TRANSPORTATION RESEARCH PART F

Transportation Research Part F 1 (1998) 137160

Aggressive driving: the contribution of the drivers and the situation$

David Shinar *
Ben Gurion University of the Negev, Department of Industrial Engineering and Management, PO Box 653, Beer-Sheva 84105, Israel
Received 14 August 1998; received in revised form 1 December 1998; accepted 12 December 1998

Abstract Aggressive driving is dened in terms of the frustrationaggression model. In that context aggressive driving is a syndrome of frustration-driven behaviors, enabled by the driver's environment. These behaviors can either take the form of instrumental aggressionthat allows the frustrated driver to move ahead at the cost of infringing on other road users' rights (e.g., by weaving and running red lights)or hostile aggression which is directed at the object of frustration (e.g., cursing other drivers). While these behaviors may be reective of individual dierences in aggression, it is argued that the exclusive focus on the characteristics of the aggressive drivers and how to control them is short-sighted. Instead, this paper proposes a multi-factor approach to the problem. Five studies conducted so far tend to support this approach, by showing that specic aggressive behaviorssuch as honking and running red lightsare associated with cultural norms, actual and perceived delays in travel, and congestion. Ergonomics-oriented approaches that involve environmental modications are proposed. # 1999 Elsevier Science Ltd. All rights reserved.
Keywords: Agression; Road rage; Red light running; Horn honking

1. Introduction The current hot issue in driver behavior and trac safety is ``Aggressive Driving'', also known as ``Road Rage''. I believe that what we are dealing with is an old issue but on a scale that is new and bigger than ever before. The old issue is that of aggressive drivers, and the new phenomenon is that of ubiquitous aggressive driving that we see all around us. Aggressive driversthough not in these wordswere already identied in a study published nearly half a century ago by two Canadian psychiatrists,Tillman and Hobbs (1949). They compared the non-driving background of high-crash rate taxi drivers to that of crash-free taxi drivers.
$

Keynote address presented at the International Congress of Applied Psychology, August 13, 1998, San Francisco, CA. * Fax: +972-7-647-2215; e-mail: shinar@bgumail.bgu.ac.il

1369-8478/99/$ - see front matter # 1999 Elsevier Science Ltd. All rights reserved. PII: S 1369-8478(9 9 ) 0 0 0 0 2 - 9

138

D. Shinar/Transportation Research Part F 1 (1998) 137160

They found that crash-involved drivers exhibited a high degree of socio-pathology as evidenced in a much greater involvement with social agencies, courts, police, jail, collection agencies, etc., than crash-free drivers; and they concluded that ``man drives as he lives''. Although not stated in these terms, I think this study was a precursor to our current focus on aggressive driving and the aggressive driver. Psychologistswho by denition study peoplehave focused primarily on who is he/she? Why does he/she behave that way? And how should we ``treat'' him/her? Thus, although ``aggressive driving'' is a current buzz word, the problem of aggressive drivers has been around for a long time. In fact, already three decades ago it merited a whole book``Aggression on the Road'' (Parry, 1968), and was argued to be the reason behind 85% of all crashes in Great Britain (Whitlock, 1971). It is important to distinguish between aggressive driving and aggressive drivers. While aggressive drivers have been identied for a long time, mass aggressive driving may in fact be a new phenomenon. This is true at least in the sense that it is a common observable behavior on the road todaymuch more common than it was a few years ago. The term ``aggressive driving'' is used rather loosely by dierent people, and in some cases it is used synonymously with ``road rage''. The National Highway Trac Safety Administration (NHTSA) distinguishes between them by classifying aggressive driving as a trac oense and dening it as ``the operation of a motor vehicle in a manner which endangers or is likely to endanger people or property'' (Martinez, 1997). In contrast, road rage is classied as a criminal oense and is dened as ``an assault with a motor vehicle or other dangerous weapon by the operator or passenger(s) of one motor vehicle on the operator or passenger(s) of another motor vehicle or vehicles precipitated by an incident which occurred on a roadway''. The problem with such a denition is that it is still too general and somewhat tautological. It is general in the sense that it encompasses just about any crash-producing driver action or inaction: driving too fast or driving too slowly, falling asleep at the wheel as well as driving under the inuence of a stimulant. It is also tautological because it denes the behavior (aggression) by its consequence (crash damage). For both of these reasons, it is very unlikely to lead to solutions of the problem, though it may make it easier for an ocer to cite a driver for non-specic dangerous driving other than ``reckless driving''. 1.1. The frustrationaggression model The denition I propose is based on the frustrationaggression model originally proposed by Dollard, Doob, Mowrer, Miller, and Sears (1939) in their classic book ``Frustration and Aggression''. They dened aggression as a ``sequence of behavior, the goal-response to which is the injury of the person toward whom it is directed'' (p.9), and postulated (and demonstrated very convincingly) that ``aggression is always a consequence of frustration'' (p.2). Interestingly, Dollard et al. (1939) chose to illustrate the frustrationaggression model with the behavior of a hypothetical college student who is stopped and berated by a police ocer (the frustration source) in front of his girlfriend. Once he drove away, according to Dollard et al., the student ``grated the gears frequently in shifting, refused to let other cars pass him, and made insulting comments about every policeman who came in sight'' (p. 12). This link of aggression to frustration is very important, since it also implies that all aggressive behaviors are instigated by a frustrating situation, behavior, or event.

D. Shinar/Transportation Research Part F 1 (1998) 137160

139

Factors can reduce or increase overt aggression include the following: (1) level of frustration the greater it is the greater the frustration and the more aggressive the response (Dollard, Doob, Mowrer, Miller, & Sears, 1939); (2) penalty for aggressive behaviorthe greater the penalty and the greater likelihood to be penalized, the lower the aggression (though it could be displaced to another time and location) (Dollard et al., 1939); (3) arbitrariness or legitimacy of the frustratoraggression will be greater when the frustrator is perceived is unfair, or inappropriate (Buss, 1963). Brown and Hernnstein (1975) stated it best by saying that aggression is instigated by ``illegitimate (behavior that results in) disappointment of legitimate expectations'' (p. 274). In this denition ``legitimate'' is synonymous with ``fair'' or ``appropriate''. This denition implies that we have norms about what is ``fair'' or ``legitimate'' and ``appropriate'' vs what is not. Aggression is directed towards another person only when his/her behavior is considered illegitimate: unfair, contrary to the norm, or contrary to the expected behavior. The aggressive behavior can assume one of two general forms: instrumental or hostile (Baron & Byrne, 1994, p. 461). In the context of driving, instrumental behavior includes all of the driving behaviors that the aggressor assumes will help him/her move ahead and overcome the frustrating obstacle. Typical behaviors can be honking the horn at drivers blocking the path, weaving in and out of trac, ``cutting'' in front of other drivers, and running red lights. Hostile behaviors are the kind that make us ``feel better'' without really solving the problem. They are a means to vent anger and are more consistent with Dollard et al.'s original denition. They are actually aimed at hurting the frustrator, and in the context of driving they fall under the category of road rage. Obviously the dichotomy is not always clear-cut: honking the horn at a pedestrian or another driver may serve both the instrumental and the hostile aspects of aggression. 1.2. Aggression (and frustration) on the road For greater highway safety, I believe that the greatest benet of embracing the frustration aggression model is that it provides a systems approach to the problem. As such it can be used to show that aggressive driving can be reduced not only by changing driver behavior directly (e.g. through enforcement) but also through changes in the environment that breeds aggression. Furthermore, eorts that focus exclusively on restraining driver instrumental aggression through enforcement may actually contribute to road (or o-road) rage as displaced aggression. Using the above concepts, I propose that we dene aggressive driving as a syndrome of frustration-driven instrumental behaviors which are manifested in: (a) inconsiderateness towards or annoyance of other drivers (tailgating, ashing lights, and honking at other drivers), and (b) deliberate dangerous driving to save time at the expense of others (running red lights and stop signs, obstructing path of others, weaving). I would dene road rage as hostile (vs instrumental) behaviors that are purposefully directed at other road users. These can be either driving behaviors (e.g., purposefully slowing a following vehicle or colliding with a lead vehicle) or non-driving behaviors (e.g., physically attacking someone). Unlike the all-encompassing NHTSA denition, according to this denition neither falling asleep at the wheel, nor speeding through a dangerous curve or weaving at excessive speed in the absence of trac are manifestations of aggressive driving. The rst example is obvious, but the latter is less obvious. Although the behavior is dangerous and purposeful, it is not caused by any

