Expert drivers are better than non-expert drivers at rejecting unimportant information in static driving scenes

Transportation Research Part F 59 (2018) 389–400

Contents lists available at ScienceDirect

Transportation Research Part F journal homepage: www.elsevier.com/locate/trf

Expert drivers are better than non-expert drivers at rejecting unimportant information in static driving scenes Kristen Pammer a,⇑, Alexandra Raineri a, Vanessa Beanland b, Jason Bell c, Maria Borzycki a a

Research School of Psychology, The Australian National University, Canberra, ACT 2601, Australia1 Department of Psychology, University of Otago, Dunedin, New Zealand c School of Psychological Science, University of Western Australia, Perth, WA 6009, Australia

b

a r t i c l e

i n f o

Article history: Received 29 March 2018 Received in revised form 24 September 2018 Accepted 24 September 2018

Keywords: Driving Road safety Situational awareness Inattentional blindness Hazard detection Expert drivers

a b s t r a c t Safe driving is predicated on a driver’s ability to prioritise scene information to segregate hazards and potential hazards from other information, and allocate attention accordingly. Previous research has demonstrated that expert drivers are superior at detecting potential hazards when compared with non-expert drivers. However, hazard perception is a multifaceted skill involving at least three components: drivers must look at the hazard, then detect it, and finally appraise it and respond appropriately. In the current study, we explored how expert drivers (paramedics, n = 151) and non-expert drivers (n = 189) detect hazards of different threat value. To explore this question, we used a static, driving-related inattentional blindness (IB) task, in which an unexpected object in a critical trial varied from high threat (child running onto the road) to medium threat (pedestrian standing by the road) to low threat (garbage bin next to the road). We hypothesised that experts would have heightened awareness of hazards, which could be reflected as either generally higher rates of noticing objects in the driving scene (lower IB overall), or a heightened ability to prioritise the threat value of objects in the scene (lower IB for high threat, but not low threat objects). The results demonstrated that approx. 90% of drivers, irrespective of expertise, detected high threat objects placed on the side of the road. However, experts were more likely than non-experts to detect medium threat objects (55% of expert drivers vs. 40% of non-expert drivers), whereas the opposite pattern occurred for low threat objects (almost 20% of non-expert drivers noticed low-threat objects, compared with none of the expert drivers). We argue that expertise allows drivers to calibrate a hierarchy of attentional filtering to not only direct attentional resources to locations of interest, but also to explicitly prioritise objects of interest when driving. Importantly, this appears to be due to training rather than years of experience. These results point to the importance of not just increasing awareness while driving, but to develop discriminative capacity to filter out what is unimportant to facilitate safe driving. Crown Copyright Ó 2018 Published by Elsevier Ltd. All rights reserved.

⇑ Corresponding author. 1

E-mail address: [email protected] (K. Pammer). Now at the University of Newcastle, NSW 2308, Australia.

https://doi.org/10.1016/j.trf.2018.09.020 1369-8478/Crown Copyright Ó 2018 Published by Elsevier Ltd. All rights reserved.

