category: literaturenote citekey: prechteffectsdrivingangerdriver2017 title: Effects of driving anger on driver behavior – Results from naturalistic driving data authors: "Precht, Lisa; Keinath, Andreas; Krems, Josef F." year: 2017 date: 2017-00-00 2017 doi: 10.1016/j.trf.2016.10.019 publication: "Transportation Research Part F: Traffic Psychology and Behaviour" url: "https://doi.org/10.1016/j.trf.2016.10.019" zotero_key: F9VBTU3E zotero_storage: VMM749IP collections: transpordi_planeerimine folder: 001_artiklid firstAuthor: "Precht, Lisa"

status: converted

Transportation Research Part F

journal homepage: www.el sevier.com/locate/trf

Effects of driving anger on driver behavior – Results from naturalistic driving data ^q

Lisa Precht a,⇑ , Andreas Keinath ^a , Josef F. Krems ^b

^a BMW Group, 80788 München, Germany
^b Technische Universität Chemnitz, 09107 Chemnitz, Germany

article info

Article history: Received 11 July 2016 Received in revised form 21 October 2016 Accepted 24 October 2016 Available online 27 December 2016

Keywords: Driving anger Driver behavior Driver distraction Naturalistic driving Aggressive driving

abstract

There is a positive relationship between driving anger and near-crash or crash risk. However, it remains unclear if anger in fact contributes to traffic accidents and whether this happens due to cognitive overload or aggressive driving behaviors. This study investigated how anger affects driving behavior based on naturalistic driving data from the second Strategic Highway Research Program (SHRP 2). Ten-minute trip segments were analyzed in which drivers exhibited anger with regard to driving errors, violations, and aggressive expressions. This data was compared to a matched baseline consisting of the same drivers not exhibiting anger. Results showed that anger resulted in more frequent aggressive driving behaviors but did not increase driving error frequency. Anger consequently creates danger due to deliberate behaviors rather than because of cognitive overload. In congruence with this finding, only anger triggered by threats, provocations, and frustrations increased the frequency of deliberate infringements. In contrast, anger due to having conflicts with someone on the phone or with a passenger was not linked to any type of aberrant driving behavior. Finally, severe displays of anger were accompanied by more violations as compared to slight or marked anger.

1. Introduction

According to the World Health Organization in 2015, the number of deaths on the road globally plateaued at 1.25 million per year, making road traffic injuries a leading cause of death (Toroyan, 2015). A central factor in crash risk is road user behavior (Rowe, Roman, McKenna, Barker, & Poulter, 2015). Especially inattention and driver distraction due to secondary tasks are among the leading causes of motor vehicle accidents (e.g., Klauer, Dingus, Neale, Sudweeks, & Ramsey, 2006; Klauer et al., 2014). Distractions while driving can stem from a competition for visual processing, manual interference, or cognitive sources of distraction, such as speaking with a passenger, conversing on the phone, or in the context of voice control (Strayer, Turrill, Coleman, Ortiz, & Cooper, 2014). It is controversial how cognitive distraction affects driving performance. Some findings suggest that cognitive distraction may impair traffic safety as a result of a diversion of attention (Strayer et al., 2013). However, especially during long drives, it can also reduce the risk of driving errors caused by fatigue (Jellentrup, Metz, & Rothe, 2011). Therefore, it is not surprising that cognitively distracting tasks have been found to be asso-

E-mail address: Lisa.Precht@bmw.de (L. Precht).

^q The findings and conclusions of this paper are those of the authors and do not necessarily represent the views of the VTTI, SHRP 2, the Transportation Research Board, the National Academies, or the BMW Group, which funded the current work.

^⇑ Corresponding author.

Nomenclature

GLMM Generalized Linear Mixed Model LMM Linear Mixed Model

NDS Naturalistic driving study

SHRP 2 Second Strategic Highway Research Program VTTI Virginia Tech Transportation Institute

ciated with a decreased crash and near-crash risk in several naturalistic driving studies (Hickman, Hanowski, & Bocanegra, 2010; Klauer et al., 2006; Olson, Hanowski, Hickman, & Bocanegra, 2009).

1.1. Driving anger

A particularly interesting special case of cognitive distraction is driving anger. Driving anger is considered to be a situation-specific emotional construct comprised of anger-related feelings and thoughts that occur while driving (Nesbit & Conger, 2011). Cognitive distraction, in turn, is a withdrawal of attention from the processing of information necessary for the safe operation of a vehicle (Strayer et al., 2014). The feelings and thoughts that accompany driving anger are likely to withdraw attention from the driving related information processing. Driving anger can therefore constitute a mental impairment, particularly when it arises in connection with a conversation. Several studies indicate that anger is among the most important factors involved in unsafe driving (e.g., Dahlen, Martin, Ragan, & Kuhlman, 2005). In a survey by Joint (1995), about 90% of 526 car drivers reported that they had been victims of road rage incidents during the past year. Moreover, a recent analysis of naturalistic driving data from the second Strategic Highway Research Program (SHRP 2) found that driving while observably emotionally agitated, for example due to anger, increased the risk of a crash by 9.8 times compared with apparently alert, attentive and sober driving episodes (Dingus et al., 2016).

1.2. Anger and aberrant driving behavior

Even though several studies found a connection between driving anger and increased risk of near-crash or crash events (e.g., Deffenbacher, Deffenbacher, Lynch, & Richards, 2003; Dingus et al., 2016), it is not entirely clear how anger is associated with these situations. It is possible that anger does not even precede them but instead is a result of these events (Underwood, Chapman, Wright, & Crundall, 1999). On the other hand, if anger indeed contributes to traffic injuries, it is unclear whether this happens due to increased cognitive load, leading to unintentional failures, or because of aggressive driving and therefore deliberate unsafe driving behaviors.

It is highly relevant to examine the effects of driving anger on driving behavior, as the behaviors of road users explain much more variance of traffic accidents than do vehicle, roadway, or environment factors (Singh, 2015). In a survey investigating factors leading up to crashes, the driver was found to be the critical reason in about 74% of the crashes due to driving errors or violations (Singh, 2015). These driving behaviors can be referred to as ''aberrant driving behaviors" (Zhang, Chan, & Zhang, 2015, p. 124). Until now, aberrant driving behaviors have not received satisfactory attention in studies investigating the impact of driver distraction, particularly in naturalistic driving studies (Young, 2015). Indeed, these behaviors are a crucial factor threatening road safety (Zhang et al., 2015). Distraction risk estimates are likely confounded and erroneously increased when aberrant driving behaviors are not taken into account in driver distraction investigations (Young, 2015).

According to Reason (1990), in order to appropriately assess aberrant driving behaviors it is necessary to distinct between errors and violations. In fact, these behaviors appear to have different psychological origins (Reason, Manstead, Stradling, Baxter, & Campbell, 1990). Errors are defined as unwitting deviations of action from intention or the failure of planned actions to achieve their intended consequence (Reason, 1990; Reason et al., 1990). In contrast, violations are ''deliberate deviations from those practices believed necessary to maintain the safe operation of a potentially hazardous system" (Reason et al., 1990, p. 1316). Thus, in order to investigate the connection between driving anger, driving behavior and near-crash or crash risk, it is necessary to consider both the effect of anger on unintentional cognitive failures and on deliberate infringements of accepted principles of safe driving.

1.3. Anger and aggressive driving

There is evidence that angry drivers engage in deliberate unsafe driving behaviors. For example, Parker, Lajunen, and Stradling (1998) found in a questionnaire study that 89% of 270 drivers admitted committing violations such as chasing other road users or sounding the horn to communicate annoyance to other road users. Indeed, several studies noted that angry drivers engage more often in aggressive and dangerous driving behaviors (e.g., Berdoulat, Vavassori, & Sastre, 2013; Deffenbacher et al., 2003; Nesbit, Conger, & Conger, 2007; Zhang & Chan, 2016). These aggressive driving behaviors, in turn, have been shown to increase the crash risk by 11.1 times (Dingus et al., 2016).

However, aggressive behaviors are not necessarily a response to anger. According to Stangor (2011), there are two fundamental types of aggression which are caused by different psychological processes. Emotional or impulsive aggression is the result of negative emotions such as anger. Instrumental or cognitive aggression, however, is intentional, planned, and can occur without any connection to a particular emotion (Stangor, 2011). Under real driving conditions, it remains unclear if aggressive driving behaviors are normally triggered by anger or are merely used to obtain a goal.

1.4. Anger and cognitive overload

Regardless of the possible triggering of aggressive behavior, driving anger results in increased cognitive load. It may withdraw attention from the information processing necessary for the safe operation of a vehicle and could therefore cause driving errors as opposed to deliberate violations. Some researchers claim that the experience of anger impairs mental processes, for example, due to a loss of concentration while driving (e.g., Deffenbacher, Lynch, Oetting, & Yingling, 2001; Deffenbacher et al., 2003). In fact, research has demonstrated adverse effects of anger on driving performance, such as an increase in driving errors or the reaction time to hazardous events (Jeon, Walker, & Yim, 2014; Stephens, Trawley, Madigan, & Groeger, 2013; Zhang & Chan, 2016). On the other hand, previous investigations have reported weak or non-significant relationships between driving anger and errors (Nesbit et al., 2007; Zhang et al., 2015) or even reduced driver distraction as a result of angry as compared to neutral speech (Hsieh, Seaman, & Young, 2010). Against the background of the above-mentioned ongoing controversial discussion on the relationship between cognitive distraction and errors, the diverging findings regarding the effects of anger on mental processes are not surprising. Thus, it is necessary to examine the connection between driving anger and cognitive overload under genuine driving conditions.

