment investigated age differences in gap traffic scenarios. Distance co-varied as a function of these
selection during road-crossing tasks in a simulated traffic two manipulations. Levels of time gap were chosen based
environment in which time gap and vehicle speed were sys- on group average walking speeds in observational studies
tematically varied. On the basis of findings of on-road obser- (Oxley, 2000; Oxley et al., 1997a), and represented theoret-
vations (Oxley, 2000; Oxley et al., 1997a,b) which showed ically safe and unsafe time gaps. Simulated traffic scenes
that some of the oldest pedestrians left larger distance gaps were downloaded onto three video tapes, each containing the
than their younger counterparts, but not longer time gaps and scenes in a different random order. Images were projected
consequently made risky crossing decisions, it was expected onto a large white curved screen measuring approximately
that the road-crossing responses of older adults would be 2.5 m in width and 1.5 m in height. The size of projection
based on distance, not time gap, and that they would leave was roughly equivalent to the visual angle of vehicles when
smaller safety margins than younger adults. viewed from the side of the road. Participants were able to
scan the traffic scene freely, hence any undetected periph-
2.1. Method eral field loss would not have affected judgements. Each
participant viewed the traffic scenes on each video tape and
2.1.1. Participants presentation was counterbalanced. Thus, they viewed a total
There were three groups of participants: 18 younger adults of 45 traffic scenes.
aged between 30 and 45 years, 18 young–old adults aged Responses were made on a computer keyboard on the desk
between 60 and 69 years, and 18 old–old adults aged 75 years in front of participants. Most of the keys were blackened and
and over. All participants were volunteers, licensed drivers covered. Two keys (‘D’ and ‘J’) labelled ‘YES’ and ‘NO’
and in good health. Assessment of functional performance were available for participants to indicate whether they would
revealed group differences, with the oldest group performing ‘cross’ the road or not. The keys for numbers 1–9, with labels
more poorly than younger adults on cognitive, perceptual, ‘very unsafe’ below the 1 key and ‘very safe’ below the 9 key,
visual and physical measures, but all performances were provided a nominal rating scale on which participants were
within the normal range for the relevant age groups. In partic- asked to rate their safety of the road-crossing.
ular, no significant peripheral visual field losses were detected
in the older participants. Details of the outcomes of these 2.1.3. Procedure
assessments and their relationship to road-crossing decisions Participants were seated at a desk in a darkened, quiet room
are reported elsewhere (Oxley et al., 2001). 2 m in front of the projection screen with their right index fin-
ger resting on the ‘YES’ key and their left index finger on the
2.1.2. Stimulus materials ‘NO’ key and looking down towards the keyboard. A buzzer
A simulated road environment of moving traffic generated sounded as the first ‘trigger’ vehicle passed the point of cross-
from data files from a mid-range driving simulator was used ing and at the same time activated a timer. Participants were
in these experiments (Fig. 1). It showed an undivided, straight instructed to look at the traffic scene as soon as they heard the
two-way residential road (with visual and audio features to buzzer and to decide whether or not they would ‘cross’ in front
make the environment as realistic as possible) from the per- of the second vehicle (walking normally across the street),
spective of a pedestrian waiting at the kerb on a mid-block responding as quickly as possible by pressing the ‘YES’ or the
‘NO’ key. This deactivated the timer and the time interval was
recorded as decision time. After this, participants were asked
to rate how safe or unsafe they thought the ‘crossing’ would
have been by pressing the appropriate key (1–9). No time lim-
its were imposed for this response. Practice trials were given
until participants indicated that they fully understood the
task.
Walking time over a distance equivalent to the width of
an average road lane (5.6 m) at two walking paces (normal
walking pace and fast walking pace, i.e., walking as fast as
Fig. 1. Stimulus traffic scenarios presented in the road-crossing simulation
they could) was also measured.
(Note: the road appeared straight when projected onto a curved presentation Total testing time took approximately 60 min (with small
screen). breaks for older participants, if required).