390

K. Pammer et al. / Transportation Research Part F 59 (2018) 389–400

1. Introduction Safe driving is predicated on attending to objects that are important in the environment, but also filtering out what is unimportant. For example, it is essential to process a child standing on the side of the road, but less important to note a garbage bin by the road. Failing to detect critical objects when driving are estimated to constitute approximately 5% of all crashes (Stutts, Reinfurt, Staplin, & Rodgman, 2001), and around 9% of crashes involving serious injury (Beanland, Fitzharris, Young, & Lenné, 2013). Research has investigated which qualities of the visual scene are important for capturing attention (McCarley, Steelman, & Horrey, 2014), and how performance is mediated by experience (Chapman & Underwood, 1998; Crundall et al., 2012; Crundall, Stedmon, Crundall, & Saikayasit, 2014; Underwood, Crundall, & Chapman, 2002). Relatively less research has been devoted to understanding the role of specific expertise – such as that demonstrated by emergency first responders – in managing attentional filtering when driving (Johnston & Scialfa, 2016). Identifying and describing differences in the attentional filtering of expert drivers is likely to be enormously helpful in the development of driver training programs. The use of hazard detection or hazard perception tests has become ubiquitous in the driving literature to investigate an observer’s ability to detect hazards and critical objects in a driving scene. However, hazard perception is a complex and multi-faceted process, which involves several stages. First, drivers must look at the location of a potential hazard, then they must detect it, and finally they must appraise it correctly to formulate an appropriate response (Crundall, Clarke, Ward, & Bartle, 2008). Hazard perception tests provide an overall measure of drivers’ ability to appraise the hazard, but if a participant fails to respond correctly, it is unclear which stage of processing was inadequate. For this reason, complementary techniques have been adopted to explore driver’s ability to detect expected or unexpected objects in the driving environment, such as change blindness (e.g., Beanland, Filtness, & Jeans, 2017; Galpin, Underwood, & Crundall, 2009; Harms & Brookhuis, 2016) and visual search tasks (e.g., Beanland, Lenné, & Underwood, 2014). These methods all focus on the ‘‘detection” stage of hazard perception, by placing potential hazards in full view. One potential issue is that in these paradigms, participants are explicitly instructed to seek out and identify objects or changes. This creates a slightly artificial experience, as participants respond to a greater number of hazardous objects than they would during real driving. An alternative experimental technique is the inattentional blindness (IB) paradigm (Mack & Rock, 1998), which involves presenting an unexpected object in the participant’s field of view. Because the critical stimulus of interest is unexpected, IB paradigms provide a nice complement to other paradigms in which the observer is deliberately searching for something. In a laboratory-based IB tasks, using abstract geometric shapes as stimuli, up to 100% of participants may fail to detect an unexpected stimulus on the screen if they are engaged in another task (Mack & Rock, 1998; Most et al., 2001). The exact rate of participants who experience IB will vary depending on task parameters and other factors such as the participant’s age and expertise (Furley, Memmert, & Heller, 2010; Horwood & Beanland, 2016; Memmert, 2006). In a driving version of IB, participants are presented with photographs of driving scenes and asked to make a drivingrelated judgement, such as assessing the safety of the driving situation. IB occurs when the participant fails to detect an additional object placed in the scene that has not been present in any of the other scenes (Pammer & Blink, 2013; Pammer, Bairnsfather, & Burns, 2015; Pammer, Sabadas, & Lentern, 2017). The participant is not told to look for an additional hazard; it simply appears as part of the driving scene while the participant is making other driving-related judgements. This could be considered analogous with real driving during which hazards sometimes occur unexpectedly, such as a child running out onto the road, or someone in a parked car suddenly opens a door toward oncoming traffic. The participant’s ability to report the presence of an object is a good indicator of whether an object has attracted the participant’s attention, and most importantly, IB tasks mimic looked-but-failed-to-see (LBFTS) errors (Pammer & Blink, 2013; Pammer et al., 2015, 2017). 1.1. Looked-but-failed-to-see-crashes and inattentional blindness LBFTS errors involve a situation experienced by almost all drivers at some point, where they look at oncoming traffic for a clear manoeuver but fail to see an oncoming vehicle that was plainly in their line of sight. Sometimes considered an example of inadequate surveillance, looking but failing to see an oncoming vehicle has been attributed to 21% of crashes at intersections (Brown, 2002), 71% of surveillance errors made by elderly drivers (Cicchino & McCartt, 2015), and have been cited as one cause of crashes at rail level crossings (Rudin-Brown, George, & Stuart, 2014; Salmon, Read, Stanton, & Lenné, 2013). LBFTS crashes are also believed to be a major cause of incidents with bicycles (Herslund & Jørgensen, 2003) and powered two-wheelers such as motorbikes (ACEM, 2009; Clabaux et al., 2012; Pammer et al., 2017). Brown (2002) reports that most LBFTS incidents occur in the daytime under clear conditions, rather than at night-time. This is important as it suggests that LBFTS crashes are less to do with conspicuity and more to do with cognition; if a driver can look directly at a hazard in the driving situation, under clear, daylight conditions, and still fail to see it, then it implies that it is something about the way a driver thinks and approaches a driving situation that makes them more vulnerable to LBFTS crashes. If this is the case, then it stands to reason that a cognitive model offers the best way to understand LBFTS crashes, and subsequently mitigate the risk of them occurring.

