Horswill 2008

7
The Hazard Perception Ability of Older Drivers Mark S. Horswill, 1 Shelby A. Marrington, 1 Cynthia M. McCullough, 1 Joanne Wood, 2 Nancy A. Pachana, 1 Jenna McWilliam, 1 and Maria K. Raikos 1 1 School of Psychology, University of Queensland, St. Lucia, Brisbane, Australia. 2 School of Optometry, Queensland University of Technology, Kelvin Grove, Brisbane, Australia. We investigated the hazard perception ability of older drivers. A sample of 118 older drivers (65 years and older) completed a video-based hazard perception test and an assessment battery designed to measure aspects of cognitive ability, vision, and simple reaction time that might plausibly be linked to hazard perception ability. We found that hazard perception response times increased significantly with age but that this age-related increase could be accounted for by measures of contrast sensitivity and useful field of view. We found that contrast sensitivity, useful field of view, and simple reaction time could account for the variance in hazard perception, independent of one another and of individual differences in age. Key Words: Contrast sensitivity—Hazard perception—Useful field of view. H AZARD perception ability has been suggested to be an important factor in determining crash risk in older adults (Watzke & Smith, 1994). However, there has been relatively little research examining this skill in older individuals. In the context of driving, hazard perception ability, defined as the ability of individuals to anticipate potentially dangerous situations on the road ahead, has been identified as one of the few measures of driving-specific skills that correlate with crash risk (Horswill & McKenna, 2004). It has typically been assessed by measuring response times to hazardous traffic situations presented on film or video (Deery, 1999; McKenna & Crick, 1991; Pelz & Krupat, 1974; Sagberg & Bjornskau, 2006; Watts & Quimby, 1979). Hazard perception tests have been incorporated into licensing procedures in the United Kingdom and in some states in Australia, and the validity of such tests has been demonstrated by means of correlations with crash risk (Hull & Christie, 1992; McKenna & Horswill, 1999; Pelz & Krupat; Quimby, Maycock, Carter, Dixon, & Wall, 1986; Transport and Road Research Laboratory, 1979), differences between novice and experienced drivers (McKenna & Crick; Quimby & Watts, 1981; Sexton, 2000), and correlations between drivers’ hazard perception video test scores and expert ratings of drivers’ performance on real roads (Mills, Hall, McDonald, & Rolls, 1998). Theoretically, how would hazard perception ability be expected to vary among older drivers? The multifactorial model for enabling driving safety proposed by Anstey, Wood, Lord, and Walker (2005) provides a framework for making such predic- tions. The model states that driving behavior is predicted by both a driver’s capacity to drive safely and his or her self-monitoring and beliefs about driving capacity. In turn, a driver’s capacity to drive safely is determined by cognition, vision, and physical function. If hazard perception ability maps onto a driver’s capacity to drive safely, then we propose that it is likely to be determined by elements of cognition and vision. This leads to the prediction that, within the older driver population, hazard perception ability is likely to decline with age because of age- related decreases in cognitive and visual function. What elements of cognition or vision are likely to account for individual differences in hazard perception in older drivers, and why? Horswill and McKenna (1999) found that novice drivers’ reactions to road hazards were slowed significantly by the presence of a secondary verbal task that did not involve either a visual or motor response component, suggesting that hazard perception requires higher level cognitive resources. Taking this one step further, McKenna and Farrand (1999) found that experienced drivers’ hazard perception reaction times were slowed more than novices’ reaction times when the drivers were required to perform a secondary verbal task. This could indicate that experienced drivers gained their hazard perception advantage over novices by applying cognitive resources to a sophisticated and proactive visual search for hazards—an advantage that was diminished once those cognitive resources were diverted elsewhere. If this is an appropriate model of hazard perception ability, then we would expect age-related deficits in cognitive functioning to have an impact on this ability. The predominant theories to explain these declines include general cognitive slowing (Salthouse, 1996), inhibitory deficits (Hasher & Zacks, 1988), and task-switching deficits (Mayr & Liebscher, 2001), all of which may have consequences for hazard perception. For example, slowing of performance will have a direct impact on hazard perception as it is a dynamic task; inhibitory deficits could be important because hazard perception involves iden- tifying relevant cues from irrelevant cues; and task-switching deficits may create problems as hazard perception often in- volves attending to multiple sources of information. Visual performance in a population becomes more variable with advancing age, both through the normal aging process as well as through an increase in the prevalence of ocular disease later in life (Haegerstrom-Portnoy, Schneck, & Brabyn, 1999). Hazard perception involves the processing of visual stimuli, and so factors such as visual acuity and contrast sensitivity may mediate hazard perception ability in older drivers. Drivers are less likely to be able to anticipate hazards effectively if they have difficulty seeing the cues associated with them. Only two studies that we are aware of (Olson & Sivak, 1986; Underwood, Phelps, Wright, van Loon, & Galpin, 2005) have specifically investigated hazard perception skills in older Journal of Gerontology: PSYCHOLOGICAL SCIENCES Copyright 2008 by The Gerontological Society of America 2008, Vol. 63B, No. 4, P212–P218 P212 at University of Queensland on January 25, 2013 http://psychsocgerontology.oxfordjournals.org/ Downloaded from

description

cercetare

Transcript of Horswill 2008

Page 1: Horswill 2008

The Hazard Perception Ability of Older Drivers

Mark S. Horswill,1 Shelby A. Marrington,1 Cynthia M. McCullough,1 Joanne Wood,2

Nancy A. Pachana,1 Jenna McWilliam,1 and Maria K. Raikos1

1School of Psychology, University of Queensland, St. Lucia, Brisbane, Australia.2School of Optometry, Queensland University of Technology, Kelvin Grove, Brisbane, Australia.

We investigated the hazard perception ability of older drivers. A sample of 118 older drivers (65 years and older)completed a video-based hazard perception test and an assessment battery designed to measure aspects ofcognitive ability, vision, and simple reaction time that might plausibly be linked to hazard perception ability. Wefound that hazard perception response times increased significantly with age but that this age-related increasecould be accounted for by measures of contrast sensitivity and useful field of view. We found that contrastsensitivity, useful field of view, and simple reaction time could account for the variance in hazard perception,independent of one another and of individual differences in age.

