You are on page 1of 13

Pergamon

Safety Science, Vol. 22, No. 1-3. pp. 63-75, 1996 Copyright 0 1996 Elsevier Science Ltd Printed in Great Britain. All rights resewed
0925-7525/96 $15.00 + 0.00

SO9257535(96)00006-9

DRIVER APPROACH BEHAVIOUR AT AN UNPROTECTED RAILWAY CROSSING BEFOREANDAFTERENHANCEMENT OF LATERAL SIGHT DISTANCES: AN EXPERIMENTAL INVESTIGATION OF A RISK PERCEPTION AND BEHAVIOURAL COMPENSATION HYPOTHESIS
Nicholas John Ward al*, Gerald J.S. Wilde b
a HUSAT Research Leicestershire, UK b Department Institute, Loughborough Queens University of Technology, Loughborough,

of Psychology,

University, Kingston,

Ontario, Canada

Abstract-Common sense dictates that railway crossing with restricted lateral visibility should be more hazardous than sites with unrestricted visibility (Schoppert and Hoyt, 1968). However, evidence of any relationship between lateral sight distance and accident history has not been consistently demonstrated. Based on the tenets of risk homeostasis theory, Wilde et al. (1987) explain this disassociation in terms of motorist compensatory behaviour (e.g. speed reduction) in response to the perceived risk associated with restricted visibility. This regulatory mechanism is expected to sustain a more-or-less constant safety margin. This proposition was explored in this study by examining the effect of enhancing lateral sight distances at an unprotected crossing. Parallel observations were made at an untreated site to control for secular confounding. As expected, improvement of lateral sight distances resulted in an upstream shift toward longer search durations and a tendency toward faster approach speeds, but failed to produce a calculated net safety benefit. A survey of local residents suggests that the enhancement treatment reduced perceived risk. Results are discussed with respect to the merits of sight distance enhancement and alternative interventions. Copyright 0 1996 Elsevier Science Ltd

1. Introduction A motorist approaching train. Uncertainty will be based on past experience, schedule, or (ii) there is a crossing may be uncertain about the probability of encountering a high without reliable knowledge of the scheduling of train arrivals either because (i) the trains at the site do not adhere to a regular no prior experience with the site. This uncertainty is likely to be

* Corresponding author. Department of Psychology, University of Leeds, Leeds LS2 9JT UK. Phone: 0113 233 5747, Fax: 0113 233 5749 (E-mail: NICW@Psychology.leeds.ac.uk)

63

64

N.J. Ward and GJS.

Wilde

greatest at a passive crossing where there are no automated signals to alert motorists to an approaching train. Ehrlick (1989) commented that the approach to a passive crossing is experienced by motorists as a context with a connotative meaning whereby an encounter with a train is perceived probabilistically as a foreseeable yet unlikely hazard. In this sense, motorists approach behaviour is based on a conditional protocol to give way to a probabilistic hazard which is rare but extremely adverse. It is reasonable to expect motorists to engage in search activities in response to perceived situational uncertainty about the existence of an approaching hazard. Such efforts have been observed in numerous field studies of crossing approach behaviour (e.g. Wilde et al., 1976, 1987). Imagine the case of an approach to a crossing that has limited visibility due to dense vegetation obscuring the view of the track and any approaching train. Because detection of approaching trains is impeded in this case, such conditions embody a greater degree of uncertainty than when there is a relatively clear view of the track. To the extent that this uncertainty about the presence of hazards impedes motorist capacity to decide upon appropriate actions, this condition is probabilistically more hazardous than conditions permitting a clear and unobstructed view of the track. Intuitively, restriction of lateral sightline distance should be associated with a greater incidence of accidents because of the greater hazard associated with the obstruction of lateral visibility. However, evidence that restricted lateral visibility at railway crossings is hazardous has not been forthcoming in spite of its apparent plausibility (Schoppert and Hoyt, 1968). Specifically, sight distance has not been found to relate reliably to crossing accident rates (van Belle et al., 1975; Zalinger, et al., 1977; Russell, 1974, pp. 16-23; for a review of predictive formulae see Hauer and Wilde, 1987). The lack of evidence to link restricted lateral visibility with higher accident rates seems paradoxical in light of the intuitive relationship between these two variables. Schoppert and Hoyt (1968) comment on this paradox in their treatise on driver behaviour at railway crossings:
This does not seem logical: sight distance should be one of the most important variables. If the driver canrwt see the crossing ana down the track an adequate distance, then he and his vehicle are being expected to perform beyond their physical limitations (p. 29).