964 J.A. Oxley et al. / Accident Analysis and Prevention 37 (2005) 962–971
Table 1
Mean walking times (normal and fast-paced) and decision times by age group (with standard deviation)
Normal-paced walking time (s) Fast-paced walking time (s) Decision time (s)
Mean (S.D.) Minimum Maximum Mean (S.D.) Minimum Maximum Mean (S.D.) Minimum Maximum
Young group (n = 18) 3.87 (0.31) 3.22 4.38 2.72 (0.32) 2.07 3.13 0.66 (0.20) 0.18 1.27
Young–old group (n = 18) 4.60 (0.61) 3.69 5.78 3.57 (0.55) 2.75 4.72 0.88 (0.15) 0.15 3.59
Old–old group (n = 18) 6.96 (1.57) 4.56 9.63 5.74 (1.72) 3.85 9.25 1.45 (0.42) 0.23 8.01
Fig. 2. Proportion of actual (connected data points) and predicted (unconnected data points) yes responses as a function of age group, distance gap, vehicle
speed and time gap.
J.A. Oxley et al. / Accident Analysis and Prevention 37 (2005) 962–971 965
group, vehicle speed and walking time also predicted crossing ences between actual and predicted responses, p’s < 0.01. In
decisions, χ2 (2) = 153.54, p < 0.001, R = 0.20, χ2 (2) = 90.76, time gaps of 4 s, it was predicted that 44% of young–old
p < 0.001, R = 0.16 and χ2 (1) = 32.33, p < 0.001, R = 0.10, participants could have responded yes, but that no old–old
respectively. A significant time gap by age group interaction participant should have done so. In the time gap of 7 s, all
was also found, χ2 (8) = 152.53, p < 0.001, R = 0.21. No other young–old participants, but only 61% of the old–old partic-
interaction was significant in this model. Model 2 further ipants could have responded yes. Actual response rates of
revealed that distance gap was a strong predictor of cross- the young–old participants were lower than predicted for all
ing decisions, χ2 (14) = 426.27, p < 0.001, R = 0.36. As in critical time gaps, indicating that their decisions were overly
Model 1, age group and walking time predicted crossing deci- conservative. In contrast, actual response rates of the old–old
sions, χ2 (2) = 152.26, p < 0.001, R = 0.21 and χ2 (1) = 36.25, participants were generally higher than they should have been
p < 0.001, R = 0.10, respectively. The interaction between age (Chi-square p’s < 0.05). There was one exception from this
group and distance gap was not significant. These findings general trend: In the 7 s time gap condition, where vehicle
indicate that responses in general were primarily based on distance was close, fewer old–old participants than expected
the distance of the vehicle and to a much lesser extent on indicated that they would have crossed.
time gap, and that groups responded differently to increases
in time gaps. 2.2.3. Safety margins
Fig. 2 shows the proportion of positive crossing responses Safety margins were defined as the sum of the mean walk-
by vehicle conditions for each age group. These data show ing time (i.e., the average of normal and fast walking times)
that all participants were less likely to indicate that they plus decision time, being subtracted from the time gap of the
would cross when time and distance gaps were small than approaching vehicle for each trial in which a yes crossing
when these were larger. At distances of 22 m and below response was made. This measure, calculated for each indi-
(time gap of 1 s) the response rate was close to zero for all vidual separately for every trial, therefore, represented the
three groups. Beyond that, the proportion of yes responses time interval between the participant reaching the centre of
increased rapidly for the young participants who reached the road and the vehicle arriving at the crossing line. Individ-
close to 100% asymptote when the vehicle was more than ual safety margins were then plotted against individual mean
100 m or 7 s away from them. Response rates for the other walking times for the critical time gaps of 4 and 7 s and sub-
two groups increased more gradually and they reached close jected to regression analysis. These distributions are shown
to 100% asymptote later (young–old group at about 150 m or separately for age groups in Fig. 3. Also shown on each dis-
10 s; old–old group at about 200 m or 13 s). It is noteworthy tribution is the zero-line that denotes where safety margins
that between 10 and 40% of old–old participants decided to were at the zero level. Any points below this line indicate an
cross when time gaps were only 4 s (depending on vehicle unsafe response, while points above the line indicate a safe
distance) even though most took longer to walk the distance response.
