Supporting drivers in forming correct expectations about transitions between rural road categories

Accident Analysis and Prevention 43 (2011) 101–111

Contents lists available at ScienceDirect

Accident Analysis and Prevention journal homepage: www.elsevier.com/locate/aap

Supporting drivers in forming correct expectations about transitions between rural road categories Agnieszka Stelling-Konczak ∗ , Letty Aarts, Kirsten Duivenvoorden, Charles Goldenbeld SWOV Institute for Road Safety Research, P.O. Box 1090, 2260 BB Leidschendam, The Netherlands

a r t i c l e

i n f o

Article history: Received 1 October 2009 Received in revised form 15 July 2010 Accepted 29 July 2010 Keywords: Road safety Road design Expectations Information

a b s t r a c t In order to support drivers in forming the right expectations on the road, road categories are being made recognisable and predictable in the Netherlands. The present study investigated which of the selected road layouts can make rural road categories most recognisable for road users, especially in transitions from one road category to another. A second objective was to study whether explicit information could contribute to a better recognisability of transitions. The experiment was performed with a series of photographs showing sections of two road categories with an intersection in between. The road layout of road categories varied in markings and separation of driving direction (within-subjects factor). Informed and non-informed participants (between-subjects factor) had to indicate their expectations regarding speed limit and access restriction of each road section, before and after a transition. The results show that for transitions between distributor and through roads, the physicality of separation of driving direction is a better distinctive characteristic than the currently used edge marking. The green centre marking on through roads also enhances recognisability, but only with additional information. As far as transitions between distributor and access roads are concerned, the results demonstrate that this type of transitions is better recognised when no markings on access roads are present. Physical separation of driving directions on distributor roads also improves recognisability, although this layout is associated with higher speed limits. Providing explicit information has in general a positive effect on the reconisability of transitions. Implications are discussed in the light of potential safety effects. © 2010 Elsevier Ltd. All rights reserved.

1. Introduction Having the right expectations about the road one is driving on is important for safe driving as expectations influence driving behaviour (e.g. speed) and help to anticipate events (e.g. van der Hulst et al., 1999). Experimental studies show that drivers can overlook objects when they do not expect them (e.g. Theeuwes, 1991; Theeuwes and Hagenzieker, 1993). According to Malaterre (1990) a substantial number of crashes (59%) are the result of inappropriate expectations. One way to support drivers in forming correct expectations is by making road layout predictable and recognisable. A recognisable road layout is assumed to support the correct expectations of drivers about what road type they are driving on, which road users they can encounter and what driving behaviour (e.g. driving speed) is expected of the driver. By means of evoking correct expectations, a recognisable road layout can make the traffic behaviour more predictable, prevent indecisive behaviour of road users, and allow them to act more on routine. As routine behaviour

∗ Corresponding author. Tel.: +31 70 317 33 72; fax: +31 70 320 12 61. E-mail address: [email protected] (A. Stelling-Konczak). 0001-4575/$ – see front matter © 2010 Elsevier Ltd. All rights reserved. doi:10.1016/j.aap.2010.07.017

is related to less serious error types (Reason, 1990), a recognisable road layout can in the end prevent errors that could lead to crashes (e.g. Wegman et al., 2008). The important role of a predictable and recognisable road environment forms the core of one of the principles of the Dutch road safety vision Sustainable Safety (Koornstra et al., 1992; Wegman et al., 2008): the principle of predictability. This principle builds on two other ones: the principle of functionality of roads and the principle of homogeneity of mass, speed and driving direction. Ideally, roads have only one function: a flow function or an exchange function (functionality); and their layout facilitates homogeneous use in speed, mass, and direction (homogeneity). Sustainable Safety distinguishes three road categories according to their function: (1) through roads (TR): high speeds, physical separation of driving directions, slow vulnerable road users are not allowed (access restrictions for mopeds, bicyclists, and agricultural vehicles), (2) distributor roads (DR): intermediate speeds on road sections and low speeds at intersections, physical separation of slow and vulnerable road users and fast traffic is preferred (access restriction at road sections for mopeds, bicyclists, and agricultural vehicles), and

102

A. Stelling-Konczak et al. / Accident Analysis and Prevention 43 (2011) 101–111

(3) access roads (AR): the mixture of all traffic types requires low speeds, which is also enforced by the road layout. 1.1. Elaboration of Sustainable Safety principles in the Netherlands An important issue of the Sustainable Safety vision is speed management in relation to road layout and access restrictions. In the Netherlands, two types of TRs can be found: (a) motorways with speed limits of 100 or 120 km/h and (b) regional TRs with a speed limit of 100 km/h. Rural DRs have a speed limit of 80 km/h, which is the rural default limit. A sustainably safe road design requires physical separation of driving directions and safe shoulders were speeds are 70 km/h or higher (Wegman et al., 2008). These requirements are not always met, especially for rural DRs and regional TRs. In practice, there are also many DRs in the Netherlands where agricultural traffic is allowed, due to lack of alternative route options. Rural ARs mostly have a speed limit of 60 km/h. A speed limit of 80 km/h is also common, particularly when separate bicycle paths are available. In fact, the speed limit of rural ARs is the result of a negotiation between the safety-minded who plead for 40 km/h due to the mixture of vulnerable and high-speed traffic, and the flow-minded who wanted to keep the 80 km/h limit situation. Ideally, predictability is an integral characteristic of the functionality and homogeneity principles and can be found in a corresponding layout. As incorporating all three principles into understandable and uniform road layout turned out not to be that simple, a number of so-called ‘Essential Recognisability Characteristics’ (ERCs. CROW, 2004) were selected as a first structured attempt to make Dutch roads recognisable and predictable. These ERCs consist of road characteristics that are continuously visible to drivers and can provide them cues about the road type they are driving on. The ERCs mainly consist of variations in edge marking (e.g. broken and continuous) and the type of separation of driving directions (e.g. no separation, broken, continuous and/or coloured markings, physical separations) in unique combinations of patterns per road category. Table 1 presents a number of examples of rural road layouts with ERCs for the three road types. It has to be noted that drivers (can) rely on far more than the defined ERCs, such as road width and surface, curvature, and road environment characteristics (e.g. Davidse et al., 2004; Goldenbeld and Van Schagen, 2007; Kaptein et al., 1998; van Schagen et al., 1999; Weller et al., 2008) when recognising a road category. Furthermore, recognition of categories is enhanced when (1) the differences between categories are sufficiently large and (2) the variation within each category is small (Kaptein and Claessens, 1998; Theeuwes and Godthelp, 1995). Given the current variation within the layout of road categories in the Netherlands, and the relative small distinction in ERCs between the layouts of the different categories, it is not surprising that some road categories are not correctly recognised, i.e. the expectations of road users regarding speed limits, manoeuvres and road user types allowed on the road are not always correct (e.g. Aarts and Davidse, 2007). 1.2. Enhancing recognisability A recognisable and predictable road layout as a prerequisite for predictable road user behaviour is related to the internationally more familiar concept of ‘self-explaining roads’ (SER). The concept is used for a layout that does not need any additional explanation or learning process to know what it means and what to expect. The predictability principle is more specifically meant to support drivers in their expectations about speed, manoeuvres and vehicle types allowed. In both concepts a question remains whether it is really possible to give clear information to road users via road layout. It has been suggested that roads can be made at least