140

D. Shinar/Transportation Research Part F 1 (1998) 137160

impediments or frustrations to movement. It is most likely a reection of a personal tendency such as risk taking or sensation seeking (Jonah, 1997). Fig. 1 provides framework. This gure also includes predisposing factors that contribute to an aggressive disposition and ultimately increase the likelihood of aggressive driving. These are: (1) reduced communication between frustrating and frustrated drivers; (2) reduced threat of penalty from other drivers because of the shield and anonymity provided by the aggressor's vehicle (Ellison-Potter, 1997). We can illustrate the relevance of each of the concepts in Fig. 1 by considering a driver stopped at a signalized intersection behind another car. While the trac light is red, the driver's desire to move ahead is frustrated, but the stopping of the car ahead is legitimate so no aggressive behavior ensues. If the lead driver does not proceed to move once the light changes to green then there is

Fig. 1. Schematic representation.

D. Shinar/Transportation Research Part F 1 (1998) 137160

141

no legitimacy to the frustration, and the driver then has a disposition to honk. In a cultured neighborhood with a police car in sight, the lead driver will refrain from acting on the aggressive disposition. But in an alienated city with time pressure and no police in sight, that driver is likely to honk at the impeding driver and then cross the intersection after the light has changed to red. This model of aggressive behavior does not negate the existence of individual dierences in aggression, but it does highlight the contribution of the environment (in frustrating the drivers' goals) to aggressive behavior. Overt behavior is always the end product of individual tendencies in the presence of environmental catalysts. The role of personality and predisposing tendencies that a driver brings with him/her to the driving environment is illustrated by Retting and Williams' (1996) nding that red light violators are three times as likely to have previous multiple violations in comparison to non-violators. 1.3. Aggressive driving vs aggressive drivers The scientic concern with aggressive driving is very topical and the public concern with it is much greater than ever before. In a survey conducted in England in 1995, 62% of the drivers queried felt that the behavior of motorists has changed for the worse ``in recent years'', and 88% stated that they had experienced at least one of the listed manifestations of road aggression (being subjected to aggressive tailgating, ashing lights, rude gestures, deliberately obstructed, verbal abuse, physical assault, and unspecied others) (Connell & Joint, 1997). In a 1996 survey conducted in the US, Washington DC area, drivers said they felt more threatened by aggressive drivers (40%) than by drunk drivers (33%) (AAA Foundation for Trac Safety, 1997). Finally, in focus groups of Washington DC Beltway drivers, conducted in 1997, aggressive driving was perceived as the number one cause of crashes (Preusser Research Group, 1998). Summarizing the situation before a special meeting of the US House of Representatives, Committee on Transportation and Infrastructure, Williams (1997) suggested that what may have worsened over the past ve decades are the ``conditions that encourage this behavior''. Consequently, I believe that one key to the solution of aggressive drivingand the one that I want to focus on hereis to study the conditions that cause aggressive driving. I propose that over the past few decades drivers have not changed their personality and have not become more aggressive people. Instead, the conditions that elicit aggressive behaviors especially on the roadhave changed, so that the level of frustration is above the ``aggression threshold'' for more and more drivers. To illustrate, one common source of frustration and aggression is congestion. If we measure congestion in terms of the number of vehicles per mile of roadway we can assume that the more vehicles/mile, the higher the level of congestion, and the higher the expected level of frustration. Based on FHWA (1997) statistics, in the US in 1980 there were a total of 3.86 million miles of public roads and a total of 155.8 million vehicles registered in the US. By 1995 the total mileage increased to 3.91 million and the total number of vehicles registered increased to 201.5 million. In other words, while the length of the roadways increased by less than 1.5% the number of vehicles increased by 29%. The change over time in terms of vehicles/miles of roadway is plotted in Fig. 2. On the same gure, on the right Y axis is a scale of the ``threshold of aggression''; i.e. the amount of frustration needed to elicit overt aggressive behavior. This axis relates to the three identical hypothetical distributions of aggression. We can

142

D. Shinar/Transportation Research Part F 1 (1998) 137160

Fig. 2. A hypothetical relationship between aggressive driving and congestion.

now see that there is no need to propose that the number of aggressive people has increased in order to explain the increasing prevalence of aggressive driving. What we witness is simply an increase in the level of societal frustration, that exceeds the threshold of overt aggressive behavior of more and more people. Other synergetic factors that I believe contribute to the increase in frustration are those associated with the general increase in the stress of life. Instant communications and global competition have made time more valuable than ever beforeand its loss, equally more frustrating than ever before. In light of this, I think it is rather unfortunate that the focus of the aggressive driving issue has been almost exclusively on who are the aggressive drivers and what aggressive behaviors they display, rather than on why drivers in general are more aggressive now than before and what can be done (not necessarily to the drivers) to ameliorate the situation. In accordance with the frustrationaggression model, the driver's goal is to achieve mobility with minimum disruptions (and possibly with some pleasure). The negative feelings of frustration and aggressive behaviors that are generated when the path to this goal are blocked are most likely to be directed toward the perceived illegitimate source of the frustration, but may be generalized to other drivers. Given this approach, to reduce or eliminate aggressive driving by treating its causes rather than its carriers (especially when driver surveys suggest that aggressive driving characterizes the majority of the driving population) we must rst identify the source of the frustration. It is a fair assumption that the frustrated goals in driving are ecient mobility and pleasure. Automobile manufacturers and highway designers certainly believe these are relevant goals since they are promoted in advertisements, in cars people buy, and in routes people choose to drive. Yet, with more registered vehicles than licensed drivers (1.11 vehicles per drivers in the US), and when most drivers want to use the roads at similar times, we observe how both goals are frustrated. The situational sources of the frustration are the same factors that cause congestion: blocked path of