of a carriageway, even at their fastest pace. For all age groups, the safety margin decreased with
Fig. 2 also illustrates the findings of regression analyses increasing walking plus decision times and this was partic-
that distance, not time gap, was the main determinant of cross- ularly so for the oldest group. The distribution of the young
ing decisions for all groups. Vehicle speed was also taken into participants (Fig. 3a) shows little variability and relatively
account, but to a lesser extent. For instance, for the three time safe crossing decisions. Only 19% of young participants
gap conditions of 4 s, the proportion of positive responses indicated that they would have crossed in an unsafe man-
of all groups increased as distance gap increased. This also ner. Moreover, the unsafe crossings for this group were only
occurred for the 7 and 10 s time gap conditions for the two marginally unsafe, generally falling between 0 and −1 s of
older groups (the young group was already at or close to 100% safety margin. The distributions of the older groups (Fig. 3b
asymptote). Time gap also seemed to contribute to crossing and c) showed more variability, especially that of the oldest
decisions. In the 88.9 m/4 s and 77.7 m/7 s conditions, partic- group. While only 18% of the young–old participants indi-
ipants in all groups were more likely to respond positively in cated unsafe crossings, a much larger proportion (70%) of
the 7 s condition even though the corresponding distance was the old–old participants did so. The unsafe responses of the
smaller than that in the 4 s condition. If decisions had been young–old group fell between 0 and −2 s of safety margin. In
made solely on the basis of distance, the response rate should contrast, the unsafe responses of the old–old group extended
have decreased for the 77.7 m/7 s gap. to −10 s. This confirms that many of the old–old participants
An additional analysis was carried out to examine whether who decided to cross with a 4 and 7 s time gap should have
any adjustment in response by older participants was large said ‘no’.
enough to compensate for increased walking times. Actual
proportions of yes responses of young–old and old–old par-
ticipants in critical time gaps (4 and 7 s) were compared 3. Experiment two
with those predicted from the mean of individual normal and
fast walking times (shown in Fig. 2 as unconnected/unfilled In the first experiment, participants were asked to make
data points). Chi-square analysis revealed significant differ- decisions under time pressure and consequently mean deci-
966 J.A. Oxley et al. / Accident Analysis and Prevention 37 (2005) 962–971
3.1. Method
3.1.1. Participants
Forty-five participants took part in this experiment con-
sisting of 15 young adults aged between 30 and 45 years,
15 young–old adults age between 60 and 69 years and 15
old–old adults aged 75 years and over. All participants had
taken part in the first experiment and were familiar with the
task.
3.2. Results
Fig. 3. Distribution of safety margins for all positive crossing decisions at 4
and 7 s time gaps plotted as a function of walking time.
3.2.1. Yes/no responses
Hierarchical modelling of the data was employed to exam-
sion times were much lower than in the on-road observational ine the effect of independent variables on yes/no responses.