1.5. Anger trigger and intensity

It has been suggested that the relationship between driving anger and driver behavior depends on different natures of anger or situational characteristics, respectively (Lajunen & Parker, 2001; Li, Yao, Jiang, & Li, 2014; Sullman, Gras, Cunill, Planes, & Font-Mayolas, 2007; Zhang et al., 2015). Zhang and Chan (in preparation) found that the 14 anger-provoking traffic scenarios of the short version of the Driving Anger Scale (Deffenbacher, Oetting, & Lynch, 1994), can be classified into three subscales according to their goal-blocking features. They concluded that there are distinct dimensions of anger and reported the three driving anger factors: ''hostile gesture", ''arrival-blocking" and ''safety-blocking". These anger triggers were found to have differing impacts on aberrant driving behaviors (Zhang et al., 2015). This finding indicates that the anger trigger variable moderates the interaction between anger and driver behavior. In previous studies, especially the frustration anger trigger was associated with aggressive driving (Li et al., 2014; Shinar, 1998; Zhang et al., 2015). According to the frustrationaggression hypothesis by Dollard, Miller, Doob, Mowrer, and Sears (1939), aggressive driving is a result of anger or frustration caused by thwarted goals. This assumption is consistent with the driving anger factors reported by Zhang and Chan (in preparation) which either block the goal to drive safely, the ''general survival goal" (Zhang et al., 2015, p. 125), or the arrival goal. These are considered the main goals in driving (Cnossen, 2000). It is therefore likely that provocations, frustrations, and threats are common triggers for anger in the driving context.

Additionally, naturalistic driving data shows that drivers spend a large proportion of time talking to a passenger or on the phone. Within the SHRP 2 dataset, 14.3% of the secondary tasks carried out during trips were interactions with a passenger; 3% were phone conversations (retrieved February 2, 2016). It is likely that these conversations can cause anger as well. Driving anger due to conflicts with someone over the phone or with a passenger is usually not linked to a driving related, thwarted goal. Therefore, its effects on driver behavior should differ from the effects of anger triggers due to provocations, frustrations, or threats.

Beyond anger triggers, according to Frijda, Ortony, Sonnemans, and Clore (1992), it is primarily the emotional intensity, and therefore the emotional impact of an event, that determines whether an emotion causes inappropriate behavior. According to the appraisal theory of emotions, it is the subjective assessment of a given situation – in terms of its consequences for well-being, the satisfaction of goals, motives, or concerns – that determines whether an emotion is elicited or not, as well as its quality and intensity (Frijda, 1986; Lazarus, 1991). It is likely that the anger intensity variable moderates the relationship between anger and driver behavior as well. However, it is possible that the intensity of anger, which is accessible via video coding, does not always correspond to the driver's actual experienced emotion. In a study by Mesken, Hagenzieker, Rothengatter, and de Waard (2007), participants indicated their experienced emotions and the corresponding emotional intensity during a test drive. Afterward, the drivers' facial expressions were coded in video recordings of the respective participants. Mesken et al. (2007) found no association between the stated emotional strength and the coded facial expressions. Nevertheless, in order to fully understand the implications of driving anger under genuine driving conditions, it is necessary to investigate the interaction between anger and behavior by taking anger intensity as well as anger triggers into consideration.

1.6. Driving anger – investigative approaches

It is important to note that most of the previous findings in regard to anger and driving outcomes stem from questionnaire studies. A recent meta-analysis by Zhang and Chan (2016) investigated the relationships between driving anger and different types of driving outcomes in 51 studies published over the past two decades. Approximately 88% of these studies were questionnaire-based, whereas five studies employed driving simulation and one study was conducted in the field. Given that driving anger and its effects on driver behavior have never been investigated in-depth based on naturalistic driving data, the relationship between real experienced anger and the actual accompanying driver reactions under genuine driving conditions remains unclear.

In comparison to driving simulation and field experiments, naturalistic driving studies (NDSs) offer researchers a unique view into the vehicle and an invaluable opportunity to study driver performance and behavior in the real world with actual consequences and risks (Carney, McGehee, Harland, Weiss, & Raby, 2015). The largest and most comprehensive naturalistic driving study to date is the second Strategic Highway Research Program (SHRP 2), therefore representing the current gold standard for driver distraction data. The analysis of this data is essential to improve the understanding of cognitive distraction in general and the effects of anger on driver behavior in particular.

1.7. The current study – aims and hypotheses

The aim of the present investigation was to examine how driving anger, anger triggers, and anger intensity affect driver behavior based on naturalistic driving data. Therefore, ten-minute trip segments of SHRP 2 video footage were analyzed in which drivers exhibited driving anger with regard to driving errors, violations, and aggressive expressions. This data was compared to a matched baseline consisting of the same drivers not exhibiting anger. A main goal of this study was to determine whether anger is linked to driving errors and thus an increase in cognitive load, or to deliberate unsafe driving behaviors and therefore aggressive driving. Consistent with previous research, our first hypothesis was that (1) anger would increase the frequency of aggressive expressions. Moreover, we assumed that (2) anger would be associated with deliberate unsafe driving behaviors, such as violations, in contrast to unintentional driving errors.

Furthermore, this study aimed at investigating the reasons for anger under real driving conditions and the potentially different impacts of these anger triggers on driving behavior. We expected that (3) we would find anger triggers pursuant to the three driving anger factors suggested by Zhang and Chan (in preparation): provocation (hostile gesture), frustration (arrivalblocking), and threat (safety-blocking). In addition, anger due to conflicts with a passenger or on the phone was anticipated. We hypothesized that (4) the different anger triggers would result in varying driver behaviors: only anger in response to a thwarted driving goal was expected to be associated with driving violations. In order to investigate genuine anger triggers, it was necessary to not only analyze driving anger effects based on naturalistic driving data but to analyze relatively long segments of that data, too. Most studies working with naturalistic driving data analyzed trip segments lasting less than a minute (e.g., Dingus et al., 2016; Hickman et al., 2010; Klauer et al., 2006; Olson et al., 2009; Young, 2015). Indeed, in order to assign a certain anger trigger to exhibited anger, it is often necessary to observe the driver and his/her surroundings for a longer period of time. Therefore, in the present study, trip segments lasting for ten minutes were coded, making it possible for the first time to examine the impact of real anger triggers under natural driving conditions.

Moreover, the influence of observed anger intensity on the interaction between anger and driver behavior was investigated. Against the background of the earlier discussed findings by Mesken et al. (2007), we predicted that (5) there would not be a significant impact of anger intensity on the interaction between anger and driver behavior.

Finally, it was examined whether observed aggressive driving behaviors were triggered by anger and thus emotional/ impulsive aggressions or, in contrast, triggered by instrumental aggressions without the connection to a particular emotion. We hypothesized that (6) emotional aggressions would occur more frequently than instrumental aggressions.

2. Method and materials

2.1. Data

In this study, naturalistic driving data from the second Strategic Highway Research Program (SHRP 2) was analyzed. It is the largest study of its kind, including data from 50 million vehicle miles and 5.4 million trips. Data from instrumented vehicles was collected from more than 3500 participants during a 3-year period in the U.S. (Dingus et al., 2016; Victor et al., 2014). Recorded data consisted of driving parameters such as speed, acceleration, and braking, all vehicle controls, forward radar, and lane position. In addition, video views forward, to the rear, on the driver's face, and on the dashboard were captured (Blatt et al., 2015). The Virginia Tech Transportation Institute (VTTI) developed reduced or aggregated files to provide easier access to the SHRP 2 naturalistic driving study data. The VTTI's video data reduction allows filtering events according to comprehensive event data. This data includes, for example, driver behaviors and impairments, secondary tasks, and environmental/situational factors. The video reduction data, as defined in the SHRP 2 event data dictionary (available at https:// insight.SHRP2nds.us/), is provided for six-second data segments at events of interest. Such events include crashes, nearcrashes (e.g., defined by sudden braking or steering), and baseline events (defined in various ways: random, match of driver;

Fig. 1. Data and sampling procedure summary.

Hedlund, 2013). A more detailed description of the data in SHRP 2 can be found at the study's data access website ''InSight". See Fig. 1 for a summary of the data and the sampling procedure.

2.1.1. Test and control group data

The subset of SHRP 2 data used in this study consists of ten-minute segments, extracted from 108 different trips. A trip segment was defined as the duration of the crash, near-crash, or baseline, plus/minus 5 min (beginning 5 min before the crash, near-crash, or baseline starts, and ending 5 min after the crash, near-crash, or baseline event). Fifty-four of these trip segments represent the test group. The trip segments of the test group include crash, near-crash, and baseline events selected based on the following data specifications (defined in the SHRP 2 event data dictionary):

(1) The trip duration is greater than or equal to 10 min.
(2) The locality is either ''Moderate Residential", ''Business/Industrial" or ''Interstate/Bypass/Divided Highway with no traffic signals".
(3) The traffic density is either ''Level-of-service A1: Free flow, no lead traffic", ''Level-of-service A2: Free flow, leading traffic present", ''Level-of-service B: Flow with some restrictions", or ''Level-of-service C: Stable flow, maneuverability and speed are more restricted".
(4) The driver impairment ''Angry" is present.

According to the SHRP 2 dataset, available from the study's data dissemination website, VTTI coded 103 events of ''Anger" in 95 different trips (retrieved February 24, 2016). The requested data specifications resulted in the 54 trip segments used as the test group in this study. The data specifications are based on the six-second trip segments for which event data is provided by VTTI. The locality and traffic density specifications were intended to ensure a certain level of homogeneity in the data. Nevertheless, it is important to reiterate that ten-minute trip segments were analyzed in this study.

The control group data consists of one matched trip segment for each test group trip segment. The control group segments were intended to be as similar as possible to the associated test group segments, the only difference being that emotional impairment was not present. Thus, VTTI matched the baselines individually and manually for each event. The baseline segments were matched for the following variables (defined in the SHRP 2 event data dictionary):

(1) Car driver (had to be the same).
(2) Number of front seat passengers.
(3) Number of rear seat passengers.
(4) Secondary task 1.
(5) Weather.
(6) Lighting.
(7) Surface condition.
(8) Traffic flow.
(9) Traffic density.
(10) Relation to junction.
(11) Locality.
(12) Intersection influence.
(13) Alignment.
(14) Construction zone.
(15) Weekday vs. Weekend.