D. Shinar/Transportation Research Part F 1 (1998) 137160

143

travel by other cars, red signal lights, insucient number of lanes for the volume, discourteous drivers, and the tendency of many people to organize their workrecreation cycles around the same times and similar routes. In this paper I describe ve studies that evaluated the eects of some potential situational sources of frustration in modern drivingall related to time pressure: (1) the duration of green and red of trac signals, (2) the objective and subjective trac congestion, (3) time-of-day and day-of-week as they relate to rush vs non-rush hours, and (4) dierences in location reecting dierent cultural norms. Because we know that aggressive behavior and crash-involvement varies with age and sex men in general being more aggressive than women and younger persons being more aggressive than older persons (Dollard et al., 1939)we also looked at the eects of age and sex of the frustrated drivers. Gender dierences in aggression and violent behavior have been studied ad-nausea. In a metaanalysis of 143 studies that examined gender dierences in aggressive behavior, Hyde (1984) found that while none of the studies showed that women were more aggressive than men, and many studies showed that men are more aggressive, the gender dierences explained only 5% of the variance, once other individual characteristics were taken into account. The discrepancy between these studies and the over-involvement of males in violent crimes is due to the fact that in most psychological studies, the situation is rigged and the aggressive behavior that is measured is rather mild. This suggests that women are less likely to exhibit extreme aggression, but possibly just as likely to exhibit more subtle aggression, and exhibit aggression in more subtle verbal terms (e.g., the expressions ``shoot!'' and ``sugar!''). In driving, aggressive behaviors span a wide range from muttering, through yelling and making obscene gestures, and all the way to violent actions with the car. Thus, it is prudent to specify the gender role in terms of concrete specic behaviors. Hauber (1980) studied the aggressive reactions (including horn honking) of drivers to pedestrians who stepped o the curb to cross the street in non-signalized locations, and found no dierences between male and female drivers. Doob and Gross (1968) found that male drivers were quicker than female drivers to honk at vehicles stopped at the green light, but did not report dierences in the likelihood of honking. Deaux (1971) in his attempt to replicate Doob and Gross' ndings, obtained no gender dierences in the likelihood of honking at a stopped car, though drivers of both genders honked more at stopped female drivers than male drivers. Interestingly, Hauber obtained an opposite result: both male and female drivers were more aggressive towards crossing male pedestrians than towards female pedestrians. Retting and Williams (1996) did not nd any dierences between males and females in the likelihood of driving through red light, but Lawson (1991) reported that crash-involved red light running drivers were more likely to be males than the non-violating drivers in these collisions. In conclusion, the driving aggression studies are not conclusive with respect to the eect of gender. The relationship between age and aggressive driving seems to be more consistent. Retting and Williams (1996) found that driving through the red light is more common among young drivers. Lawson (1991) in his analysis of red light running crashes, found that the red light violators were more likely to be under 35 years old than the non-violators. Hauber (1980) reported that younger drivers are more likely than older drivers to honk at pedestrians. The two aggressive behaviors that we studied were running red lights (instrumental aggression) and horn honking at stopped vehicles that obstructed trac movement at signalized intersections

144

D. Shinar/Transportation Research Part F 1 (1998) 137160

during the green phase (hostile or instrumental aggression). Despite the prominence of the aggressive driving issue in the media, there are still very few empirical studies that directly assessed variables that may aect aggressive driving, and fewer still that used these two behaviors that we studied; namely driving through red lights, and horn honking at drivers blocking trac at signalized intersections. A recent study on driving through red lights by Retting and Williams (1996) compared the characteristics of 462 drivers who violated the red light signal with over 900 drivers who did not violate the red light at the same intersection. They found that drivers who drove through the red light were younger, were less likely to use seat belts, and had worse driving records than the drivers who did not drive through the red light. These results are consistent with the numerous studies of crash analyses that nd that young male drivers are over-involved in crashes relative to women and older men (e.g. Cerrelli, 1995; NHTSA, 1996). Horn honking was rst used as an observable dependent measure of aggressive driving behavior in a eld study published by Doob and Gross almost 30 years ago (1968). In their experimental paradigm the experimenter's car was stopped at a signalized intersection, and remained stopped when the light changed to green. They then studied the horn-honking behaviors of the drivers stopped behind the experimenter's car. They rationalized that this is a realistic situation for observing frustration-aggression, since it provides a ``clearly identiable frustrator and a fairly typical response for the blocked driver'' (p. 213). Also they noted that horn honking has both an instrumental value (since it may quicken the movement of the blocking car) and an emotional tension-reducing value (since it is an unpleasant stimulus to the driver of the blocking car). The eects of several independent variables on horn honking have been studied by dierent researchers. In all of these studies the experimenter's car was stopped at a red signal. When the light changed to green a timer was started and it was stopped when either the driver behind honked the horn or 12 s had elapsed. Special care was taken in these studies to conduct all trials during non-rush hours. In the rst study of the kind, Doob and Gross (1968) showed that drivers more readily honked at a low-status car (old Rambler sedan) than at a high-status car (new Chrysler Imperial). In light Sunday trac the mean delay (for male drivers) was 6.8 s for the former and 8.5 s for the latter. Also males were quicker to honk than women by 12 s. Deaux (1971) partially replicated Doobs and Gross' ndings and also showed that drivers more readily honked at female drivers than at male drivers, supportingwhat they termedthe ``damn female driver'' stereotype. Turner, Layton and Simons (1976), measured honking behavior on Saturdays and found a greater likelihood of honking at a blocking pickup truck when the curtain behind the driver was drawn than when it was removed, suggesting that drivers (all males) are more likely to honk when their anonymity is assured than when it is not. Finally, in the most recent study using this measure, Ellison, Govern, Petri and Figler (1995) observed honking tendencies by drivers of convertible cars and 44 vehicles with the top up (providing anonymity) and with the top down (with no anonymity). They found that the time to honk was signicantly shorter when the driver behind the stopped car was with the top up (6.36.5 s) than when he/she were in an open top (9.0 9.6 s). They did not nd any dierences between men and women. In all of the above studies, care was taken to conduct the research in non-rush hours. Thus, the very common situation of congestion and (possibly) time pressure was not a factor. The eects of congestion, rush hours, or crowding on driving behavior, to the best of my knowledge, have not been studied empirically. However there is ample evidence from social psychology and urban

