Riding at Risk: A Comprehensive Analysis of Motorcycle Accidents and Contributing Factors

Chapter 1: The nature and bases of motorcycle accidents

1.0 Introduction and scope

Driving has become an essential part of modern life and is the most common form of human transportation (Ball, Owsley, Sloane, Roenker, & Bruni, 1993). Reimer, D’Ambrosio and Coughlin (2007) asserted that despite the significant technological progress achieved in terms of car safety, accompanied by advances in driver training programmes (see also, Hedlund, 2007), traffic accidents remain a predominant cause of death among people under the age of 40 in developed countries (Plainis et al., 2003). Cars are considered to be one of the safer forms of automotive transportation, largely due to the development of crash zones, which provide drivers with a level of protection. Motorcycles, on the other hand, continue to provide little protection to motorcyclists, increasing the risk of serious injury or fatality in the event of an accident. According to the Department of Transportation (2009), while motorcycles account for 4% of all registered vehicles and serve only 1% of all transportation needs, 21% of all traffic fatalities involve motorcyclists. Motorcyclists clearly face a substantially higher risk of being involved in traffic accidents, and are more likely to suffer serious injury or death as a result. Moreover, although the overall number of traffic accidents has declined in recent years, the injury and death toll levels remainhigh, which has provoked interest in the reasons for such high rates of motorcycle accidents (Eurostat, 2007). These traffic accidents may be attributed to several potential causes. For example, they may be classified as resulting from adverse weather (Edwards, 1998), difficult road conditions (Crundall & Underwood, 1998), and a lack of driving experience (Crundall, Underwood, & Chapman, 1999). However, it has been argued that these and other causes fundamentally involve a failure by the driver to spot a potential hazard and respond to the risks it involves in a timely manner (Grayson, Maycock, Groeger, Hammond, & Field, 2003). Additionally, those studying the field of criminology may seek criminology dissertation help to explore the social implications of traffic safety and accident prevention.

On the basis of experimental analysis, Grayson et al. (2003) have argued that drivers go through a risk event in every journey, and have suggested that they go through four steps or components in order to prevent a potentially dangerous situation. These steps are: hazard detection, threat appraisal, action selection, and implementation. According to this model of risk, drivers must first be aware of the hazard, evaluate whether the hazard is sufficiently dangerous to merit a response, select an appropriate response to the danger and, finally, perform the necessary actions required by the response. Within this framework, failure at any stage of the risk assessment could result in an accident. The principal concern of the research presented in this thesis was to investigate the origin of accidents involving collisions between cars and motorcycles. The next sections consider the evidence that is directly relevant to this concern, and to the four components of the model presented by Grayson et al. (2003). At the end of Chapter 1 I will return to this model as a framework for the specific objectives to be addressed by the research conducted in this thesis.

1.1 Overview of the literature

As noted in the preceding section, motorcyclists are at a higher risk (relative to car drivers) of being involved in accidents. The Department of Transport (2004) conducted an in-depth study on motorcycle accidents. Their methodology largely relied on human interpretation of road accident case reports. They drew a heterogeneous sample of police road accident files and examined them. They analysed the road accidents database and other converging evidence, such as photographs, maps and statements of vehicle examiners. They also examined the attitudes of motorcyclists quantitatively by means of a questionnaire. Upon analysis they found that these accidents could be attributed to issues including a lack of attention, losing control at bends in the road, unwise overtaking by motorcyclists, and the relative rarity of motorcycles relative to other vehicles on the road. On exploring the nature of these accidents, the Department of Transport found that 28% of them involved the car driver pulling out when the motorcycle was very close, which implies that the driver did not see the motorcycle. The study attributed the rest of the accidents to either a failure to detect the motorcycle or a poor time-of-arrival judgement on the driver’s part. They found that 38% of the accidents involved right of way violations, in which the motorcycle is travelling straight ahead on a road while another vehicle is trying to enter that same road in front of the motorcycle. However, it was found that less than 20% of these involved a motorcyclist who rated as either fully or partly to blame for the accident. The majority of motorcycle right of way violation accidents were found to primarily be the fault of other motorists. Other investigations of motorcycle accidents have also revealed similar findings (see also, Clarke, Ward, Bartle, & Truman, 2004). Peek-Asa and Klaus (1996) examined descriptions of this type of accidents. They found that 96% of motorcycle accidents at junctions occur due to right of way violation.

In-depth on scene examinations of motorcycle accidents by the Traffic Safety Centre in California through found that majority of these accidents occurred in broad daylight, under no adverse weather conditions, which suggests that inconspicuousness of the motorcycle may not be the only cause (Hurt, Ouelett, & Thom, 1981). These results highlight the importance of other factors that also lead to the failure to detect the presence of motorcycles, such as ‘looked but failed to see’-type errors and/or distractions (Crundall, Humphrey,& Clarke, 2008). Rumar (1990) postulated that a lack of attention by the driver to relevant driving events is one of the main contributing factors. The lack of attention could be a product of a variety of factors (e.g., auditory distraction from passengers or mobile phone conversations) that are likely to be exacerbated by the complex nature of the visual scene. It has also been argued that such distractions might disproportionately influence the processing of unexpected or low frequency objects such as motorcycles (Hancock, Oron-Gilad & Thom, 2005; Wolfe, Horowitz, Van-Wert & Kenner, 2007). The next sections will focus on why drivers are especially likely to fail to detect motorcycles.

1.2 Visual detection of cars and motorcycles

The visual system is obviously a driver’s key source of information while driving (Sivak, 1996) and visual attention is frequently portrayed as the spotlight for navigating the visual scene (e.g., Erikson & Erikson, 1974). Some studies of direct relevance to this thesis have shown that information outside the focus of attention is neglected. For instance, Galpin, Underwood and Crundall (2009) first presented participants with an image of a road, which was replaced by the brief presentation of a blank blue screen. The participants were then presented with a similar image of a road, except that in the latter image the road markings were removed. Most participants showed great difficulty in identifying the difference between the target image and the original image. Furthermore, other studies have shown that when individuals engage in a particular task they often neglect surrounding stimuli: a phenomenon referred to as ‘inattentional blindness’ (see also, Crundall, Shenton & Underwood, 2004). In one study by Galpin, Underwood and Crundall (2009), participants were asked to play a driving game while their eye movements were recorded. There were two types of games: some participants were instructed to free-drive, while others were told to follow a particular car. Participants taking part in the intentional car-following task produced less horizontal eye movements, had longer fixations, neglected pedestrians, and were more likely to be involved in crashes. The results revealed that taking part in such a task narrows attention, with poor processing of visual information from the peripheral areas of the visual field.