As in the case of the test group selection, the ten-minute control group segments were matched based on the six-second trip segments for which event data is provided by VTTI.

2.1.2. Final data sample

The final data set used for analyses comprised (1) 54 ten-minute test group segments in which the car drivers exhibited anger and (2) 54 ten-minute matched control group segments of the same drivers without showing anger. The 108 trip segments stem from 38 different drivers. The video reduction data provided by VTTI is categorized according to its ''event severity", such as ''crash", ''near-crash", ''crash-relevant", and baseline. Pursuant to this categorization, the test group data included 24 near-crash, 3 crash, and 27 baseline events. The matched control group segments originated completely from baseline events or data that was not previously coded.

The 38 drivers (18 female) were between 16 and 79 years of age (N = 13 were between 16 and 24 years; N = 10 were between 25 and 34 years; N = 9 were between 35 and 59 years; N = 6 were between 60 and 79 years). Participants' vehicles included 20 ''Cars", 3 ''Pickup Trucks", 13 ''Crossover SUVs", and 2 ''Vans/Minivans". The participants stated that they drove approximately 5000–60,000 miles during the last year (M = 15797.30, SD = 11314.604). Moreover, they stated that they had been driving for 2–60 years (M = 20.41, SD = 17.186).

2.2. Variables and data coding

2.2.1. Dependent variables

Appendix A shows a list of the coding scheme used including all variables, definitions, and examples. The main dependent variable of this analysis was driver behavior. Driver behavior was operationalized through the variables: (1) driving error, (2) traffic violation, (3) unclassified unsafe driving behavior, (4) other driver behavior, and (5) aggressive expression.

The majority of driver behaviors were coded on the basis of the original ''SHPR 2 Researcher Dictionary for Video Reduction Data" (Version 3.4, pp. 46–51; available at https://insight.SHRP2nds.us/). The driver behavior categories ''Aggressive driving, specific, directed menacing actions" and ''Aggressive driving, other" were replaced by 15 aggressive expression variables, based on the 49-item Driving Anger Expression Inventory (DAX; Deffenbacher, Lynch, Oetting, & Swaim, 2002). Only clearly observable items of the DAX were used. These variables were included to allow for a more detailed evaluation of aggressive driving behavior than what had previously been possible with the original SHRP 2 coding scheme.

2.2.2. Independent variables

Independent variables consisted of driving anger, anger intensity, and anger trigger. In addition, the apparent reason for aggressive expression was recorded in order to classify the aggressive expressions as either emotional/impulsive or instrumental/cognitive (see Appendix A for the variable definitions). Driver emotions and emotional intensity were coded based on a coding scheme developed by Fitch, Hanowski, Burnett, and Crundall (2013; see Tables 1 and 2). Categories of apparent reasons for the exhibited emotions included: (1) Threat to safety, (2) Frustration/Thwarted goals, (3) Driving behavior perceived as provocative, (4) Talking to passenger/on the phone, (5) Other reason, and (6) No apparent reason. These triggers were generated based on an extensive literature research and comply with the three-factor model (safety-blocking, arrival-blocking, and hostile gesture) of anger provoking events (based on the 14 anger-provoking traffic scenarios of the short version of the Driving Anger Scale; Deffenbacher et al., 1994) reported by Zhang and Chan (in preparation). Furthermore, the category ''Talking to passenger/on the phone" was included given the large proportion of time drivers spent engaging in those tasks while driving. It is likely that conversations can be triggers for anger as well.

2.2.3. Control variables

Given the control group trip segments were matched only on the basis of six-second sections for which event data was already provided, several control variables were coded as well. These control variables allowed for the comparison of the test and control group data regarding influencing factors. To assure a high consistency between the data groups, the following control variables were coded: (1) traffic density, (2) intersection influence, (3) lighting, (4) weather, and (5) surface condition. All control variables were coded on the basis of the original ''SHPR 2 Researcher Dictionary for Video Reduction Data" (Version 3.4, available at https://insight.SHRP2nds.us/).

2.2.4. Data coding procedure

For each ten-minute trip segment, four video views were coded: (1) video view forward, (2) to the rear, (3) on the driver's face, and (4) a view of the driver's interactions with the steering wheel and center console of the vehicle. The video coding took place at VTTI's secure data enclave. A data reduction software tool, developed by VTTI, was used. The data was coded in collaboration with a professional VTTI data reductionist. The start and end time of each dependent, independent, and control variable were coded. Each rater coded different variables, and therefore both raters coded all segments. The professional VTTI data reductionist coded in particular the original SHRP 2 driver behavior categories.

Inter-rater reliability of the coding schemes used to measure emotion and aggression related variables was examined. Accordingly, two raters independently coded the variables of driver emotion, emotion intensity, emotion trigger, aggressive expression, and apparent reason for aggressive expression before the actual coding started. Cohen's kappa was calculated based on the emotion and aggression variables, which were coded independently in seven ten-minute trip segments (6.5 percent of total segment sample). The resulting kappa values indicated an appropriate level of reliability (see Table 3).

Following the data reduction, coded driver behaviors (except for aggressive expressions) were categorized according to Reason et al.'s (1990) differentiation between a violation and an error. The classification of the aberrant driving behaviors as either deliberate or a misjudgment was mostly straightforward with the aid of the video recordings. When a driver behavior was considered unsafe but could not be assigned according to Reason et al.'s (1990) definition to either violation or error, it was categorized as unclassified unsafe driving behavior (behavior was neither error nor violation but still unsafe, or it was not possible to assess whether behavior was deliberate or a misjudgment). Driver behaviors not identified as a violation, error, or unsafe, were categorized as other driver behavior and excluded from the analysis.

Aggressive Expressions were coded according to three categories as specified by the DAX (Deffenbacher et al., 2002): (1) ''Verbally Aggressive Expression", (2) ''Physically Aggressive Expression", and (3) ''Using the Vehicle for Aggressive Expression". However, if an aggressive expression was considered a violation as defined by Reason et al. (1990) and was not already a part of the other coded driver behaviors, it was additionally categorized as a violation. See Appendix B for a list of the behaviors which were effectively coded in the dataset and their corresponding assignments to the driver behavior categories.

For the analyses, the total number of driver behaviors – errors, violations, and unclassified unsafe driving behaviors – per trip segment was collected. The same method was applied for the variables anger intensity and anger trigger. Each driving anger occurrence in the trip segments was categorized according to its intensity and apparent trigger. An anger occurrence started when the driver exhibited one or more of the operational definitions of anger and, in addition, the driving context clearly indicated that these facial expressions emerged due to anger or frustration and not because of, for example, a concentrated driving maneuver. The anger occurrence ended when the corresponding facial expressions were no longer exhibited, which necessitated the coding of another emotion. However, when the anger intensity as defined in the emotional intensity reduction categories (see Table 2) changed throughout a single anger occurrence, two or more anger occurrences were coded, pursuant to the respective anger intensities.

2.3. Data analysis

Prior to the main analysis, checks of the test and control groups were performed. Linear Mixed Models (LMM) were applied to ensure that drivers exhibited more anger in test group trip segments compared to the control group segments. The total time in seconds was compared in which the emotion variables were exhibited throughout the trip segments. Statistical calculations were performed in IBM SPSS v.20.0. (IBM Corp. Released., 2011). A Kolmogorov–Smirnov test and a visual inspection were carried out to test normal distribution of the emotion variables (anger/frustration, neutral/no emotion shown, other emotions). None of the variables was normally distributed, therefore, the dependent variables were rank-

Table 1 Driver emotion reaction definitions developed by Fitch et al. (2013).

| Emotion | Operational definition | | | |--------------------------|-------------------------------------------------------------------------------|--|--| | Unable to determine | Cannot tell what emotion the driver is showing | | | | Neutral/no emotion shown | The driver has a straight face, does not smile or laugh, does not gesture | | | | Happy | The driver smiles or laughs | | | | | The driver gestures in excitement | | | | Angry/frustrated | The driver lowers/squeezes eyebrows, wrinkling forehead | | | | | The driver clenches his/her teeth | | | | | The driver yells (opens mouth wide with eyebrows lowered) | | | | | The driver gestures in anger/frustration | | | | | The driver raises his/her upper lip or tightens lips | | | | Sad | The driver has droopy eyebrows (raises inner eyebrows, lowers outer eyebrows) | | | | | The driver frowns by lowering the outer corners of his/her lips | | | | Surprised | The driver's eyebrows raise | | | | | The driver's mouth opens | | | | | The driver gestures | | | | Other | Emotional reaction that does not fit into any other category | | | | | Please define in the Notes section on the Excel Log | | |

Table 2 Driver emotional intensity reduction definitions developed by Fitch et al. (2013).

Intensity of emotion	Operational definition
Unable to determine	Cannot tell the intensity of the emotion
Neutral/no emotion shown	The driver has a straight face, does not smile or laugh, does not gesture
Slight (emotion somewhat shown)	The driver no longer has a straight face
	However, no gesturing or head movement is observed
Marked or pronounced (emotion very much shown)	The driver no longer has a straight face
	The driver gestures one time in a reserved manner
	The driver moves his head one time
Severe (emotion extremely shown)	The driver has wide eyes and a wide open mouth
	The driver is screaming
	The driver gestures wildly, or the driver moves his head frequently

Table 3 Inter-rater reliabilities of the emotion and aggression related variables.