K. Pammer et al. / Transportation Research Part F 59 (2018) 389–400

391

1.2. Inattentional blindness, driving and object relevance The likelihood of experiencing IB is predicted by several factors, including what the observer expects to see in the scene (Most, Scholl, Clifford, & Simons, 2005). Attention can be conceptualised as a filter that allows the progress of information that is important for a task at hand and blocks irrelevant information. Such an attentional filter is predicated on our knowledge of the environment and is constantly updated as we experience more exemplars of a situation (Neisser, 1976). Therefore when driving, we mentally collect information so as to quickly filter what is important versus unimportant. For example, experienced drivers are more likely to focus on events that might lead to a hazard, compared with novice drivers (Borowsky, Shinar & Oron-Gilad, 2010; Crundall et al., 2012; Pollatsek, Narayanaan & Fisher, 2006), suggesting that an attentional sweep of the driving environment allows experienced drivers to prioritise the relevance of information and predict upcoming events (Pammer et al., 2017). This is consistent with the finding elsewhere that drivers implicitly process information such as road signs (Charlton, 2006) in a way that changes their subsequent driving behaviour. The speed at which this filtering is done suggests that the unexpected object, irrespective of its importance or salience, undergoes some degree of attentional processing before we are consciously aware we have seen it. Thus, if an object or event is considered not relevant, it may never reach conscious awareness. Pammer et al. (2015) provided evidence for the occurrence of pre-attentive filtering in a driving-related IB study in which participants were asked to rate an image of a real driving scene as either safe or unsafe. After a series of images, one – the critical image – had an unexpected object placed on the side of the road, which had not been present in any of the preceding images. The level of danger each unexpected object posed to the driving environment was evaluated using questionnaires, and varied from high (a child running towards the road) to low (a garbage bin on the median strip). Experimental results revealed that rates of IB correlated with perceived threat, as assessed through the questionnaires: the child running towards the road was noticed by nearly all observers, whereas the garbage bin was rarely noticed. These results support the notion of pre-attentive processing to differentiate important and unimportant information. In a driving scenario, objects that pose a threat to safety are highly relevant to the driving task. Objects posing little to no threat to the safety of road users are ignored as unimportant (Borowsky & Oron-Gilad, 2013; Vlakveld et al., 2011). Nevertheless, although people were more likely to detect an ‘‘important” object on the side of the road compared with an ‘‘unimportant” object, up to 20% of people still failed to detect a child running by the side of the road, which is consistent with reported real-world lapses of attention where drivers fail to detect hazards and threats (Beanland, Fitzharris, et al., 2013; Stutts et al., 2001). Similarly, drivers are likely to miss road signs when they are placed on the ‘‘wrong” side of the road in an unexpected location (Borowsky, Shinar, & Parmet, 2008), and are more likely to miss motorbikes compared with other, more common vehicles (Beanland et al., 2014; Pammer et al., 2017). Charlton and Starkey (2013) demonstrated that as drivers became increasingly more familiar with a driving scene, they were commensurately less likely to notice changes or additions to the environment, with drivers often failing to detect other cars, police car warning signs and even buildings. However, again, whether a driver detected an object or change in the environment was related to its importance or relevance, with changes to line markings detected far more often than changes to surrounding buildings. 1.3. Expertise in hazard detection The literature does not distinguish well between ‘‘experienced” and ‘‘expert” drivers (e.g., see Borowsky & Oron-Gilad 2013 for discussion). Many studies refer to ‘‘experts” as drivers who have had experience in driving, even if they have not been tutored, mentored and assessed in this skill. Whilst clearly there is a relationship between experience and expertise, there are independent elements, such that one can have considerable driving experience but lack skill, and conversely, considerable skill can be acquired quite quickly (Shanteau, Weiss, Thomas, & Pounds, 2002). Moreover, evidence suggests that there are distinct differences between an experienced driver and an emergency responder, such as paramedics and police pursuit drivers, who have been specifically taught advanced driving skills and had their competence assessed (Crundall, Chapman, France, Underwood, & Phelps, 2005; Johnston & Scialfa, 2016; Pammer & Blink, 2018). In terms of expert drivers, an Australian government report found that paramedics have the lowest crash rate among emergency response drivers, constituting only 156 (10%) of emergency service vehicle accidents from 1996 to 2000 (Symmons et al., 2005). Similar to other emergency service drivers, ambulance officers’ attention to driving must compete with operational challenges including operating a hand-held radio, and multiple competing auditory signal (sirens, dispatch communications, pagers), and visual cues (mobile data terminals, vehicle occupants), all while they navigate at high speed to unfamiliar locations. The evidence suggests that expert drivers are particularly skilled at perceiving potential hazards on or around the roadway (Crundall et al., 2005; Crundall, Chapman, Phelps, & Underwood, 2003; Horswill, Taylor, Newnam, Wetton, & Hill, 2013). In one study, police drivers were compared with age and experience matched controls, and also with novice drivers. The police drivers demonstrated a much broader range of scanning (Crundall et al., 2003). Moreover, the police drivers had higher electrodermal responses to hazards (a proxy measure used to identify physiological, involuntary responses to arousing stimuli), suggesting that they were subconsciously processing a greater range of potentially important stimuli than the matched controls and novices. As drivers gain more experience on the roads it is likely they will have negotiated a broader range of events, each one differing in levels of hazardousness (Underwood, Ngi, & Underwood, 2013). As a result, the driver becomes equipped with the skills to predict when and where a potential hazard might occur (Fisher, Pollatsek, & Pradhan, 2006; Johnston & Scialfa, 2016; Pradhan, Pollatsek, Knodler, & Fisher, 2009). However, a typical experienced driver may have