partly ‘self-explaining’ at low cost through speed colour coding (Campagne, 2005). Such a way of coding may enhance the distinctiveness of road types, but the meaning of each colour in terms of speed limit, manoeuvres and vehicle types allowed remains abstract and does not necessarily fit the expectations elicited by the total appearance and affordance of the road (e.g. Gibson, 1986; Goldenbeld and Van Schagen, 2007; Weller et al., 2008). From a holistic safety perspective, however, SER should ideally meet psychological as well as physical requirements. Whereas colour coding may be sufficient for the first one, it is certainly not enough for the second one: colour-coded roads do not, for instance, prevent drivers from colliding at high speeds. Taking both psychological and physical safety requirements into account, only a few road types in the Netherlands turn out to be self-explaining. Theeuwes (1994) found that motorways, equipped with emergency lanes, safety barriers and gantries were nearly always and very quickly correctly recognised by drivers. Aarts and Davidse (2007) found similar results when rural roads were equipped with a safety barrier: participants associated safety barriers with motorways and thus with high speed limits. Low speed zones, on their turn, may easily be recognised as such, provided that they are characterized by a credible low-speed design by a narrow and curved road, with uneven surface, and a built-up area near to the road (e.g. Martens et al., 1997). For other road categories, recognisability or self-explainingness is not so obvious in the sense that drivers do not always classify them in the same way as intended by the road authority. A study by Weller et al. (2008) for instance, found that the subjective impression of road users about road pictures had either to do with a kind of monotony, comfort, or demand. These classifications were, for instance, influenced by horizontal alignment, road width and surface quality, presence or absence of a centre line and sight distance. In the Netherlands, the enhancement of predictability has, until now, mainly been dealt with by implementing ERCs leaving other important characteristics more or less intact. The effects of this intermediate step in the Sustainable Safety approach have been evaluated in a rural roads photo-classification study of Aarts and Davidse (2007). The results showed that DRs and TRs were often mixed up. It has been concluded that the edge marking, which is used for the distinction between these road types, may not have a relevant meaning to road users. Another finding was that roads with a physical separation of driving directions (particularly those with a safety barrier) were often classified as roads where higher speed limits applied. Apparently, road users pay more attention to the separation of driving direction than to the edge marking. This seems logical as the former layout characteristic provides road users with information that is relevant for possible behavioural choices (i.e. overtaking), whereas the latter does not. Finally, ARs were better recognised as low speed roads when equipped with non-compulsory red cycle lanes or when no markings were present at all. This finding is in line with some previous studies where red cycle lanes were found to be a self-explaining layout characteristic (Kaptein and Theeuwes, 1996). Based on the evaluation study of Aarts and Davidse (2007) it can be concluded that many road layouts in the Netherlands are not yet well recognisable and predictable for road users. Not all layout variants have been evaluated so far. However, based on indications from previous studies, it can be expected that some of the layout characteristics can be helpful in forming the right expectations. First, the green centre line marking (see Table 1 variants 3 and 4) may turn out to be of benefit for recognisability of through roads. On the one hand, psychological research shows that the perception of colour takes place directly, i.e. it does not need cognitive interpretation (Kolb and Whishaw, 2003). It has been suggested that SER may be partly realised at low cost through speed colour coding (Campagne, 2005). On the other hand, the meaning

A. Stelling-Konczak et al. / Accident Analysis and Prevention 43 (2011) 101–111

103

Table 1 Examples of variants of road layout based on the ‘Essential Recognisability Characteristics’. Road type

Variants of rural road layout with Essential Recognisability Characteristics (ERCs)

Through roads (TRs)

Distributor roads (DRs)

Access roads (ARs) 1 and 7 single carriageway with a broken centre line marking; 2 and 8 single carriageway with a continuous centre line marking; 3 single carriageway with a broken centre line marking filled with green; 4 single carriageway with a continuous centre line marking filled with green; 5 and 9 single carriageway with a curb; 6 and 10 single carriageway with a central reservation; 11 brick road without road marking; 12 asphalt road without marking; 13 asphalt road with side marking to the edge; 14 asphalt road with side marking towards the centre.

of a colour may not have an intuitive link with functionalities such as a speed limit or particular road users to be expected on the road. In that case the green marking can only be helpful provided that the road users are well informed about its meaning. Second, as for the road layout of access roads, alternative philosophies to that of Sustainable Safety, e.g. Shared Space, are becoming increasingly popular in the Netherlands and abroad. When designing the traffic system, the starting point of the Sustainable Safety approach is allowing for human errors and limitations by offering protection (Wegman et al., 2008). Shared Space, on its turn, claims to employ human capacities to assure one’s own safety. The Shared Space philosophy is a reaction to the increasing demand to return to the natural beauty of the environment, where road design forms an integrated part of the local social and cultural context and where social rules rather than traffic rules apply (Keuninginstituut and Senza-Communicatie, 2005). The advocates of Shared Space claim to encourage social behaviour by regulating less with signs and markings and calling upon the cognitive capacities, communication skills and self-regulating ability of people. A manifestation of the philosophy is a reduction in the number of traffic regulations, such as traffic signs, speed humps, traffic lights, markings, separation of road users. If the layout of Shared Space locations is indeed able to evoke expectations of low speed and convey the message that a mixture of all kinds of traffic is present, the final result can perfectly contribute to a sustainably safe design, as long as a number of preconditions are taken into account (e.g. sharing space is not advocated from a safety point of view in traffic situations that are meant to facilitate high speeds). 1.3. Recognisability of roads in case of category transitions Up until now, research on recognisable road design has particularly concentrated on individual road sections. Because drivers move from one road to another, traversing different categories,