According to Itti and Koch (2000) attention is drawn to the most salient region of the visual field, with salience being determined by an object’s low-level features. For example, Hughes (1996) postulated that spatial frequency might determine scene processing, with low frequency objects extracted first, followed by objects with higher spatial frequencies. In terms of moving objects, such as cars and motorcycles, spatial frequency is represented by the width of the object. In general, cars tend to have a greater width compared to motorcycles; therefore, cars can be viewed as large blocks moving through the visual field with a low spatial frequency. Conversely, motorcycles have relatively high spatial frequencies due to their smaller width. Therefore, drivers would be expected to extract information about cars first before directing their attention to objects with higher spatial frequencies, such as motorcycles. In essence, cars are easier to detect than motorcycles because attention is determined by the spatial frequency of objects. This is unlikely to be the complete explanation for the relative frequency of accidents involving cars and motorcycles. As I will show in the next two sections, there is evidence that experience and distraction can affect the detection of motorcycles and crashes in general.

1.3Experience and the detection of motorcycles

Crundall et al. (2008) reviewed literature to assess evidence on factors which are important to design an intervention, targeting car drivers with the aim of improving car–motorcycle interactions. Their report proposed a framework for interpreting evidence on car drivers’ skills and attitudes towards motorcyclists. Their review postulated that motorcycle experience (especially experience of driving a motorcycle) helps inform drivers about motorcycles and their movement on the road. This knowledge refines their understanding and enhances their ability to deploy strategies and skills aimed at avoiding accidental collisions with motorcycle drivers. Therefore, the negative impact of motorcycles’ low-level characteristics associated with bottom-up processing - notably their higher spatial frequency and lower salience - has a limited impact on car drivers with motorcycle experience.

Magazzu, Comelli and Marinoni (2006) evaluated the effect of the type of motorcycle licence which was held by car drivers in responsibility for motorcycle-car crashes. They obtained data through multicentric case-control study of the Motorcycle Accidents In-Depth Study (MAIDS) regarding the risk of crash and serious injuries of motorcyclists. The study in used a non-parametric method, classification and regression tree (CART). It was then compared to standard unconditional logistic regression. It was found that drivers with a motorcycle licence were less responsible for motorcycle-car crashes than drivers who did not have one. Both CART and unconditional logistic regression consistently reported this. They attributed their findings to the fact that “expectations may also play a role by lowering the threshold for motorcycle detection” explaining that “those drivers who also ride motorcycles have greater exposure to motorcycles and are more aware of the potential dangers at junctions”. They reasonably assumed that, “such dual drivers are less likely to cause motorcycle crashes”.

The role of expectation in detecting vehicles is also discussed by Brooks and Guppy (1990). On the basis of experimental analysis, they found that drivers whose family members and close friends rode motorcycles were less likely to collide with motorcycles. These drivers were found to have better observation of motorcycles than drivers who did not have family or close friends who rode motorcycles. They explained these results by arguing that “the greater exposure to motorcycles that these drivers receive may reduce thresholds for spotting them.” Therefore, the role of expectation (of seeing a particular vehicle type) in vehicle detection could prove to be of great significance in increasing the driver’s ability to detect a vehicle of that type. In the context of motorcycles, this would mean that when a driver has a more realistic expectation of seeing a motorcycle s/he will be less likely to miss its appearance when it does occur. As the findings of Magazzu et al. (2006) and Brooks and Guppy (1990) suggest, drivers who are sufficiently aware of the possibility of a motorcycle approaching, show similar detection ability when judging approaching motorcycles and cars.

1.4 Distraction

The National Highway Traffic Safety Administration analysed two databases: the 100-Car Study event database that consists of near-crashes, and incidents; and the baseline database. This study calculated the population attributable risk percentages. This calculation found an estimate of the percentage of crashes and near-crashes in the population where the specific inattention-related activity was a contributing factor. The results of this analysis indicated that driving while drowsy contributed to 22 to 24 percent of the crashes and near-crashes. Secondary-task distraction was found to contribute to 22 percent of all crashes and near-crashes. Overall, driver inattention was found to contribute to, approximately, 25 percent of reported traffic accidents. Moreover, Stutts, Reinfurt, Staplin and Rodgman (2001) conducted a descriptive analysis of five years of the National Accident Sampling System (NASS) Crashworthiness Data System (CDS) data and also analysed the narratives for two years for both data. The descriptive analyses and the narrative analysis provided input so that a more comprehensive taxonomy of driver distractions could be developed. The analyses of CDS data revealed that driver distraction as a form of inattention is responsible for over half of all inattention crashes.

Regan (2004) reviewed literature concerning driver distraction and posited that with the development of wireless communication, entertainment systems and the introduction of route navigation systems into vehicles, preoccupation with the impact of electronic devices’ in-vehicle usage has grown considerably. Indeed, engaging with these devices competes for the driver’s attention, which could potentially increase the risk of a distraction-related accident occurring. Wickens’ (2002) multiple resource theory proposed that if two tasks are performed concurrently, then dual-task interference will occur, especially if the cognitive demands generated by one or both of the tasks are high. Therefore, engaging in secondary tasks will likely compete for the driver’s cognitive resources and thereby interfere with his/her driving performance. Moreover, it has been argued that as such in-car electronic devices proliferate the rate of distraction-related crashes would rise.

Young and Regan (2007) assert that “drivers must continually allocate their attentional resources to both driving and non-driving tasks.” Since many aspects of the driving task become automated with experience, experienced drivers may engage in secondary tasks without any serious consequences for their driving performance or safety. They also posit that drivers are also capable of adapting their driving to meet the demands of the driving environment or exhibit various compensatory behaviours to account for a decrease in their attention to the driving task (see also, Haigney, Taylor & Westerman, 2000). However, under certain conditions, “the driver may fail to allocate sufficient attention to the driving task” to an extent that compromises their driving performance. This indicates that driver distraction occurs when a driver’s normal cognitive processes (attention-sharing) and compensatory strategies fail and the driver is no longer able to adequately allocate attention to more than one task while simultaneously maintaining a satisfactory level of driving.