Based on the tenets of risk homeostasis theory (RHT), Wilde et al. (1987) offer an explanation for this apparent disassociation between accident risk and lateral sightline visibility. The proposition is based on motorist appraisal of the prevailing hazard associated with the degree of restricted lateral visibility. In essence, it is proposed that motorists realise that their view of the track is restricted, and thus engage in compensatory modification of their approach behaviour. Because detection of an approaching train is impeded, it is not possible to reduce situational uncertainty as to the probability of a collision. Approach behaviour is modified to sustain a degree of perceived situational risk. By regulating approach behaviour, in a homeostatic manner which sustains this target level of incurred risk, motorists are expected to realise an approximately constant safety margin (i.e. probability of an accident). Motorists are assumed to be motivated to search for hazards on approach to a crossing with lateral restricted visibility in order to facilitate judgements about prevailing risk i. Motorists

It was assumed in this study that the irregular operation of trains at the sites (see section on Site Selection) prevented formulation of accurate predictions by motorists about the scheduling of train arrivals which may otherwise dictate motorist actions (Sanders, Kolsrud and Berger, 1973) rather than prevailing uncertainties about the presence of
trailIS.

Reslricred fated

sight distance

65

are also assumed to realise that their visual search is impeded by conditions of restricted lateral sight distance. As a result, it is expected that approaching motorists will: (1) Perceive the approach to be risky because, without visual access, the perceived likelihood of colliding with an approaching train remains uncertain. (2) Reduce approach speed to decrease stopping distance and to avail themselves of more time for visual search, thereby sustaining a safety margin sufficient to accommodate the situational uncertainty. Wilde et al. (1987) conclude that if motorists make such behavioural adjustments to variations in visual obstruction of the track, then there is no reason to expect a significant association between visibility characteristics of crossings on the one hand and their accident frequency on the other (Wilde et al., 1987, p. 35). The field study reported here provided an empirical test for the RI-IT proposition posited by Wilde et al. (1987) to explain the lack of association between accident rates and sightline distances. An empirical test of this proposition was attempted by designing a study which permitted a test of the conuerse to the original proposition by examining the effect of enhancing lateral sight distances. Three hypotheses can be derived from the implications of the RHT proposition (Wilde et al., 1987) when applied to the case of improved lateral sightline distances. (1) Visual search behaviour will be displayed earlier, in accordance with viewing opportunities, resulting in a perceived increase in approach safety. (2) Approach speed will increase commensurate with the perceived safety benefit of the sightline enhancement. (3) As a result of these behavioural changes, no net safety benefit will be realised. Previous field studies that have investigated these prospects in relation to lateral sight distance variations have yielded inconsistent findings (Aberg, 1988; Ehrlich, 1989; King, 1989; Parsonson and Rinalducci, 1982; Sanders et al., 1973; Wigglesworth, 1978). This inconsistency may be due to shortcomings of the methodologies applied. Apparently, only two experimental designs have previously been used to compare motorists behaviour at railway crossings under conditions of restricted and unrestricted lateral visibility: (1) between different crossings (e.g. King, 1989, Sanders et al., 1973) and (2) between opposite approaches to the same crossing (e.g. Parsonson and Rinalducci, 1982; Wigglesworth, 1978). These between-sire designs are unable to control for potentially confounding site attributes. In contrast, a within-site design in which sightline distance is manipulated at a site (an A-B design) is able to eliminate contamination between treatment effects and other site attributes. A comparis.on of motorist behaviours before and after enhancement of lateral sight distance at a site will (in principle) hold site attributes constant except for the sightline treatment, and can therefore identify treatment effects more clearly. However, it is also necessary to control for the effect of secular trends (e.g. gas prices, seasonal influences) which may co-vary with the treatment manipulation. This can be achieved by making parallel comparisons at a non-modified site during the same time periods. Another problem with earlier studies is that behavioural data for lateral head movements (which serve as a proxy-measure of searching behaviour) have predominately been operational&d as categorical (yes or no) data captured over a single (unspecified) region within the approachway and recorded as frequency counts (e.g. Aberg, 1988; King, 1989; Parsonson and Rinalducci, 1982; Wilde et al., 1976, 1987; Wigglesworth, 1978). When the temporal and spatial co-ordinates of lateral head movements are ignored, it is not possible to assess the significance of the search behaviour. For example, two motorists with the same total number