the question is whether they will notice changes in road layout sufficiently and adapt their expectations correctly and smoothly enough. There are indications that this can be true (Janssen et al., 1999). More fundamental psychological research however, points to the phenomenon of ‘change blindness’ raising the question whether just the change in road layout can be sufficient, or whether additional cues are needed to draw extra attention to the category change. As research has shown that change blindness can take place during driving (e.g. Martens, 2007; Velichkovsky et al., 2002), this phenomenon may also occur when a driver encounters a transition between road categories. Category changes often take place in the vicinity of intersections, where the driver has to process a lot of information. One can expect that salient category changes, which directly elicit the right expectations, may contribute to higher levels of safety because they are more easily noticed by the driver. The present study focuses on this topic because the subject of recognisable layouts in relation to road category transitions has received only very limited attention. 1.4. The present study The present study explores the recognisability of rural roads in transitions from one road category to another by focusing on recognisable layout of the road categories themselves. The first objective was to investigate whether and which of the current ERC road layouts (1) are distinctive enough for road users to recognise a change in the road category and (2) elicit the right expectations. A second objective was to examine whether explicit information about the layout of rural roads with ERCs can serve as an extra aid for enhancing recognisability. The study adopts a similar methodology to that of Janssen et al. (1999), namely asking participants about their expectations after showing sequences of photographs of a first road followed by photographs of a second road. The following more detailed hypotheses were tested (see Table 1 for road layouts):

104

A. Stelling-Konczak et al. / Accident Analysis and Prevention 43 (2011) 101–111

Table 2 Quota per province. Age

Male

18–24 25–39 40–49 50–64 65–79

1 2 1 1 1

Female 1 2  1

1.4.1 Transitions between TRs and DRs (1) Edge line marking alone will not be sufficient to distinguish a TR (continuous edge line marking) from a DR (broken edge line marking). (2) The more physically the driving directions are separated (central reservation or curb), the more often roads will be classified as higher order roads (in terms of speed limit and access restriction) than roads with no physical separation of the driving directions (centre markings). (3) With additional information of the meaning, green centre marking on TRs will lead to a better distinction between TRs and DRs compared to a similar centre marking layout without green. 1.4.2 Transitions between ARs and DRs (4) ARs will be more often classified as lower order roads than DRs when AR is characterized by a basic design without any marking (layout variant 6 and 7) than when the AR is equipped with road marking (variant 8 and 9). (5) The more physically the driving directions of DRs are separated, the more often DRs are correctly classified as higher order roads than ARs. (6) Providing explicit information about road layout assists road users in recognising a transition between DRs and ARs. 2. Methods 2.1. Participants Forty car drivers participated in the experiment and were recruited in four Dutch provinces: two quite advanced in implementing the ERC and two less advanced. Per province, a stratified sample of 10 participants was used to reflect the main driving population with regard to gender and age (see Table 2). The participants were aged 20–71 years (M = 39 years; SD = 15 years). Forty-six percent of the participants drove less than 15 000 km a year, 28% drove more than 20 000 km a year and 26% between 15 000 and 20 000 km

Fig. 2. Answer options (text is translated from Dutch) used in the preparatory and experimental task.

a year. The mean duration of car driving licence ownership was 17 years (SD = 14 years). The participants were recruited by mailings and via informal contacts. They received a gift voucher of D 25, and travelling expenses were repaid. 2.2. Material and task The test material used in the study consisted of sets of photographs, each comprising five succeeding photographs (see Fig. 1). Each set showed two photo’s of a first road, succeeded by a photograph of an intersection and followed by another two photographs of a second road. In the intersection photograph, the road layout of both the first and second road category could be seen, however, the design of the intersection itself gave no clues about the previous and succeeding road. Every photograph stayed on the screen for 1 s. After each two images of the first and second road, participants had to indicate their expectations about speed limit and access restrictions of the road they had seen just before. Responses of expectations were given by clicking a virtual button showing a pictogram of the legal road situation (see Fig. 2). The participants were then asked to indicate the certainty of these expectations concerning the previously shown road. Responding was self-paced. Errors could not be corrected. During the test, the participants were informed about their progress; the test showed which set they were about to begin (out of the total amount of sets). Information was manipulated by providing the participants with either relevant or irrelevant information: - relevant information: leaflet explaining the rules for rural road layout (speed limit and applicable access restrictions); the information in the leaflet, especially the pictures of the road layout were copied from information leaflets that local authorities provide to their inhabitants.

Fig. 1. Schematic illustration of the sequence of events in each individual trial (set).

A. Stelling-Konczak et al. / Accident Analysis and Prevention 43 (2011) 101–111

- irrelevant information: text about general traffic safety issues. The instruction included a schematic illustration of the sequence of events in each individual trial (similar to Fig. 1) and showed the moments at which the participants would be asked about their expectations and the certainty of these expectations concerning the previously shown road. Participants also completed a questionnaire on familiarity with the ERC-markings. The questionnaire contained only the main pictures that are used as campaign material in the provinces, showing two variants of road layout per road category. For each picture of road layout, the participants were requested to indicate (1) if they had already seen it on Dutch roads, (2) the speed limit they supposed to be correct, and (3) the type of road users allowed on such a road. Furthermore, the participants were asked if they had previously received any information regarding the meaning of the ERC-markings, and if so, the source of this information.

2.3. Procedure The session leader welcomed the participants and told them that the experiment aimed to study their opinion about Dutch roads. Participants were randomly assigned to one of the two information conditions and dependent on this, they received relevant or irrelevant information. Next, written instructions about the experimental task were distributed. Upon reading the instructions, participants performed a preparatory road classification task that was a similar but shorter version of the main experimental task. This preparatory task consisted of ten sets of photographs (see Fig. 1). The aim of the task was to familiarise participants with the type and speed of photograph sequences and with the use of the answering format. When finished with the preparatory task, participants completed the experimental task, consisting of 64 sets of rural road category changes. At the end of the experiment, participants were asked to fill in a final questionnaire on familiarity with the ERC-markings. The total experiment took approximately 75 min.