1.5 Secondary tasks: Mobile phones

Over the past two decades, the popularity of mobile phones has grown dramatically (Allen Consulting Group, 2004). Unsurprisingly, there has been a related increase in the usage of mobile phones in cars. One study examined the effects of hands-free cell phone conversations on simulated driving (Strayer, Drews & Johnston, 2003). It reported that an estimated 85% of mobile phone owners used their phones whilst driving, with most users deeming it acceptable to use their phones during a car journey (as cited in Goodman et al., 1999). Strayer et al. (2003) found that “these conversations impaired driver’s reactions to vehicles braking in front of them.” The study examined whether this impairment could be attributed to a withdrawal of attention from the visual scene, which would yield a form of inattentional blindness. Cell phone conversations were found to impair explicit recognition memory for roadside billboards. This interpretation received support from data, which indicated cell phone conversations to be impairing of implicit perceptual memory for the items presented at fixation. Their results suggested that “the impairment of driving performance produced by cell phone conversations is mediated, at least in part, by reduced attention to visual inputs.” These findings are in line with numerous studies that have examined the relative effects of using hands-free and hand-held mobile phones on driving performance. The results of such studies suggested that hand-held phones usage significantly degrades driving performance. The conclusion that the main risk associated with using a hand-held mobile phone pertained to the physical interference caused by interacting with the device resulted in many countries prohibiting the use of hand-held mobiles while driving (Goodman et al., 1997; see also Matthews, Legg & Charlton, 2003).

However, subsequent research has shown that, while the physical obstruction does present significant safety hazards, the cognitive distraction associated with being engaged in a conversation can have an equally damaging impact on driving performance. A substantial amount of research has since confirmed that using a hands-free phone provides no more advantages in terms of driving performance than a hand-held device (Redelmeier & Tibshirani, 2003; Strayer et al., 2003).

Briem and Hedman (1995) investigated effects on driving performance in a simulated driving pursuit tracking task of using a hands-free, mobile telephone. The primary task of the experiment was to drive safely. The participants were subjected to three secondary task blocks each, while they drove for 20 minutes. These task blocks were – a simple telephone conversation about a familiar topic, a difficult telephone conversation, incorporating a test of working memory, and car radio tuning and listening. The driving was divided into two halves – a simulated firm road surface and half on a slippery road surface. The participants' behaviour was subsequently observed. It was classified in four categories, two without and two with a secondary task, with driving – on a clear road, and with obstacles, and with driving involving the secondary task components of communication, and instrument manipulation. They found that a difficult conversation affected the driving adversely. Any prolonged manipulation of the telephone was found to be liable to produce a performance decrement, particularly under conditions that put heavy demands on the driver's attention and skill.

Furthermore, a study by Haigney et al. (2000) revealed no safety advantages for the hands-free phone over a hand-held one. They examined the relative effects on driving performance of each type of phone in a simulator study in which participants were required to complete four simulated drives while dealing with incoming calls on either hands-free or hand-held mobile phones. The results indicated that, when using a mobile phone, the vehicle speed and standard deviation of accelerator pedal travel was lower, and the mean heart rate was higher. However, no significant differences were found between the two different types of usages (hands-free or hand-held). Further studies have shown that a similar deterioration in driving performance was produced by both hands-free and held-held mobile phone usage (Strayer et al., 2003). In these studies, both hands-free and hand-held phones led to greater misses and slower responses to traffic signals compared to when drivers were not conversing on the phone. A further study also revealed that conversing on a mobile phone while driving imposed an increased workload demand on drivers regardless of the type of phone used (hands-free vs. hand-held) and that drivers tended to overestimate the advantages of using hands-free mobiles while driving.

Studies have also investigated how differing levels of cognitive distraction affect driving performance, notably how the complexity or emotionality of a phone conversation can influence driving performance. McKnight and McKnight (1993) explored the differences in drivers’ ability to attend to the simulated driving task when engaged in either simple or complex hand-held phone conversations. Five distraction conditions were included: placing a call through dialling on a mobile phone, holding a simple conversation, holding a complex phone conversation, turning the radio on, and no distraction. The results showed that all of the three conditions involving mobile phone usage were found to increase failure in adequately responding to traffic situations, such as vehicles slowing down or pedestrians crossing the road. Complex conversations led to the greatest errors and poorest driving performance. A similar study has found that the response to visual targets (e.g., noticing boards and signs) was significantly slower under more cognitively complex phone conversations compared to simple ones (Al-Tarawneh et al., 2004). Patten et al. (2004) found that during peripheral detection tasks, drivers took longer to react when engaged in complex phone conversations rather than simple ones. These studies highlight the important impact on driving performance of cognitive distraction. More specifically, they demonstrate that the deterioration in driving performance associated with mobile phone usage may be more directly related to the cognitive demands of engaging in conversation while simultaneously performing the driving task.

Mobile phone use studies (see also, Nabatilan, Aghazadeh, Nimbarte, Harvey, & Chowdhury, 2012; National Safety Council, 2012) have also begun to explore the effects of phone use on visual behaviour. Visual behaviours predominantly assessed in these studies are those of visual scanning and eye fixations. For example, in a study by Harbluk, Noy and Eizenman (2002) drivers rode an instrumented car along a city route while carrying out secondary tasks of varying complexity using a hands-free mobile. Each participant was subjected to one of three conditions: no secondary task, solving an easy arithmetic addition task, and solving a complex arithmetic task. Measures of visual scanning, vehicle control, drivers’ subjective assessment of workload, safety and distraction were all recorded. Visual scanning measures revealed that “under increased cognitive demands, drivers made significantly less saccadic eye movements and spent more time looking centrally and less time looking at the periphery for impending hazards.” Participants were also found to be spending less time checking their mirrors and more time searching up and down the road. Hard-braking was found to increase during the complex addition task. Increase in complexity of the task was found to increase the drivers’ perception of the workload, distraction level and perception of their driving as being less safe.

