1 Accident Analysis and Prevention 40 (2008) Cyclists perception of risk in roundabouts Mette Møller, Tove Hels Danish Transport Research Institute, Knuth-Winterfeldts Allé, Building 116, DK-2800 Kgs. Lyngby, Denmark Received 21 April 2007; received in revised form 23 October 2007; accepted 29 October 2007 Abstract Converting an intersection into a roundabout improves motor vehicle safety, but a similar safety effect is not found for car bicycle collisions. Very little is known about the reasons behind these collisions. In this study a first step towards an understanding of the reasons behind these collisions is taken. The study focuses on cyclists perceived risk in specific situations, factors influencing the perception of risk and cyclists knowledge about traffic rules regulating the interaction between road users in roundabouts. One thousand and nineteen cyclists aged participated in the study. Data were collected using structured interviews conducted in five Danish roundabouts. Underestimation of risk and lack of knowledge about relevant traffic rules may contribute to car bicycle collisions in roundabouts. Cyclists prefer road designs with a clear regulation of road user behaviour. A need to increase knowledge about traffic rules regulating road user behaviour in roundabouts is identified Elsevier Ltd. All rights reserved. Keywords: Roundabout; Risk perception; Bicycle 1. Introduction In Denmark, after single-vehicle accidents, the second most common type of accident is an accident involving a left turning vehicle (Danmarks Statistik, 2006). When an intersection is converted to a roundabout left turns no longer occur and the traffic flow is slowed down. Consequently, the number of motor vehicle accidents decline. A safety effect of converting an intersection to a roundabout has been found in Denmark as well as in other countries. In a review study Elvik and Vaa (2004) found a mean reduction in serious injury accidents of 10 40%. The studies included were from Northern Europe, Australia and the US. In a Danish study it was found that the average number of injured per accident in roundabouts was 1.06 whereas it was 1.35 in intersections (Jørgensen and Jørgensen, 2002). Due to the fact that there is a high degree of underreporting of Danish bicycle accidents (Bach, 2004) it is difficult to estimate the size of the problem. According to the official Danish accident statistics cyclists are killed and 1,500 cyclists are injured every year (Hemdorff and Lund, 2005). Approximately 110 of these accidents took place in a roundabout. However, a new Corresponding author. Tel.: ; fax: addresses: (M. Møller), (T. Hels). Danish study found that the degree of underreporting of bicycle accidents in roundabouts is 75% (Hels and Orozova-Bekkevold, 2007). Thus, bicycle accidents in roundabouts appear to be a much larger problem than hitherto believed. In accordance with this, safety effects of converting an intersection to a roundabout are primarily found in relation to car car collisions and do not apply to car bicycle collisions to the same extent (Herslund and Jørgensen, 2003). Consequently, in Denmark 81% of killed or injured road-users in roundabouts are cyclists or moped riders (Jørgensen and Jørgensen, 2002). The most frequent types of police registered bicycle accidents in roundabouts are accidents involving a cyclist circulating in the roundabout and a car that either enters or exits the roundabout. Only little is known about the reasons behind these accidents, although results from a few studies focusing on the behaviour of the driver indicate, that the phenomenon looked-but-failed-to-see is involved (Rääsänen and Summala, 1998; Herslund and Jørgensen, 2003). Looked-but-failed-to-see regards the behaviour of the driver just before the accident and refers to situations where the driver, although looking in the direction of the cyclist, fails to perceive it. Even less is known about the behaviour of the cyclist in these situations although it has been suggested that misinterpretation of the intention of the driver possibly is a contributing factor (Rääsänen and Summala, 1998). Misinterpretation of the intention of the driver may cause /$ see front matter 2007 Elsevier Ltd. All rights reserved. doi: /j.aap