Key Words: Contrast sensitivity—Hazard perception—Useful field of view.

H AZARD perception ability has been suggested to be animportant factor in determining crash risk in older adults

(Watzke & Smith, 1994). However, there has been relativelylittle research examining this skill in older individuals. In thecontext of driving, hazard perception ability, defined as the abilityof individuals to anticipate potentially dangerous situations onthe road ahead, has been identified as one of the few measures ofdriving-specific skills that correlate with crash risk (Horswill &McKenna, 2004). It has typically been assessed by measuringresponse times to hazardous traffic situations presented on film orvideo (Deery, 1999; McKenna & Crick, 1991; Pelz & Krupat,1974; Sagberg & Bjornskau, 2006; Watts & Quimby, 1979).Hazard perception tests have been incorporated into licensingprocedures in the United Kingdom and in some states inAustralia, and the validity of such tests has been demonstrated bymeans of correlations with crash risk (Hull & Christie, 1992;McKenna & Horswill, 1999; Pelz & Krupat; Quimby, Maycock,Carter, Dixon, & Wall, 1986; Transport and Road ResearchLaboratory, 1979), differences between novice and experienceddrivers (McKenna & Crick; Quimby & Watts, 1981; Sexton,2000), and correlations between drivers’ hazard perception videotest scores and expert ratings of drivers’ performance on realroads (Mills, Hall, McDonald, & Rolls, 1998).

Theoretically, how would hazard perception ability beexpected to vary among older drivers? The multifactorial modelfor enabling driving safety proposed by Anstey, Wood, Lord, andWalker (2005) provides a framework for making such predic-tions. The model states that driving behavior is predicted by botha driver’s capacity to drive safely and his or her self-monitoringand beliefs about driving capacity. In turn, a driver’s capacity todrive safely is determined by cognition, vision, and physicalfunction. If hazard perception ability maps onto a driver’scapacity to drive safely, then we propose that it is likely to bedetermined by elements of cognition and vision. This leads to theprediction that, within the older driver population, hazardperception ability is likely to decline with age because of age-related decreases in cognitive and visual function.

What elements of cognition or vision are likely to account forindividual differences in hazard perception in older drivers, and

why? Horswill and McKenna (1999) found that novice drivers’reactions to road hazards were slowed significantly by thepresence of a secondary verbal task that did not involve eithera visual or motor response component, suggesting that hazardperception requires higher level cognitive resources. Takingthis one step further, McKenna and Farrand (1999) found thatexperienced drivers’ hazard perception reaction times wereslowed more than novices’ reaction times when the driverswere required to perform a secondary verbal task. This couldindicate that experienced drivers gained their hazard perceptionadvantage over novices by applying cognitive resources toa sophisticated and proactive visual search for hazards—anadvantage that was diminished once those cognitive resourceswere diverted elsewhere.

If this is an appropriate model of hazard perception ability,then we would expect age-related deficits in cognitivefunctioning to have an impact on this ability. The predominanttheories to explain these declines include general cognitiveslowing (Salthouse, 1996), inhibitory deficits (Hasher & Zacks,1988), and task-switching deficits (Mayr & Liebscher, 2001),all of which may have consequences for hazard perception. Forexample, slowing of performance will have a direct impact onhazard perception as it is a dynamic task; inhibitory deficitscould be important because hazard perception involves iden-tifying relevant cues from irrelevant cues; and task-switchingdeficits may create problems as hazard perception often in-volves attending to multiple sources of information.

Visual performance in a population becomes more variablewith advancing age, both through the normal aging process aswell as through an increase in the prevalence of ocular diseaselater in life (Haegerstrom-Portnoy, Schneck, & Brabyn, 1999).Hazard perception involves the processing of visual stimuli,and so factors such as visual acuity and contrast sensitivity maymediate hazard perception ability in older drivers. Drivers areless likely to be able to anticipate hazards effectively if theyhave difficulty seeing the cues associated with them.

Only two studies that we are aware of (Olson & Sivak, 1986;Underwood, Phelps, Wright, van Loon, & Galpin, 2005) havespecifically investigated hazard perception skills in older

Journal of Gerontology: PSYCHOLOGICAL SCIENCES Copyright 2008 by The Gerontological Society of America2008, Vol. 63B, No. 4, P212–P218

P212

at University of Q

ueensland on January 25, 2013http://psychsocgerontology.oxfordjournals.org/

Dow

nloaded from

Page 2: Horswill 2008

drivers. Both studies found no significant difference in reactiontimes to hazards between older and younger drivers, but bothstudies had low sample sizes (15 and 12 older drivers,respectively). When drivers older than the age of 65 yearshave been tested in larger numbers, it has been as part ofa cross-age sample and the data have indicated that hazardperception ability peaks at age 55 and then declines (Quimby &Watts, 1981). However, the decrease in hazard perceptionability shown on aggregate for older drivers in this study couldbe due to specific pathologies in certain individuals rather thanrepresenting a normal age-related decline in all older adults.As most older drivers are highly experienced, and drivingexperience may compensate for age-related deficits, it ispossible that healthy older drivers may be able to maintaintheir hazard perception performance.

The Current ProjectIn this article we describe research in which we developed

and validated a test of hazard perception response time, whichwas completed by a sample of older drivers who also completeda battery of cognitive and vision tests. Our aims were (a) toconfirm that hazard perception declines with age in a sample ofolder drivers, (b) to determine whether cognitive ability orvision can account for age-related variance in hazard perceptionin such a sample, and (c) to determine whether cognitive ability,vision, or simple reaction time can account for individualdifferences in older drivers’ hazard perception. We favoredmeasuring hazard perception response time by using a video-based test rather than an on-road assessment because hazardsare relatively infrequent events; participants would have todrive for many hours to encounter the number of hazards thatcan be presented in a few minutes on a video, and video-basedsimulation allows for a much higher degree of experimentalcontrol than is possible in the real world.