Variable	Sample size of comparisons	Cohen's kappa value
Driver emotion	N = 34	k = 0.82
Emotional intensity	N = 31a	k = 0.81
Emotion trigger	N = 31a	k = 0.71
Aggressive expression	N = 46	k = 0.72
Reason for aggressive expression	N = 38b	k = 0c

^a Inter-rater reliability of emotional intensity and emotion trigger could only be calculated from the consistently coded emotions by both raters.

transformed and the LMMs were applied on the rank-transformed data (Conover & Iman, 1981). Ties occurring during the rank transformation were replaced with their mean rank. As at least two trip segments were sampled from the same driver, this variable was treated as a random effect to take the autocorrelation of emotion variables within driver into account. Fixed ''test group" (categorical variable with reference group ''control group") effects for all models were tested to analyze whether the test and control group drivers differed substantially with respect to the duration of anger demonstrated (showing of no emotion, respectively). The same procedures were used to verify whether other emotions and the control variables were evenly distributed in the test and control groups.

To evaluate the effects of driving anger on driver behaviors, the frequency of driver behaviors (driving errors, unclassified unsafe driving behaviors, traffic violations, and aggressive expressions) was analyzed as a function of test and control group, anger intensity, and anger trigger, respectively. As a result, the dependent variables were discrete counts and the use of a normal distribution was not adequate (Coxe, West, & Aiken, 2009; Gardner, Mulvey, & Shaw, 1995; Long, 1997). Typical distributions for count data include the Poisson and negative binomial distribution. When the dataset contains many zeros, more than expected based on either the Poisson or the negative binomial distribution, it is ''zero-inflated" (Zuur, Ieno, Walker, Saveliev, & Smith, 2009). In this case, the zero-inflated Poisson or the zero-inflated negative binomial model is appropriate. Additionally, in the current study, at least two trip segments were sampled from the same driver. Driver behaviors exhibited by the same driver may thus be correlated and violate the independence assumption of traditional regression models (Hubbard et al., 2010). To take account of the Poisson/negative binomial distributed data with repeated measures, Generalized Linear Mixed Models (GLMM) with a log link function were applied. The application of GLMM techniques allows to address elements such as nested and zero inflated count data (Zuur et al., 2009). They combine the properties of two statistical frameworks – linear mixed models (which incorporate random effects and therefore the autocorrelation of driver behavior within each driver in the current study) and generalized linear models (which analyze non-normal data by using link functions, such as Poisson or negative binomial distributed data; Bolker et al., 2009).

The glmmADMB package (Fournier et al., 2012; Skaug, Fournier, Bolker, Magnusson, & Nielsen, 2016) in R 3.2.3 (R Core Team, 2015) was used to investigate how each type of driving behavior was predicted by the driving anger variables. The response variable was driver behavior (error, violation, unclassified unsafe driving behavior, aggressive expression). Separate models were fit for each of the six driver behavior variables. A Poisson distribution was used for the outcome variables error and unclassified unsafe driving behavior. The outcome variables verbally aggressive expression, physically aggressive expression, and using vehicle for aggressive expression followed a negative binomial distribution. The variable violation followed a zeroinflated negative binomial distribution. Driver was treated as a random effect to take into account the autocorrelation of driver behavior within driver. Fixed effects were either test/control group, anger intensity (slight, marked, severe), or anger trigger (threat to safety, frustration/thwarted goals, driving behavior perceived as provocative, talking to passenger/on the phone).

Finally, a data analysis was conducted to exclude the possibility that anger effects were confounded as a result of the data sampling procedure. The test group segments (anger group segments) stem from 24 near-crash, three crash, and 27 baseline

^b Inter-rater reliability of reason for aggressive expression could only be calculated from the consistently coded aggressive expressions by both raters.

^c High agreement (p⁰ = 0.95) with simultaneous high expected (chance) agreement (p^e = 0.95).

events (see Fig. 1). The control group segments, on the other hand, stem completely from baselines/not previously coded data and therefore do not include near-crash or crash-events. Only the ''SHRP 2 baseline events" were selected randomly from the raw naturalistic driving data set as the near-crash/crash events were specifically sought out in the SHRP 2 data in order to make them assessable to researchers (see Section 2.1). Therefore, as an additional verification of driving anger effects on driver behavior, anger effect analyses were conducted based exclusively on the randomly sampled test group and respective control group segments. These data samples did not include near-crash or crash events and consisted of a test group (anger group) including 27 ten-minute segments and the respective 27 matched control group segments from 22 different drivers. Apart from the different trip segment samples, data analyses were conducted accordingly to the above described approach.

3. Results

Fig. 2 shows the descriptive statistics for the driver behavior variables as a function of the test and control groups. There were more violations (M = 4.02, SD = 4.95), unclassified unsafe driving behaviors (M = 0.63, SD = 1.12), other driver behaviors (M = 0.20, SD = 0.45), and aggressive expressions (M = 4.39, SD = 5.12) in the test group data compared to the control group data (M = 1.85, SD = 3.12; M = 0.33, SD = 0.75; M = 0.09, SD = 0.35; M = 0.39, SD = 1.02). Driver errors were relatively evenly distributed between the test group (M = 0.46, SD = 0.88) and the control group (M = 0.54, SD = 1.08).

Overall, in 39 trip segments (16 in test group) zero violations occurred, in 31 trip segments (10 in test group) one to two violations occurred, in 16 trip segments (11 in test group) three to five violations occurred, in 16 trip segments (13 in test group) six to ten violations occurred, and in 6 trip segments (four in test group) eleven or more violations occurred (M = 2.94, SD = 4.26). Furthermore, there were 77 trip segments (40 in test group) with zero errors, 25 trip segments (11 in test group) with one to two errors, and six trip segments (three in test group) with three or more errors (M = 0.5, SD = 0.98). There were 78 trip segments (35 in test group) with zero unclassified unsafe driving behaviors, 25 trip segments (16 in test group) with one to two unclassified unsafe driving behaviors, and five trip segments (three in test group) with three or more (M = 0.48, SD = 0.96). In 58 trip segments (15 in test group) zero aggressive expressions occurred, in 17 trip segments (eight in test group) one to two aggressive expressions occurred, in 17 trip segments (16 in test group) three to five aggressive expressions occurred, in 11 trip segments (10 in test group) six to ten aggressive expressions occurred, and in five trip segments (all five in test group) eleven or more aggressive expressions occurred (M = 2.39, SD = 4.19).

In the analyzed data set, 39 near crash and three crash events were observed. Of these 39 near crash events, 16 events were deliberate ''brake checks", executed by either the observed driver or another road user driving in front of the observed driver. Two of the three crash events involved hitting the curb.

3.1. Test and control group checks

3.1.1. Emotions in test and control group

Three LMMs were carried out to ensure that drivers exhibited more anger/frustration in the test group trip segments compared to the control group segments. The total time in seconds in which the emotion variables were exhibited throughout the trip segments was compared. The results showed a significant difference between the two groups with respect to the duration of exhibited anger (F(1, 85.16) = 148.21, p < 0.001; see Fig. 3). Thus, it was not surprising either that there was a significant difference in the duration of the exhibition of no emotion (F(1, 75.97) = 118.7, p < 0.001). The presentation of other emotions, such as happy, sad, and surprised, did not differ between the test and control group (F(1, 64.68) = 1.37, p = 0.246).

Fig. 2. Total number of coded driver behaviors as a function of test and control group.

Fig. 3. Mean demonstration of driver emotions as a function of test and control group. Standard error bars are included.

Table 4 Results of LMM analyses for models predicting the control variables.

Model		Fixed effect	F-value (df)	p-value
Traffic density	Level-of-service A1 to B	Test groupa	1.519 (1, 84.422)	0.221
	Level-of-service C to D	Test groupa	1.559 (1, 83.734)	0.215
	Level-of-service E to F	Test groupa	2.123 (1, 91.345)	0.149
Intersection influence	Traffic signal	Test groupa	0.016 (1, 68.550)	0.901
	Uncontrolled	Test groupa	1.122 (1, 65.944)	0.293
	Interchange	Test groupa	0.345 (1, 69.112)	0.559
	Parking lot	Test groupa	3.091 (1, 67.761)	0.083
	No intersection influence	Test groupa	0.100 (1, 67.203)	0.752
Lighting	Dawn	Test groupa	1.000 (1, 106)	0.320
	Daylight	Test groupa	0.813 (1, 69.691)	0.370
	Dusk	Test groupa	0.143 (1, 66.692)	0.706
	Darkness	Test groupa	0.139 (1, 73.361)	0.711
Weather	No adverse conditions	Test groupa	0.151 (1, 88.383)	0.698
	Mist/light rain	Test groupa	0.000 (1, 106)	0.990
	Raining	Test groupa	0.230 (1, 106)	0.633
Surface condition	Dry	Test groupa	0.153 (1, 85.527)	0.697
	Wet	Test groupa	0.001 (1, 87.067)	0.981
	Snowy	Test groupa	1.206 (1, 52.182)	0.277

^a The reference group for the categorical variable ''test group" is ''control group".

3.1.2. Control variables in test and control group

In order to verify that the control variables were evenly distributed in the test and control groups, 18 LMMs were implemented. The total time in seconds in which the control variables occurred throughout the trip segments was compared. The results showed that there were no significant differences in the control variables (see Table 4). Hence, the test and control group drivers were equally exposed to traffic density, intersection influence, lighting, weather, and surface condition.

3.1.3. Driving anger effects on aggressive expressions

Of all the aggressive expressions coded, 143 were categorized as a verbally aggressive expression, 71 as a physically aggressive expression, and 44 as a use of vehicle for aggressive expression. Anger was coded 253 times as the apparent reason for aggressive expressions. Other reason was coded four times and no apparent reason once.

Three GLMMs were conducted to investigate how aggressive expressions were predicted by the test and control groups and therefore by the demonstration of driving anger. Dependent variables included aggressive expression frequencies displayed by the drivers, such as: verbally aggressive expressions, physically aggressive expressions, and using vehicle for aggressive expressions. The findings from the corresponding GLMM models are presented in Table 5. For verbally aggressive expressions, there was a significant effect for group condition (z = 7.32, p < 0.001), indicating that there were 9.35 times as many verbally aggressive expressions displayed in the test group segments (or anger group segments) as in the control group segments on average. There was also a significant effect for physically aggressive expressions (z = 4.84, p < 0.001), suggesting

Table 5 Results of GLMM analyses for models predicting the aggressive expression variables.