D. Shinar/Transportation Research Part F 1 (1998) 137160

145

anthropology that crowding and congestion lead to increase in violence. The urban plight of most large cities is a testimonial to that. A related issue, that to the best of my knowledge also has not been tested directly, is that people drive more aggressively when under pressure than when they are not under pressure [though in a related study, McMurray (1971) found that people just prior to divorce have a higher crash rate than either much before or after it]. In the trac scene, time pressure, congestion, and alienation (or anonymity) are often confounded; occurring at the same times and in the same places. Thus we are probably more likely to observe some aggressive driving in crowded cities during the rush hours than in small towns during the weekends. Some direct evidence for this was obtained by Hennessy and Weisenthal (1997). In their study, drivers were queried about their mood and behaviors through cellular phones in congested and noncongested trac. Using structured questionnaires, they found that ``state'' stress (as distinguished from ``trait'' stress) was much higher in high congestion than in low congestion, and aggressive behaviors (including ``swearing'', ``purposeful tailgating'', and ``horn honking'') were twice as frequent in the high congestion than in the low congestion. 2. Five empirical studies The purpose of the studies reported below was to investigate the behavioral eects of some common frustrating situations and mediating variables on aggressive driving, and determine if the eects of these situations are more pronounced on some types of drivers (e.g. young males) than on others. Two paradigms were used in all the studies: running red lights and observing the behaviors of drivers blocked at signalized intersections. 2.1. Experiment I: driving through red lightsa pilot study (with Avner Ben Yaakov) This study used running through red lights as a measure of instrumental aggressive behavior, and tested the hypothesis that the longer the expected delay from waiting for the light to change, the more likely drivers will be to drive through the red light. Thus, it was expected that for drivers approaching the intersection as the light changes to red, the expected delay as measured by the duration of the red phase will be directly related to the drivers' likelihood of driving through the red ght. Conversely, we hypothesized that the longer the green phase, the less likely the driver will be to run through the red light. 2.1.1. Method 2.1.1.1. Subjects and locations. The subjects in the study were a convenience sample of drivers who drove through 10 pre-selected intersections. No data were collected on any vehicle or driver characteristics. Ten intersections in metropolitan Tel-Aviv were chosen. The speed limit at all locations was 50 kph. The green phase durations in the 10 intersections varied from 10 to 50 s, and the red phase durations varied from 30 to 70 s (in Israel, the green phase is typically protected for all turns, meaning that each leg has its own dedicated green phase. Thus in a standard fourlegged intersection, with equal time for all directions, the duration of the red phase is three times as long as the green phase). Total cycle time varied from 60 to 100 s, and the Pearson correlation between the green and red phase was r=0.74. All measurements were made during the rush

146

D. Shinar/Transportation Research Part F 1 (1998) 137160

hours of 7:00-9:00 am and 2:30-4:30 pm. This ensured, that on all trials, there were cars in the queue when the trac light changed from green, to yellow, and to red. Observations were limited to the straight-through lanes only. One hundred cycles were observed at each of the 10 intersections. 2.1.1.2. Design. The independent measures were the durations of the green and red phase and the dependent measure was the number of cars running through the red light per lane of trac per 100 signal light cycles. For example, in a two-lane intersection in 100 trials 154 cars ran through the red light. Therefore there was an average of 77 cars per lane, over 100 observations of light changes. Thus the number of cars running the red light per lane per cycle in that intersection was noted as 0.77. 2.1.2. Results and discussion The results showed a moderate inverse monotonic relationship (Pearson r=0.41, p<0.001) between the number of drivers running the red light and the duration of the green phase, and a slightly weaker direct relationship (r=0.35, p<0.001) between the number of cars running the red light and the duration of the red phase. Figs 3 and 4 illustrate these relationships. These patterns suggest that when people anticipate a brief ``window of opportunity'' to move ahead, they are more likely to be aggressive in order to exploit it, to the point where they feel justied in driving through the red phase. In a like manner, the foreknowledge of waiting a long time for the red light phase to end, is also suciently frustrating to induce many drivers to violate it. Whether the duration of the green phase is a more powerful inuence than the duration of the red phase remains to be studied with more signalized intersections and more trials. It is interesting to also note the very high actual frequency of the red light violations. In their studies in multiple lane roads in suburban Washington, DC, Retting, Williams and Greene (1996) recorded an average of 5.6 such violations per hour! 2.2. Experiment II To further understand the psychological and physical variables associated with running red lights, a larger experiment was designed. In this study both physical parameters associated with

Fig. 3. The relationship between running red lights and the duration of the green phase.

D. Shinar/Transportation Research Part F 1 (1998) 137160

147

Fig. 4. The relationship between running red lights and the duration of the red phase.

delays at intersections as well as perceived congestion were evaluated. In addition potential differences reecting dierent cultures and norms (by comparing a major city environment with a rural town environment, and night driving with daytime driving) were also investigated. 2.2.1. Method 2.2.1.1. Subjects locations. The subjects were all of the drivers who drove through the red light in the course of 8000 signal light cycles observed at 40 dierent signalized intersections selected for the study. Half the intersections were located in the heart of Israel's largest urban center (Tel Avivwith a population of over 1,000,000) and half were in a large rural town (Beer Sheva with a population of approximately 150,000). There were a total of 227 drivers who entered the intersection after the red light came on. However, only one of these drivers entered after more than a few seconds had elapsed. The rest, 226 drivers, entered within 13 s after the light had turned to red. Consequently only those 226 drivers were considered aggressive and only they were included in the analysis. Eighty percent of the drivers were males, and the modal age group was 3040 years old (as estimated by the researchers). Twenty-eight percent of the cars also had passengers. 2.2.1.2. Design. Forty urban signalized intersections were selected for the study, ranging in duration of the green phase from 6 to 49 s. At each site 100 signal-light cycles were observed during the day and 100 signal light cycles were observed during the night hours. All observations were limited to one through lane only. The major independent measures were (1) the red phase duration, (2) the green phase duration of the signal, (3) the average waiting time for cars arriving at the intersection at the time of observations, (4) the perceived congestion (on a scale of 110), (5) the level of urbanization (major hub city vs peripheral town), and the time of observation (day vs night). The dependent measure was the number of cars going through red per 100 cycles. 2.2.1.3. Procedure. The experimenters positioned themselves so that they would be partially hidden from the view of the drivers approaching in the selected leg of the intersection.

148

D. Shinar/Transportation Research Part F 1 (1998) 137160

2.2.2. Results and discussion Pearson r and phi correlations showed that the frequency of red-light running was signicantly related to the perceived congestion (r=0.75), the average waiting time at the intersection (r=0.60), the day/night dierence (r=0.42), the duration of the complete signal cycle (r=0.32), the duration of the red phase (r=0.30), and the urbanization level (r=0.23). Interestingly, the correlation between frequency of running red lights and the duration of the green phase was not signicant. A step-wise regression analysis with the physical variables only, indicated that the two signicant predictor variables were waiting time and day/night, which together yielded R=0.69; i.e. slightly less than the prediction based on perceived congestion. The relationship between running the red light and the average waiting time is presented in Fig. 5, and between running the red light and the perceived congestion in Fig. 6. The correlation between perceived congestion and waiting time was r=0.65, and the correlation between waiting time and red phase duration was r=0.48. To further examine the eects of the dierent predictor variables and potential interactions among them, the continuous variables of waiting time and congestion were redened as categorical variables, and ANOVA was performed on their eects on running red lights The ANOVA revealed that although waiting time and perceived congestion have a consistent eect, they are mediated by the cultural norms in the area. This is illustrated in Fig. 7, which reects the signicant interaction between City and Waiting time [F(2.68)=3.87, p=0.023]. A similar interaction was observed for the eects of congestion, where the eects of congestion were much higher in Tel Aviv than in Beer Sheva. Interestingly, in both cities, going through red light was twice as common during the day (3.96) as during the night (1.91) [F(1.68)=8.00, p=0.006], despite the reduced enforcement at night. Although most of the drivers running through red lights were males, their proportion in the sample was no dierent than the proportion of males driving through these intersections. In fact the two proportions were nearly identical. Males constituted 80.5% of those running through red lights, and 80.0% of those driving through those intersections. However, men ran through red lights with a lower average wait time than women, 85.1 vs 109.6 s [F(1.224)=5,81, p=0.017].