STUDY 1One key feature of driving is that the context is highly

familiar, especially to experienced drivers. It is thereforepossible that although older drivers suffer problems with thenovel, unfamiliar stimuli generally presented in both vision andpsychological tests, they may have fewer problems withfamiliar driving scenarios.

We considered it important to develop a hazard perceptiontest that was based on situations and locations that would befamiliar to our sample of older drivers. Therefore, our first stepwas to develop a new video-based hazard perception test usingeveryday road scenes filmed in the vicinity from which oursample was taken and featuring genuine traffic conflicts.However, it was also important to demonstrate that this newhazard perception test yielded the same properties as previoustests. In order to do this, we used McKenna and Crick’s (1991)strategy of determining whether the test could discriminatebetween a high crash risk and a lower crash risk group, namelynovice and experienced drivers (McKnight & McKnight,2003). This is not a trivial step: Some hazard perception testshave failed to discriminate between these groups, arguablyraising questions as to their validity (Crundall, Chapman,Phelps, & Underwood, 2003; Sagberg & Bjornskau, 2006).

We defined traffic conflicts as situations in which a collisionor near collision with another road user (including stationeryvehicles, cyclists, or pedestrians) would occur unless the drivertook some type of evasive action (slowing, steering, etc).

METHODS

ParticipantsWe recruited 17 experienced drivers (over 10 years of

driving experience; 10 women and 7 men) and 16 novicedrivers (up to 3 years driving experience; 13 women and3 men); all drivers were from the university (and receivedcourse credit) or were friends of ours (no payment given). Theexperienced drivers (age, M ¼ 41.94 years, SD ¼ 8.51) weresignificantly older than the novice drivers (age, M ¼ 19.00years, SD¼ 1.86); t(30)¼ 10.54, p , .001. Note that becauseage and experience are naturally confounded in this sample (asis the case in most studies in the field), it would be inap-propriate to draw conclusions regarding the effect of experienceindependent of age from these data.

MaterialsWe filmed the scenes for the hazard perception test from the

driver’s perspective during daylight hours and under clearweather conditions (we used a Sony DSR-PDX10P videocamera, Sony Corporation, New York, NY). We edited 33potential traffic conflicts into a 23-minute sequence (1.5incidents per minute). Because our test was designed asa latency measure, we selected all the traffic conflict situationsto become eventually unambiguous in order to minimizemissing data. In other words, everyone would be expected torespond by the end of the scenario. We used speciallydeveloped software to measure participants’ reaction times tothese events.

ProcedureParticipants were seated 2.7 m from a screen onto which

a 1.12 m 3 0.93 m image of the hazard perception test footagewas displayed by means of a Sony VPL-CS5 data projector.Participants wore headphones through which they could hearthe sound recorded from inside the car. First, participants wereshown a preliminary video sequence of traffic conflicts thatwere not used in the final test. We did this to familiarizeparticipants with the footage as well to allow the use of thesedata as a control group for a hazard perception training packagenot analyzed in the present article. Participants then completedthe hazard perception test, in which they were instructed topress the response button as quickly as possible whenever theyanticipated a potential traffic conflict. Traffic conflicts weredefined as ‘‘situations in which a collision or near collision withanother road user (including stationery vehicles, cyclists, andpedestrians) would occur unless you take some type of evasiveaction (slowing, steering, etc.).’’ Participants also completeda questionnaire that included demographic information anddriving history.

RESULTS

We replace missing values for reaction times to the trafficconflicts (where participants failed to respond to traffic

OLDER DRIVERS’ HAZARD PERCEPTION ABILITY P213

at University of Q

ueensland on January 25, 2013http://psychsocgerontology.oxfordjournals.org/

Dow

nloaded from

Page 3: Horswill 2008

conflicts) by the group means for that conflict. There was nosignificant difference between the number of hazards missed bynovices (3.7 out of 33) and experienced drivers (3.3 out of 33);t(31)¼0.41, ns. A Kolmogorov–Smirnov test indicated that thedistribution of mean reaction times did not differ significantlyfrom a normal distribution; Kolmogorov–Smirnov z¼ 0.50, ns.Cronbach’s alpha for the test was a ¼ 0.78. The mean hazardperception reaction time of the experienced drivers (2,745 ms,SD ¼ 381) was significantly faster than that of the novices(3,168 ms, SD ¼ 359), with t(31) ¼ 3.28, p ¼ .003, and itconstituted a large effect size (Cohen’s d ¼ 1.14; see Cohen,1992); this provided evidence for the validity of our test. Theeffect size was similar to that found for previous tests(McKenna & Crick, 1991).

STUDY 2In Study 2, we examined three key hypotheses: (a) that

hazard perception declines with age; (b) that cognitive ability orvision can account for age-related variance in hazard perceptionin older drivers; and (c) that cognitive ability and vision andperformance on a simple reaction time test can account forindividual differences in older drivers’ hazard perception. Toexamine these hypotheses, we asked a sample of drivers aged65 years or older to complete our hazard perception test,together with a battery of measures of cognition and vision, aswell as simple reaction times.

The measures of cognitive ability that we included weredesigned to cover aspects of cognition that have been found tocorrelate with driving performance or crash risk in previousstudies or were considered important for theoretical reasons.Measures of cognitive functioning, such as the Mini-MentalState Examination (Folstein, Folstein, & McHugh, 1975), havebeen found to predict both on-road performance (Fitten et al.,1995; Odenheimer et al., 1994) and crash risk (Owsley,Ball, Sloane, Roenker, & Bruni, 1991; Stutts, Stewart, &Martell, 1998). The Trail-Making Tests (Spreen & Strauss,1991), considered to measure executive function and visualattention, have been found to correlate with crash risk (Goodeet al., 1998; Stutts et al.) as well as on-road driving performance(Odenheimer et al.). Both of these measures could influencehazard perception if it is a task that involves running a proactivemental model of the traffic environment and therefore requireshigh-level cognitive resources (Horswill & McKenna, 2004).Along the same lines, we might also predict that individualdifferences in measures of fluid intelligence may influencehazard perception.