2 1056 M. Møller, T. Hels / Accident Analysis and Prevention 40 (2008) the cyclist to expect the driver to yield. However, not having seen the cyclist, the driver has no intention to yield. Consequently, both parties continue and thus increase the risk of collision. In order to verify this suggestion as well as being able to reduce the number of bicycle accidents in roundabouts, studies focusing specifically on the cyclists are necessary. Perception of risk has been shown to influence decision making as well as actual behaviour in different situations (see for instance Deery, 1999; Lam, 2001; Chaudhary et al., 2004; Lam, 2005). Since the 1960s a number of different theories and models of driving behaviour that include perceived risk as a central motivational factor have been suggested (for an overview, see Ranney, 1994). Thus, although the authors of the different theories and models disagree with regard to the specific influence of perceived risk, perceived risk is generally recognised as an important influential factor in relation to driving behaviour. However, within the field of traffic safety research, studies on perceived risk have primarily focused on drivers. Therefore, the influence of risk perception on the behaviour and accident involvement of cyclists is less well documented. In particular, knowledge about risk perception in specific situations is sparse. Increased knowledge about perceived risk in specific situations would be useful in at least two respects. Firstly, such knowledge enables a comparison of perceived and actual risk. This is valuable knowledge in the process of clarifying whether or not the cyclists perception of risk is a contributing factor in car bicycle collisions in roundabouts. Low levels of perceived risk could discourage the use of caution in specific situations. Secondly, knowledge about perceived risk in specific situations would be useful for the development of road designs leading to lower levels of perceived risk. Studies have found that high levels of perceived risk influence travel mode in cases where persons have the possibility to choose (Noland, 1995). Thus, when bicycling seems too dangerous road users choose different modes of transportation or even refrain from taking a trip. This may lead to a decline in car bicycle collisions but using a more integrated approach as suggested by Racioppi et al. (2004) it becomes clear that a negative outcome in terms of lower level of public health outweighs the apparent positive effect caused by a decrease in car bicycle collisions. This is particularly relevant in relation to elderly bicyclists because independent outdoor mobility is an essential part of the quality of life of older persons (Farquhar, 1995; Siren and Hakamies-Blomqvist, 2006). However, health issues are relevant for the general population. In Denmark physical inactivity is estimated to cause 7 8% of all fatalities (Juel et al., 2006). Thus, from a public health perspective, as well as from a traffic safety perspective, increased knowledge about cyclists perception of risk in specific situations is highly relevant. As a first step towards a better understanding of the reasons behind car bicycle collisions in roundabouts this study focuses on cyclists perception of risk in a number of selected situations in roundabouts. In addition cyclists knowledge about traffic rules regulating the interaction between different road users is clarified. The results contribute to an understanding of factors influencing car bicycle collisions in roundabouts. Implications of the results are discussed. Fig. 1. A diagram identifying very general components of the roundabouts included in the study. 2. Objectives The aim of this study was to describe cyclists perception of risk in different situations. A second aim was to identify factors influencing the perception of risk with a particular focus on the influence of the existence of a cycle facility. A third aim was to clarify whether or not cyclists know the traffic rules regulating the interaction between road users in roundabouts. 3. Methods 3.1. Data Data were collected in five Danish roundabouts. The roundabouts were selected based on the following three criteria: (1) design feature, (2) traffic volume and (3) location. The roundabouts should be located centrally in towns to ensure a certain number of passing bicyclists per day. Very general components of the roundabouts included in the study can be seen in Fig. 1. All roundabouts had sidewalks along the roundabout as depicted in Fig. 1, but only three roundabouts had zebra crossings as depicted. Two roundabouts had a cycle facility in the roundabout as well as along the sidewalks. This is not shown in Fig. 1. The goal was to select roundabouts with and without cycle facilities and roundabouts that were as similar as possible when considering the three selection criteria. Details of the roundabouts included in the study are shown in Table 1. Table 1 Roundabouts included in the study Roundabout Number of legs Cycle facility Entering cyclists per day Entering cars per day 1 3 Yes Yes No No No Number of respondents