Fig. 5. The relationship between running red lights and waiting time.

D. Shinar/Transportation Research Part F 1 (1998) 137160

149

Fig. 6. The relationship between running red lights and rated congestion.

Fig. 7. The relationship between running red lights, waiting time and location.

2.3. Experiment III: horn honking at stopped cars at signalized intersectionsa pilot (with Avner Shevach) The purpose of this study was to evaluate the eect of both situational variables (trac signal duration, and time pressure) and individual dierences (age and sex of drivers) on aggressive behavior (horn honking) when waiting behind a car that does not begin to move when the trac signal light changes to green. The duration of the trac signal green and red phases are directly

150

D. Shinar/Transportation Research Part F 1 (1998) 137160

related to the expected time delays, and were shown in the rst study to aect aggressive behavior. Therefore signal duration was included as a situational variable that eects frustration level. ``Time pressure'' was manipulated by the day/time of travel, assuming that driving in weekday rush-hour trac would be more stressful and generate more aggression than weekend driving (as would be suggested by Hennessy & Wiesenthal's, 1997, ndings). 2.3.1. Method 2.3.1.1. Procedure. The study was conducted at two signalized intersections. One with a long green phase (35 s) and one with a short green phase (10 s). The male experimenter positioned himself in the trac stream so as to be the rst car stopped at the red trac signal, and turned his head to the right as if distracted by something on the front passenger seat. Each trial began when the signal changed to green and terminated when the driver behind the experimenter honked his/ her horn. Driver horn reaction time (RT) was measured by an inconspicuous assistant who was standing on the curb, partially obscured from the sight of the following driver. The RT consisted of the lag between the onset of the green light and the onset of the horn honking. As soon as the driver honked, the trial ended, and the experimenter drove around the block to return to the same site. In addition to the reaction time, the assistant also noted the sex and apparent age of the honking driver. Two dierent compact cars were used by the experimenters, one at each site. Neither car had any bumper stickers, commercial logos, or any other messages printed on them that might have aected the attitudes of the drivers behind them. 2.3.1.2. Design. Two situational independent variables were manipulated in this study: (1) time pressurehalf the data were collected on weekdays during rush hour trac and half on Saturday trac (in Israel Saturday is the day of rest, there is no public transportation, and with exception of places of entertainment, all businesses are closed); (2) signal durationshort green phase (10 s) vs long green phase (35 s). With 18 observations at each time pressuresignal duration combination there were a total of 72 trials. In addition to the two situational variables, the eects of age ( as estimated by the assistant on the curb) and sex were also analyzed. 2.3.2. Results and discussion None of the driversin all 72 trialswaited for the green phase to end before honking. Consequently there was no need to provide an articial time limit for drivers who failed to honk (as was done in the US studies). An Analysis of Variance conducted on the honking delay data showed that mean honking delay was shorter at the short signal duration than at the long one [F(1.68)=5.32, p<0.01], and during the weekday rush hours than on the weekend [F(1.68)=9.44, p<0.01]. Interestingly, even during the weekend hours, the honking delay times were approximately half as long as the ones obtained in the US studies. This probably reects a cultural difference. The interaction between these variables was not signicant. The average honking delay times for the four conditions are presented in Fig. 8. The analysis of the individual dierences showed that males were more impatient to honk than females (with an average delay of 2.89 vs 3.94 s [F(1.70)=5.86, p<0.01)], and there was a low but signicant correlation (r=0.25, p=0.03) between honking delay and age: older drivers being slightly more patient than younger drivers.

D. Shinar/Transportation Research Part F 1 (1998) 137160

151

Fig. 8. The relationship between honking delay, trac congestion, and green-phase duration.

2.4. Experiment IV: multiple situational factors that inuence honking at intersections Although the pattern of results from the pilot study was consistent with the hypothesis and previous research, the actual horn honking delays were much shorter than those observed in the US studies. Therefore, the purpose of this study was to validate the results obtained in Experiment III on a broad range of dierent intersections and to extend the ndings to additional variables. 2.4.1. Method 2.4.1.1. Procedure. The procedure was similar to that of Experiment III. The experimenter positioned himself in the trac stream so as to be the rst car stopped at a trac signal. Each trial began when the signal changed to green and terminated when the driver behind the experimenter honked his/her horn. Driver horn honking time was measured by the assistant sitting next to the driver, and consisted of the lag between the onset of the green light and the onset of the horn honking. As soon as the driver honked the trial ended. However, unlike Experiment II, in this study each trial was conducted at a dierent intersection, and the progression from one trial to the next consisted of driving to the next signal light in the predetermined route, consisting of 15 short signals and 15 long signals. In addition to the horn honking time, the assistant also noted the sex and apparent age of the honking driver. (Because there were four dierent experimenter+assistant teams in the study, an adjustment for individual dierences in the RT was made by measuring each assistant's simple RT on 10 trials and then adjusting each one's RT measures relative to the dierence between his mean and the group mean). All four experimenter cars were compact passenger cars without any bumper stickers (that might have aected the attitudes of the drivers behind them; c.f., Turner, Layton, & Simmons, 1976).

152

D. Shinar/Transportation Research Part F 1 (1998) 137160

2.4.1.2. Design. Four situational independent variables were manipulated in this study: (1) congestionbased on measures made during work hours trac (07:0017:00 on weekdays) vs Saturday (09:0018:00) trac (in Israel, Saturday is the day of rest, there is no public transportation, and with exception of some places of entertainment, all businesses are closed); (2) road typerural roads vs urban downtown Tel Aviv streets; (3) signal durationshort green phase (10 s or less) vs long green phase (30 s or longer); (4) distractionthe experimenter appeared either distracted (looking towards and conversing with his assistant) vs not distracted (looking straight ahead). These four dichotomous variables, constituted an experimental design of 16 combinations. Fifteen trials were conducted on each combination for a total of 240 trials. In addition to these four variables, the eects of estimated age and sex were also analyzed. 2.4.2. Results and discussion On all 240 trials, the blocked drivers honked before the green phase ended. Across all conditions, the average reaction time to honk the horn was 3.3 s. Thus there was no need to measure incidence of honking as a dependent measure, andunlike in the case in the US studiesthere was no need to make assumptions about honking delays, when the drivers did not honk. An analysis of variance of the 4 independent variables yielded a signicant main eect of signal duration [F(1.224)=5.94, p=0.008], congestion [F(1.224)=3.20, p=0.038], and distraction [F(1.224)=6.61, p=0.006]. Also signicant were the two way interactions road typecongestion [F(1.224)=7.09, p=0.004], and signal durationdistraction [F(1.224)=3.08, p=0.040], and the three way interaction of road typecongestiondistraction [F(1.224)=11.75, p<0.001]. These signicant eects and interactions are represented in Figs 9 and 10. As can be seen from these gures drivers are more impatient and are quicker to honk when the green phase is short than when it is long: 3.0 s vs 3.7 s. Honking delays are also shorter during

Fig. 9. Honking delay as a function of congestion and urbanization.