2.4. Dependent variables The dependent variables in this study were constructed as follows: first, the expectations concerning the speed limit and access restriction of each participant per individual road were scored one (correct) if they matched the road categorization according to ERCs or otherwise zero (incorrect). Two scores were determined: (1) the correctness of the speed limit and (2) the correctness of both speed limit and access restriction. Second, the correctness score for each transition per participant was determined, using the correctness scores per individual roads making up a transition. These scores were determined differently for transitions between distributor and through roads and for transitions between distributor and access roads.

2.4.1. Transitions between DRs and TRs Each DR–TR transition and each DR–DR transition was scored one (correct) or zero (incorrect) based on whether or not: (1) Both roads (before and after a transition) had elicited correct expectations with regard to speed limit. (2) Both roads (before and after a transition) had elicited correct expectations with regard to both speed limit and access restrictions.

105

(3) The second road had elicited expectations of a ‘higher order road’ (higher speed limit and more access restrictions).1 (4) the second road had elicited expectations of a ‘lower order road’ than the first road (lower speed limit and less access restrictions).1 Subsequently, the percentages of these measures per combination of road type layout (transition type) were calculated forming four dependent variables: A1, A2, A3 and A4. 2.4.2. Transitions between DRs and ARs Each AR–DR transition was scored one (correct) or zero (incorrect) based on whether or not a DR had elicited expectations of a ‘higher order road’ than an AR. Based on this score, the dependent variable B1 was calculated: the correctness-percentage of this measure per road layout variant. 2.5. Design The study used a within-subjects design for the effects of road layout characteristics. They were as follows (see also Table 3): 2.5.1. Transitions between distributor and through roads Edge marking effect: measures A1 and A2 were used to compare (1) transitions between DRs and TRs with the same type of SDD (edge marking changed from broken to continuous and vice versa) and (2) transitions between two DRs with the same type of SDD (no change in edge marking). Physicality of SDD effect: measures A3 and A4 were used to compare three groups of (DR–DR and TR–DR) transitions consisting of (1) a road with a double continuous centre marking and a road with a curb and vive versa, (2) a road with a double continuous centre marking and a road with a central reservation and vice versa and (3) a road with a curb and a road with a central reservation. Green centre marking effect: measures A1 and A2 were used to compare transitions between (1) DRs with double centre marking and TRs with a double centre marking filled with green (and vice versa) and (2) DRs with double centre marking and TRs with centre marking without green (and vice versa). 2.5.2. Transitions between DRs and ARs Edge marking effect: measure B1 was used to compare DR–AR transitions where ARs (1) had no markings (brick road or asphalt road) and (2) had edge markings (roads with broken edge marking close to the edge or closer to the middle of the road). Physicality of SDD effect: measures B1 and B2 were used to compare (1) transitions between ARs and DRs with no physical separation (double broken centre line), (2) transitions between ARs and DRs with a physical SDD (low concrete curb) and (3) transitions between ARs and DRs with a highly physical SDD (grass central reservation). To limit the duration of the experiment, only the before mentioned factors were varied in a semi-randomised way. Other factors that are known to influence driver’s expectations and behaviour on the road, such as road width, presence of buildings and/or other vertical elements near to the road, varied randomly over all situations but were not very different from each other. A between-subjects design tested the effect of explicit information about road layout. The participants were randomly assigned to one of two experimental conditions (20 participants per condition).

1 Roads with a speed limit of 100 km/h with access restrictions for all types of slow moving traffic formed the highest category and roads with a speed limit of 60 km/h open to all traffic formed the lowest category (order as shown in Fig. 2).

106

A. Stelling-Konczak et al. / Accident Analysis and Prevention 43 (2011) 101–111

Table 3 Overview of transitions and groups of transition that were compared to one another to test each hypothesis. Road category

Type of separation of driving direction

First (1) or second (2) section within the transition

DR

Broken centre line DR

Broken centre line Continuous centre line Ridge Central reservation

TR

Broken centre line Continuous centre line Green broken centre line Green continuous centre marking Ridge Central reservation

1 2 1 2 1 2 1 2 1 2 1 2 1 2 1 2 1 2 1 2

a, d c

e No e

Central reservation

c

c

c a, d

e c No

No e

No

b, d, f b, d, f

e c

e c

c, g e, g c e c h

6 5

b, d, f

c, g e, g

Characteristics

First (1) or second (2) section of transition

AR

Brick surface

1 2 1 2 1 2 1 2

e No b, d, f b, d, f

c e

DR

Broken centre line

Edge markings close to road side Edge markings distanced from road side

Ridge

a, d

Road category

Asphalt surface

Continuous centre line

Continuous centre line

No No i, j i, j h, j h, j h, j h, j

Ridge

Central reservation

No i, k i, k i, k h, k h, k h, k h, k

i, l i, l i, l i, l h, l h, l h, l h, l

No: transitions presented to subjects but not used in analysis due to irregularities in road scenes. Within-subject hypotheses (N = number of transitions with these characteristics): 1: ‘same edge marking vs. different edge marking before and after transition’: a (N = 3) vs. b (N = 5). 2: ‘to less physical SDD’, ‘same before and after transition’, ‘to more physical SDD’: c (N = 12) vs. d (N = 8) vs. e (N = 11). 3: ‘same centre line (non-green only)’ vs. ‘same centre line (TR with green, DR non-green)’: f (N = 5) vs. g (N = 4). 4: ‘AR with edge marking’ vs. ‘AR without edge marking’: h (N = 12) vs. i (N = 9). 5: ‘DR centre line’ vs. ‘DR curb’ vs. ‘DR central reservation’: j (N = 6) vs. k (N = 7) vs. l (N = 8).

2.6. Analysis

3. Results

2.6.1. Transitions between DRs and TRs The effects of various road category layout pairs on driver expectations were tested with separate repeated-measures ANOVAs per relevant measure, with information condition (information/no information) as between-subjects variable. Additional post hoc analysis was performed to determine the effects of characteristics with more than two levels.