Groeger (2000) offered a memory-based theoretical accountof hazard perception, in which drivers compare a developingtraffic situation with previously encountered situations in orderto evaluate the level of danger. Alternatively, hazards could benoticed because they are distinctive events that do not map ontoan experienced driver’s vast knowledge of nonhazardousdriving. Consistent with these predictions, a test of workingmemory was found to correlate with performance on a video-based measure of drivers’ decisions as to when to pull outacross a stream of oncoming traffic (Guerrier, Manivannan, &Nair, 1999), which could be argued to be a hazard perceptionjudgment. In addition, Odenheimer and colleagues (1994)found substantial correlations between on-road driving assess-

ments and measures of both visual and verbal memory indrivers who were older than 60 years of age. Goode andassociates (1998) reported that a measure of visual memorycould distinguish between crashers and noncrashers. Finally,given the importance of experience to hazard perception(consistent with the memory-based model), it is also possiblethat crystallized intelligence could play a role in effectivehazard detection.

The Useful Field of View Test, in which participants arerequired to detect peripherally presented targets while attendingto a central task, is considered to measure (among other things)cognitive factors such as visual attention (both divided andselective) and speed of processing. It has been found tocorrelate with crash risk (De Raedt & Ponjaert-Kristoffersen,2000; Goode et al., 1998; Owsley et al., 1998; Owsley et al.,1991; Sims, McGwin, Allman, Ball, & Owsley, 2000), on-roadperformance (De Raedt & Ponjaert-Kristoffersen), and closed-road performance (Wood, 2002; Wood & Troutbeck, 1995).One could argue that the most plausible causal mechanism isthat reductions in the useful field of view result in a situation inwhich individuals are less likely to detect traffic hazards,leading in turn to unsafe driving behaviors and an inflatedcrash risk.

We also included measures of vision such as contrastsensitivity, which has been found to correlate with crash risk(Owsley et al., 1991) and both closed-road (Wood, 2002;Wood & Carberry, 2006) and on-road (McKnight & McKnight,1999) driving performance. In addition, we included a measureof simple reaction time because this has been found to correlatewith older driver performance in some on-road studies(McKnight & McKnight; Odenheimer et al., 1994).

METHODS

ParticipantsWe tested 131 community-dwelling drivers who were 65

years of age or older (and who had a minimum of 10 years ofdriving experience). We recruited the participants either froman older adult research participant pool or from advertisements.We excluded 13 participants from the analysis because ofincomplete or absent hazard perception test data. The missingdata were the result of either technical problems or aninsufficient number of viewed hazards: Some participantsdeveloped motion sickness, so their tests were terminated early.These participants viewed less than 75% of the hazards. Notethat motion sickness in simulators is known to be a particularproblem for older individuals (Edwards, Creaser, Caird,Lamsdale, & Chisholm, 2003).

We also stopped the hazard perception test early for threeindividuals who became motion sick but completed over 75%of the test. For these participants, we included the data but wereplaced missing values with overall sample means (rather thangroup means). The analyzed sample (n ¼ 118) included 51women and 67 men (age¼ 65–84 years; see Table 1 for meanand standard deviation). In response to a question on overallhealth, all participants rated themselves as at least ‘‘fair,’’ with88% indicating that their health was either ‘‘good,’’ ‘‘verygood,’’ or ‘‘excellent.’’ All participants scored at least 75 out of

HORSWILL ET AL.P214

at University of Q

ueensland on January 25, 2013http://psychsocgerontology.oxfordjournals.org/

Dow

nloaded from

Page 4: Horswill 2008

100 on the Modified Mini-Mental State Examination (3MS;Teng & Chui, 1987).

Materials

Hazard perception test. —We selected the scenes from thehazard perception test that best discriminated between experi-enced and novice drivers. We included 17 of the original 33scenes and filmed 14 new traffic conflicts in an attempt toimprove the test stimuli further (31 incidents in total). We usedsome of the scenes that failed to discriminate betweenexperienced and novice drivers in Study 1 as a 1.5-minutepractice. The new test was 22 minutes long (1.4 incidents perminute). The instructions and response mode remained thesame, although we printed the instructions in a larger font.

Cognitive measures. —We had participants complete the3MS (Teng & Chui, 1987) to measure cognitive status, Trail-Making Tests A and B (Spreen & Strauss, 1991) to tap intoexecutive function and visual attention, and the MatrixReasoning and Vocabulary subtests from the WechslerAbbreviated Scale of Intelligence (The Psychological Corpo-ration, 1999) to measure fluid and crystallized intelligence,respectively. We included the Digit Span (attention), Letter–Number Sequencing (working memory), and Digit–Symbol(processing speed and working memory) subtests from theWechsler Adult Scale of Intelligence (Wechsler, 1997). Wealso used the Useful Field of View Test developed by Woodand Troutbeck (1995), which we administered by using a SonyVPL-CS5 data projector. Participants were required to detectthe presence of a central target and indicate the position ofa peripheral target. The task was completed with and without anarray of visual distracters.

Vision measures. —We tested binocular static visual acuityby using the National Vision Research Institute (Melbourne,Australia) logMAR chart (logarithm of the minimum angle ofresolution); we tested binocular letter contrast sensitivity by

using the Pelli–Robson Letter Sensitivity chart (Pelli,Robson, & Wilkins, 1988).

Simple reaction time measure. —We devised a test of simplereaction time, in which participants were required to pressa response button as quickly as possible whenever a blue circleappeared in the center of the video screen. There were 16 trialsover 1.5 minutes and the circle appeared at random intervals(though always in the same location). We recorded participants’reaction times in milliseconds; we excluded responses morethan 2 SD from an individual’s mean over the 16 trials asoutliers. The overall reaction time measure was the mean of allthe responses not excluded as outliers.