D. Shinar/Transportation Research Part F 1 (1998) 137160

153

Fig. 10. Honking delay as a function of distraction and signal duration.

congested weekday hours than during the uncongested weekend hours (3.1 vs 3.6 s) and when the driver appears distracted than when he is not: (3.0 vs 3.7 s). The signicant interactions further reveal that the eect of congestion is limited to city driving, where during the weekend drivers are willing to wait nearly 50% longer before they honk their horns relative to their wait during the weekday rush-hours. The interaction of signal duration with distraction indicates that when the driver ahead appears distracted (and therefore assumed to be unaware of the light change) the honking is almost immediate, regardless of the signal duration (2.9 vs 3.0 s), but when the driver ahead is not given that benet of the doubt (does not appear to be distracted), then the honking RT is nearly 50% longer for long green phases than for short ones. With respect to individual dierences, the perceived driver age ranged from 18 to 70, with a median of 38, and the majority of the drivers (80%) were males. Analysis revealed that young drivers (appearing to be 30 years old or less) honked signicantly faster than older drivers (50 or older): 3.04 vs 4.5 s. On the average, men and women did not dier signicantly in their horn honking delays [t(238)=0.60, ns]. The correlation between perceived age and honking delays was low but positive and signicant (r=0.17, p=0.007), and when the correlation between age and honking delays was calculated separately for men and women, it was essentially zero for women (r=0.03), but positive and signicant (r=0.21, p=0.004) for men, indicating that the age eect is relevant for men only. In short, young males are more impatient than older ones. Finally, drivers of commercial vehicles (trucks, buses, and taxicabsconstituting 10% of the vehicles) and drivers of vans, pickups, and 4-wheelers (constituting an additional 24%) did not dier signicantly from drivers of passenger cars. 2.5. Experiment V: demographics and honking at intersections (with Ayner Shevach) The primary goals of the present study were (a) to validate the measure of honking delay relative to other behavioral indicators of aggression, and (2) to determine the eect of cultural dierences as reected in the behavior of drivers in two residential communities that are in close proximity to

154

D. Shinar/Transportation Research Part F 1 (1998) 137160

each other, and an industrial zone that is half way between them. Ellison et al. (1995) obtained moderate correlations between honking delays, number of honks, and duration of honks, and also noted that some drivers revved their engines instead of honking. Doob and Gross (1968) also noted an inverse relationship between honking delay and the tendency to honk more than once. In this study we sought to extend the validity of honking delay to othermore directbehavioral manifestations of aggression. With respect to cultural dierences, our previous studies in Israel already demonstrated that there is a large dierence between our results and those obtained in the US. Also both the horn honking and the red light running studies yielded signicant dierences between cities within Israel. These dierences are also consistent with stereotypes of drivers from dierent geographies (e.g. New York and Boston drivers vs Midwest drivers). Therefore, the second objective of this study was to examine the eects of cultural dierences and control, for time and location dierences by obtaining all the data at the same time and from locations of close proximity to each other. 2.5.1. Method 2.5.1.1. Subjects and locations. The subjects were 142 drivers, of whom 21% were females, who arrived behind the research vehicle while it was the rst car stopped at a red light. The three intersections chosen were from demographically dierent locations, although they were all within a 15 km radius residential communities, one representing a high socioeconomic community (HSEC) and one representing a low socioeconomic community (L-SEC) in Israel. To illustrate the dierence between the two communities, the H-SEC has a median monthly income of $2255 while the L-SEC has $767; the modal occupation in the H-SEC is professional/academic and in the LSEC it is blue collar; the H-SEC is made up exclusively of single homes while the L-SEC is almost exclusively apartments; in a national rating of the quality of life in 188 residential communities the H-SEC was rated 183 while the L-SEC was rated 106. The third intersection was at a busy light-industry area that becomes a pub/disco center at night. 2.5.1.2. Procedure. The experimenter, a female driver in an open pickup truck, positioned herself as the rst vehicle at the signalized intersection while the light was red. When the light turned to green, she remained in place while an assistant watched the behavior of the driver behind the truck. Regardless of the behavior of the driver behind the experimenter, after 8 s elapsed the experimenter drove ahead. 2.5.1.3. Design. The primary independent measure of the study was the location: L-SEC, H-SEC, Industrial, and entertainment. By varying the times of the data collection we were also able to assess the eects of congestion. Finally presence of passengers, driver sex and apparent age of the blocked driver were also noted. Four dierent measures were used to record the behavior of the following driver: (1) Whether or not the driver honked within the 8 s wait; (2) honking delay based on time to rst honk the horn; (3) type of honking (single short honk, repeated honks, prolonged honk); (4) patience/ impatience based on whether or not the driver cursed and/or made hand gestures. 2.5.2. Results and discussion In the previous two studies, the experimenter started to drive away either immediately after the blocked driver honked or after one complete cycle. In this studyprobably because on all trials

D. Shinar/Transportation Research Part F 1 (1998) 137160

155

the experimenter started to drive away after 8 s11% of the blocked drivers did not honk or display any signs of impatience. Consequently the honking delay time was set at 9 s for these drivers and they were classied as ``patient''. No signicant relationships were found between whether or not the person honked and any of the independent variables. The principal analyses focused on two issues: the relationship among the remaining three measures of aggression, and the eects of the location on these measures of aggression. The relationship between type of honking, honking delay, and overt behavior and type of honking behavior is illustrated in Table 1. As can be seen from the table those who only use the horn for one short honk also wait longer to honk than those who honk repeatedly or continuously [F(2.124)=15.0, p<0.001], and the percent of drivers who show signs of impatience (by cursing or waving their hands) is signicantly higher among those who honk repeatedly or continuously than among those who only honk once and briey (12 64X26, p<0.001). Finally, honking delay was also longer for those classied as patient than for those classied as impatient (5.0 vs 2.5 s, t=7.36, p<0.001). Of the dierent independent measures, the most consistent in its eects was the location. Its eect was noted on both the honking RT [F(4.122)=10.90, p<0.001], and on the percent of drivers showing impatience (12 7X99, p<0.05). These eects are presented in Table 2. The similarity of the honking delay and the behavioral indications of impatience in the industrial zone and the L-SEC is probably not incidental since the drivers in both areas share similar socioeconomic backgrounds. An unexpected nding was the longer honking delays obtained in the industrial zone during the pubs/discos hours. The drivers during these hours were mostly young men and seemed to be unhurried and actually less aggressive. One possibility is that most of the drivers in that situation were probably in no hurry to get anywhere. Still this result is at odds with the many ndings that show alcohol using young males as an aggressive group. To relate our ndings to other measures of individual dierences (beyond age and sex), with the assistance of the trac records department, we were able to track and interview by phone 14
Table 1 Relationship between type of honking, honking delay, and overt measure of impatience Type of honking Measure of aggression Honking delay % impatient Single short honk (n=51) 4.1 s 15.7% Repeated honking (n=34) 2.9 s 16.8% Continuous honking (n=42) 2.4 s 92.9%