Nevertheless, there are several concerns regarding previous research that examined the effects on driving performance of in-car mobile phone use. Young and Regan (2007) have argued that adopting artificial mathematical and verbal tasks to simulate phone conversations may lead to an overestimation of the damaging effects of mobile phone use. Rakauskas, Gugerty and Ward (2004) explored the relationship between the level of conversational difficulty and driver distraction. While driving in a simulated environment, the participants were asked to answer a set of both easy and difficult questions (e.g. “What are you doing tomorrow?” and “Do you think the world would be a better place in 100 years?”). The results revealed that although engaging in a phone conversation did degrade driving performance, changes in the task’s level of difficulty had no additional effects on driving performance in terms of mean speed or subjective workload ratings. One explanation for these findings is that “naturalistic conversations require less cognitive effort than that demanded by the verbal reasoning and mathematical tasks used in previous studies” (such as, Shinar, Tractinsky, & Compton, 2005). An alternative explanation is that the two complex tasks did not sufficiently differ in their level of difficulty to reveal any differential effects on driving performance.

Another concern of this thesis is that previous studies of the effects of distraction did not sufficiently take the effects of learning into consideration, as they used only a limited number of trials. It has been argued that had the participants been given more time to interact with the devices, they would have learned to more effectively to time-share between their non-driving and driving tasks. Shinar et al. (2005) examined whether repeated experience of conversing on a mobile phone would lead to a learning effect and reduce the impact on driving performance of the secondary task. As expected, mobile phones were found to have a negative effect on driving performance. It was noted that phone-using drivers exhibit lower mean speed and greater steering variability. However, over the course of the trials, the negative effects of the phone task diminished. The research suggests that previous studies, which have used only a limited number of experimental trials or have used artificial phone tasks, may have thus been overestimating the detrimental effects of mobile phone use on driving performance. This thesis will investigate the influence of learning on distraction and provide theoretical implications on the same.

1.6Distraction and driving demands

Strayer et al. (2003) argued that “changes in the demands of the driving task itself, such as during great traffic density or adverse weather conditions, can affect the distracting effects of engaging in a non-driving task.” Accordingly, engaging in a non-driving task may have a different effect on driving performance depending on whether it takes place on a quiet or busier road. On a busy road, drivers must pay more attention to incoming traffic. This may place a greater cognitive demand on the driver which would then lead to reduction of the spare cognitive capacity for the performing secondary tasks, such as conversing on a mobile phone. A number of recent studies have begun to investigate the relationship between the performance of an in-vehicle non-driving task and the complexity of the driving environment (Brookhuis, de Vries & de Waard 1991; Horberry, Anderson, Regan, Triggs & Brown, 2003; Young, Regan & Hammer, 2003).

Strayer and Johnston (2001) explored how the driving environment affects pursuit-tracking performance while using a mobile phone. They hypothesised that under difficult driving conditions, the drivers’ ability of dividing attention between the driving and non-driving tasks were less. This resulted in a greater degradation of their driving performance. Participants in the experiment were required to converse on either a hand-held or hands-free mobile phone while performing an easy and predictable simulated driving task, and a difficult and unpredictable driving simulation. The results revealed that “when using a mobile phone, participants failed to detect twice as many tracking targets compared to when they were not using a mobile phone.” Additionally, this was found to be more pronounced in the difficult driving task. The findings of this study support the notion that mobile phone usage, whether hands-free or hand-held, reduces driving performance significantly. The results also suggest that when cognitive demands of the driving tasks are high, the ability of the individual to allocate attention between the driving and non-driving tasks is further diminished. As such, the distracting effects of the non-driving task are amplified, thus leading to even greater degradations in driving performance. The present study aims to investigate this in more detail and in the context of accidents involving cars and motorcycles.

Strayer et al. (2003) found that “conversing on a hands-free mobile phone while driving led to an increase in response latencies to a lead vehicle breaking.” Furthermore, this increase in reaction time (RT) was more pronounced as the density of the traffic increased. The impact of adverse weather conditions have also been shown to influence the distracting effect of in-car mobile phone usage on drivers’ ability to make safe turning decisions (Cooper & Zheng, 2002). The participants pressed down on the accelerator pedal when they felt it was safe to turn in front of the approaching vehicles. In half the trials, the driving circuit was wet while in the other half the circuit was dry. For half the trials, participants were asked to listen and respond to a complex message (the distraction). Meanwhile, participants in the other half experienced no distractions. The results of the study demonstrated that when participants were distracted by the mobile phone task, they failed to take into account the road surface conditions when deciding whether to accept or reject the opportunity to turn into a gap in the traffic. According to the data, participants initiated twice as many potential collisions when distracted by the mobile phone task under wet road conditions. The researchers concluded that the task reduced the participants’ ability to adequately consider and process all the information necessary to make a safe driving decision. In other words, under difficult driving conditions, the effects of distraction are enhanced.

1.7 Distraction and driver experience

Over the past few years, there has been a growing body of evidence suggesting that driver experience can influence the distracting effects of mobile phones and other in-vehicle devices (Lam, 2002; McPhee, Scialfa, Dennis & Caird, 2004). For instance, research has consistently shown that young novice drivers, with less driving experience, are more susceptible to the distracting effects of engaging in a secondary task compared to more experienced drivers: They register up to nine times higher crashing rates than those of more experienced drivers (National Highway Traffic Safety Association, 2000; Pradhan et al., 2005). According to road accident data, young novice drivers are among the highest population segments to be involved in an accident (Deery, 1999; Underwood & Crundall, 2003). These results have been taken to suggest (Regan, Deery & Treegs, 1998) that novice drivers have not yet acquired the driving skills necessary to operate a vehicle while engaging in secondary tasks. Due to this lack of driving skills, they must allocate significant attentional resources exclusively to the driving task, which means that they do not have sufficient attentional resources to devote to non-driving tasks such as speaking on the phone (see also, Underwood, Crundall & Chapman, 2002).