3 M. Møller, T. Hels / Accident Analysis and Prevention 40 (2008) Data on car and bicycle volume were collected following a standard procedure used by the road authorities as the total number of cars and bicycles entering the roundabout per 24 h. The traffic volume was assessed in different years for different roundabouts. Therefore, data on traffic volume were standardized to 2001-numbers. For further details regarding this procedure see Hels and Orozova-Bekkevold (2007). Data were collected using interviewer administered questionnaires with mostly closed questions cyclists aged participated in the study. Interviews were carried out Tuesdays through Thursdays between 7:30 in the morning and 16:30 in the afternoon. Each interview took about 10 min. Cyclists younger than 18 years old and cyclists driving so fast that trying to stop them would create a dangerous situation were not asked to participate in the study. The response rate varied according to time of day. During morning rush-hour (7:30 9:00) the response rate was at times as low as 20%. During the rest of the day the response rate was 60%. The questionnaire used included questions about risk perception and background factors (travel behaviour, familiarity with the roundabout and use of a bicycle helmet, bicycle accident involvement and near-accident involvement) and cyclist characteristics (age and gender). In relation to all questions about perceived risk the cyclists were asked to answer according to the particular conditions in the roundabout where the interview took place Perceived risk Perceived risk was rated based on 12 questions. Five of the 12 questions rated perceived accident risk in different traffic situations on a semi-quantitative scale including very high (4), high (3), low (2) and very low (1). Seven of the 12 questions rated perceived danger in different traffic situations on a semi-quantitative scale including very much (4), to a large extent (3), to some extent (2) and not at all (1). For both sets of questions a total score was calculated. The numbers in the brackets above indicate the points given for each answer. A high score indicate a higher level of perceived risk than a low score. On the scale measuring perceived accident risk the highest possible score was 20 and the lowest possible score was 5. On the scale measuring perceived danger the highest possible score was 28 and the lowest possible score was 7. In order to measure the internal consistency of the questions used and thus of the reliability of each of the two scales Cronbach s alpha was calculated. The reliability of the scale measuring perceived accident risk was 0.70 (Cronbach s alpha). The reliability of the scale measuring perceived danger was 0.69 (Cronbach s alpha) Statistical analysis The analysis of perceived risk was performed in several steps. The first step included a descriptive analysis used to identify the level of perceived risk in each of the situations introduced as well as in the roundabouts in general. The second step included a number of chi-square tests to analyse the association between different background variables, cyclist characteristics and design features. For this purpose only, data on perceived risk were grouped in two groups with the categories low and very low in one group and high and very high in the second group. Using multiple linear regression the third step of the analysis was performed to analyse the effect of each background variable, cyclist characteristics and roundabout design feature (Table 2). The aim of the analysis was to construct a simple linear model that explained as much of the variation in the cyclists perception of risk as possible. The model building strategy was backwards elimination: all variables were included initially and non-significant variables were subsequently removed. The final model consisted of significant variables only (Table 3). We aimed at keeping the model simple; thus, we only included first order components and no interaction terms. Thirteen out of the 14 variables regarding background information, cyclist characteristics and roundabout features listed in Table 2 were included initially. The one variable measuring the purpose of the trip at the time of the interview was not included in the analysis. This variable was not expected to be associated with perception of risk. The analysis of perceived risk was performed using the same three steps in relation to perceived danger and perceived risk of accident involvement. The total individual score on the relevant scale of perceived risk was