Model	Fixed effect	Estimate Exp(b); incident rate ratio	SE	z-value
VerballyAEa	Test groupd	9.350	0.306	7.32
PhysicallyAEb	Test groupd	18.876	0.607	4.84
VehicleAEc	Test groupd	14.488	0.641	4.17

^a Verbally aggressive expression.
^b Physically aggressive expression.
^c Use of vehicle for aggressive expression.
^d The reference group for the categorical variable ''test group" is ''control group".

that there were 18.88 times as many physically aggressive expressions displayed in the anger group as in the control group on average. Finally, for using vehicle for aggressive expressions, there was a significant effect for group condition as well (z = 4.17, p < 0.001), implying that there were 14.49 times as many aggressive expressions with use of the vehicle displayed in the anger group as in the control group.

The incident rate ratios of verbally and physically aggressive expressions, as well as use of vehicle for aggressive expressions predicted by the test and control groups have to be interpreted with caution. Firstly, it is likely that the three categories of aggressive expressions affected each other. Secondly, the operational definitions of the verbally and physically aggressive expressions partly overlap with the definition of driving anger in the current study. Anger was coded, inter alia, when the driver yelled (verbally aggressive expression) and when the driver ''gestured in anger/frustration" (physically aggressive expression). Therefore, these variables are partially confounded and can be considered another manipulation check of the anger distribution within the test and control groups.

3.2. Driving anger effects on errors and unclassified unsafe driving behaviors

Two GLMMs were conducted to investigate how errors and unclassified unsafe driving behaviors were predicted by the test and control groups and therefore by exhibited driving anger. Dependent variables consisted of the frequencies of coded driving errors and unsafe driving behaviors in the data. The findings from the corresponding GLMM models are presented in Table 6. Regarding driving errors, there was not a significant effect for group condition (z = 0.54, p = 0.59). This result indicates that the anger demonstrated while driving does not lead to more errors compared to a similar, but ''emotionally neutral" driving situation. For unclassified unsafe driving behaviors, there was a significant effect for group condition (z = 2.18, p = 0.03), suggesting that drivers displayed 1.89 times as many unsafe driving behaviors in the anger group as in the control segments on average.

3.2.1. Anger intensity and trigger effects on errors and unclassified driving behaviors

Four GLMMs investigating whether anger intensity or trigger predicted errors or unclassified driving behaviors showed no significant effects.

3.3. Driving anger effects on violations

A GLMM was conducted to investigate how test and control groups and thus exhibited driving anger predicted violations. The dependent variable was the frequency of violations in the trip segments. The findings from the GLMM models regarding the variable violations are presented in Table 7. In terms of violations, there was a significant effect for group condition (z = 4.85, p < 0.001), suggesting that drivers committed 2.21 times as many violations in the anger group as in the control group on average.

3.3.1. Anger trigger effects on violations

A GLMM was conducted to investigate how an anger trigger considered either a threat to safety, frustration/thwarted goal, driving behavior perceived as provocative, or talking to passenger/on the phone predicted violations (the anger triggers other reason and no apparent reason were excluded from the analyses.) Regarding the analyzed anger triggers, violations were best described by the anger trigger threat (z = 2.96, p = 0.003). This result suggests that each anger trigger categorized as threat

Table 6 GLMM models predicting errors and unclassified unsafe driving behaviors as a function of test and control group.

Model	Fixed effect	Estimate Exp(b); incident rate ratio	SE	z-value	p-value
Errors	Test groupb	0.862	0.273	0.54	0.587
UnsafeDBa	Test groupb	1.889	0.291	2.18	0.029

^a Unclassified unsafe driving behavior.

^b The reference group for the categorical variable ''test group" is ''control group".

Table 7 Results of GLMM analyses for models predicting violations.

Fixed effect		Estimate Exp(b); incident rate ratio	SE	z-value	p-value
Test groupa		2.205	0.163	4.85	0.000
Anger intensity	Slight	1.081	0.114	0.68	0.498
	Marked	0.950	0.085	0.60	0.550
	Severe	1.198	0.069	2.64	0.008
Anger trigger	Threat	1.570	0.152	2.96	0.003
	Frustration	1.086	0.031	2.63	0.008
	Provocation	1.328	0.094	3.03	0.003
	Conversation	0.994	0.059	0.09	0.925

^a The reference group for the categorical variable ''test group" is ''control group".

Fig. 4. Frequencies of anger triggers.

increased violations by a factor of 1.57. There were also significant effects for the anger triggers frustration (z = 2.63, p = 0.008) and provocation (z = 3.03, p = 0.003), implying that each anger trigger categorized as frustration increased violations by a factor of 1.09 and of 1.33 when categorized as provocation. There was not a significant effect for the anger trigger conversation (z = 0.09, p = 0.93). See Fig. 4 depicting the descriptive statistics for the anger trigger variables.

3.3.2. Anger intensity effects on violations

Of all coded anger occurrences, 73 were categorized as slight anger intensity, 131 as marked, and 73 as severe. A GLMM was performed to investigate how these anger intensity categories predicted violations. Violations were best described by severe anger occurrences (z = 2.64, p = 0.008). This result indicates that each anger occurrence categorized as severe increased violations by a factor of 1.2. There were not any significant effects for the anger intensities slight (z = 0.68, p = 0.50) or marked (z = 0.60, p = 0.55).

3.4. Additional verification of driving anger effects on driver behavior

An additional data analysis was conducted to exclude the possibility that anger effects were confounded as a result of the data sampling procedure (see Section 2.3). Only the ''SHRP 2 baseline events" were selected randomly from the raw naturalistic driving data set. Therefore, eight GLMMs, corresponding to the preceding analyses, were conducted based exclusively on the randomly sampled test group and respective control group segments stemming from SHRP 2 baseline events. Therefore, these analyses constitute an additional verification of the above reported relationships between anger and driver behaviors. The data samples, which did not include near-crash or crash events, consisted of a test group (anger group) including 27 ten-minute segments and the respective 27 matched control group segments from 22 different drivers. In these segments, 81 violations (51 in test group) occurred (M = 1.5, SD = 2.26), 31 errors (17 in test group) occurred (M = 0.57, SD = 1.11), 14 unclassified unsafe driving behaviors (9 in test group) occurred (M = 0.26, SD = 0.62), and 55 aggressive expressions (47 in test group) occurred (M = 1.02, SD = 2.05). Of the coded anger occurrences, 45 were categorized as slight anger intensity, 84 as marked, and 38 as severe. Of the coded anger triggers, 85 were categorized as frustration, 6 as provocation, and 70 as conversation. There was no anger trigger categorized as threat in this data subset.

However, it is important to note that this data subset is not representative of the interaction between anger and aberrant driving behavior, as most of the near-crash and crash events in this study were directly related to driving anger. The findings are presented in Table 8.

Table 8 Results of GLMM analyses for the trip segments stemming exclusively from the SHRP 2 baseline events.

Model	Fixed effect		Estimate Exp(b); incident rate ratio	SE	z-value	p-value
Violations		Test groupa	1.700	0.230	2.31	0.021
Errors		Test groupa	1.214	0.361	0.54	0.591
UnsafeDBb		Test groupa	1.800	0.558	1.05	0.292
VerballyAEc		Test groupa	10.001	0.606	3.8	0.000
PhysicallyAEd		Test groupa	3.000	0.577	1.90	0.057
VehicleAEe		Test groupa	4.998	1.100	1.47	0.142
Violations	Anger triggerf	Frustration	1.066	0.0302	2.12	0.034
		Provocation	0.000	277,000	0.00	1.000
		Conversation	0.926	0.0784	0.99	0.323
Violations	Anger intensity	Slight	1.072	0.160	0.44	0.660
		Marked	0.980	0.083	0.24	0.810
		Severe	1.094	0.072	1.26	0.210

^a The reference group for the categorical variable ''test group" is ''control group".
^b Unclassified unsafe driving behavior.
^c Verbally aggressive expression.
^d Physically aggressive expression.
^e Use of vehicle for aggressive expression.
^f The anger trigger ''threat" did not occur in the SHRP 2 baseline event trip segments.

4. Discussion

The current study investigated how driving anger affects driver behavior based upon naturalistic driving data. Tenminute trip segments in which drivers exhibited driving anger were analyzed with regard to driving errors, violations, unclassified unsafe driving behaviors, and aggressive expressions. This data was compared to a matched baseline consisting of the same drivers not exhibiting anger/frustration. In summary, the current findings showed that anger resulted in more frequent aggressive driving behaviors without increasing the frequency of driving errors (see Tables 5–7). Therefore, even though anger may withdraw attention from the processing of driving relevant information, it creates danger due to deliberate behaviors rather than because of cognitive overload. This becomes even more evident in regards to the impacts of driving anger depending on the different anger triggers. Only demonstrated anger in response to a threat, frustration, or provocation significantly increased the frequency of aberrant driving behaviors in terms of deliberate infringements (see Table 7). In contrast, anger triggered by a conversation as opposed to a reaction to another driver was not linked to any type of aberrant driving behavior (see Table 7). The cognitive distraction, however, was expected to be highest when anger was caused by a conflict with a passenger or while on the phone. In these situations, drivers were emotionally impaired and experienced an additional demand due to the conversation. These results provide compelling evidence for the relationship between anger and deliberate dangerous driving behaviors targeted at other road users as opposed to errors due to cognitive distraction. The substantial increase in aggressive expressions in connection with demonstrated anger constitutes further support of this assumption (see Table 5).