ProcedureParticipants were initially mailed a questionnaire that

gathered driving history, health, and demographic information.During the testing session (which took place either at theuniversity or at community centers), participants first completedthe cognitive tests in the following order: 3MS, Trail-MakingTests A and B, the Vocabulary subtest, the Matrix Reasoningsubtest, and the Digit Span, Letter–Number Sequencing, andDigit–Symbol subtests. Then we administered the visual acuity,contrast sensitivity, and useful field of view measures (first withno distracters and then with distracters). Finally, participantscompleted the simple reaction time test and the hazardperception test. The entire session took approximately 2 hoursto complete and participants were not paid. We offered theparticipants breaks between the tests as a way to minimizefatigue effects.

RESULTS

As in Study 1, missing values in the hazard perception testwhere participants failed to press the response button during thetime window for a hazard were replaced by sample means (3.8out of 31 hazards were missed on average). There were also 38missing values (0.01%) in the data set, which were either due toequipment problems or to the 3 participants who terminated the

Table 1. Mean Raw Scores and Intercorrelations for the Variables in Study 2 (N ¼ 118)

Variable M (SD) 1 2 3 4 5 6 7 8 9 10 11 12 13 14

1 HPRT (ms) 3634.23 (656.66) —

2 Age (years) 73.23 (5.66) 0.27 —

3 3MS (out of 100) 95.34 (4.06) �0.22 �0.47 —

4 TMT Part A time (s) 40.31 (14.45) 0.05 0.32 �0.28 —

5 TMT Part B time (s) 93.02 (34.4) 0.17 0.38 �0.33 0.53 —

6 Vocabulary (out of 80) 67.35 (7.51) �0.06 �0.15 0.45 �0.14 �0.29 —

7 Matrix Reasoning (out of 28) 20.7 (5.99) �0.04 �0.33 0.30 �0.29 �0.40 0.47 —

8 Digit Span (out of 30) 18.61 (4.16) �0.17 �0.19 0.20 �0.16 �0.32 0.31 0.29 —

9 LNS (out of 21) 10.08 (2.67) �0.11 �0.35 0.47 �0.29 �0.51 0.36 0.43 0.56 —

10 Digit–Symbol (out of 133) 56.86 (13.54) �0.19 �0.45 0.48 �0.57 �0.67 0.34 0.42 0.34 0.53 —

11 logMAR �0.01 (0.12) 0.25 0.10 0.03 0.05 �0.04 0.09 0.07 �0.12 0.00 0.05 —

12 Pelli–Robson 1.71 (0.1) �0.42 �0.34 0.22 �0.06 �0.17 0.09 0.14 0.17 0.21 0.25 �0.29 —

13 UFOV, no distracters (out of 24) 19.16 (3.96) �0.30 �0.38 0.37 �0.24 �0.29 0.24 0.41 0.15 0.32 0.36 0.00 0.24 —

14 UFOV, with distracters (out of 24) 14.3 (4.37) �0.17 �0.53 0.50 �0.36 �0.41 0.29 0.33 0.18 0.40 0.46 0.10 0.14 0.60 —

15 Simple reaction time (ms) 300.19 (66.51) 0.30 0.17 �0.15 0.14 0.05 0.04 �0.04 �0.03 �0.07 �0.17 0.08 �0.10 �0.16 �0.07

Note: HPRT¼ hazard perception reaction time; 3MS¼Modified Mini-Mental State Examination; TMT¼ Trail-Making Test; LNS¼ Letter–Number Sequencing

subtest; logMAR ¼ logMAR chart, used for static visual acuity (logarithm of the minimum angle of resolution); UFOV ¼ Useful Field of View Test. The Pelli–

Robson chart is used for contrast sensitivity. For the TMT, HPRT, and logMAR, a lower score is better; for the 3MS, the UFOV Test, the Pelli–Robson chart, and

the Vocabulary, Matrix Reasoning, Digit Span, LNS, and Digit–Symbol subtests, a higher score is better. For the reaction time, a lower time is faster.

OLDER DRIVERS’ HAZARD PERCEPTION ABILITY P215

at University of Q

ueensland on January 25, 2013http://psychsocgerontology.oxfordjournals.org/

Dow

nloaded from

Page 5: Horswill 2008

test early because of motion sickness but completed over 75%of the test. These missing values were also replaced by samplemeans (note that we carried out all analyses by excluding allparticipants with any such missing values, and it made littledifference to the pattern of correlations). The raw scores for allvariables are found in Table 1. Cronbach’s alpha for the hazardperception test was a¼ 0.85.

We transformed the variables to maximize normality wherenecessary (which, given the sample size, we checked byinspecting histograms) by using either an inverse, square root,or logarithmic transformation. To aid interpretation, wereflected any scales that were reversed by the transformationsback to their original orientation.

Does Hazard Perception Decrease With Age inOlder Drivers?

We found a correlation between age and hazard perceptionresponse time, indicating that older drivers were slower atresponding to traffic conflicts (Table 1). This correlation wassignificant (p , .05) after we controlled for 30 multiplecomparisons (corresponding to the first two columns ofcorrelations in Table 1, i.e., all correlations involving hazardperception, age, and the other variables) with a Bonferronicorrection (a ¼ 0.05, one-tailed tests).

To determine whether the variance between hazard percep-tion and age could be accounted for by our other measures, weperformed a hierarchical regression, in which any variable thatcorrelated significantly with both age and hazard perception(after we applied a Bonferroni correction for 30 multiplecomparisons) was entered in Block 1, and age was added inBlock 2. Two variables, that is, useful field of view with nodistracters and contrast sensitivity, fulfilled these criteria.Assumptions required for multiple regression were satisfied.We found that age could not account for unique variance inhazard perception ability once individual differences in usefulfield of view (no distracters) and contrast sensitivity had beenaccounted for (Table 2). That is, the declines in hazardperception with age could be accounted for by individualdifferences in measures reflecting both cognitive ability andvision.

What Accounts for Individual Differences in HazardPerception Ability in Older Drivers?