used as the dependent variable. With regard to desired changes to reduce the accident risk for cyclists in the roundabouts descriptive analyses including a number of chi-square tests were performed. In all analyses a dichotomous coding of the traffic volume variables were used. With regard to bicycle traffic volume the dichotomous coding reflects the level at which the number of cyclists entering a roundabout significantly influences the risk of being involved in a bicycle accident (Hels and Orozova-Bekkevold, 2007). With regard to car traffic volume the dichotomous coding reflects the two groups found in the data (Table 1). All analyses were performed using SPSS version 14.0 for windows. The type 1 error criterion was set to 5%. 4. Results One thousand and nineteen cyclists were interviewed for the study (59% women and 41% men). The mean age of the interviewed cyclists was 42.9 years (S.D. ± 15.6, range 18 85). The large majority of the cyclists were familiar with the roundabout in which they were interviewed. Fifty-eight percentage (n = 588) bicycled through the roundabout 4 days a week or more. Eighty-four percentage (n = 852) bicycled through the roundabout at least once a week. Forty-one percentage (n = 421) had been involved in a near-accident in the roundabout where the interview took place Risk perception in different situations The cyclists perception of risk in different situations followed a very similar pattern in all roundabouts included in the study. In all roundabouts one situation was perceived as particularly dangerous measured as perception of accident risk as well as perception of danger. This particular situation was the situation in which a cyclist is circulating in the roundabout and a
4 1058 M. Møller, T. Hels / Accident Analysis and Prevention 40 (2008) Table 2 Questions included in the questionnaire with regard to background factors and cyclist characteristics Variables Category N Background Wore helmet at time of interview 0 (yes) 75 Binary 1 (no) 942 Destination at time of interview 1 (work/education) 526 Qualitative 2 (leisure time activities) (other) 69 How often do you ride your bike through this particular roundabout? 1 (6 7 days per week) 228 Semi-quantitative 2 (4 5 days per week) (1 3 days per week) (1 3 days per month) (<1 day per month) 62 How often do you ride your bike? 1 (6 7 days per week) 786 Semi-quantitative 2 (4 5 days per week) (1 3 days per week) 50 4 (1 3 days per month) 8 5 (<1 day per month) 2 How many kilometres do you bike per week? 1 (1 15 km per week) 152 Semi-quantitative 2 (16 30 km per week) (31 45 km per week) (46 60 km per week) (61 75 km per week) ( km per week) ( km per week) 70 8( 151 km per week) 43 Drivers licence 0 (yes) 780 Binary 1 (no) 239 How often do you drive a car (as driver) 1 (6 7 days per week) 66 Semi-quantitative 2 (4 5 days per week) 79 3 (1 3 days per week) (1 3 days per month) (<1 day per month) never 331 Involvement in near-accident 0 (no) 546 Binary 1 (yes) 421 Bicycle accident involvement in a roundabout 0 (yes) 47 Binary 1 (no) 970 Cyclist characteristics Age Numerical, discrete Gender 0 (male) 416 Binary 1 (female) 600 Roundabout design feature Cycle facility 0 (yes) 451 Binary 1 (no) 505 Traffic volume bicycles 0 ( 1000) 41 Binary 1 (>1000) 915 Traffic volume cars 0 (>10000) 387 Binary 1( 10000) 569 car is exiting the roundabout. Eighty-two percentage of all interviewed cyclists perceived this situation as very much or to a large extend dangerous. Sixty-five percentage of all interviewed cyclists perceived the accident risk in this particular situation as very high or high. The level of perceived risk in specific situations as well as the perception of the general level of risk was significantly higher in roundabouts without a cycle facility than in roundabouts with a cycle facility (Figs. 2 and 3). Thus, for instance the number of cyclists perceiving the accident risk in a situation with a circulating cyclist and a car exiting the roundabout was significantly larger in roundabouts without a cycle facility than in roundabouts with a cycle facility [χ 2 (1, n = 1015) = , p <.001]. Similarly the number of cyclists that perceived the situation with a