66 Table 1 Attributes of test and control sites Crossing Passive (Test site) Passive (Control site) Tracks ,N-S ,E-W Angle 80 53 Road Paved Packed Lanes 2 2 Volume 9/30 13/77

NJ. Ward and G J.S. Wilde

Speed (km/h) 95/60~4,) gO/60,4,,

Note: All data provided by Transport Canada-Railway Safety Office. Sites were located in central Ontario, Canada. Angle: is the angle between approach road and track in right quadrant. Tracks: superscript notation refers to track direction (from left to right as travelling toward crossing). Volume: units per day for trams/road vehicles. Speed: regulated speed for trams/road vehicles (subscript value refers to advisory speed).

of lateral head movements have not evidenced equivalent search behaviour in the case in which one motorist engages search activities later in the approach or engages in searches of shorter duration. Based on these considerations, this field study was designed to provide a real-world examination of the risk compensation proposition (Wilde et al., 1987) by adopting a within-site design to examine the effect of sightline enhancement on time and space referenced approach behaviours. Observations were made at an unprotected crossing before and after enhancement of existing lateral visibility. Parallel observations were made at an untreated site. To summarise the hypotheses tested, it was expected that sightline enhancement would result in (1) earlier search behaviour, (2) increased perceived approach safety, (3) increased approach speed, and (4) no net change in safety margin.

2. Method 2.1. Site selection Table 1 lists the attributes of the unprotected (passive) railway crossing for which lateral sight distances were increased by the removal of vegetation in the obstructing quadrant. A second passive crossing was identified in the same locality to serve as a control site. Both test and control sites were on track lines used by freight trains. These trains operated intermittently and were not subject to a regular schedule. Accordingly, in the absence of predictability of train arrivals at these sites, it is unlikely that (local) motorists were basing their approach behaviour on expectations realised from past experience. That is, with respect to the issues raised in the Introduction, motorists were presumed to be uncertain as to the presence of a train upon approach to the crossings such that the hypothesised uncertainty reduction motives (risk assessment) and activities (visual search behaviours) were instigated 2. A group of independent Civil Engineers was commissioned to measure the lateral sight distances at the test site with an electronic distance meter (EDM). The EDM was set upon a tripod at a height approximating the eye-level of a motorist (1.5 m) from each measurement position within a specified approach distance (see Fig. 1). The EDM was aimed at a reflective target approximating the height of the train (2.5 m>. The target was moved down the track

Refer to footnote 1.

Restricted

lateral

sight distance

67

_ ____ _ __ - Before enhancement


Left quadrant
QI

After enhancement

<
120 80 40m 40m 1 (5m) 2 (10m) 3 (15m) 4 (25m) 5 (35m) -e-*

Right

quadrant 160

240

200

160

80

120

200

I I

260

/....;;:::--::--;

E t

6(60m)

7 (85m)

Fig. I. Engineering survey results vegetation at the passive test site.

of lateral

sight distances

before

and after

removal

of obstructing

quadrant

away from the crossing until detection by the EDM was obscured. The greatest discernible lateral distance recorded by the EDM was then denoted by the lateral sight distance from that point in the approachway. Figure 1 illustrates the extent of lateral sight distance in both approach quadrants at the passive test site before and after the obstructing vegetation was cleared. It is apparent that the removal of the obstructing vegetation was successful in increasing lateral sight distances in both quadrants. 2.2. Observation protocol and procedure Motorist. behaviour was observed during summer weekdays (Monday-Friday) over a five-week period. At the test site, the first and third week constituted the first pre-enhancement observation period and the second pre-enhancement observation period, respectively. The approach quadrants were cleared of obstructing vegetation in Week 4, followed by the post-enhancement observation period in Week 5. The control site was not modified between the second pre-enhancement period and the post-enhancement period. Thus, the control site provided an opportunity to identify the occurrence of possible secular confounds at the test site between the second pre-enhancement period and the post-enhancement period. 2.3. Data collection As illustrated in Fig. 2, vehicles were tracked through an observation array of seven sonar units stationed along the roadside at incremental area comprising an distances from the

68 Hazard zone

NJ. Ward and GJS.