A number of studies have also demonstrated age-related deterioration in driving performance, with older drivers more susceptible to be affected by distractions (Lam, 2002). In particular, studies on stimulators have shown that compared to younger drivers, old drivers are more likely to miss traffic signals, and showed greater reduction in their ability to maintain speed and lane position when talking on a mobile phone (McKnight & McKnight, 1993; Reed & Green, 1999; Schreiner, Blanco & Hankey, 2004). However, it remains the case that, according to car-crash data, younger drivers were at the highest risk of being involved in a fatal car accident while using a hand-held phone. However, a study conducted by Strayer and Drews (2004) failed to demonstrate any significant age-related differences in driving performance when conversing on a mobile phone. The absence of an effect on driving performance of drivers’ age might be attributable to the study comparing older drivers to young novice ones who, unlike young experienced drivers, are particularly susceptible to distracting effects. In support of this view, Shinar et al. (2005) demonstrated that both older and younger inexperienced drivers were more negatively affected by phone conversations compared to middle-aged participants.

Studies of visual search and attention offer a compelling explanation for the reduction in driving performance often associated with novice drivers. They show that task-related visual search patterns are learned (van der Gijp et al., 2016), with adequate learning resulting in a proactive allocation of visual attention (Hayhoe & Ballard, 2005). Furthermore, although the link between attention and performance is not always straightforward, task experience usually results in more efficient visual search patterns (as seen in Brockmole, Hambrick, Windisch & Henderson, 2008; Charness, Reingold, Pomplun & Stampe, 2001; Pashler, Johnston & Ruthruff, 2001).

Ball et al. (1993) assessed several aspects of vision and visual information processing in drivers between the ages 55-90 years found that “the size of the useful field of view, a test of visual attention, had high sensitivity (89%) and specificity (81%) in predicting which older drivers had a history of crash problems. Older adults with substantial shrinkage in the useful field of view were six times more likely to have incurred one or more crashes in the previous 5-year period. Eye health status, visual sensory function, cognitive status, and chronological age were significantly correlated with crashes, but were relatively poor at discriminating between crash-involved versus crash-free drivers.” These findings suggest that proficient visual attention allocation has been linked to better driving performance and safety. A study reviewing frameworks on the role of attention while driving posited that a lack of attention, or a failure to allocate attention efficiently, has been associated with adverse consequences such as a greater risk of traffic accidents (Trick, Enns, Mills & Vavrik, 2004).

Crundall and Underwood (1998) have argued that reduced performance often displayed by novice drivers is a result of inefficient driving strategies. They proposed, on the basis of evidence presented in the next paragraph, that through experience drivers allocate their attention more effectively, thereby reducing the cognitive demands of the driving task. According to this view, more experienced drivers have greater attentional resources available to them for secondary tasks. Furthermore, experienced drivers often have better visual search patterns, which enable them to spot potential hazards more effectively than novice drivers.

Crundall and Underwood (1998) investigated differences in visual search patterns between experienced and novice drivers. They found that experienced drivers have greater sampling rates of the visual scene, with a greater number of short fixations. In addition, experienced drivers exhibited greater horizontal scanning of the visual scene. In another study, participants were shown a driving video clip while their visual search patterns were assessed. The results indicated that, compared to experienced drivers, novice drivers have longer fixations while watching the video, which seems to suggest novices take longer to process the visual scene. Moreover, under dangerous driving conditions, these fixations became even longer for novice drivers. When the demands of the driving task increase, novice drivers are less able to process the visual scene effectively, leading to visual attention being concentrated on a very specific area of the visual scene. Consequently, novice drivers are less able to scan the visual scene for potential hazards and may therefore be more likely to be involved in traffic accidents.

Eye-tracking studies have also demonstrated a difference between novice and experienced drivers. Recarte and Nunes (2000) examined the impact of performing verbal and spatial-imagery tasks on visual search when driving. The participants drove 84 km on 2 highways and 2 roads and performed 2 verbal tasks and 2 spatial-imagery tasks on each route. Their eye movements were recorded on all routes. Pupillary dilation was indicative of similar effort for each task. The results showed that “visual functional-field size decreased horizontally and vertically, particularly for spatial-imagery tasks. Fixations were found to be longer during the spatial-imagery task as compared with ordinary driving. On assessing driving performance, glance frequency at mirrors and speedometer decreased during the spatial-imagery task.” Results were interpreted in terms of multiple attention-resource theories and were attributed to the fact that specific regions of the visual scene attract attention differently based on the driver’s experience.

.Underwood et al. (2002) showed that novice drivers have a greater number of fixations on the rear-view mirror, while experienced drivers tend to focus on the near-side mirror. In addition, novice drivers tend to direct their attention more to in-car objects than their experienced counterparts. Crundall et al. (1999) investigated differences in eye movements between experienced, novice and non-drivers. Participants were required to watch a video containing at least one hazardous event. The primary task was to assess how hazardous each clip was. In addition, participants had to respond to lights that randomly appeared on the four corners of the screen. Experience as a variable was chosen since it was indicative of better performance in driving. The results further validated this by showing that experienced drivers scored the most correct identifications of peripheral targets, with non-drivers performing worst. Again, driving experience played a significant role in determining participants’ ability to attend to visual targets outside the central field of vision. In effect, the more efficient search patterns, learned through driving experience, do not only allow drivers to detect potential hazards more easily but also reduce the cognitive demands of the driving task. Therefore, more experienced drivers have greater cognitive resources available to them to devote to other tasks - such as speaking on mobile phones. Greater driver experience, in other words, reduces degradations in driving performance associated with distraction. Other ways in which experience, for example, might affect driving performance have been the subject of further investigation.

1.8 Compensatory behaviours

Poysti, Rajalin and Summala (2005) have demonstrated that compensatory behaviours can manifest themselves at a number of levels, ranging from the strategic (avoiding secondary tasks completely) to the operational (reducing speed). At the highest level, drivers could choose not to engage in a potentially distracting task, thereby moderating their exposure to the risk. For example, older drivers experience a greater degree of impaired performance than young drivers when using a mobile phone. This, results in compensatory behaviour at the highest level, and older drivers are thus more likely to avoid using mobile phones when driving (Alm & Nilsson, 1995). Burns, Parks, Burton, Smith and Burch (2002) examined driving performance on a stimulator. Their study comprised of four conditions – 1) motorway with moderate traffic, 2) car following, 3) curving road 4) and dual carriageway with traffic lights. The drivers were supposed to answer a standard set of questions and converse with the experimenter over a mobile phone during each condition. The design of the study was repeated measures. The independent variables were, normal driving, alcohol impaired driving, and driving while talking on Hands-free or Hand-held phone. They found that drivers had a tendency to slow down when talking on Hand-held or Hands-free phones, even when they were specifically instructed to maintain a set speed. They concluded that drivers often attempt to reduce their workload and moderate their level of risk by decreasing their speed. Similar studies have found that drivers also tend to increase the inter-vehicle distance (Jamson, Westerman, Hockey & Carsten, 2004; Strayer & Drews, 2004).