5 M. Møller, T. Hels / Accident Analysis and Prevention 40 (2008) Table 3 Factors significantly influencing perception of danger Variable Parameter estimate 95% CI t Constant *** Gender 1.426*** Near-accident 2.007*** Traffic volume cyclists 3.613*** Traffic volume cars.788** Cycle facility 2.462*** N = 932, R-square = 0.41, **p < 0.01, ***p < circulating cyclist and an exiting car as very much or to a large extent dangerous, was significantly larger in roundabouts without a cycle facility than in roundabouts with a cycle facility [χ 2 (1, n = 1014) = 7.382, p =.007]. With regard to the perception of the general level of risk the number of cyclists perceiving the roundabout as dangerous [χ 2 (1, n = 1007) = , p <.001] or involving a high level of risk [χ 2 (1, n = 1013) = , p <.001] was significantly higher in roundabouts without a cycle facility Factors influencing perception of risk Multiple regression analyses showed that the cyclists perceptions of risk were significantly influenced by a combination of roundabout design factors, traffic intensity and cyclist characteristics. Table 3 shows the results of the analysis using perception of danger as the dependent variable. Generally the analysis showed agreement between factors influencing the level of perceived risk measured as level of accident risk as well as level of danger. This was true in relation to roundabout design features and cyclist characteristics. The significant roundabout design features included the existence of a cycle facility; the significant cyclist characteristics included gender and involvement in a near accident. Thus, the existence of a cycle facility, being a male cyclist and not having been involved in a near-accident was associated with a significantly lower level of perceived risk on both scales. Travel behaviour measured as how often the cyclist went through the roundabout was not associated with perception of danger. It was, however, associated though not significantly with perception of accident risk (p = 0.18). Thus, cyclists who seldom cycled through the roundabout perceived the accident risk as greater than cyclists who went through the roundabout on a more regular basis. Perception of accident risk decreased with age (p = 0.02). As shown in Table 3 perceived danger was also associated with traffic volume. Thus, traffic volumes of more than 1000 entering cyclists per day and less than 10,000 entering cars per day were associated with increased perception of danger. In order to check for a possible interaction effect of bicycle and car traffic volumes, the analysis was rerun including this interaction. The interaction term was significant (p < 0.01) while the car traffic volume variable became insignificant and was consequently excluded. This indicates that some of the variation in the dependent variable covered by the car traffic volume variable was also covered by the interaction term. Thus, car traffic volume influences the perception of risk, but the influence decrease as the bicycle volume increases. See Section 5 for further discussion. Fig. 2. Percentage of cyclists in each roundabout, who has answered that the risk of being involved in a bicycle accident is large or very large in each situation. The cyclist is assumed to circulate in the roundabout in all situations. Roundabout no. 1 and 2 have a cycle facility. Roundabout no. 3, 4 and 5 do not have a cycle facility. Fig. 3. Percentage of cyclists in each roundabout, who has answered that each situation very much or to a large extent is dangerous. Roundabout 1 and 2 are with a cycle facility and 3, 4 and 5 are without a cycle facility.
6 1060 M. Møller, T. Hels / Accident Analysis and Prevention 40 (2008) Fig. 4. Percentage of cyclists in each roundabout, who has answered that each of the suggested change, will improve safety. Roundabout 1 and 2 are with a cycle facility and 3, 4 and 5 are without a cycle facility Desired changes to reduce the accident risk for cyclists The cyclists were asked to indicate which of six possible changes that they believed would reduce the risk of bicycle accidents in the roundabout where the interview took place. Three changes regarded design features regulating the interaction between different road users. Three changes regarded the road user behaviour and the number of road users. The majority of cyclists indicated that reducing the number (67%, n = 676) and the speed (66%, n = 672) of the cars in the roundabout would improve bicycle safety. An effect of cycle facility was found in three situations. In roundabouts without a cycle facility the cyclists were more likely to indicate that more cyclists [χ 2 = (1, n = 1017) , p <.001] and more space to the cyclists [χ 2 = (1, n = 1017) , p <.001] would improve safety (Fig. 4). In addition they were more likely to indicate that changing the roundabout to an intersection controlled by traffic lights would improve safety [χ 2 = (1, n = 1017) , p <.001]. Eighty-two percentage (n = 459) of the cyclists interviewed in a roundabout without a cycle facility indicated that building a cycle facility would reduce the accident