Wilde

Crossing control device

t
1-E E
10m 04 10 m 83 -

x 2 :onc
F 2:one

z:oae E
2Lone D

Zone C

25 m

Zone B

82

25 m

Zone A

31-

Fig. 2. Generic set-up of sonar array and observation equipment at railway crossings.

crossing. Each successive sonar pair defined a specific observation zone of a pre-determined distance (zones A, B, C, D, E, F, XI. The sonar units emitted a sonar pulse perpendicularly across the approach lane and sent a signal to a central data processor to indicate detection of a vehicle at that position. This process derived estimates of space-mean vehicle speed through-

Restricted lateral sight distance

69

Table 2 Number of cases in each treatment

level with complete

data throughout

approachway

Passive test crossing First pre-enhancement period Second pre-enhancement period Post-enhancement period

Measure (number of observed motorists) Right head Left head Speed (km/h) 18 18 25 27 26 32 23 23 32 motorist gender, estimated age,

Note: all treatment levels at each crossing were comparable in terms of observed observed number of passengers, and categorised vehicle type.

out the sonar array. Brake light activation was recorded by one observer and lateral head movements by another observer as vehicles drove past these observers (stationed in a roadside van) toward the crossing. These data were linked to the microprocessor through remote switches. Because the microprocessor continuously monitored which observation zone the vehicle occupied, all data values were referenced to their spatial and temporal position within the defined zones of the approachway. The base unit for quantifying lateral head movement and brake light duration was seconds per metre (s/m) 3. This reflects the duration of observed activity per unit of distance, and thereby provides a measure which is standardised for the different lengths of the individual observation zones. 2.4. Subject criteria Only single free moving vehicles travelling within the observation array were considered. Vehicles travelling within a traffic stream were excluded. In accord with the low train volume at the test and control sites, no train approached the crossing during the observation of any subject case. Passenger cars, vans (including passenger, work, and panel vans), and truck-style vehicles (e.g. pick-up trucks, jeeps) were included as valid cases. Professional drivers (e.g. police), drivers with out-of-province license plates, and drivers operating vehicles subject to specific traffic regulations at crossings (e.g. school buses) were excluded. In order to satisfy the requirements of sightline enhancement as a between-subject factor, only the first recorded instance of any repeat case was included (based on a matching of licence plates and observable driver characteristics). The distributions of motorists by age, gender, vehicle type, and number of accompanying passengers (estimated by observation and not tabulated here) were observed to be similar across treatment levels. Table 2 includes the final number of accepted participants involved in the repeated-measure analyses with valid data for each position in the approachway. 2.5. Perceptions of approach safety (manipulation check) Observed motorists could not be stopped within the approach because data collection was intended to be inconspicuous and unobtrusive. In order to obtain perceptions of approach

3 Values presented in the result figures are multiplied than one second.

by 1000 (s/km)

to avoid discussions

of fractional

times less

70

NJ. Ward and G.J.S. Wilde

safety, 26 local residents who lived in the proximity of both the control and test sites were surveyed. Some of these residents were included in our motorist samples.

3. Results Several findings from local residents survey are relevant. First, most respondents (18/26) had noticed that lateral sight distances had been enhanced at the passive test site. Few reported no apparent change (8/26) and none reported decreased lateral visibility, Second, of those respondents cognisant of the sightline enhancement, a majority (13/ 18) reported a commensurate decrease in perceived risk (i.e. an increase in safety). A few reported no apparent change (5/ 18) and none reported an increased perceived hazard. As an additional check, Ward (1992) showed colour slides of both approach quadrants of the test site before and after lateral sightline enhancement of lateral sight distances to a group of respondents (enrolled in an undergraduate course in psychological ergonomics). Restricted and enhanced slide pairings for each quadrant were presented side-by-side to permit comparative judgements. Conditions of enhanced sight distance were rated to be significantly less hazardous on a composite scale: How good is the visibility of the track at this crossing? How difficult would it be to detect a train as it approaches the crossing? How likely is it that a train approaching this crossing would not be seen? How risky do you believe it is to drive your car across this crossing? How anxious would you feel driving your car across this crossing? 3.1. Motorist approach behaviour The analysis took the form of a mixed-design ANOVA with position within the approachway (ZONE) treated as a repeated-measure factor with seven observation zones (zones, A, B, C, D, E, F, X). There were three levels of sightline treatment (first pre-enhancement period, second pre-enhancement period, post-enhancement period). The period of observation repreb)

. ..-.........

prc_Enhanccment

I), ;f

.._ . . . . . . _ Pre-Enhancement -Post-Enhancement

#2 j

Pre-Enhancement

%2

$L.@ +6@$gP 8 &8


Observation Zone (Approachway)

4+ci@&G80RN 6@
Observation Zone (Appronchw:

Test Site

Control Site

Fig. 3. Profiles of time-based measure of right lateral head movement at the test and control sites.