studies (close to six times shorter for all age groups). It is Walking time was held constant while the impact of age
possible that under time constraints, decisions are made pri- group, time gap, vehicle speed, distance and presentation
marily on the most immediately available, and hence, most time were examined in two separate analyses. Model 1
easily accessible information which in this case is the distance included age group, time gap, vehicle speed and presen-
of the vehicle. Judgements about speed require information tation time. This analysis revealed that age group, vehicle
about vehicles over time and this may require longer time to speed and time gap were significant predictors of crossing
process with advancing age. It is, therefore, possible that, if decisions: age group, χ2 (2) = 51.17, p < 0.001, R = 0.30, time
given more time to inspect the stimulus display and to make gap, χ2 (2) = 131.45, p < 0.001, R = 0.30, and vehicle speed,
a crossing decision, participants are more likely to base their χ2 (1) = 102.01, p < 0.001, R = 0.27. Model 2 included age
decisions on the time gap (integrating distance and speed) group, vehicle distance and presentation time and revealed
rather than the distance alone. This may be especially impor- that distance gap was a significant predictor of crossing deci-
J.A. Oxley et al. / Accident Analysis and Prevention 37 (2005) 962–971 967
with a change from 7 to 10 s, despite a decrease in distance predicted levels, for larger distances and fast vehicle speed,
and vehicle speed. Interestingly, in the 5 s presentation time regardless of presentation time. It seems, therefore, that even
conditions, asymptote was reached at the same point as for when old–old adults had sufficient time to process the time
the other two groups and earlier than in the 1 s presentation gap of oncoming vehicles, many of them still made unsafe
time condition and in Experiment 1. crossing decisions based on vehicle distance.
A comparison of the actual and predicted response rates
for critical time periods of 4 and 7 s (predictions were based 3.2.2. Safety margins
on walking times and are shown as unconnected data points The analysis of safety margins (Fig. 5) shows that, as
in Fig. 4b and c) suggests that, as in Experiment 1, young–old in Experiment 1, only a small proportion of young and
participants responded conservatively for all critical time and young–old participants decided to cross with insufficient
distance gaps under both presentation conditions. Proportions intervals between them and the vehicle arriving at the cross-
of yes responses by old–old participants were generally above ing line. In the 1 s presentation time condition only 14.5% of
Fig. 5. Distribution of safety margins for all positive crossing decisions at 4 and 7 s time gaps plotted as a function of walking time.
J.A. Oxley et al. / Accident Analysis and Prevention 37 (2005) 962–971 969
the young group fell below the zero line, in the 5 s condition it such as ‘the further the car is away from me, the safer it is
was 15.9%. The corresponding proportions for the young–old to cross’. This finding suggests an immediacy effect where a
group were 23.4 and 10.6%, respectively. However, just as in vehicle far away, irrespective of its travelling speed, is judged
Experiment 1, many of the oldest group who indicated that to be less threatening than one close up.
they would have crossed did so when it was clearly not safe. While making crossing judgements primarily on vehicle
In the 1 s condition, 76.9% made unsafe decisions, compared distance may be a strategy that works reasonably well for the
with 64.2% in the 5 s condition. These differences in propor- young and young–old participants, who can walk fast enough
tions as a function of presentation time were evaluated by to avoid even relatively close vehicles, it clearly leads to
Chi-square and found not to be significant (p = 0.08 and 0.09 many risky crossing decisions by slower old–old participants.
for young–old and old–old participants, respectively). This However, considering that all groups use the same informa-
confirms the earlier conclusion that the unsafe behaviour of a tion on which to base their decisions, group differences in
large proportion of old–old participants was not affected by the processing of traffic information cannot fully explain the
the time they had to inspect the traffic scene. high proportion of risky behaviour by old–old participants
observed in previous studies (Carthy et al., 1995; Oxley et
al., 1997a) and in the experiments reported here.
4. Discussion There are some alternative explanations. One, it is possi-
ble that old–old participants resort more frequently than their
The current study confirmed the findings of the observa- younger counterparts to making crossing decisions which
tional studies, showing that older adults made risky crossing are based on vehicle distance, as such decisions often have
decisions (Oxley, 2000; Oxley et al., 1997a). A large propor- to be made quickly. The findings of Experiment 2 sug-
tion of the oldest group (>75 years) decided that they could gest that under time-constrained conditions, old–old par-
cross the virtual road when in fact it was unsafe for them to do ticipants seemed to have difficulties integrating speed and
so, given their walking times. Negative safety margins sug- distance information of approaching vehicles. When presen-
gest that in an on-road environment, drivers would have had tation times were short, time gap had little effect on the
to take evasive action to avoid a collision. Increased presenta- crossing decisions of this group. This may lead to danger-
tion time of the traffic scene did not significantly change the ous road-crossing behaviour.