risk in the roundabout Give way rules In all roundabouts included in the study car drivers entering or exiting the roundabout must yield to road users circulating in the roundabout. The large majority of the cyclists (88%, n = 904) were aware of these traffic rules. However, 11% (n = 114) were not. Of these 114 cyclists, 86 did not know the yield rule for drivers entering the roundabout. Nineteen did not know the rule regulating road users exiting the roundabout whereas nine cyclists did not know any of the relevant traffic rules. Sixty percentage (n = 68) of the cyclists who did not know the traffic rules were 45 years old or older. No association between gender and knowledge about traffic rules was found. Forty-seven percentage (n = 54) of the cyclists who did not know the relevant traffic rules had a drivers licence and drove a car on a regular basis. 5. Discussion The aim of this study was to increase knowledge about cyclists perception of risk in roundabouts. Risk perception in specific situations, factors influencing perceived risk, and attitude towards suggested safety improvements were studied as well as knowledge about traffic rules regulating the interaction between road users was studied. Before discussing the results, a few methodological issues should be considered. The aim was to include roundabouts that were as similar as possible according to the selection criteria. However, it is possible that minor design feature differences have influenced the risk perception of the cyclists. Further studies including a larger number of roundabouts are needed. Until then, the results of this study should be regarded as indicative of central issues worth studying further in relation to cyclists perception of risk in roundabouts. The sampling of the cyclists may also influence the results. All interviews were conducted in roundabouts. Consequently cyclists who find the roundabouts too dangerous to use may be underrepresented in the sample. Additional studies focusing on this group of cyclists are relevant, in order to identify the barriers that keep those cyclists away from the roundabouts. Despite the limitations of the study results indicate that underestimation of risk and lack of knowledge about relevant traffic rules may be contributing factors in car bicycle collisions in roundabouts. The cyclists perceived the situations involving a circulating bicycle and an exiting or entering car as involving the highest level of risk. This result indicates an accordance between the perceived and actual risk as the majority of the Danish police recorded car bicycle collisions in those two situations (Herslund and Jørgensen, 2002). However, according to the police recorded data more entering than exiting cars are involved in collisions. Thus, although the largest numbers of the cyclists perceive risk in accordance with the actual risk, some cyclists do not. Therefore, it is possible that underestimation of the risk related to entering cars is a contributing factor in car bicycle collisions in roundabouts. Underestimation of risk could discourage sufficient use of caution. In situations where cyclists misinterpret the intention of the driver underestimation of risk may contribute to a collision possibly due to small safety margins. Further studies focusing on the relation between perceived risk and actual behaviour are needed to clarify this. The results regarding perceived risk in specific situations also suggest that the level of perceived risk increases in situations where the level of perceived control and predictability is low.
7 M. Møller, T. Hels / Accident Analysis and Prevention 40 (2008) In the enter-situation the car is in front/side of the cyclist and thus most likely visible. In the exit-situation the car comes from behind, and consequently, the cyclist may feel less able to control and predict the interaction with the exiting cars leading to an increase in perceived risk in that situation. This is supported by the fact that the situation perceived as dangerous by the lowest number of cyclists, is the situation where the cyclist either enters or exits the roundabout. Those situations involve a very low level of interaction with other road users leaving the cyclist in almost full control of the situation. These results are in accordance with other studies of perceived risk that have found that a low level of perceived control in a situation is related to an increased level of perceived risk in that situation (Sjöberg, 2000). Results of this study show that perceived risk is influenced by a combination of factors regarding the individual cyclist (age and gender), design features