Resrricred

lateral

sight distance

71

100

(b)

Observation

Zone (Approachway)

Observation

Zone (Approacbway)

Test Site
Fig. 4. Profiles of cumulative control sites. percentage of motorist sample engaging

Control Site
in right lateral head movements at the test and

senting the sightline treatment factor (pre- and post-enhancement) was treated as a betweensubject factor because too few cars appeared across all conditions (inferred by observation of license plate numbers). For repeat cases, only the first encounter was included in the data set. Separate analyses were conducted between the post-enhancement period and each of the two pre-enhancement periods considered individually. Although this treatment resulted in a set of non-independent (non-orthogonal) tests, the diagnostic advantage of these comparisons was deemed more important than the preservation of test independence. Figures 3(a) and (b) illustrate duration of right lateral head movements at the passive test and control sites. There was a general trend across all treatment periods at the test site towards an increase in the duration of right head movements with increased proximity to the crossing [F(6, 390) = 23.22, p < 0.000011. As expected, motorists engaged in search behaviours of longer duration earlier in the approachway after sightline enhancement compared with the second pre-enhancement period [F(6, 390) = 2.29, p < 0.051. The parallel comparison at the paired control site was nor significant. No significant difference existed with regard to the first pre-enhancement period. Figures 4(a) and (b) show the cumulative percentages of the motorist sample engaging in right lateral head movements at the passive test and control sites. It was apparent that a higher percentage of the motorist sample engaged in search behaviour toward the right quadrant sooner in the approachway after the sightline enhancement [Kolmogorov-Smimov, DCri,= 38.5%, p < 0.051. The parallel comparisons at the paired control site were nor significant. There was a general trend across all treatment periods at the test site towards an increase in the duration of lef head movements with increased proximity to the crossing [ F(6, 384) = 17.16, p < 0.00011. There was no statistically reliable evidence that the frequency and/or duration of left lateral head movements was affected by the enhancement of lateral sightline distance.

72

NJ. Ward and G.J.S. Wilde

3.1. I. Per$ormance benefit (vehicle speed) Figures 5(a) and (b) report on approach speed for the passive and control sites. There was a general trend across all treatment periods at the test site towards a decrease in speed with increased proximity to the crossing [F(6, 516) = 453.72, p < O.OOOOl]. Although there was no evidence that the magnitude of speed change affected by the enhancement of lateral sightline distance was statistically reliable, a performance benefit after sightline enhancement was evident in terms of a consistent positive direction (sign) of speed change. Approach speed was consistently faster throughout the approachway after sightline enhancement [sign test, p < 0.011. By comparison, there was no comparable trend in terms of either the magnitude of change or the sign of change in speed throughout the approachway at the control site. 3.1.2. Net safety benefit (safety margin) Motorists were categorised as safe or unsafe, based on their approach behaviour. As stated by Schoppert and Hoyt (1968) the basic assumption made is that the driver at all points on his approach must be able to see a train in time to stop or proceed across the crossing ahead of the train (p. 3). Accordingly, motorists were classified as safe if, at the distance from the approachway they were first observed to search approach quadrants, it was possible to (1) stop before the crossing, or (2) proceed ahead of a (hypothetical) train approaching from the furthest observable point on the track. The safe-unsafe calculation was based on (i) vehicle speed; (ii) distance from crossing at which first search activity was noted; (iii) maximum expected train speed; (iv> maximum expected train distance; and (v) prevailing lateral sight distance. Minimum stopping distance of a vehicle was derived from equations provided by Baerwald (1965) assuming a zero grade approach and a friction coefficient of 0.60 for the road surfaces and weather conditions encountered, and a perception-reaction time of 2.5 s (Olson et al., 1984). The ability to stop or proceed was referenced to a standard hazard zone distance (see Fig. 2); the near-rail boundary of the hazard zone served as the reference stop line and the far-rail boundary served as the minimum safe distance to traverse.
(a)
1 ..--....-Pre-Enhancement
.__..--.Pre-Enhancement

80,

70 ,

(b)
-------.Pre-Enhancement _ Post-Enhancement #2

Y1 Y2

Observation

Zone (Approachway)

Observation

Zone (Approachway)

Test Site
Fig. 5. Profiles of time-based measure of approach

Control Site
speed at the test and control sites.