In addition to the effects of mobile phone use on driving behaviour, research has also been conducted on the effects of other in-vehicle devices on driving performance. One of the main findings of such studies has been that drivers tend to reduce their vehicle speed when engaging with other devices. For instance, Chiang et al. (2001) showed that drivers reduced their speed when entering coordinates into a route navigation system (see also, Horberry et al., 2003). Another compensatory behaviour such drivers display is to increase their inter-vehicle distance. In a driving simulator study, Jamson et al. (2004) showed that, when processing emails using a speech-based system, drivers increased the distance between their car and a lead vehicle. Strayer and Drews (2004) also demonstrated a 12 percent increase in inter-vehicle distance when drivers engaged in a mobile phone conversation. Interestingly, in both studies, the drivers’ compensatory behaviour was often insufficiently adequate to avoid collisions with other road users. In other words, although drivers engage in compensatory behaviours, these attempts to reduce the risk of an accident are often insufficient.

Finally, research has also shown that drivers could alter the amount of attention allocated to the non-driving and driving tasks in response to changes in their driving environment. Brookhuis et al. (1991) studied the effects of telephoning while driving in three different traffic conditions: in light traffic on a quiet motorway, in heavy traffic on a four-lane ring-road, and in city traffic. They recruited 12 participants who were unfamiliar with mobile telephones. The participants drove an instrumented vehicle for one hour each day during three weeks. They were expected to operate the mobile phone for a short while in each of the three traffic conditions. The results indicated that there was a significant effect of telephoning while driving as opposed to normal driving. The effort was subjectively measured by an effort scale and objectively measured by heartrate indices and on some of the measured parameters of driving performance. One half of the subjects operated the telephone manually and the other half performed used a handsfree mobile telephone set. The study found that “subjects who operated the handsfree telephone showed better control over the test vehicle than the subjects who operated the handheld telephone. This was measured by the steering wheel movements.” In the course of the 15 test days, a clear improvement was found for some of the measurements. However, this study found that on a busy road engaging in a phone task did not affect the amount of attention paid to other traffic. These results suggest that the level of attention assigned to the secondary task is situation-dependent and will change according to both the driving conditions and the demands of the task.

1.9 Overview, objectives and experiments

The main aim of this thesis is to understand the causes of motorcycles accidents through the use of experimental scenarios that allow the characteristics of the relevant stimuli to be readily manipulated. The thesis will focus on the first component of the hazard model (Brockmole et al., 2008), which is hazard detection. The key manipulations were motivated by factors known to contribute to car accidents, but ones that have not been the subject of detailed analysis in the context of accidents involving motorcycles. These manipulations and the experiments in which they are listed below:

Experiment 1 investigated the effects of distracting stimuli on components of virtual driving performance, notably the effect of distractions on the perception of oncoming vehicles (when a motorist is making a decision about attempting to pull out of a junction onto another road). Experiments 2, 3 and 4 examined the effect of visual and oral distractors on the driver’s performance. Experiments 5 and 6 used the same images as in previous experiments except that a highly busy road was chosen in order to assess the generality of the previous observations. Experiment 7 In this experiment, there was no secondary task (sounds). The results confirm that when you ask the drivers about the type of approaching vehicle, they miss spotting the motorcycle in the far position and their reaction time is longer as well. Experiment 8 examined the differences between novice drivers and experienced drivers in ……... The results of this study show that experienced drivers tend to spot the approaching vehicle more accurately than novice drivers, but they took more time to spot the unexpected object (motorcycle) in the far position.

Chapter 2 Detecting vehicles at T-junctions: The role of distraction

2.1 Introduction

2.1.1 Background

As noted in Chapter 1, according to Department of Transport (2004) figures, 96% of all road accidents involved ‘right of way’ violations at junctions: this is when a vehicle travelling straight along a road collides near a T-junction with another vehicle attempting to join that same road. Studies indicate that 28% of accidents involving a car-motorcycle collision seem to result from the car driver not seeing the approaching motorcycle (Lehtonen, Lappi & Summala, 2011); with the remaining 72% of accidents seeming to result from a time-of-arrival calculation error by the driver in estimating the speed of approach of the motorcycle and, as a result, pulling out of the junction at the wrong time. Motorcyclists have been shown to be particularly vulnerable to being involved in a fatal or dangerous traffic accident, with a ‘killed and serious injury’ (KSI) rate that is approximately twice that for pedal cyclists, and more than fifteen times that for automobile drivers/passengers. Moreover, although motorcyclists comprise 1% of road users in the UK, they account for 13% of all injuries and fatalities. In their examination of the relation between the travelled distance and injuries sustained, Uchida, De Waard and Brookhuis (2011) found that, in 2010, a motorcyclist was thirty times more likely to be killed or critically injured in a road traffic accident compared to a car driver involved in a similar incident.

The aim of Experiment 1 was to attempt to model these differences in the incidence of accidents involving motorcycles and cars, and to assess one potential contributor to accidents, namely distraction. The procedure was adapted from a previous study on motorcycle detection, conducted by Crundall et al (2008). , Experiment 1 investigated the impact of different forms of distraction and examined whether they had different effects on participants’ ability to detect motorcycles and cars. The procedure involved briefly presented (250ms) snapshots of motorcycles and cars at various distances from a T-Junction at which the notional driver was positioned in a car. The participant’s task was to detect whether the snapshot contained a vehicle. The vehicles could be presented near to the junction, in the mid distance or far away from the junction. The nature and presentation time was intended to mimic a driver’s activity at a junction where brief inspection of oncoming traffic from the right might form part of the basis for pulling out. The participants’ task was to decide if the image they were presented with contained an approaching vehicle. While viewing the images, they were subjected to distracting stimuli.