regulating the interaction between road users (cycle facility) as well as the traffic volume. Additional individual and design features such as geometry, lighting, pavement surfaces etc. may influence perception of risk. The results regarding the influence of car traffic volume are counterintuitive and controversial as they contradict with results found in other studies (see for instance Parkin et al., 2007). Although it is possible, that the influence of car traffic volumes changes as the car or bicycle traffic volumes increase, it is more likely that car traffic volume in this study unexpectedly interacts with one or more factors present in the roundabouts but not taken into account in the study. Rather than putting too much emphasis on this particular part of the results further studies are recommended. The study showed that roundabouts with a cycle facility generally are perceived as safer than roundabouts without a cycle facility. This may reflect a sense of control and predictability created by a clear separation of the cyclists from other road users. Clear evidence of a safety effect of a cycle facility has not yet been established. Schoon and van Minnen (1994) found a safety effect of a cycle facility in roundabouts with vehicle volumes above 8000 vehicles per day but no effect in roundabouts with lower traffic volumes. In a more recent Danish study no effect of cycle facility in roundabouts was found on the number of cyclist accidents (Hels and Orozova-Bekkevold, 2007). This may partly be due to small sample sizes but may also be a result of risk-compensation (Peltzman, 1975; Wilde, 1988). In roundabouts with a cycle facility cyclists are very likely to compensate the decreased perception of risk by a behavioural change that involves increased risk taking. Consequently the possible safety effects of cycle facilities may be reduced. The behavioural change reflecting risk-compensation may or may not be a result of a fully conscious decision. In terms of behaviour it may appear as decreased use of caution, reduced safety margins and as pointed out by one of the reviewers of this article, reduced vigilance of subtle cues about the driver s imminent actions. In this study perceived risk was measured based on two dimensions risk of being involved in an accident and perceived danger. The two dimensions ask for a cognitive judgement and an emotional response, respectively. This corresponds to the two fundamental ways in which people comprehend risk (Slovic, 1987; Slovic et al., 2004). However, in the result section emphasis was put on results related to the emotional response. This was motivated by the fact that research on perceived risk increasingly recognises the emotional reaction as primarily responsible for the behavioural response rather than the cognitive judgement (Sjöberg, 1998; Sjöberg, 2000). The cyclists answers with regard to what changes they think would improve the cyclists safety indicate that irrespective of the existence of a cycle facility, cars are generally seen as a major threat to cyclists safety. However, the existence of a cycle facility does influence perception of the changes needed. In roundabouts without a cycle facility cyclists want the drivers to slow down whereas the number of drivers are considered the major problem in roundabouts with a cycle facility. Results also indicate that cyclists believe that more space would improve cyclist safety. With regard to roundabouts without a cycle facility it is not surprising that cyclists may feel trapped by motor vehicle traffic. However, results suggest that cyclists also feel trapped by other cyclists. This indicates that initiatives that clarify and regulate the interaction between road users in a roundabout, including interaction between cyclists, may lead to a decreased the level of perceived risk among cyclists using the roundabout. With this study a step towards a better understanding of factors influencing cyclists perception of risk in roundabouts is taken. A highly relevant next step would be to undertake a study in which the relation between perceived risk and actual behaviour in different situations is studied. The number of intersections converted into a roundabout has increased rapidly since the beginning of the 1990s. This has been the case in Europe (Brüde and Larsson, 2000; Rääsänen and Summala, 2000) as well as in the US (Perseaud et al., 2001). Knowledge about traffic rules is generally achieved through driver education. Many road users using the roads today got their drivers licence before the roundabout boom and thus before the rules regulating interaction in a roundabout were widely