4.1. Anger effects on aberrant driving behavior

The present results are in good agreement with previous driving anger investigations. For example, the meta-analysis by Nesbit et al. (2007) found stronger relationships between anger and violations, as well as weaker, non-significant relationships between anger and errors. Especially the correlation between anger and violations (e.g., Berdoulat et al., 2013; Deffenbacher et al., 2003; Zhang & Chan, 2016), in addition to the relationship between anger and aggressive expressions (e.g., Berdoulat et al., 2013; Deffenbacher et al., 2003; González-Iglesias, Gómez-Fraguela, & Luengo-Martín, 2012) has been shown in numerous studies. These results suggest that driving anger causes aberrant driving behaviors in the sense of deliberate infringements that are mostly aggressive and targeted at other road users, in contrast to driving errors. Angry drivers therefore use their vehicle in order to react to frustrations, threats or provocations due to other road users. They attempt to influence other drivers, either to achieve their driving goals or to take revenge for other road users' driving behaviors such as ''brake checks". As opposed to the results of the current study, simulator-based investigations have demonstrated adverse effects of anger on driving performance, such as an increase of driving errors (Jeon et al., 2014; Stephens et al., 2013). Moreover, a recent meta-analysis reported a significant zero-order correlation of 0.179 between driving anger and errors, based on three questionnaire studies (Zhang & Chan, 2016). However, previous investigations have also reported weak or nonsignificant relationships between driving anger and errors (Nesbit et al., 2007; Zhang et al., 2015) or even a reduced level of driver distraction as a result of angry compared to neutral speech (Hsieh et al., 2010). Indeed, driving errors seem to be more related to anxiety (Briggs, Hole, & Land, 2011; Matthews, 2002; Shahar, 2009) or stress (Kontogiannis, 2006) experienced by drivers rather than driving anger.

Furthermore, the operational definition of ''error" deviates strongly between different studies. Across several investigations, error definitions also contained deliberate infringements and hence driver behaviors categorized as violations as defined in the current study (e.g., Jeon et al., 2014). There does appear to be a distinction between the psychological origins of driving violations and errors (Reason et al., 1990). Therefore, a valid comparison of the present findings to previous study results which summarized aberrant driving behaviors regardless of error or violation is difficult and questionable.

More importantly, as shown in a recent meta-analysis, most of the previous findings in regard to anger and driving outcomes stem from questionnaire-based studies (Zhang & Chan, 2016). The present investigation is the first to analyze the relationship between driving anger and driver behavior using naturalistic driving data, thus providing compelling evidence that neither driving anger alone, nor in combination with a conversation, increases driving error frequency. Clearly fighting with a passenger or with someone on the phone constitutes an extreme case of cognitive distraction. It is remarkable that the combination of emotional impairment, speaking, and listening was not linked to any type of aberrant driving behavior while driving in real traffic. Thus, it is unlikely that either anger in particular or even cognitive distraction in general endangers the driver under real driving conditions as a result of an inadequate amount of remaining attentional resources. Deliberate dangerous driving behaviors as opposed to cognitive distraction were mainly responsible for the hazardous situations observed in this study.

4.2. Anger trigger effects on driving behavior

The relationship between driving anger and driving behavior depends on different natures of anger or situational characteristics respectively (Lajunen & Parker, 2001; Li et al., 2014; Sullman et al., 2007; Zhang et al., 2015). The anger triggers encountered in the current study corresponded to a large extent with the three-factor model suggested by Zhang and Chan (in preparation) and complemented the frequently coded anger trigger conversation (see Fig. 4). Furthermore, the present findings provide convincing evidence that the reason for feeling anger determines whether the driver will engage in hazardous driving behaviors. In contrast to anger expressed due to conflicts on the phone or with a passenger, anger triggers associated with thwarted driving goals, usually due to other road users, were likely to induce aggressive driving. The analysis of the respective video material clearly demonstrated that anger initiated by a threat, provocation or frustration was often associated with ''retaliations" in the form of aggressive behaviors against the driver who was the catalyst for the corresponding anger. The fact that a substantial proportion of the near-crash events in the coded data sample were deliberate ''brake checks" further supports this observation.

Apart from its significant impact on violations, frustration was the most frequently observed anger trigger by a considerable amount. This result corresponds to earlier findings (Li et al., 2014; Shinar, 1998; Zhang et al., 2015) and supports the frustration-aggression hypothesis (Dollard et al., 1939). According to this theory, driving aggression is a result of anger or frustration caused by thwarted goals. An important goal of drivers in traffic settings is to arrive at the destination. Furthermore, the main priority of drivers is the prevention of collisions and therefore to drive safely (Cnossen, 2000). Anger or frustration are likely to occur when these goals are blocked, for example, due to a slowly moving car ahead or threatening behavior from other traffic participants. This in turn can lead to aggressive behaviors, such as sounding the horn or blocking other drivers (Zhang & Chan, 2016). Even though the claim by Dollard et al. (1939) that frustration always leads to aggression may be an overstatement (Berkowitz, 1989; Tek, 2014), the frustration-aggression hypothesis provides an explanation for the anger trigger effects on violations in the current study. Only the (driving-related) goal-blocking anger triggers including threat (safety goal), provocation (general survival goal; Zhang et al., 2015), and frustration (arrival goal) were associated with an increase in violation frequency. Interestingly, Anderson and Bushman (2002) suggested that ''most provocations can be seen as a type of frustration in which a person has been identified as the agent responsible for the failure to attain the goal" (p. 37).

In addition to these assumptions, there are two fundamental types of aggression according to Stangor (2011) that are caused by different psychological processes – emotional/impulsive aggression and instrumental/cognitive aggression. The vast majority of observed aggressive expressions in the current study occurred in connection with anger. To this end, almost all coded aggressions were emotional or impulsive in contrast to instrumental or cognitive. The emotional form of aggression occurs as a reaction to a frustration or a perceived threat, which can lead to anger and may cause a driver to seek revenge through demonstrating aggressive behaviors, such as ''brake checking" another road user (Berdoulat et al., 2013). This claim is in strong congruence with the present findings. Aggressive driving behavior in the form of violations was best described by the anger trigger threat. Although this interaction was not systematically investigated in this study, it was observed that many coded violations occurred in the context of retaliations.

Contrary to the findings of the current study and the described theories, Zhang et al. (2015) found that anger caused by safety-blocking had a negative impact on deliberate violations and that the anger trigger hostile gesture was a non-significant predictor. On the other hand, Li et al. (2014) reported a correlation between anger toward hostile gestures and ''aggressive actions", which included violations and aggressive expressions. However, the current survey was the first to investigate driving anger effects on driver behavior in naturalistic driving data segments. Moreover, it was the first study to analyze naturalistic driving data segments lasting as long as ten minutes. Therefore, this approach allowed for the observation of real driver behavior in the context of the traffic environment as well as the examination of the impact of anger in regard to different genuine anger triggers for the first time. Most studies working with naturalistic driving data have analyzed trip segments lasting less than a minute (e.g., Dingus et al., 2016; Hickman et al., 2010; Klauer et al., 2006; Olson et al., 2009; Young, 2015). Indeed, in order to assign a certain anger trigger to an occurrence of anger, it is often necessary to observe the driver and his/her surroundings for a longer period of time.

4.3. Anger intensity effects on driving behavior

Importantly, the impact of anger on violations differed as a function of its intensity as well. Severe driving anger incidents were accompanied by more violations compared to slight or even marked anger. This finding could imply that it is necessary for driving anger to be of a certain strength in order to cause aggressive driving behaviors. This result is consistent with the claim by Frijda et al. (1992) that primarily emotional intensity determines whether an emotion causes inappropriate behavior. According to the appraisal theory of emotions, the subjective assessment (appraisal) of a given situation – in terms of its consequences for well-being, the satisfaction of goals, motives, or concerns – determines whether an emotion is elicited or not, as well as its quality and intensity (Frijda, 1986; Lazarus, 1991). Another interesting hypothesis in this context is that the experienced frustration is the primary determinant of the occurrence and intensity of anger rather than the appraisal of an event (Berkowitz & Harmon-Jones, 2004). In the vein of these assumptions, the present anger intensity effects suggest that anger triggers perceived as highly goal-blocking (safety goal, general survival goal, arrival goal), and therefore exceedingly frustrating, are particularly likely to increase the frequency of deliberate driving violations.

4.4. Driving anger effects on driver behavior after the exclusion of near-crashes and crashes

The results of the analyses of driving anger effects on driver behavior in trip segments, which did not include near-crash or crash events, are in good agreement with the findings discussed above.

Although this data subset was smaller in scale, the relationship between anger and deliberate dangerous driving behaviors remained significant as opposed to an effect on driving errors. The most striking difference between the complete data set and the ''near-crash/crash-free" data subset was, that the anger trigger threat did not occur in the latter case. This anger trigger was mainly coded in connection with abrupt braking maneuvers of preceding vehicles. Furthermore, for the anger trigger provocation there was no significant effect in the near-crash/crash free data subset in contrast to the complete dataset. Only approximately 4% of anger occurrences in the near-crash/crash free data subset were coded as provocations, compared to approximately 11% in the complete dataset. As described in Section 4.2 in more detail, anger initiated by a threat, provocation or frustration was often associated with ''retaliations" in the form of aggressive behaviors against the driver who was the catalyst for the corresponding anger. These retaliations were often equivalent to the behaviors which triggered the anger. This means that drivers who were ''brake checked" by another road user often attempted to do the same. In the same vein, drivers who felt provoked by the behavior of other road users often reacted accordingly.

The fact that the physically aggressive expression and use of vehicle for aggressive expression variables were not significantly increased in the anger group of the near-crash/crash-free data subset compared to the control group constitutes further support of this observation. On the one hand, this finding indicates that without the threat of abrupt braking maneuvers and fewer provocations these aggressive expression categories were less frequently triggered. On the other hand, drivers in the current study also provoked near-crashes, for example in the form of ''brake checks", by exhibiting these aggressive expressions toward other drivers. Anecdotally, one participant was ''giving the finger" to an overtaking vehicle, which in turn, briefly after this physically aggressive expression, crossed 3 lanes in order to almost ram the car of the study participant laterally.