Experiment 1

All participants were presented with a sequence of images featuring a motorcycle, a car, or neither. The vehicles were presented at three distances (near, mid, far). After each image, the participants’ task was to indicate whether it contained an approaching vehicle or not. In order to study the effect of distraction on their ability to detect approaching vehicles, the participants were divided into four groups: Control, Sound, Image and Verbal. Participants in the control group received no distraction; those in the sound group received presentations of a stream of auditory words; participants in the image group answered questions involving mental visualisation (e.g., “what is bigger, a car or a bus?”); and those in the verbal group received an auditory stream of words and had to indicate, on hearing each word, whether it contained the ‘ch’ syllable.

2.2 Experiment 1

2.2.1 Method

2.2.2 Participants

Sixty participants (14 males, 46 females) were recruited the majority of which were students from Cardiff University. The mean age of the participants was 21.9 years, with mean driving experience of 4.5 years since passing the driving test. They all reported having normal or corrected-to-normal vision. Two of the participants had experience of riding a motorcycle. The participants were randomly assigned to one of the four groups described above.

2.2.3 Materials

Primary Task

For the primary task, involving the detection of approaching vehicles, the stimuli presented to all participants consisted of a sequence of ten T-junction scenes interspersed with instruction screens. These T-junction scenes, presented for 250 milliseconds each, were taken from the viewpoint of a car driver who had just approached the T-junction and is looking out to his/her right for approaching vehicles. Some of the scenes were digitally edited so as to include an approaching vehicle (either a car or a motorcycle) positioned in either the far, mid, or near distance, and travelling towards the T-junction (see Figure 1). Editing the ten scenes in this way produced 60 pictures in total. An additional set of 10 T-junction scenes featuring no approaching vehicle was created, bringing the total number of pictures presented to each group to 70 pictures. These pictures were presented on a standard monitor using E-Prime presentation software. A standard computer keyboard was used to collect the responses of the participants.

Secondary tasks

The secondary tasks required the following additional materials:

Group Control: Participants only had to perform the primary task and were not subjected to any additional distractions.

Group Sound: An audio recording of a stream of words was presented to the participants while they were performing the primary task. The task was delivered to the participants via the speakers in the lab, and those word been chosen from the psychology linguistic website(for example house, book, car.....etc).

Group Image: An audio recording of questions involving mental spatial visualisation was presented to the participants as they performed the primary task. These questions featured comparisons between objects. For instance, “is ‘object’ bigger than a bus?” Participants responded verbally, - with either a ‘yes’ or ‘no’. All of the responses were recorded. Both tasks were presenting in the same time. Tthe participant responded to for the first task on the keyboard and atin the same time they answered the question verbally.

Group Verbal: An audio recording of a stream of words was presented, including words that contain the syllable ‘ch’ (e.g. ‘chair’). While engaging in the primary task, participants had to indicate (by saying ‘yes’ or ‘no’) whether each word contained the ‘ch’ syllable. All of the responses were recorded. The same word steam used in the sound and imaginary group used in the verbal group except that the participants have to answer them verbally, these words were presented via speakers in the lab while they werare responding to the computer task via the keyboard.

For the three Groups Sound, Image and Verbal2,3 and 4, some of the same set of words was used, which included ‘boy’, ‘church’, ‘tower’, ‘bus’, ‘car’, ‘bible’. The , and the list of words was generated using adapted from a psychological linguistics website. All words are the same in all three groups except the task is different.

2.2.4 Design

There were three independent variables, two within-subjects and one between-subjects. The within-subjects variables were the nature of the ‘Vehicle’ (motorcycle or car), and the second was the ‘Distance’ at which the vehicle was presented in the image (‘near’, ‘mid’ or ‘far’). The between-subjects variable was group (Control, Sound, Image, Verbal). There were two dependent measures: ‘accuracy’ and ‘speed’. The accuracy with which participants detected the approaching vehicles was measured in terms of d’ in accordance with signal detection theory (Macmillan & Creelman, 2004). Speed was measured in terms of the time taken to respond (in ms) from the respond to the image.

2.2.5 Procedure

Prior to the start of the experiment, participants were shown a practice run of 10 images, in order to give them an opportunity to familiarise themselves with the experimental format. During these experimental trials, each participant was instructed to fixate a “+” on the screen for 250 ms, and then the image of a T-junction was presented for 250 ms. These timings were adopted from the experimental setup used in Crundall and& Underwood, 2008 (see also, Crundall, Underwood & Chapman, 2010). Participants were then required to identify whether a vehicle was present (or not) by pressing the appropriate keyboard key (see Figure 2) as quickly as possible. The participants were instructed to press ‘0’ to indicate that the shown picture contained no approaching vehicle and to press ‘2’ to indicate that they detected the presence of an approaching vehicle in the picture. While they performed the primary task, participants in three of the groups (Sound, Image and Verbal) received a secondary task. Both tasks were presenting in the same time. Tthe participants responded for the first task on the keyboard for the first task and atin the same time they answered the question verbally. all word steam are the same in all there group except that task is different. .

Results

The results were pooled over trials of the same type for the purpose of statistical analysis, and the analysis is presented separately for measures of accuracy and speed.

2.3.1 Accuracy:

The mean d’ scores are shown in Figure 3. Taking both panels together, the scores for cars appear to be more accurate than for motorcycles and participants are less accurate when the vehicles are depicted in the far distance than when they are depicted as close to the junction. This effect of distance appeareds to be more marked for motorcycles than for cars. Statistical analysis broadly confirmed this description of the pattern of results in Figure 3. A group by vehicle by distance ANOVA was conducted on the accuracy of vehicle detection. This analysis revealed no significant effect of group (F(3, 56) = 1.13, MSE = 90.578, p> 0.05). There was, however, a significant effect of vehicle (F(1, 56) = 137.1, MSE = 34.178, p< 0.001) and distance (F(2, 112) = 126.5, MSE = 109.556, p< 0.001). There were also two-way interactions between: group and vehicle (F(3, 56) = 3.3, MSE = 81.422, p< 0.05), group and distance (F (6,112) = 2.98, MSE = 109.556, p< 0.01), and vehicle and distance (F(2,112) = 82.7, MSE = 34.178, p< 0.001). There was no three-way interaction (F (6,112) =1.55, MSE = 90.578, p< 0.05).