used. Therefore, it is possible that some road users are not aware of these traffic rules. And in fact results from this study show, that this is the case. Although the large majority of cyclists know the rules, a minority does not. This may causes road users to behave in unpredictable ways leading to increased collision risks. In this study, about half of the cyclists who showed a lack of knowledge of traffic rules, had a driver s licence and drove a car on a regular basis. It is easy to imagine that such drivers may be a major threat to the safety of cyclists circulating in the roundabout. Thus, further studies focusing on knowledge about traffic rules particularly among older road users are highly needed in order to clarify the prevalence of the problem. In addition the results identify a need to increase knowledge among road users about traffic rules in roundabouts particularly among older road users. References Bach, U., Færdselsuheld registreret på Sydvestjysk Sygehus [Traffic accidents recorded at the hospital in South West Jutland]. Carl Bro, 30 (in Danish). Brüde, U., Larsson, J., What roundabout design provides the highest possible safety? Nordic Road Transport Res. 2,
8 1062 M. Møller, T. Hels / Accident Analysis and Prevention 40 (2008) Chaudhary, N.K., Solomon, M.G., Cosgrove, L.A., The relationship between perceived risk of being ticketed and self-reported seat belt use. J. Safety Res. 35, Danmarks Statistik, Færdselsuheld 2005 [Road traffic accidents 2005]. Danmarks Statistik (in Danish). Deery, H.A., Hazard and risk perception among young novice drivers. J. Safety Res. 30, Elvik, R., Vaa, T., The Handbook of Road Safety Measures. Elsevier. Farquhar, M., Elderly people s definitions of quality of life. Social Sci. Med. 41, Hels, T., Orozova-Bekkevold, I., The effect of roundabout design features on cyclist accident rate. Accident Anal. Prev. 39, Hemdorff, S., Lund, H., Trafikuheld Året 2004 [Traffic accidents, 2004]. Notat, Vejdirektoratet (in Danish). Herslund, M.B., Jørgensen, N.O., Uheldet skyldtes, at bilisten ved indkørsel i krydset overså cyklisten [The cause of the accident was that the driver did not see the bicyclist]. Dansk Vejtidsskrift 12, (in Danish). Herslund, M.B., Jørgensen, N.O., Looked-but-failed-to-see-errors in traffic. Accident Anal. Prev. 35, Jørgensen, E., Jørgensen, N.O., Trafiksikkerhed i rundkørsler i Danmark. En analyse af uheld i danske rundkørsler i årene [Traffic safety in 82 Danish roundabouts] Vejdirektoratet, rapport 235. Institut for Veje, Trafik og Byplan, DTU (in Danish). Juel, K., Sørensen, J., Brønnum-Hansen, H., Risikofaktorer og folkesundhed i Danmark. [Risk factors and public health]. Statens institut for folkesundhed (in Danish). Lam, L.T., Parental risk perception of childhood pedestrian road safety. J. Safety Res. 32, Lam, L.T., Parental risk perceptions of childhood pedestrian road safety: a cross cultural comparison. J. Safety Res. 36, Noland, R.B., Perceived risk and modal choice: risk compensation in transportation systems. Accident Anal. Prev. 27, Parkin, J., Wardman, M., Page, M., Models of perceived cycling risk and route acceptability. Accident Anal. Prev. 39, Peltzman, S., The effects of automobile safety regulation. J. Polit. Econ. 83, Perseaud, B.N., Retting, R.A., Garder, P.E., Lord, D., Safety effect of roundabout conversions in the United States: empirical bayes observational before-after study. Transport Res. Rec. 1751, 1 9. Racioppi, F., Eriksson, L., Tingvall, C., Villaveces, A., Preventing Road Traffic Injury. A Public Health Perspective for Europe. WHO. Ranney, T.A., Models of driving behaviour: a review of their evolution. Accident Anal. Prev. 26, Rääsänen, H., Summala, H., Attention and expectation problems in bicycle-car collisions: an in-depth study. Accident Anal. Prev. 30, Rääsänen, H., Summala, H., Car drivers adjustments to cyclists at roundabouts. Transport. Hum. Factors 2, Schoon, C., van Minnen, J., The safety of roundabouts in The Netherlands. Traffic Eng. Control 35, Siren, A., Hakamies-Blomqvist, L., Does gendered driving create gendered mobility? Community-related mobility in Finnish women and men aged 65+. Transport. Res. Part F 9, Sjöberg, L., Worry and risk perception. Risk Anal. 18, Sjöberg, L., Factors in risk perception. Risk Anal. 20, Slovic, P., Perception of risk. Science 236, Slovic, P., Finucane, M.L., Peters, E., MacGregor, D.G., Risk as analysis and risk as feelings: some thoughts about affect, reason, risk and rationality. Risk Anal. 24, Wilde, G.J., Risk homeostasis theory and traffic accidents: proportions, deductions and discussion of dissension in recent reactions. Ergonomics 31,