Applying a Bonferroni correction for 15 comparisons (thenumber of correlations between all variables and hazardperception; see Table 1), we found that the variables that

correlated with hazard perception were contrast sensitivity,useful field of view (no distracters), simple reaction time, andvisual acuity (a ¼ 0.05, one-tailed). It is worth noting that thecorrelations between hazard perception and the 3MS, Trail-Making Test B, Digit Span and Digit–Symbol subtests, andUseful Field of View Test (full distracters) would also havebeen significant without the Bonferroni correction for multiplecomparisons (many studies in the literature do not appear toapply this correction). This indicates that it may be worthfollowing up the relationship between hazard perception andthese variables in a more powerful study (either by using agreater sample size or by limiting the number of comparisons).Note that although the useful field of view measure withdistracters appeared to correlate less with hazard perceptionthan the measure without distracters, the difference betweenthe correlation coefficients was not statistically significant,t(117)¼�1.63, ns.

To determine whether the relationship between hazardperception and its significant correlates, that is, contrastsensitivity, useful field of view (no distracters), simple reactiontime, and visual acuity, could be accounted for by individualdifferences in either age or each other, we performed a directentry regression by using hazard perception score as thedependent variable and age, visual acuity, contrast sensitivity,useful field of view (no distracters), and simple reaction time asindependent variables. Assumptions required for multipleregression were satisfied. We found that contrast sensitivity,useful field of view (no distracters), and simple reaction timecould account for variance in hazard perception once individualdifferences in age and all other variables were adjusted for (seeTable 3). This is consistent with our proposal that individualdifferences in hazard perception in older drivers are due tocognitive and vision processes; it also indicates the importanceof simple reaction time in this age group. Note that the over-lapping 95% confidence intervals for the magnitude of the stan-dardized regression weights of these three variables (Table 3)suggest that there were no significant differences between thesizes of their independent relationship with hazard perception.That is, we cannot conclude that any one of these threepredictors had a greater effect on hazard perception than any ofthe others in the present study.

GENERAL DISCUSSION

We found that hazard perception ability declines withincreasing age, even in a sample of older adults who consideredthemselves relatively healthy, and that a significant proportion

Table 2. Hierarchical Regression, With Hazard Perception as the Dependent Variable

Model 1 Model 2

Standardized Coefficients Standardized Coefficients

95% CIs for b 95% CIs for b

b p Lower bound Upper bound Partial Correlations b p Lower bound Upper bound Partial Correlations

Age .082 .376 �.099 .264 .083

Contrast sensitivity �.369 ,.001 �.536 �.203 �.376 �.347 ,.001 �.521 �.174 �.345

UFOV (no distracters) �.212 .014 �.378 �.046 �.227 �.187 .040 �.362 �.011 �.191

Note: CI ¼ confidence interval; UFOV ¼ Useful Field of View Test. For hazard perception, lower values are better. For contrast sensitivity and the UFOV

Test, a higher score is better. The significance of R2 change from the null model to Model 1 is ,.001; the significance of R2 change from Model 1 to Model 2 is

.376. For the overall model, the Model 1 R2 value is .218; the Model 2 value is .224.

HORSWILL ET AL.P216

at University of Q

ueensland on January 25, 2013http://psychsocgerontology.oxfordjournals.org/

Dow

nloaded from

Page 6: Horswill 2008

of individual differences in hazard perception in this sample canbe attributed to cognitive and vision factors.

One of the variables that had a significant impact on hazardperception performance, independent of other measures, wascontrast sensitivity. This finding has implications for a numberof issues. These include the timing of cataract removaloperations, because a major effect of cataracts is loss ofcontrast sensitivity and many individuals with cataracts drivefor months before having surgery whereas others may elect notto have surgery at all. The finding is also relevant to driverscreening, as there have been calls for a minimum level ofcontrast sensitivity to be required for driving. Finally, therecould be road-design implications, because it is possible thatcreating high-contrast driving environments may improvehazard detection for those with poor contrast sensitivity. Forexample, Horberry, Anderson, and Regan (2006) reportedevidence that higher visibility road markings (which wouldincrease image contrast) improved the driving performance ofolder adults at night. It could be the case that such marking mayhave a differentially greater benefit for drivers with poorcontrast sensitivity (see McGregor & Chaparro, 2005, for otherdesign improvements that could aid older drivers).

We also found evidence that the useful field of view isimportant for hazard perception. This provides a mechanism toexplain the strong relationship found between crash risk andperformance on the Useful Field of View Test (De Raedt &Ponjaert-Kristoffersen, 2000; Goode et al., 1998; Owsley et al.,1991, 1998; Sims et al., 2000), in which a reduction in theuseful field of view makes traffic conflicts harder to detect,which in turn impacts crash risk.

As with any simulator study, it is possible to argue that therelationships found between our hazard perception test scoreand the other measures are an artifact of our testing proceduresand may not generalize to hazard perception on the real road,despite the strong evidence for the validity of video-basedhazard perception tests. However, we have already outlined theproblems involved with the real-world measurement of hazardperception, which we believe makes our approach moreappropriate.

If poor hazard perception is a problem for some older adults,then what can be done about it? One possibility is to exploit thehazard perception test as a screening tool to assess fitness todrive in older adults. At the moment, this assessment is oftenperformed by the use of intuitive judgments by medicalpractitioners. In contrast, the hazard perception test, by usingtask-specific and realistic stimuli, provides an objective

assessment of drivers’ ability to perceive and respond togenuine traffic conflicts. If older drivers are found to performpoorly on hazard perception tests, then it may be possible todevelop training interventions specifically aimed at improvingtheir hazard perception skill. Hazard perception ability has beenfound to be modifiable by both on-road and video-basedtraining (McKenna & Crick, 1997; McKenna, Horswill, &Alexander, 2006; Mills et al., 1998) in novice and moreexperienced age groups, though to date no one has attemptedthis with older adults. Such training typically involves havingparticipants watch filmed traffic situations while an expertinstructor describes strategies for anticipating hazards; it ispossible that such training might benefit older drivers eventhough they have considerable driving experience. In addition,there is evidence that speed of processing training for olderadults can successfully improve performance on a useful fieldof view test (Ball, Edwards, & Ross, 2007), which may in turnimprove hazard perception ability and allow individuals tocontinue driving without compromising road safety.