Table 2 The eects of location on honking delay and percent impatient drivers Location Measure of aggression Honking delay % impatient Industrial (n=30) 2.7 s 59% L-SEC (n=40) 2.3 s 70% H-SEC (n=29) 3.6 s 45% Pubs/discos (n=28) 4.6 s 43%

156

D. Shinar/Transportation Research Part F 1 (1998) 137160

drivers observed at the L-SEC location and 13 drivers observed at the H-SEC location. Although the samples were small they still diered signicantly in their honking RTs (2.8 vs 4.3 s for the LSEC and H-SEC, respectively; t=2.51, p<0.05). The interview further revealed dierences in honking RT as a function of the drivers' stated occupations: with average RT=5.3 for white collar workers, 4.2 for managers, 3.9 for retired/unemployed people, and 2.4 s for blue collar workers [F(3.23)=4.8, p=0.01]. All attempts to relate the drivers' self perceptions to their observed behavior, generally demonstrated no signicant relationships, though there were multiple strong and signicant correlations among self-reported measures of driver characteristics, attitudes, and behaviors. One interesting observation was that in general the drivers perceived themselves in much more positive terms than they perceived the ``typical Israeli drivers''. Fig. 10 reects this pattern. It summarizes the mean rating that drivers gave themselves and mean rating that they gave the ``typical'' Israeli drivers on a scale of 110 on aggression on the road, consideration of others on the road, and safe driving. The interviewed drivers thought that they were signicantly less aggressive [t(27)=8.75, p<0.001], more considerate of other road users [t(27)=2.65, p=0.013], and safer [t(27)=2.55, p=0.017] than the typical Israeli driver (Fig. 11). 3. General discussion Both measures of aggressive driving used in the studies above were validated in the course of the studies: the validity of driving through red lights was demonstrated by the fact that all but one of the 227 drivers observed in the red light running study entered the intersection within 13 s of the onset of the red light. The validity of the time to honk the horn was demonstrated through its inverse relationship to the tendency to make overt gestures and verbal abuse, and to honk the horn repeatedly or for long durations.

Fig. 11. Ratings of self vs ``typical'' Israeli drivers.

D. Shinar/Transportation Research Part F 1 (1998) 137160

157

These results do not refute the importance of individual dierences in aggression and the possibility of reducing aggressive driving through attitudinal and behavioral changes. However, empirically supported psychological theory suggests that dispositions of aggression that cannot be acted onwhile they may reduce driving aggressionwill result in aggressive behavior at other places and/or times. What the present results do show is the strong associations between the environmental conditions under which drivers operate and the level of manifest aggression. The ergonomic approach to problem solving assumes that (a) in a system in which many users behave ``inappropriately'', the fault is more likely to be with the system design than with the individual users, and (b) it is more ecient (and more user-friendly) to change bad design than to force people to adapt to it. In his testimony to Congress, Martinez (1997), the NHTSA administrator also noted several examples of how changes in trac travel patterns reduced trac violations and crashes. However, the systems approach is not without its costs. Trac congestion is a fact of life in the Western world. The results of the above studies directly demonstrate its eects on aggression. Yet based on US statistics we can see that the solution is not being sought in the construction of more roadways. Also, where congestion is most prevalentin the citiesthere is no room for more highways. Instead we should seek more ecient use of the existing highways. Better trac ow plans that would make people's natural behavior consistent with the desired behavior. Such solutions can be both within the driver highway system (e.g. better trac ow controls through signals, HOV lanes, changing one-way streets, etc.), as well as beyond it (e.g. exible work hours, telecommuting, etc.). All of the studies reported above are consistent in reecting the role of situational variables in two manifestations of aggressive driving: running red lights and horn honking. In all studies, overt manipulation of drivers' frustration level aected the two measures of aggressive driving. The potential for reducing such behaviors through the manipulation of the environment can be appreciated from Retting and Greene's (1995); study, where red light violations were reduced when the clearance intervals (also known as the inter-green interval) were increased. It is possible that the situations manipulated to reect varying levels of frustration are typical of the driving environment and culture in motorized countries at the end of this century. If this is so, then it may explain why such behaviors are either more common today than in yesteryear, and why we focus on them more than before. The pressured urban lifestyle of the late 1990s, (coupled with the spreading instantaneous communications that act to compress time), is such that we are under constant pressure to perform under an ever-increasing time and work pressure. In this environment, weekday driving is one of the last blocks of perceived ``wasted'' time (and we try to eliminate that by working behind the wheel as much as possiblewith cellular phones and dictating machines). Thus, it is likely that the increased value of lost time, is causing us to be much more aggressive on the road then ever before: especially during the working days in work-related travel. Some of the mediating factors studied here cannot be that easily manipulated to reduce stress: time of day, and geographically and culturally based norms. Other factorssuch as congestion producing factorsare costly to aect. Nontheless, it is critical to consider these factors if aggressive driving is to be reduced. The empirical studies reviewed above highlight the importance of these factors. Our results concerning individual dierences did not shed much light on this issue. We were unable to show consistent eects of either age or sex. Though when signicant eects were