2.3.2 Reaction Time:

The mean reaction times are shown in Figure 4. Taking both panels together, the reaction times for cars are faster than for motorcycles, and are slower for vehicles that are depicted in the far distance than those that are depicted closer to the junction. This effect of distance appears to be more marked for motorcycles than for cars. Statistical analysis broadly confirmed this description of the pattern of results in Figure 3. A group by vehicle by distance ANOVA was conducted on the speed of vehicle detection.

Reaction time is the second parameter of this study. The analysis revealed a significant effect of between group,s F (3, 56) = 6.079, MSE= 238882.603, p 0.05. A post hoc Tukey test reveals a significant increment in reaction time as the task get more difficult. The reaction time for the sound group increased but not significantly compared to the control group (842ms vs. 779ms, p˃0.05). As the task got more difficult for the image group, their reaction time increased significantly compared to the control group (1001ms vs. 779ms, p 0.05). This increment continued as the task got even more difficult for verbal group, as they spent more time significantly compared to the control group (1046ms vs. 779ms, p 0.01). Despite there was differences between the sound and the image group, this difference was not significant (842ms vs. 1001ms, p˃0.05). As the task get more difficult for the verbal group, there reaction increased significantly compared to the sound group (1046ms vs. 842ms, p 0.05). Finally, as both the image group and verbal group have the largest reaction time, the difference between these two groups was not significant (1001ms vs. 1046ms, p˃0.05).

The analysis of variance revealed a significant effect on the vehicle factor F (1, 56) = 44.299, MSE 18801.609, p< .001. The results showed that reaction time was large when the approaching vehicle was a motorcycle compared to cars indicating a more difficult condition when the approaching vehicle was a motorcycle. (965 ms vs. 869ms)

Regarding the distance factor, the analysis also revealed a significant effect factor F (2,112) = 24.470, MSE 24765.762, p< .001. A post-hoc Tukey test revealed a significant increment in reaction time for the far condition compared to mid condition ( 997ms vs. 891ms, p< .001), and a significant increment compared to near condition (997ms vs. 862ms, p< .001). The test did not show any significant effect between the mid and near conditions (891ms vs. 862ms, p˃0.05)

The analysis revealed several two-way interactions between factors. The first two-way interaction was between groups and type of vehicle F (3, 56) = 4.745, p< .01. The main effect of the type of vehicle appeared significantly in every group, as reaction time was significantly large for Motorcycles compared to Cars for all groups, except for the sound group as it was long but marginal. Another large significant difference found within the level of motorcycle, as the verbal group spend around 300ms longer than the control group (1105ms vs. 815ms, p< .05). This increment resulted in this significant interaction that showed how motorcycle condition is more difficult especially when engaging in verbal discussions

The analysis revealed another significant two-way interaction between the type of vehicle and distance factors F (2,112) = 5.073, p< .01. A post-hoc Tukey test showed the same effect pattern of the location factor for each type of vehicle, as the reaction time was significantly large for the far location compared to the near and mid location in case of a motorcycle and a car approaching; and there was no significant difference between the near and mid locations. The test also found another significant pattern that lead to this significant interaction, as the reaction time significantly increased for motorcycles compared with cars within some levels of the location factor. The results showed that reaction time increased significantly for approaching motorcycle compared to cars at the far condition (1078ms vs. 917ms, p< .001). The same effect was also found between motorcycles and cars at the mid location (934ms vs. 849ms, p< .05). On the other hand, this effect was not found in the near condition between motorcycles and cars (834ms vs. 841ms, p˃0.05)

Finally, the analysis revealed a three-way interaction between all factors F (6,112) = 2.597, p< .05. A simple main effect analysis for the factors revealed motorcycles at far and mid condition was significantly longer than other conditions F (3,112) = 3.482, p< .05, F (3,112) = 3.304, p< .05. The results also showed that image and verbal group spent significantly longer time in the far condition compared to other groups and other locations F (1,25) = 23.996, p< .001, F (1,56) = 21.829, p< .001. The results also revealed a significant increment in time for motorcycle condition with image and verbal group, and this increment did not appear with the other groups or with other type of vehicles F (2,112) = 10.590, p< .001, F (2,112) = 9.388, p< .001. These results showed how motorcycles are difficult to spot and it took long time to perceive them compared to cars. These results were very clear as the task get harder with far distance compared to close ones; and this was not clear in the car condition. These results also were very clear between groups as the group with the more difficult task needed more time to spot the vehicle especially with motorcycle. In general, motorcycles are hard to detect and this difficulties are limited for motorcycles in harder to spot condition, that resulted in the three-way interaction

2.4 Discussion

The results of Experiment 1 confirm that the procedure is sensitive to revealing effects on detection accuracy and speed of vehicle type and distance. Motorcycles appear to be perceived less readily than cars and this effect is more pronounced depending on distance, with vehicles at a distance perceived less efficiently than vehicles that are closer to the observer. The typeTypes of distraction also appeared to have an noticeable effect on perception, with stimuli that require a considered response or additional processing, reducing performance more than passive stimuli.

In general, control and sound group were more inclined to evaluate the picture compared to verbal and to the imaginary group. Within this tendency, they were significantly more cautious toward motorcycles.

Distance, level of distraction and vehicle density appeared to have a negligible effect on perception of cars in the control and verbal groups while image and verbal distractions appeared to have a significant impact on perception and reaction times in all groups for both vehicle types.

Of course, there are all sorts of other aspects – most obviously size, as is discussed in Chapter 4 – that could account for these difference in detection of cars and motorcycles.

In order to fully understand the different impacts that various distraction stimuli can have on vehicle detection ability, it is important to examine whether the primary and secondary tasks are calling upon similar cognitive resources/processes. Put simply, it seems likely that a secondary distraction task that is cognitively similar (i.e. that uses the same cognitive processes) to the primary task can have a greater impact in reducing the participant’s ability to perform the primary task.

Alternatively, it is also possible that when an individual is tasked with processing visual (primary task) and verbal (secondary task) information, the brain may prioritise verbal inputs and, as a result, cause the brain’s perception of images to become less efficient, thus increasing RTs for the primary task.

Of course, more experimental work is needed in this regard in order to establish with more certainty the nature of such relationships and mechanisms.

Continue your exploration of Research and Development CEO Share Ownership and Board Independent with our related content.