ACKNOWLEDGMENTS

This research was funded by the Australian Research Council inassociation with Queensland Main Roads (Linkage Grant LP0348431).Additional funding was provided by a University of Queensland New Staffgrant to Mark Horswill. We thank Andrew Hill and Maria Theodoros forhelp in collecting and entering the data, Paul Jackson for developing thereaction time software, Mark Wetton for technical assistance, Mark Wettonand Tom Wallis for filming some of the hazards, and Frank McKenna forhelpful comments on the data.

CORRESPONDENCE

Address correspondence to Dr. Mark S. Horswill, School of Psychology,University of Queensland, St. Lucia, Brisbane, QLD 4072, Australia.E-mail: [email protected]

REFERENCES

Anstey, K. J., Wood, J., Lord, S., & Walker, J. G. (2005). Cognitive,sensory and physical factors enabling driving safety in older adults.Clinical Psychology Review, 25, 45–65.

Ball, K., Edwards, J. D., & Ross, L. A. (2007). The impact of speed ofprocessing training on cognitive and everyday functions. Journal ofGerontology: Psychological Sciences, 62(Special Issue 1), 19–31.

Cohen, J. (1992). A power primer. Psychological Bulletin, 112, 155–159.Crundall, D., Chapman, P., Phelps, N., & Underwood, G. (2003). Eye

movements and hazard perception in police pursuit and emergencyresponse driving. Journal of Experimental Psychology: Applied, 9,163–174.

De Raedt, R., & Ponjaert-Kristoffersen, I. (2000). The relationship betweencognitive/neuropsychological factors and car driving performance in

Table 3. Direct Entry Regression, With Hazard Perception as the Dependent Variable

Standardized Coefficients

95% CIs for b

b p Lower bound Upper bound Partial Correlations

Age .049 .582 �.126 .225 .052

logMAR static visual acuity .141 .093 �.022 .304 .158

Contrast sensitivity �.297 .001 �.471 �.124 �.302

UFOV (no distracters) �.177 .044 �.347 �.006 �.189

Simple reaction time (ms) .226 .006 .067 .385 .254

Note: CI ¼ confidence interval; UFOV ¼ Useful Field of View Test; logMAR ¼ logMAR chart, used for static visual acuity (logarithm of the minimum angle

of resolution). For hazard perception and logMAR, lower values are better. For contrast sensitivity and the UFOV Test, a higher score is better. For reaction time,

a lower time is faster. For the overall model, the R2 value is .295; the significance of R2 is ,.001.

OLDER DRIVERS’ HAZARD PERCEPTION ABILITY P217

at University of Q

ueensland on January 25, 2013http://psychsocgerontology.oxfordjournals.org/

Dow

nloaded from

Page 7: Horswill 2008

older adults. Journal of the American Geriatrics Society, 48, 1664–1668.

Deery, H. A. (1999). Hazard and risk perception among young novicedrivers. Journal of Safety Research, 30, 225–236.

Edwards, C. J., Creaser, J. I., Caird, J. K., Lamsdale, A. M., & Chisholm,S. L. (2003). Older and younger driver performance at complexintersections: Implications for using perception-response time anddriving simulation. Paper presented at the Second InternationalSymposium on Human Factors in Driver Assessment, Training, andVehicle Design, Park City, UT.

Fitten, L. J., Kent, M. P., Wilkinson, C. J., Little, R. J., Burns, M. M.,Pachana, N., et al. (1995). Alzheimer and vascular dementias and driving.The Journal of the American Medical Association, 273, 1360–1365.

Folstein, M. F., Folstein, S. E., & McHugh, P. R. (1975). Mini-MentalState: A practical method for grading the cognitive state of patients forthe clinician. Journal of Psychiatric Research, 12, 189–198.

Goode, K. T., Ball, K. K., Sloane, M., Roenker, D. L., Roth, D. L.,Myers, R. S., et al. (1998). Useful field of view and otherneurocognitive indicators of crash risk in older adults. Journal ofClinical Psychology in Medical Settings, 5(4), 425–440.

Guerrier, J. H., Manivannan, P., & Nair, S. N. (1999). The role of workingmemory, field dependence, visual search, and reaction time in the leftturn performance of older female drivers. Applied Ergonomics, 30,109–119.

Haegerstrom-Portnoy, G., Schneck, M. E., & Brabyn, J. A. (1999). Seeinginto old age: Vision function beyond acuity. Optometry & VisionScience, 76, 141–158.

Hasher, L., & Zacks, R. T. Working memory, comprehension, and aging:A review and a new view. Psychology of Learning and Motivation, 22,193–225.

Horberry, T., Anderson, J., & Regan, M. A. (2006). The possible safetybenefits of enhanced road markings: A driving simulator evaluation.Transportation Research Part F—Traffic Psychology and Behaviour,9(1), 77–87.

Horswill, M. S., & McKenna, F. P. (1999). The effect of interference ondynamic risk-taking judgments. British Journal of Psychology, 90,189–199.

Horswill, M. S., & McKenna, F. P. (2004). Drivers’ hazard perceptionability: Situation awareness on the road. In S. Banbury & S. Tremblay(Eds.), A cognitive approach to situation awareness: Theory andapplication (pp. 155–175). Aldershot, England: Ashgate.

Hull, M., & Christie, R. (1992). Hazard perception test: The Geelong trialand future development. Paper presented at the National Road SafetySeminar, Wellington, New Zealand.

Mayr, U., & Liebscher, T. (2001). Is there an age deficit in the selection ofmental sets? European Journal of Cognitive Psychology, 13(1–2), 47–69.

McGregor, L. N., & Chaparro, A. (2005). Visual difficulties reported bylow-vision and nonimpaired older adult drivers. Human Factors, 47(3),469–478.