proportion of risky decision makers amongst old–old partic- Interestingly, decision times of the oldest group were
ipants. The young–old participants, on the other hand, made longer than those of younger participants both in obser-
relatively conservative decisions and very few of them left vational studies (Oxley, 2000; Oxley et al., 1997a) and in
unsafe safety margins. Indeed, many of them decided not to Experiment 1. Notwithstanding, neither longer decision nor
cross, when their walking times indicated that they could have presentation times seemed to lead to safer decisions, both
done so safely. on the road and here under experimental conditions. This
The most surprising finding was that it was not only the indicates that opportunity to process speed and distance infor-
oldest participants, but also both younger groups who made mation about oncoming vehicles may help, but does not
road-crossing decisions that were based primarily on the dis- necessarily lead to a much higher proportion of safer deci-
tance of oncoming vehicles and to a lesser extent on time sions by old–old adults. In contrast, the other two groups
of arrival. Others, too, have found that younger drivers and made crossing decisions that were generally safe, despite all
pedestrian use the same visual information when making gap indications that vehicle distance was the primary determinant
judgements (Connelly et al., 1998; De Lucia et al., 2003; of these decisions and despite indications that the young–old
Simpson et al., 2003). Therefore, while in theory consider- participants also had difficulties processing vehicle speed and
ation of both distance and speed of approaching vehicles is distance when under time constraints in Experiment 2.
highly pertinent to crossing roads, in practice this does not This leads us to consider the second alternative expla-
seem to happen. There are several reasons why crossing deci- nation. At all vehicle distances up to asymptote, old–old
sions could be primarily based on vehicle distance in these participants were less likely to indicate that they would have
experiments. It may be that the results are an artefact of the crossed than their young counterparts. This was appropri-
impoverished two-dimensional viewing conditions employed ate as their longer walking times demanded larger gaps and
here. This is, however, an unlikely explanation. First, a val- suggests that they seem to take their slower walking speeds
idation study (Oxley et al., 1997b) showed that crossing into account when making road-crossing decisions. Despite
decisions and perceptions of safety by younger and older this ability, old–old participants were not able to compen-
adults in real-world and filmed versions of traffic scenes were sate adequately for their slower walking speed to make safe
highly correlated. Second, the results clearly show that, given decisions. This possibility was also considered by Lee et al.
favourable exposure conditions, time gap was processed by (1984) who argued that, with diminished perceptual, cogni-
all groups, including the old–old group, however, this vari- tive and motor abilities, some older adults may experience
able did not seem to function as the primary determinant of difficulty in accommodating their judgments of safe gaps in
crossing decisions. It is possible that, on inspection of the the traffic to slowed walking times. Sabey (1988) and Yanik
oncoming traffic, pedestrians would make an initial decision and Monforton (1991) also suggested that losses occur in
970 J.A. Oxley et al. / Accident Analysis and Prevention 37 (2005) 962–971
Salthouse, T., 1991. Theoretical Perspectives on Cognitive Aging. Simpson, G., Johnston, L., Richardson, M., 2003. An investigation of road
Lawrence Erlbaum, New Jersey. crossing in a virtual environment. Accid. Anal. Prev. 35, 787–796.
Schiff, W., Oldak, R., Shah, V., 1992. Aging persons’ estimates of vehic- Staplin, L., Lyles, R., 1991. Age differences in motion perception and spe-
ular motion. Psychol. Aging 7 (4), 518–525. cific traffic manoeuvre problems. Transport. Res. Rec. 1325, 558–561.
Scialfa, C., Guzy, L., Leibowitz, H., Garvey, P., Tyrell, R., 1991. Age Yanik, A., Monforton, R., 1991. At fault accident patterns of older drivers.
differences in estimating vehicle velocity. Psychol. Aging 6, 60–66. J. Traffic Med. 10 (2), 87–92.