Descriptive analysis of the dependent variables revealed a response bias of one participant. This participant was removed from the analysis leaving 39 subjects in the final analysis.

2.6.2. Transitions between DRs and ARs Since the dependent variable B1 was non-normally distributed, non-parametric tests were used to test hypotheses concerning transitions between distributor and access roads. To test the effects of various AR and DR layout pairs on driver expectations, a Wilcoxon Signed Ranks test or Friedman’s ANOVA was used, depending on the amount of levels that were tested. The effect of information was tested using Wilcoxon Mann–Whitney with information condition as two-level between subject variable. Furthermore, a descriptive analysis of individual roads (through, distributor, and access) was conducted in order to find out which type of errors people make while classifying various road layouts.

3.1. Transitions between DRs and TRs 3.1.1. Edge line marking and information In 86% of the DR–DR transitions, the participants were correct about the speed limit vs. 32% of the TR–DR transitions (significant difference, F(1,37) = 81.54, p < .05; Fig. 3a). The participants were also more often correct about the speed limit and access restriction (F(1,37) = 35.74, p < .05; Fig. 3b) than in the case of TR–DR transitions. Providing information had a positive effect on the percentage correct expectations concerning the speed limit (F(1,37) = 12.56, p < .05; Fig. 3a) but not on expectations regarding both speed limit and access restriction (F(1,37) = .94; Fig. 3b). There were no interaction effects (resp. Fspeed limit (1,37) = .88; Faccess restriction (1,37) = 2.52). 3.1.2. Physicality of the separation of driving direction and information The more physical the separation of driving direction (SDD), the more often participants’ expectations corresponded with a

A. Stelling-Konczak et al. / Accident Analysis and Prevention 43 (2011) 101–111

107

Fig. 3. Mean percentage of (a) correctly adjusted expectations about the speed limit and (b) correctly adjusted expectations of both speed limit and access restrictions of DR–DR and DR–TR transitions (v.v.), for informed and non-informed participants. The separation of driving directions was in all these cases exactly similar in the roads before and after the transition. Only the edge line marking varied (TR–DR v.v.) or not (DR–DR).

Fig. 4. Mean percentage of expectations of the second road being (a) a higher order road and (b) a lower order road than the first road as the result of more or less physical separation of driving direction, for informed and non-informed participants.

Fig. 5. Mean percentage of (a) correctly adjusted expectations about the speed limit and (b) correctly adjusted expectations of both speed limit and access restrictions of DR–TR transitions (v.v.) with green and non-green TR driving direction separation, for informed and non-informed participants.

Fig. 6. Mean percentage of (a) correctly adjusted expectations about the speed limit and (b) correctly adjusted expectations of both speed limit and access restrictions of DR–TR transitions (v.v.) with different forms of TR driving direction separation compared to continuous centre line DR layout, for informed and non-informed participants.

108

A. Stelling-Konczak et al. / Accident Analysis and Prevention 43 (2011) 101–111

higher order road (F(1,37) = 2.63, p < .05; Fig. 4a) and the less physical the SDD, the more often participants’ expectations corresponded with a lower order road (F(1,37) = 18.67, p < .05; Fig. 4b). No overall difference was found in the expectations of participants who received information and participants who received no information (Ftowards more physical SDD (1,37) = .94; Ftowards less physical SDD (1,37) = 1.16, p > .05). However, the informed group adjusted their expectations more often than the noninformed group when the SDD changed into a more physical one (interaction effect: Ftowards more physical SDD (1,37) = 4.14, p < .05; Ftowards less physical SDD (1,37) = .53, p > .05). 3.1.3. Green TR centre marking and information Overall, the green TR centre marking improved correct expectations about speed limits (F(1,37) = 81.54, p < .05; Fig. 5a) but not expectations about both speed limits and access restrictions (F(1,37) = 2.94, p > .05; Fig. 5b). Participants who received information were more often correct about speed limit and both speed limit and access restrictions (respectively F(1,37) = 27.76, p < .05; F(1,37) = 9.32, p < .05). Further analysis showed that the correct expectations about speed limit and access restriction of the noninformed participants, were actually not affected by the green centre marking (resp. F(2,74) = .70, p < .05; F(2,74) = .05, p < .05). When comparing DR–TR transitions (v.v.) where the TR was equipped with (a) non-green continuous double centre marking, (b) green centre marking or (c) central reservation, the transitions with green centre marking elicited more correct expectations about the speed limit (F(1,37) = 3.54, p < .05; Fig. 6a) but not about the speed limit and access restrictions than the transitions with centre marking or central reservation (F(1,37) = 1.19; Fig. 6b). The informed participants were more often correct about the speed limit (F(1,37) = 18.52, p < .05) and about both speed limit and access restrictions (F(1,37) = 3.12, p = .09) than the noninformed participants. The informed participants profited from the received information especially when forming expectations about the speed limit (F(1,37) = 6.28, p < .05). Further analysis showed that the expectations of participants who received no information were not affected by the green TR centre marking (Fspeed (2,74) = .72; Fspeed and access (2,74) = .00). 3.2. Transitions between DRs and ARs 3.2.1. Effect of AR edge marking Changes from ARs to DRs and vice versa were more often recognised as transitions from a lower to a higher order road when the AR was not equipped with any marking than when the marking was present, z = −3.60, p < .05 (see Fig. 7). 3.2.2. Effect of DR separation of the driving directions Changes from AR to DR (v.v.) were more often recognised as transitions from a lower to a higher order road when DRs were equipped with a more physical separation of driving directions than when DRs with a less physical or no physical separation of driving directions were present (2(2) = 10.34, p < .05; see Fig. 8). Wilcoxon tests were used to follow up this finding. A Bonferroni correction was applied and so all effects are reported at a .0167 level of significance. It appeared that the recognisability of AR–DR transitions increased significantly when DR’s were equipped with a curb or a central reservation compared to AR–DR transitions consisting of DRs with marking (respectively z = −2.20 and z = −2.80). There was no difference in recognisability of AR–DR transitions when the DR was equipped with a curb or a central reservation. As far as the speed limit and the speed limit in combination with access restrictions are concerned, no effect of the separation of the driving direction was found.

Fig. 7. Mean percentage of ARs being categorized as lower order roads than DRs as a function of AR layout with and without marking in AR–DR transitions (v.v.).