Only the verbally aggressive expressions head-shaking, ''evil eye", and yelling, which can be considered the most ''cautious" of the aggressive expression categories, were significantly increased in the anger group of the near-crash/crash-free data subset. The only anger trigger which significantly increased violations in the near-crash/crash-free data subset was frustration. Moreover, in contrast to the complete dataset, there was no significant anger intensity effect on violations in the nearcrash/crash-free data subset. Interestingly, however, there was no notable difference in the distribution of the anger intensities between the two datasets: Approximately 23% of the anger occurrences coded in the near-crash/crash-free data subset were categorized as severe compared to 26% of anger occurrences in the complete dataset. Furthermore, marked and severe expressions of anger combined were almost identical in the two analyzed datasets (73% and 74%).

In summary, the results of the analyses of the near-crash/crash-free data subset provide evidence that the anger triggers threat and provocation as well as physically and use of vehicle aggressive expressions could be the main drivers behind deliberate dangerous driving behaviors which increase the risk of near-crashes and crashes – in contrast to the anger intensity, the anger trigger frustration, and verbally aggressive expressions. However, there was still a significant increase in violations in the near-crash/crash-free anger group compared to the control group.

4.5. Limitations

The study reported here has the methodological limitation that the matched baseline did not include any near-crashes or crashes in contrast to the test group (anger group). As the literature has repeatedly shown, near-crashes and crashes can occur unrelated to driving anger. It is therefore an important question, whether the results of the comparisons reported in the current study would have been different if the control group data also included crashes and near-crashes. Dingus et al. (2016), who performed the first large-scale, crash-only analysis of risk factors based on naturalistic driving data (SHRP 2 NDS), identified driver impairments, ''performance and judgment errors" (for the most part violations), as well as driver distractions as causal factors of crashes. These factors are associated with different aberrant driving behaviors. Driver distractions, for example, are particularly associated with errors, which can result in crashes (e.g., Staubach, 2009). On the other hand, other factors, such as driving intoxicated, are likely to cause crashes due to both errors and violations. However, as prevalent causal factors of crashes (e.g., fatigue, driver distractions; Dingus et al., 2016) are more closely connected to errors compared to violations, it is unlikely that an inclusion of near-crashes and crashes in the control group of the current study would have substantially affected the anger effects reported here, as these were mainly related to violations. Indeed, Dingus et al. (2016) found high crash risks in connection with violations, such as speeding or following too closely; however, it remains unclear if these violations occurred independently of driving anger.

Furthermore, interestingly enough, the significant influence of anger intensity on violations noted in this paper is contradictory to the observation of Mesken et al. (2007) that coded facial expressions did not comply with the participants' stated emotional strength. However, the current study incorporated a very different methodological approach compared to Mesken et al. (2007). It is possible that the drivers in the present study would have rated their experienced anger intensity differently than the intensity classifications made by the data reductionists. It was not possible to check this assumption in this investigation. Nevertheless, the anger intensity coding scheme used in this study differed from Mesken et al. (2007) approach. Apart from facial expressions, it included behaviors such as ''screaming" or ''wildly gesturing". It is likely that such emotional expressions would not be observed independently of strongly experienced emotions. On the other hand, the absence of these strong emotional expressions does not necessarily suggest that the person is not feeling marked or severe emotions. However, this is a problem that cannot be resolved readily in naturalistic driving studies.

Further work overall is needed in order to validate the findings suggested in this study. It is necessary to investigate larger samples of angry drivers in naturalistic driving data.

5. Conclusion

The present study analyzed how real anger occurrences affected driver behavior in ten-minute trip segments based on the largest naturalistic driving study undertaken thus far. To the best of the authors' knowledge, this study is therefore the first of its kind. Driving anger and the respective driving behaviors have never been analyzed previously in naturalistic driving video segments lasting as long as ten minutes. This investigation thus offers a special insight on the interplay between driving anger, driving behavior and other road users under real traffic conditions. Even though anger can constitute an extreme case of cognitive distraction, especially if originating from a conversation, it was not associated with an increase of driving errors in the present comprehensive video data analysis. Neither driving anger alone nor in combination with a conversation increased the error frequency under real driving conditions. Therefore, it is unlikely that anger in particular or cognitive distraction in general endangers the driver as a result of an inadequate amount of remaining attentional resources. In contrast, anger was associated with aggressive expressions and deliberate unsafe driving behaviors that were mostly targeted at other road users. Furthermore, the reason for feeling anger was a decisive factor in determining whether the drivers engaged in hazardous driving behaviors. Only the anger triggers in reaction to other road users and thwarted driving goals – threats, provocations, and frustrations – were associated with an increase in violation frequency. Even though conflicts with the passenger or with someone on the phone were frequently coded anger triggers in the present data sample as well, they were not linked to any type of aberrant driving behavior. Finally, severe demonstrations of driving anger were accompanied by more driving violations than slight or marked anger. Therefore, the impact of anger on violations also differed as a function of the anger intensity.

6. Funding

This work was supported by the BMW Group.

Acknowledgements

The authors would like to thank the employees of the Virginia Tech Transportation Institute for the very helpful cooperation.

Appendix A. Supplementary material

Supplementary data associated with this article can be found, in the online version, at http://dx.doi.org/10.1016/j.trf. 2016.10.019.

References

Anderson, C. A., & Bushman, B. J. (2002). Human aggression. Psychology, 53, 27–51.

Berdoulat, E., Vavassori, D., & Sastre, M. T. M. (2013). Driving anger, emotional and instrumental aggressiveness, and impulsiveness in the prediction of aggressive and transgressive driving. Accident Analysis & Prevention, 50, 758–767.

Berkowitz, L. (1989). Frustration-aggression hypothesis: Examination and reformulation. Psychological Bulletin, 106(1), 59.

Berkowitz, L., & Harmon-Jones, E. (2004). Toward an understanding of the determinants of anger. Emotion, 4(2), 107.

Blatt, A., Pierowicz, J., Flanigan, M., Lin, P.-S., Kourtellis, A., Jovanis, P., ..., & Hoover, M. (2015). Naturalistic driving study: Field data collection. SHRP 2 Report S2-S07-RW-1.

Bolker, B. M., Brooks, M. E., Clark, C. J., Geange, S. W., Poulsen, J. R., Stevens, M. H. H., & White, J.-S. S. (2009). Generalized linear mixed models: A practical guide for ecology and evolution. Trends in Ecology & Evolution, 24(3), 127–135. Retrieved from <http://avesbiodiv.mncn.csic.es/estadistica/curso2011/ regm26.pdf>.

Briggs, G. F., Hole, G. J., & Land, M. F. (2011). Emotionally involving telephone conversations lead to driver error and visual tunnelling. Transportation Research Part F: Traffic Psychology and Behaviour, 14(4), 313–323.

Carney, C., McGehee, D., Harland, K., Weiss, M., & Raby, M. (2015). Using naturalistic driving data to assess the prevalence of environmental factors and driver behaviors in teen driver crashes. Washington, DC: AAA Foundation for Traffic Safety.

Cnossen, F. (2000). Adaptive strategies and goal management in car drivers. Groningen: University of Groningen.

Conover, W. J., & Iman, R. L. (1981). Rank transformations as a bridge between parametric and nonparametric statistics. The American Statistician, 35(3), 124–129.

Coxe, S., West, S. G., & Aiken, L. S. (2009). The analysis of count data: a gentle introduction to Poisson regression and its alternatives. Journal of Personality Assessment, 91(2), 121–136.

Dahlen, E. R., Martin, R. C., Ragan, K., & Kuhlman, M. M. (2005). Driving anger, sensation seeking, impulsiveness, and boredom proneness in the prediction of unsafe driving. Accident Analysis & Prevention, 37(2), 341–348.

Deffenbacher, J. L., Deffenbacher, D. M., Lynch, R. S., & Richards, T. L. (2003). Anger, aggression, and risky behavior: A comparison of high and low anger drivers. Behaviour Research and Therapy, 41(6), 701–718.

Deffenbacher, J. L., Lynch, R. S., Oetting, E. R., & Swaim, R. C. (2002). The driving anger expression inventory: A measure of how people express their anger on the road. Behaviour Research and Therapy, 40(6), 717–737.

Deffenbacher, J. L., Lynch, R. S., Oetting, E. R., & Yingling, D. A. (2001). Driving anger: Correlates and a test of state-trait theory. Personality and Individual Differences, 31(8), 1321–1331.

Deffenbacher, J. L., Oetting, E. R., & Lynch, R. S. (1994). Development of a driving anger scale. Psychological Reports, 74(1), 83–91.

Dingus, T. A., Guo, F., Lee, S., Antin, J. F., Perez, M., Buchanan-King, M., & Hankey, J. (2016). Driver crash risk factors and prevalence evaluation using naturalistic driving data. Proceedings of the National Academy of Sciences, 113(10), 2636–2641.

Dollard, J., Miller, N. E., Doob, L. W., Mowrer, O. H., & Sears, R. R. (1939). Frustration and aggression. New Haven, CN: Yale University Press.

Fitch, G. M., Hanowski, R., Burnett, G., & Crundall, D. (2013). Development of a protocol to classify drivers' emotional conversation. In 3rd international conference on driver distraction and inattention.

Fournier, D. A., Skaug, H. J., Ancheta, J., Ianelli, J., Magnusson, A., Maunder, M. N., ... Sibert, J. (2012). AD Model Builder: Using automatic differentiation for statistical inference of highly parameterized complex nonlinear models. Optimization Methods and Software, 27(2), 233–249.

Frijda, N. H. (1986). The emotions: Studies in emotion and social interaction. Cambridge: Cambridge Univ. Press.