158

D. Shinar/Transportation Research Part F 1 (1998) 137160

obtained, they did tend to indicate that aggressive driving is more common among males and in younger drivers. Socio-economic dierences were however demonstrated, indicating that aggressive behavior is more common among drivers of lower socio-economic levels. 4. Conclusions and implications The general conclusion that can be drawn from the existing theory and data, and from the ve studies described above on aggressive driving, is that the frustrationaggression model is useful for understanding and predicting aggression on the road. It also provides a tool for evaluating variables that can be theoretically linked to the frustration of the driving goals. Furthermore, the specic studies demonstrated thatfrom the perspective of the frustrated drivera short green light and a car blocking trac are ``illegitimate'' events that frustrate drivers' legitimate expectations. As such they can lead to instrumental and hostile aggressive behaviors such as driving through the red light and honking at the blocking vehicle, respectively. Both of the behaviors studiedrunning red lights and honkingare valid measures of aggression, and therefore can be eliminated when the sources of frustration are removed. Although not tested directly, the model would also predict that containment of aggression through enforcement, education and public information campaigns may have some limited value, but may backre when the pent-up aggression surfaces elsewhere. The eects of individual dierences in age and sex on aggressive driving are less clear from the results of these studies. However, to the extent that dierences were obtained overt manifestations of aggressive driving seemed to decrease with age, and were less common for women than for men. Gender eects, in particular, may be very subtle since it appears that women are just as likely as men to resort to mild aggressive behaviors, and the behaviors studied hereunlike those typically labeled as ``road rage''are denitely mild. The eects of other factors such as traitand state-stress is a dierent issue that remains to be investigated. Comparisons of the results obtained in dierent environments in Israel, and the results obtained in the United States, lead to the conclusion that cultural norms inuence the amount of observable aggressive driving. These norms vary both between countries but even within a country. Dierences found here between locations in aggressionexemplied by running red lights and honkingare consistent with some stereotypes. What remains to be studied is the degree to which prevailing norms aect drivers, vs the degree that people bring the norms with them from one place to the other. Finally, despite the eort to provide a robust denition of aggressive driving, I believe that there remains a need for a more operational, useful, commonly agreed-upon denition of aggressive driving. Even the two behaviors studied here may not be specic enough. Thus, some horn honking behaviors are probably more a means of communication than a display of aggression. It appears that only certain types of horn honking (immediate/repeated/long durations) and only going through the red light immediately after its onset are true aggressive driving behaviors. The studies described above have both practical and research implications. They demonstrate that we can probably reduce a signicant amount of aggressive driving by careful user-friendly ergonomically oriented design of the driving environment. Also, from an eective public health

D. Shinar/Transportation Research Part F 1 (1998) 137160

159

perspective the way to reduce road aggression should be to control the situations that give rise to it, rather than focus exclusively on driver education and mass media campaigns. Before the results described here can be applied, they should be validated in other contexts and other countries. We need much more data on the relationship between the environmental stimuli, individual dierences, and aggressive driving, before action oriented large-scale programs can be justied. Finally, and possibly most important, since there are no operational valid denitions for aggressive driving in either police accident reports or in other crash data banks, the contribution of aggressive driving to crashesrelative to other risky behaviors such as drinking, not using safety belts, and falling asleep at the wheelremains unknown.

Acknowledgements Thanks are due to O. Ben-Shitrit, A. Guttman, G. Horowitz, T. Luft, G. Luzon, E. Mualem, E. Sharon, and E. Shirer for assistance in data collection for Experiment II; and to D. Dahan, N. Fogel, N. Gruber, M. Manor, Y. Mizrahi and A. Segev, for assistance in data collection for Experiment IV. References
AAA Foundation for Trac Safety. (1997). Aggressive driving: three studies. Washington, DC: AAA Foundation for Trac Safety. Baron, R. A., & Byrne, D. (1994). Social psychology (7th ed.). Boston: Allyn and Bacon. Brown, R., & Herrnstein, R. J. (1975). Psychology. (Chapter 6: Aggression). Boston: Little, Brown, & Co. Buss, A. H. (1963). Physical aggression in relation to dierent frustrations. Journal of Personality and Social Psycology, 67, 17. Cerrelli, E. C. (1995) Crash data and rates for agesex groups of drivers, 1994. (research note). Washington, DC: National Highway Trac Safety Administration, US Department of Transportation. Connell, D., & Joint, M. (1997). Driver Aggression. In AAA Foundation for Trac Safety, Aggressive driving: three studies. Washington, DC: AAA Foundation for Trac Safety. Deaux, K. K. (1971). Honking at the intersection: a replication and extension. Journal of Social Psychology, 84, 159160. Dollard, J., Doob, L. W., Miller, N. P., Mowrer, O. H., & Sears, R. R. (1939). Frustration and aggression. New Haven, CN: Yale University Press. Doob, A. N., & Gross, A. E. (1968). Status of frustrator as an inhibitor of horn-honking responses. Journal of Social Psychology, 76, 213218. Ellison, P. A., Govern, J. M., Petri, H. L., & Figler, M. H. (1995). Anonymity and aggressive driving behavior: a eld study. Journal of Social Behavior and Personality, 10, 265272. Ellison-Potter, P. A. (1997). The eects of anonymity, aggressive stimuli, and trait anger on aggressive driving behaviour: a laboratory simulation. Ph.D. Dissertation, Colorado State University. FHWA (1997). Highway summary statistics to 1995. (section V). Washington DC: Federal Highway Administration, US Department of Transportation. Hauber, A. R. (1980). The social psychology of driving behaviour and the trac environment: research on aggressive behaviour in trac. International Review of Applied psychology, 29, 461474. Hennessy, D. A., & Wiesenthal, D. L. (1997). The relationship between trac congestion, driver stress, and direct versus indirect coping behaviors. Ergonomics, 40, 348361. Hyde, J. S. (1984). How large are gender dierences in aggression? A developmental meta-analysis. Developmental

160

D. Shinar/Transportation Research Part F 1 (1998) 137160

Psychology, 20, 722736. Jonah, B. A. (1997). Sensation seeking and risky driving: a review and synthesis of the literature. Accident Analysis and Prevention, 29, 651665. Lawson, S. D. (1991). Red-light running: accidents and surveillance cameras. (AA/BCC-3), Basingstoke, England: AA Foundation for Road Safety Research and Birmingham City Council. As cited by Retting and Williams (1996). Martinez, R. (1997). Statement of the Honorable Ricardo Martinez, M. D., Administrator, National Highway Trac Safety Administration, before the U. S. House of Representatives, Committee on Transportation and Infrastructure, Subcommittee on Surface Transportation. Washington, DC, July 17. McMurray, L. (1971). Emotional stress and driving performance: the eects of divorce. Behavioral Research in Highway Safety, 1, 100114. NHTSA. (1996). Trac Safety Facts. (Report No. DOT SH 808 471). Washington DC: National Highway Trac Safety Administration, U. S. Department of Transportation. Parry, M. H. (1968). Aggression on the Road. London: Tavistock. Preusser Research Group, Inc. (1998). Capital beltway update: beltway user focus groups. (NHTSA Report DOT HS 808 705). Washington, DC: US Department of Transportation. Retting, R. A., & Greene, M. A. (1995). The inuence of trac signal timing on red-light running and potential vehicle conicts at urban intersections. Arlington, VA: Insurance Institute for Highway Safety. Retting, R. A., & Williams, A. F. (1996). Characteristics of red light violators: results of a eld investigation. Journal of Safety Research, 27, 915. Retting, R. A., Williams, A. F., & Greene, M. A. (1996). Red light running and sensible countermeasures: summary of research ndings. Arlington, VA: Insurance Institute for Highway Safety. Tillman, W. A., & Hobbs, G. E. (1949). The accident-prone automobile driver. American Journal of Psychiatry, 106, 321332. Turner, C. W., Layton, J. F., & Simons, L. S. (1976). Naturalistic studies of aggressive behavior: aggressive stimuli, victim visibility, and horn honking. Journal of Personality and Social Psychology, 31, 10981107. Whitlock, F. A. (1971). Death on the road. London: Tavistock. Williams, A. F. (1997). Causes and dangers of aggressive driving. Statement before the U. S. House of Representatives, Committee on Transportation and Infrastructure, Subcommittee on Surface Transportation. Washington, DC, July 17.