Fig. 8. Mean percentage of DRs being categorized as higher order road than ARs as a function of DR layout in AR–DR (v.v.) transitions.

3.3. Effect of information The participants who received information were better at categorizing ARs and DRs making up a transition as respectively lower order and higher order roads (z = −2.73, p < .05; see Fig. 9).

Fig. 9. Mean percentages of correctly categorized transitions for participants with and without relevant information.

A. Stelling-Konczak et al. / Accident Analysis and Prevention 43 (2011) 101–111

3.4. Individual roads in transitions Descriptive analysis of individual roads revealed that more than 80% of the DRs were correctly categorized as roads with the speed limit of 80 km/h. About three quarters of ARs were correctly categorized as roads with the speed limit of 60 km/h but only 47% were correct about the 100 km/h speed limit of TRs. As for DRs, 71% of all errors regarded categorizing DRs as TRs (expecting the speed limit of 100 km/h). Three quarters of the errors were made when a DR was equipped with a curb or a central reservation. Almost 70% of the participants who were correct about the speed limit, correctly indicated that slow traffic was not allowed on the road. The participants were less certain about the access restrictions of agricultural vehicles (only about 32% did not expect agricultural vehicles on the road). In the questionnaire, which only showed non-physical layout variants of road types, 83% of all participants correctly indicated the speed limit of DRs. Two-thirds of the participants were correct about the speed limit and the exclusion of bicycles and mopeds About 85% of the participants were familiar with the layout of the DRs. On TRs, about 52% of the participants assessed TRs as DRs (most often as roads with access restriction for all slow traffic). Most assessment errors were made when TRs had a broken double centre line (80% incorrectly assessed); the fewest errors were made when TRs had a green centre marking (64% correct) or a central reservation (51% correct). Eighty-six percent of the participants correctly indicated that slow traffic was not allowed on the road. In the questionnaire, 67% of the participants indicated to be familiar with the shown layout of TRs. In nearly 70% of the cases, the speed limit is estimated correctly and in more than 90% of the cases, bicyclists and mopeds are not expected. Nearly 30% expects agricultural vehicles when overtaking is allowed on these roads, against 13% when overtaking is not allowed. On ARs, 60% of all errors concerned categorizing these roads as roads with a speed limit of 80 km/h. Seventy-five percent of these errors were made when an AR was equipped with marking: 45% of the errors were made when ARs were equipped with the marking closer to the centre and 30% when they were equipped with edge marking at the outer edges of the road. Furthermore, in 11% of the cases the speed limit of asphalt roads without markings was incorrectly indicated, while 21% of brick roads, 28% of roads with marking near the edge of the road and 35% of the roads with edge marking closer to the centre were wrongly categorized. The majority (95%) of the participants who were correct about the speed limit, correctly indicated that slow traffic was allowed on the road. In the questionnaire, 76% of the participants were correct about the speed limit of ARs. In total, 42% of the participants were correct about the speed limit and the access restrictions for bicycles and mopeds on distributor roads, 57% were correct about the speed limit and the inclusion of agricultural vehicles and mopeds on ARs. Ninety-one percent was familiar with the layout of ARs. 4. Discussion The first objective of the current study was to investigate whether road layout of road sections themselves can be sufficiently recognisable to notice changes between road types. Specifically, it was the questions which of the road layout characteristics best enhance expectations about rural roads in a situation when the driver moves from one category to another. As roads in the Netherlands are currently equipped with Essential Recognition Characteristics (ERCs), this study addressed the ERC-layout variants of the three rural road types: through roads (TRs), distributor roads (DRs) and access roads (ARs).

109

Based on previous research (e.g. Aarts and Davidse, 2007), the type of the edge line markings (continuous vs. broken) was expected not be sufficient to distinguish TRs from DRs. The results of the present study confirm this expectation: a change in edge marking is hardly noticed as having a relevant meaning for the type of road, although this is the only characteristic on basis of which a distinction between TRs and DRs can be made. What is more, the results show that both TRs and DRs tended to be categorized as 80 km/h roads, irrespective of the edge line marking. This tendency is likely to be caused by the fact that 80 km/h is the default speed limit on Dutch rural roads. Whereas the change in the type of edge marking seems not to be used by the road users to distinguish DRs from TRs, the absence of edge markings can be of benefit for recognisability between other road types: the results show that the absence of road markings on ARs can enlarge the distinction between ARs and DRs. Since all DRs were equipped with markings (in combination with some form of separation of driving direction), an AR without marking was more in contrast with DRs than an AR with marking. It is plausible that in some cases, ARs with markings resembled DRs too closely, leading to incorrect expectations. The recognisability of ARs and DRs in terms of respectively lower and higher order roads improved with the absence of roads markings on ARs. However, even in the presence of road markings, the transitions between ARs and DRs were already recognised reasonably well. As hypothesized, a more physical separation of the driving directions (i.e. central reservation or a curb) evoked expectations that fitted higher order roads. This was true for transitions between DRs and TRs, where both road types equipped with a central reservation or a curb were more often associated with high-speed roads than when the roads were only equipped with central markings. This result confirms the finding of Aarts and Davidse (2007) that roads with a more physical separation, such as barriers, are regarded by drivers as high-speed roads, even as motorways (see also Theeuwes, 1994). But also in case of transitions between ARs and DRs, a more physical separation of the driving direction on DRs improved the recognisability of these roads in terms of respectively lower and higher order. However, in a considerable number of cases, participants categorized these roads as 100 km/h roads (TRs) instead of 80 km/h roads. This was especially evident when DRs were equipped with a central reservation. A likely explanation concerns the experience of road users. The majority of Dutch roads that are equipped with a central reservation are high-speed roads. The results of the present study show that not only physical separation of the driving directions is associated with higher order roads. TRs with a green centre marking are even better recognised than TRs with other types of separations, but only in combination with additional information. The results reveal that the green centre line marking is a very effective distinctive characteristic of a road category change; it is however not literally ‘self-explaining’ for speed limit and even less for the access restrictions. Furthermore, from the road safety point of view, it is not an ideal solution as it cannot prevent high speed frontal collisions taking place on these roads (homogeneity principle, see also Wegman et al., 2008). A physical separation of the driving directions is expected to be more effective to serve this aim. Besides investigating whether and which of the current ERC road layouts are distinctive enough for road users to recognise a road category transition, another objective of this study was to examine whether providing explicit information about the layout of the road types leads to a better recognisability. The results show that providing explicit information generally improves performance. However, even with additional information, the expectations of road users were often incorrect. Additional information particularly helped improving expectations about the perceived speed limits for