McKenna, F. P., & Crick, J. L. (1991). Hazard perception in drivers: Amethodology for testing and training (Final Report). Crowthorne,England: Transport and Road Research Laboratory.

McKenna, F. P., & Crick, J. L. (1997). Developments in hazard perception(Contractor Report 313). London: Department of Transport.

McKenna, F. P., & Farrand, P. (1999). The role of automaticity in driving.In G. B. Grayson (Ed.), Behavioural research in road safety IX.Crowthorne, England: Transport and Road Research Laboratory.

McKenna, F. P., & Horswill, M. S. (1999). Hazard perception and itsrelevance for driver licensing. Journal of the International Associationof Traffic and Safety Sciences, 23(1), 26–41.

McKenna, F. P., Horswill, M. S., & Alexander, J. L. (2006). Doesanticipation training affect drivers’ risk taking? Journal of ExperimentalPsychology: Applied, 12, 1–10.

McKnight, A. J., & McKnight, A. S. (1999). Multivariate analysis of age-related driver ability and performance deficits. Accident analysis andPrevention, 31, 445–454.

McKnight, A. J., & McKnight, A. S. (2003). Young novice drivers:Careless or clueless? Accident Analysis and Prevention, 35, 921–925.

Mills, K. L., Hall, R. D., McDonald, M., & Rolls, G. W. P. (1998). Theeffects of hazard perception training on the development of novicedriver skills. Retrieved November 8, 2006, from http://www.dft.gov.uk

Odenheimer, G. L., Beaudet, M., Jette, A. M., Albert, M. S., Grande, L., &Minaker, K. L. (1994). Performance-based driving evaluation of theelderly driver—Safety, reliability, and validity. Journal of Gerontology:Medical Sciences, 49A, M153–M159.

Olson, P. L., & Sivak, M. (1986). Perception-response time to unexpectedroadway hazards. Human Factors, 28, 91–96.

Owsley, C., Ball, K., McGwin, G., Sloane, M. E., Roenker, D. L.,White, M. F., et al. (1998). Visual processing impairment and risk ofmotor vehicle crash among older adults. Journal of the AmericanMedical Association, 279(14), 1083–1088.

Owsley, C., Ball, K., Sloane, M. E., Roenker, D. L., & Bruni, J. R. (1991).Visual/cognitive correlates of vehicle accidents in older drivers.Psychology and Aging, 6, 403–415.

Pelli, D. G., Robson, J. G., & Wilkins, A. J. (1988). The design of a newletter chart for measuring contrast sensitivity. Clinical Vision Sciences,2(3), 187–199.

Pelz, D. C., & Krupat, E. (1974). Caution profile and driving record orundergraduate males. Accident Analysis and Prevention, 6, 45–58.

Quimby, A. R., Maycock, G., Carter, I. D., Dixon, R., & Wall, J. G. (1986).Perceptual abilities of accident involved drivers (Research Report 27).Crowthorne, England: Transport and Road Research Laboratory.

Quimby, A. R., & Watts, G. R. (1981). Human factors and drivingperformance (Laboratory Report 1004). Crowthorne, England: Trans-port and Road Research Laboratory.

Sagberg, F., & Bjornskau, T. (2006). Hazard perception and drivingexperience among novice drivers. Accident Analysis and Prevention,38, 407–414.

Salthouse, T. A. (1996). The processing-speed theory of adult agedifferences in cognition. Psychological Review, 103(3), 403–428.

Sexton, B. (2000). Development of hazard perception testing. RetrievedNovember 8, 2006, from http://www.dft.gov.uk

Sims, R. V., McGwin, G., Allman, R. M., Ball, K., & Owsley, C. (2000).Exploratory study of incident vehicle crashes among older drivers.Journal of Gerontology: Medical Sciences, 55A, M22–M27.

Spreen, O., & Strauss, E. (1991). A compendium of neuropsychologicaltests: Administration, norms, and commentary. New York: OxfordUniversity Press.

Stutts, J. C., Stewart, J. R., & Martell, C. (1998). Cognitive testperformance and crash risk in an older driver population. AccidentAnalysis and Prevention, 30(3), 337–346.

Teng, E. L., & Chui, H. C. (1987). The Modified Mini-Mental State (3MS)Examination. Journal of Clinical Psychiatry, 48, 314–317.

The Psychological Corporation. (1999). Wechsler Abbreviated Scale ofIntelligence. San Antonio, TX: Harcourt Brace.

Transport and Road Research Laboratory. (1979). A hazard perception testfor drivers (Leaflet LF813). Crowthorne, England: Transport and RoadResearch Laboratory.

Underwood, G., Phelps, N., Wright, C., van Loon, E., & Galpin, A. (2005).Eye fixation scanpaths of younger and older drivers in a hazardperception task. Ophthalmic and Physiological Optics, 25(4), 346–356.

Watts, G. R., & Quimby, A. R. (1979). Design and validation of a drivingsimulator (Number LR 907). Crowthorne, England: Transport and RoadResearch Laboratory.

Watzke, J., & Smith, D. B. D. (1994). Concern for and knowledge of safetyhazards among older people—Implications for research and prevention.Experimental Aging Research, 20(3), 177–188.

Wechsler, D. (1997). Wechsler Adult Intelligence Scale, Third Edition:Administration and scoring manual. San Antonio, TX: The Psycho-logical Corporation.

Wood, J. M. (2002). Age and visual impairment decrease drivingperformance as measured on a closed-road circuit. Human Factors,44, 482–494.

Wood, J. M., & Carberry, T. P. (2006). Bilateral cataract surgery anddriving performance. British Journal of Ophthalmology, 90,1277–1280.

Wood, J. M., & Troutbeck, R. (1995). Elderly drivers and simulated visualimpairment. Optometry and Vision Science, 72, 115–124.

Received March 22, 2007Accepted December 19, 2007Decision Editor: Thomas M. Hess, PhD

HORSWILL ET AL.P218

at University of Q

ueensland on January 25, 2013http://psychsocgerontology.oxfordjournals.org/

Dow

nloaded from