Restricted lateral sighi distance

73

It was apparent from the analysis of the test site data that (i) the approach of approximately 75% of motorists in each sightline condition could not be classified as safe, (ii) no motorist was deemed able to stop safely before the crossing, and (iii) the distribution of safe and unsafe motorists was not affected by the change in sightline condition.

4. Discussion 4.1. Evidence of hypothesised risk perception and behavioural compensation

This study addressed the safety-related issue of lateral sightline restrictions at railway crossings. The results obtained in this study are consistent with expectations derived from RHT and the (converse) of the proposition by Wilde et al. (1987) that motorists perceive the risk of restricted lateral visibility and adopt a more conservative approach behaviour to sustain a target safety margin. In response to sight distance improvement, it was presumed that motorists perceived the approach to be safer as a result of instigating search behaviour of longer duration sooner in the approachway, and subsequently utilised this realisation of improvement for higher speeds. The reduction of perceived risk may explain the consistently faster approach speeds after sightline enhancement. So, it appears that the site treatment that was intended as a safety improvement was used by drivers to enhance performance while they maintained their preferred safety margin. This trade-off, evidenced between the initial safety benefit of the improved lateral visibility and the ensuing consumption of this benefit by the adoption of higher approach speeds as a performance benefit, resulted in no demonstrable net safety benefit. In comparison, motorists at the control site, unaffected by any modification of lateral sightline distances, demonstrated no comparable change in search behaviour, approach speed, or any change in accepted safety margin. 4.2. Reconciling conflicting evidence

As noted in the Introduction, previous investigations of the relationship between restricted lateral visibility and motorist visual search behaviour have yielded conflicting results. Findings from the present study suggest that the source of the conflict is largely methodological and resides in the choice of location in the approach to the crossing in which motorist search behaviour is recorded. As demonstrated (see Fig. 3(a)), greater restriction of lateral sightline visibility may appear to produce (i) more search behaviour if observations are made within the later part of an approachway; (ii) less search behaviour if observations are made within the early part of an approachway, or (iii) similar levels if observations are made at some intermediate position in an approachway. Although in the current study the trend in search behaviour was evident for search behaviours towards both approach quadrants, only search behaviour toward the right quadrants evidenced a shift that was statistically reliable. This may be a measurement artefact. Right lateral head movements may have been recorded with greater sensitivity and reliability; for example, these movements may have been less impeded by vehicle obstructions (e.g. headrest) or evidenced greater movement amplitude relative to the observers line of view. Larger sample sizes would have facilitated comparisons of greater power. It is also possible that the apparent asymmetry of search behaviour was indicative of a belief held by participants that trains only approached at these sites from the right. Searching of the right quadrant may also

74

NJ.

Ward and G.J.S. Wilde

have been more sensitive to environmental contingencies provided by the improvement of sightlines. Searching was perhaps less dependent on a rigid search protocol founded on an intemalised representation of search necessity which dictated a habitual search pattern independent of environment conditions. Because of the short time lapse between clearing the quadrants and the post-enhancement period, it is not possible to rule out that the observed behaviour changes were the result of the novelty of the sightline intervention. Furthermore, the relatively short duration of the post-enhancement treatment periods make any comment on the durability of the observed effects purely speculative. Longer-term observations would be desirable. However, the results do provide a basis for tentative conclusions about the safety merits of sightline enhancement 4,

5. Recommendations Despite the absence of any calculated safety benefit based on theoretical collision avoidance scenarios, it can be argued that sightline improvement produces a more informed motorist because it provides pertinent information sooner in the approachway with the net result of providing more time (and distance) to decide upon and initiate action. The enhancement of lateral sightlines as a safety measure can be contrasted with alternative traffic engineering solutions such as posting warning or advisory signs at sites with restricted lateral sight distances (see Ward and Wilde, 1995). A local motorist population is likely to be aware of prevailing site hazards and formal notification would be superfluous. The exhortation of a prescribed response to the hazard (e.g. a stop sign) may be contrary to locally validated and accepted norms leading to poor compliance and disdain that may generalise to similar sites or signage. Admittedly, these signs may be useful to non-local motorists unaware of prevailing hazards imposed by restrictions of visibility at the site. Second, the prescribed response may have unforeseen negative consequences for other types of traffic conflict, such as rear-end accidents when motorists do not respond uniformly within the traffic flow. Third, unless the environmental hazard is remedied (i.e. lateral sightlines improved), any modification of behaviour that is solicited by signage may be insufficient to secure an acceptable margin of safety. While sightline enhancement may not produce safer motorists due to their inclination to exploit the introduction of physical safety measures for the purpose of enhancing performance, it should at least produce conditions conducive to more informed driving decisions.