Frijda, N. H., Ortony, A., Sonnemans, J., & Clore, G. L. (1992). The complexity of intensity: Issues concerning the structure of emotion intensity. In M. Clark (Ed.). Emotion: Review of personality and social psychology (Vol. 13, pp. 60–89). Beverly Hills, CA: Sage.

Gardner, W., Mulvey, E. P., & Shaw, E. C. (1995). Regression analyses of counts and rates: Poisson, overdispersed Poisson, and negative binomial models. Psychological Bulletin, 118(3), 392–404.

González-Iglesias, B., Gómez-Fraguela, J. A., & Luengo-Martín, M. Á. (2012). Driving anger and traffic violations: Gender differences. Transportation Research Part F: Traffic Psychology and Behaviour, 15(4), 404–412.

Hedlund, J. (2013). SHRP 2 naturalistic driving study overview. In GHSA Annual Meeting. Retrieved from <http://www.ghsa.org/html/files/meetings/2013am/ pres/wed.hedlund.pdf>⁄ (Original work published August 28, 2013).

Hickman, J. S., Hanowski, R. J., & Bocanegra, J. (2010). Distraction in commercial trucks and buses: Assessing prevalence and risk in conjunction with crashes and near-crashes (Report No. FMCSA-RRR-10-049). Washington, DC: Federal Motor Carrier Safety Administration.

Hsieh, L., Seaman, S., & Young, R. (2010). Effect of emotional speech tone on driving from lab to road: fMRI and ERP studies. In Proceedings of the 2nd international conference on automotive user interfaces and interactive vehicular applications (pp. 22–28).

Hubbard, A. E., Ahern, J., Fleischer, N. L., van der Laan, M., Lippman, S. A., Jewell, N., ... Satariano, W. A. (2010). To GEE or Not to GEE. Epidemiology, 21(4), 467–474.

IBM Corp. Released (2011). IBM SPSS statistics for windows, version 20.0. Armonk, NY: IBM Corp.

Jellentrup, N., Metz, B., & Rothe, S. (2011). Can talking on the phone keep the driver awake? Results of a field-study using telephoning as a countermeasure against fatigue while driving. In Proceedings of the 2nd international conference on driver distraction and inattention. .

Jeon, M., Walker, B. N., & Yim, J.-B. (2014). Effects of specific emotions on subjective judgment, driving performance, and perceived workload. Transportation Research Part F: Traffic Psychology and Behaviour, 24, 197–209.

Joint, M. (1995). Road rage. Basingstoke: Automobile Association.

Klauer, S. G., Dingus, T. A., Neale, V. L., Sudweeks, J. D., & Ramsey, D. J. (2006). The impact of driver inattention on near-crash/crash risk: An analysis using the 100-car naturalistic driving study data (Report No. DOT HS 810 594). Washington, DC: National Highway Traffic Safety Administration.

Klauer, S. G., Guo, F., Simons-Morton, B. G., Ouimet, M. C., Lee, S. E., & Dingus, T. A. (2014). Distracted driving and risk of road crashes among novice and experienced drivers. New England Journal of Medicine, 370(1), 54–59.

Kontogiannis, T. (2006). Patterns of driver stress and coping strategies in a Greek sample and their relationship to aberrant behaviors and traffic accidents. Accident Analysis & Prevention, 38(5), 913–924.

Lajunen, T., & Parker, D. (2001). Are aggressive people aggressive drivers? A study of the relationship between self-reported general aggressiveness, driver anger and aggressive driving. Accident Analysis & Prevention, 33(2), 243–255.

Lazarus, R. S. (1991). Emotion and adaptation. New York: Oxford University Press.

Li, F., Yao, X., Jiang, L., & Li, Y. (2014). Driving anger in China: Psychometric properties of the Driving Anger Scale (DAS) and its relationship with aggressive driving. Personality and Individual Differences, 68, 130–135.

Long, J. S. (1997). Regression models for categorical and limited dependent variables advanced quantitative techniques in the social sciences (Vol. 7) Thousand Oaks: Sage Publications.

Matthews, G. (2002). Towards a transactional ergonomics for driver stress and fatigue. Theoretical Issues in Ergonomics Science, 3(2), 195–211.

Mesken, J., Hagenzieker, M. P., Rothengatter, T., & de Waard, D. (2007). Frequency, determinants, and consequences of different drivers' emotions: An onthe-road study using self-reports, (observed) behaviour, and physiology. Transportation Research Part F: Traffic Psychology and Behaviour, 10(6), 458–475. Nesbit, S. M., & Conger, J. C. (2011). Evaluation of cognitive responses to anger-provoking driving situations using the articulated thoughts during simulated

situations procedure. Transportation Research Part F: Traffic Psychology and Behaviour, 14(1), 54–65. Nesbit, S. M., Conger, J. C., & Conger, A. J. (2007). A quantitative review of the relationship between anger and aggressive driving. Aggression and Violent

Behavior, 12(2), 156–176. Olson, R. L., Hanowski, R. J., Hickman, J. S., & Bocanegra, J. L. (2009). Driver distraction in commercial vehicle operations (Report No. FMCSA-RRR-09-042). Washington, DC: Federal Motor Carrier Safety Administration.

Parker, D., Lajunen, T., & Stradling, S. (1998). Attitudinal predictors of interpersonally aggressive violations on the road. Transportation Research Part F: Traffic Psychology and Behaviour, 1(1), 11–24.

R Core Team (2015). R: A language and environment for statistical computing Vienna, Austria. Retrieved from https://www.R-project.org/.

Reason, J. (1990). Human error. Cambridge: Cambridge University Press.

Reason, J., Manstead, A., Stradling, S., Baxter, J., & Campbell, K. (1990). Errors and violations on the roads: A real distinction? Ergonomics, 33(10–11), 1315–1332.

Rowe, R., Roman, G. D., McKenna, F. P., Barker, E., & Poulter, D. (2015). Measuring errors and violations on the road: a bifactor modeling approach to the driver behavior questionnaire. Accident Analysis and Prevention, 74, 118–125.

Shahar, A. (2009). Self-reported driving behaviors as a function of trait anxiety. Accident Analysis & Prevention, 41(2), 241–245.

Shinar, D. (1998). Aggressive driving: the contribution of the drivers and the situation. Transportation Research Part F: Traffic Psychology and Behaviour, 1(2), 137–160.

Singh, S. (2015). Critical reasons for crashes investigated in the national motor vehicle crash causation survey (Traffic Safety Facts CrashStats. Report No. DOT HS 812 115). Washington, DC: National Highway Traffic Safety Administration.

Skaug, H., Fournier, D., Bolker, B., Magnusson, A., & Nielsen, A. (2016). Generalized linear mixed models using AD model builder. R package version 0.8.3.3. Stangor, C. (2011). Principles of social psychology. Irvington, NY: Flat World Knowledge.

Staubach, M. (2009). Factors correlated with traffic accidents as a basis for evaluating advanced driver assistance systems. Accident Analysis & Prevention, 41 (5), 1025–1033.

Stephens, A. N., Trawley, S. L., Madigan, R., & Groeger, J. A. (2013). Drivers display anger-congruent attention to potential traffic hazards. Applied Cognitive Psychology, 27(2), 178–189.

Strayer, D. L., Cooper, J. M., Turrill, J., Coleman, J., Medeiros-Ward, N., & Biondi, F. (2013). Measuring cognitive distraction in the automobile.

Strayer, D. L., Turrill, J., Coleman, J. R., Ortiz, E. V., & Cooper, J. M. (2014). Measuring cognitive distraction in the automobile II: Assessing in-vehicle voice-based interactive technologies. Washington, DC: AAA Foundation for Traffic Safety.

Sullman, M. J. M., Gras, M. E., Cunill, M., Planes, M., & Font-Mayolas, S. (2007). Driving anger in Spain. Personality and Individual Differences, 42(4), 701–713. Tek, Ö. B. (2014). The stance logistics: An attempt to adapt logistics principles on human and inter-human movement behaviors. Logistics and Transport, 23 (3), 13–36.

Toroyan, T. (2015). Global status report on road safety 2015. Geneva, Switzerland: World Health Organization.

Transportation Research Board of the National Academies of Science. (2013). The 2nd strategic highway research program naturalistic driving study dataset. Available from the SHRP2 NDS InSight Data Dissemination web site: https://insight.shrp2nds.us.

Underwood, G., Chapman, P., Wright, S., & Crundall, D. (1999). Anger while driving. Transportation Research Part F: Traffic Psychology and Behaviour, 2(1), 55–68.

Victor, T., Dozza, M., Bärgman, J., Boda, C.-N., Engström, J., & Markkula, G. (2014). Analysis of naturalistic driving study data: Safer glances, driver inattention, and crash risk: SHRP 2 safety project SO8. Transportation Research Board of the National Academy of Sciences, Wahsington, D.C. (2015) ISBN: 978-0-309- 27423-4. SHRP 2 SO8A.

Young, R. (2015). Revised odds ratio estimates of secondary tasks: A re-analysis of the 100-Car naturalistic driving study data. In Proceedings of Society of Automotive Engineers, Detroit, MI. .

Zhang, T., & Chan, A. H. S. (2016). The association between driving anger and driving outcomes: A meta-analysis of evidence from the past twenty years. Accident Analysis & Prevention, 90, 50–62.

Zhang, T., & Chan, A. H. S. (ang and Chan, in preparation). Driving anger among Chinese drivers and its relationships with aggressive driving. Accident Analysis & Prevention (in preparation).

Zhang, T., Chan, A. H. S., & Zhang, W. (2015). Dimensions of driving anger and their relationships with aberrant driving. Accident analysis and prevention, 81, 124–133.

Zuur, A. F., Ieno, E. N., Walker, N. J., Saveliev, A. A., & Smith, G. M. (2009). Mixed effects models and extensions in ecology with R. New York: Springer.

prechteffectsdrivingangerdriver2017.md 104 KB Histórico Raw