110

A. Stelling-Konczak et al. / Accident Analysis and Prevention 43 (2011) 101–111

TRs with a green centre marking. The effects were weaker or not present for the edge line marking as road type code. Being informed had also little effect on the expectations about access restrictions. This may again be a confirmation of the suggestion that participants attend less to the edge line marking and more to the centre of the road (see also Aarts and Davidse, 2007). This interpretation seems plausible as the separation of driving direction often informs road users whether overtaking is prohibited or allowed which is relevant for driving manoeuvres. Edge line marking on the other hand, is probably less meaningful for drivers and thus less salient for them. Furthermore, a green centre marking may attract attention due to its uniqueness (see the requirements for distinctiveness of roads by Theeuwes and Godthelp, 1993). Whereas the two types of edge line marking (continuous and broken) do not differ radically from each other, adding a colour makes a difference. The simplicity of the rule: ‘green means 100 km/h’ may also play a role as this rule is easier to remember than a rule concerning differences in edge line marking. As with every study, also this study had some limitations that have to be discussed. Firstly, photographs give only a “snapshot” view of a road at a single point, while the real driving task is performed in a continuously changing road environment, including the view from the road, road signs and other road users. Therefore, the issue of external validity merits further attention. As argued by Goldenbeld and Van Schagen (2007), external validity is not an allor-nothing phenomenon. Photographs of the road environment to a certain degree represent real-life situations. This is evident from photo-based self-report studies which partly confirm findings from behavioural or simulator studies. Secondly, as already mentioned, expectations (mainly concerning speed) are also influenced by characteristics such as road width, pavement and road environment characteristics (e.g. Martens et al., 1997). To not make the test too boring, some variation in the roads and surroundings was present, but varied more or less randomly over the different types of transitions. The variations were not so systematic that they could be analysed separately. Although this might add interesting results, this would have increased the duration of the test mainly because of the number of interactions that would have had to be tested. Here, a compromise was made from a participant’s point of view. Finally, some of the findings are based on a limited number of road category transitions. This could lead to distortions and low power in some of the findings. Also this is partly due to the compromise that had to be made between enough variation to make the test not too boring, and having sufficient trials from similar layouts. 4.1. Implications and recommendations Three layout characteristic appeared to improve the distinctiveness of transitions: the absence of edge markings on ARs for AR–DR transitions, a green centre line marking on TR for DR–TR transitions, and the physical separation of driving directions on the higher road for both AR–DR and DR–TR transitions. However, before recommending these characteristic for implementation, a few issues should be taken into account. As stated above, from the point of view of a recognisable road layout, ARs should preferably not be provided with lineation. This recommendation is not only relevant due to the increased recognisability of such roads, but also because of speed-reducing effect of the absence of the edge marking (van Driel et al., 2004). On the other hand, it should be kept in mind that edge marking can improve safety during darkness or poor visibility by providing some cues about the road course. The question is whether these positive effects of edge markings should be traded off against the improved recognisability in the absence of markings. Since the net effects of the absence of edge marking are not yet known, it is to

early to consider the removal of edge marking for implementation. A similar dilemma applies to the green centre line marking on TRs. On the one hand, the green marking substantially improves the recognisability of transitions between DRs and TR, the transitions which are generally badly recognised. The green centre line marking turns out to be a salient characteristic, however its meaning is not self-explaining and does not fit the homogeneity principle. At speeds of above 70 km/h, Sustainable Safety promotes the use of some physical form of directional separation, especially on rural roads (Wegman et al., 2008), in order to prevent head-on collisions which can have serious impact at high speeds. As the green centre line marking cannot prevent driving on the opposite lane, it may be implemented as an interim solution towards a sustainable safe design, as holds for the use of centre line marking on DRs and TRs in general. These results present a discrepancy between what drivers expect and recognise (i.e. no physical separation of the driving directions on DRs, no edge markings on ARs) and what is safe for other reasons. In this study, the only layout characteristic where this discrepancy does not occur is a physical separation of the driving direction. A physical driving direction separation is at least slightly more associated with higher order roads, although still not in all cases and not by all drivers as the current study shows, and is also regarded as safer at high speeds (e.g. Wegman et al., 2008). Hence, a discussion could be started about whether to implement physical forms of separation in driving direction on high-speed roads (above 70 km/h). It is important to realise that the issues of cost-effectiveness and solutions for traffic that requires overtaking (such as emergency-services or situations in which agricultural vehicles have access) will also have to be addressed. Furthermore, as both TRs and DRs have quite a high speed limit (which drivers not always comply with), this recommendation can be problematic when both the homogeneity and predictability principle are to be served at the same time. Applying the predictability principle uniformly would mean that also DRs should be equipped with physically separated driving directions. As the results of this study have shown, physical separation of the driving direction is often associated with roads with a speed limit of 100 km/h instead of 80 km/h. Physically separating the driving directions of DRs could therefore undermine the recognisability of these roads. The overestimation of the speed limit of DRs is also undesirable because of a likely rise in actual driving speeds on these roads. Since the distinction between DRs and TRs could again become unclear, it would be even more important to design each road category uniformly with its own unique separation of driving directions, ensuring that the DR and TR-variants are mutually exclusive. This will help drivers to distinguish the road categories. However, since Dutch drivers are particularly used to encounter physical separation of driving directions at motorways, drivers will need a period of adaptation before their expectations correspond to the road-type dependent forms of the physical separation. Finally, the implementation of new road characteristics should be accompanied by additional information about their meaning. It is not only important because of drivers need for being informed, but also because providing information turns out to be an essential element contributing to recognisability: SER is in fact a hypothetical concept. Information about the layout of the roads should be matched to what is essential to drivers. Also the way in which information is provided may be important. Relevant information in short understandable messages is preferred, and at locations relevant to the road user (for instance along the roads). This study supports the notion that road category transitions can be made clearly distinguishable by equipping road categories with a sufficiently recognisable layout. In many cases, however, the road category layout itself shows not to be sufficient for drivers to