Acknowledgements Funds for the support of the research have been provided by the Ontario Ministry of Transportation, Rail Safety Office. The primary author was supported by an Ontario Graduate Scholarship.

4 These results would be bolstered further by demonstrating the converse effect on motorist behaviour of restricting lateral sight distances, either as a separate (B-A) or a combined treatment (A-B-A). However, the introduction of sightline restrictions may be deemed unethical in some jurisdictions, and the removal and subsequent re-introduction of obstructions may be physically intractable.

Restricted lateral sight distance

75

Much gratitude is extended to Ron Martin for dedicated assistance in the collection of the data, as well as Martin York and Steve Ferguson for their patience and expertise in assembling the technical equipment. The authors are also grateful to the anonymous referees for their comments.

References
Aberg, L., 1988. Driver behaviour at flashing-light rail-highway crossings. Accident Analysis and Prevention, 20: 59-65. Baerwald, J.E., 1965. Traffic Engineering Handbook, Institute of Traffic Engineers, Washington, DC. van Belle, G., Meeter, D. and Farr, W., 1975. Influencing factors for railroad-highway grade crossing accidents in Florida. Accident Analysis and Prevention, 7: 103- 112. Ehrlich, D., 1989. Driver behaviour and decision making. Proceedings of the 1989 National Conference on Rail-Highway Safety, Transcom, San Diego, CA, pp. 71-78. Hauer, E. and Wilde, G.J.S., 1987. Grade Crossing Safety: A Review of Data and Methods, Department of Civil Engineering, University of Toronto. King, T.C., 1989. The effects of lateral sight distance at railway crossing on Driving Speed and Visual Search. Unpublished manuscript, Department of Psychology, Queens University, Kingston, Ontario, Canada. Olson, P.L., Cleveland, D.E., Fancher, P.S., Kostyniuk and Schneider, C.W., 1984. Parameters affecting stopping sight distance, report #NCHRP-270, Transportation Research Board, Washington, D.C. Parsonson, P.S. and Rinalducci. E.J., 1982. Positive-guidance demonstration project at a railroad-highway grade crossing. Transportation Research Record, 884: 29-34. Russell, E.R., 1974. Analysis of driver reaction to warning devices at a high-accident rural grade crossing, Joint Highway Research Project (HRP Report #74-l@. Purdue University/Indiana State Highway Commission. Sanders, J.H., Jr., Kolsrud, G.S. and Berger, W.G., 1973. Human Factors Countermeasures to improve HighwayRailway Intersection Safety, NTIS, Springfield, VA. Schoppert, D.W. and Hoyt, D.W., 1968. Factors Influencing Safety at Highway-Rail Grade Crossings (NCHRP Report #50), A.M. Voorhees and Associates, McLean, VA. Ward, N.J., 1992. Railway crossings with obstructed visibility: Perceived danger and driving behaviour. International Journal of Psychology, 27: 500. Ward, N.J. and Wilde, G.J.S., 1995. Field observation of advance warning/advisory signage for passive railway crossings with restricted lateral sightline visibility: An experimental investigation. Accident Analysis and Prevention, 27: 185% 197. Wigglesworth, E.C., 1978. The effects of local knowledge and sight restrictions on driver behaviour at open railway crossings. Journal of Safety Research 10: 100-107. Wilde, G.J.S., Cake. L.J. and McCarthy, M.B., 1976. An observational study of driver behaviour at signalised railroad crossings (Report #7.5-161, Canadian Institute for Guided Ground Transport at Queens University, Kingston, ON. Wilde, G.J.S., Hay, M.C. and Brites, J.N., 1987. Video-recorded driver behaviour at railway crossings: Approach speeds and critical incidents. (Report #87-61, Canadian Institute for Guided Ground Transport at Queens University, Kingston, ON. Zalinger, D.A., Rodgers, B.A. and Johri, H.P., 1977. Calculation of hazard indices for highway-railway crossings in Canada. Accident Analysis and Prevention, 9: 257-273.

You might also like