A. Stelling-Konczak et al. / Accident Analysis and Prevention 43 (2011) 101–111

notice a road category transition. Further research should investigate whether making a transition itself extra salient (e.g. the design of the intersection) can improve its recognisability even more. Acknowledgement This study was part of a larger project directed at transitions to come to a sustainable mobility system (TRANSUMO; www.transumo.nl). References Aarts, L.T., Davidse, R.J., 2007. Distinctiveness, Self-explainingness, and Behavioural Effects of Recognizable Rural Roads in the Netherlands ETC 2007 Congress, Noordwijk, The Netherlands, 17–19 October 2007. Campagne, D.M., 2005. Road speed colour coding and traffic speed control: an applied psychology approach. Traffic Engineering and Control 46 (8), 292– 295. CROW, 2004. Richtlijn essentiële herkenbaarheidkenmerken van weginfrastructuur: Wegwijzer voor implementatie (Directive Essential Recognizability Characteristics for Infrastracture: Manual for Implementation). CROW, Ede. Davidse, R.J., Van Driel, C.J.G., Goldenbeld, C., 2004. The Effect of Altered Road Markings on Speed and Lateral Position; A Meta-Analysis. SWOV Institute for Road Safety Research, Leidschendam. Gibson, J.J., 1986. The Ecological Approach to Visual Perception. Lawrence Erlbaum, Hillsdale (New Jersey). Goldenbeld, C., Van Schagen, I., 2007. The credibility of speed limits on 80 km/h rural roads: the effects of road and person(ality) characteristics. Accident Analysis & Prevention 39 (6), 1121–1130. Janssen, W.H., Claessens, F.M.M., Muermans, R.C., 1999. Vormgeving van duurzaam veilige wegcategorieën: Evaluatie van ‘self-explaining’ kenmerken. TNO Technische Menskunde, Soesterberg, The Netherlands. Kaptein, N.A., Claessens, F.M.M., 1998. Effects of Cognitive Road Classification on Driving Behaviour: A Driving Simulator Study. TNO Technische Menskunde, Soesterberg, The Netherlands. Kaptein, N.A., Theeuwes, J., 1996. Effecten van vormgeving op categorieindeling en verwachtingen ten aanzien van 80 km/h wegen buiten de bebouwde kom (Effects of Road Design on Categorization of 80 km/h Roads (in Dutch)). TNO Human Factors, Soesterberg. Kaptein, N.A., van Hattum, S.T., van der Horst, A.R.A., 1998. Categorization of Road Environments and Driving Speed. TNO Human Factors, Soesterberg, The Netherlands.

111

Keuninginstituut, Senza-Communicatie, 2005. Shared Space—Room for Everyone. A New Vision for Public Spaces, http://www.sharedspace.eu/nl/ publicaties/downloads/cat view/15-boeken-en-brochures/17-eigen-uitgaven. Kolb, B., Whishaw, I.Q., 2003. Fundamentals of Human Neuropsychology, 5th ed. Freeman, New York, NY. Koornstra, M.J., Mathijssen, M.P.M., Mulder, J.A.G., Roszbach, R., Wegman, F.C.M. (Eds.), 1992. Naar een duurzaam veilig wegverkeer; nationale verkeersveiligheidsverkenning voor de jaren 1990/2010. SWOV Institute for Road Safety Research, Leidschendam, The Netherlands. Malaterre, G., 1990. Error analysis and in-depth accident studies. Ergonomics 10 (11), 1403–1421. Martens, M.H., 2007. The Failure to Act upon Important Information: Where do Things go Wrong? Vrije Universiteit, Amsterdam, The Netherlands. Martens, M.H., Comte, S., Kaptein, N.A., 1997. The Effects of Road Design on Speed Behaviour; A Literature Review. Managing Speeds of Traffic on European Roads Master Deliverable d1 Report 2.3.1. TNO Technische Menskunde, Soesterberg, The Netherlands. Reason, J., 1990. Human Error. Cambridge University Press, Cambridge. Theeuwes, J., 1991. Visual Search of Traffic Scenes. TNO Technische Menskunde, Soesterberg, The Netherlands. Theeuwes, J., 1994. Self-explaining Roads: An Exploratory Study. TNO Technische Menskunde, Soesterberg, The Netherlands. Theeuwes, J., Godthelp, H., 1993. Self explaining roads. In: Stoop, J.L.D.K.J.A. (Ed.), Safety of Transportation. Delft University Press, Delft, The Netherlands, pp. 56–66. Theeuwes, J., Godthelp, H., 1995. Self-explaining roads. Safety Science 19, 217–225. Theeuwes, J., Hagenzieker, M.P., 1993. Visual search of traffic scenes: on the effect of location specifications. In: Gale, A.R. (Ed.), Vision in Vehicles, vol IV. Elsevier, Amsterdam, The Netherlands, pp. 149–158. van der Hulst, M., Meijman, T., Rothengatter, T., 1999. Anticipation and the adaptive control of safety margins in driving. Ergonomics 42 (2), 336–345. van Driel, C.J.G., Davidse, R.J., van Maarseveen, M., 2004. The effects of an edgeline on speed and lateral position: a meta-analysis. Accident Analysis and Prevention 36 (4), 671–682. van Schagen, I.N.L.G., Dijkstra, A., Claessens, F.M.M., Janssen, W.H., 1999. Herkenning van duurzaam-veilige wegcategorieën. Selectie van potentieel relevante kenmerken en uitwerking van de onderzoeksopzet (R-98-57). Stichting Wetenschappelijk Onderzoek Verkeersveiligheid SWOV, Leidschendam. Velichkovsky, B.M., Dornhoefer, S.M., Kopf, M., Helmert, J., Joos, M., 2002. Change detection and occlusion modes in road-traffic scenarios. Transportation Research Part F 5, 99–109. Wegman, F., Aarts, L., Bax, C., 2008. Advancing sustainable safety. National road safety outlook for the Netherlands for 2005–2020. Safety Science 46, 323–343. Weller, G., Schlag, B., Friedel, T., Rammin, C., 2008. Behaviourally relevant road categorisation: a step towards self-explaining rural roads. Accident Analysis & Prevention 40 (4), 1581–1588.