392

K. Pammer et al. / Transportation Research Part F 59 (2018) 389–400

many hours behind the wheel, resulting in a strong driving-related mental framework, but the similarity of the repeated driving experiences mean that their driving framework may be strong but limited in scope, thus lacking flexibility. This can be problematic in situations when driving circumstances change, and the experienced driver fails to respond to the change, relying on a limited behavioral repertoire (Salmon et al., 2013). Conversely, expert drivers are trained to drive in varied and complex environments. This is particularly the case for emergency service workers who regularly drive under varied and sometimes extreme conditions, speeds and environments. In the Australian Capital Territory (the jurisdiction where the current research was conducted), there are 25 ambulances, and each has an average annual mileage of 50,000 km. The most recent crash statistics indicate that emergency service vehicles (which also includes police and fire response vehicles) were involved in only 0.17% of all crashes (ACT Government, Justice and Community Services, 2017). Thus, it has been suggested emergency response drivers are more efficient, quicker and accurate at perceiving hazards and scanning their environment because they have developed a highly sophisticated mental representation of their driving environment (Horswill et al., 2013). These mental representations share similarities with the concept of schemata: organized mental patterns that affect how information in the world is perceived and reacted to, by providing a person with knowledge about the actions required for a specific task, based on previous experiences (Neisser, 1976). All drivers use a driving schema to provide them with the resources to interpret and navigate traffic events such as near accidents or hazardous intersections (Underwood, Chapman, Bowden, & Crundall, 2002). Experts are suggested to have a more detailed and nuanced driving schema as a function of the wealth of experience that occurs from experience in dealing with unusual or dangerous driving situations or (in the case of emergency responders) driving at high speeds. Experts can then use these schemata to direct visual search (Crundall et al., 2005), predict what will happen next (Horswill et al., 2013) and efficiently perceive hazards on or around the roadway (Johnston & Scialfa, 2016). 1.4. Current study Previous research has demonstrated drivers are reasonably good at rapidly prioritising information to attend to objects that are most important in the driving situation (e.g., Garrison & Williams, 2013). What remains unknown is precisely how expert drivers calibrate their attentional filter when driving. Is an expert driver simply good at detecting everything in the driving environment compared with a non-expert driver, or are they good at discriminating the importance of objects and re-allocating their attention accordingly? The few studies that have targeted expert drivers have not considered this question, and its resolution is important for understanding attentional allocation for all drivers. In the current study, we use a static driving-IB task in which the additional object varies according to its relevance to the driving situation (Pammer & Blink, 2013, 2018; Pammer et al., 2015, 2017). If expert drivers tune their attentional filter to be less discriminating to detect a wide range of potential hazards, then experts should be better than non-experts at detecting additional objects on the side of the road, irrespective of the object’s threat value (i.e., less IB overall). However, if experts calibrate their attentional filter to be more discriminating, then compared with non-experts, expert drivers should be more likely to detect moderate to high hazard objects, but not low-hazard objects (i.e., less IB for high threat objects only). This finding would imply that attentional calibration is potentially malleable, and therefore trainable in a way that could optimised for better hazard detection for all drivers. 2. Methods 2.1. Participants The participants consisted of 151 expert drivers and 189 non-expert drivers. The expert participants had completed required paramedic driver training. Paramedic officer driver training varies between jurisdictions but cover the same basic requirements, including skid-pan training, steering, braking and acceleration, visuospatial processes such as vision and hazard perception, as well as urgent duty driving. Paramedics in the current jurisdiction would perform approximately 2–5 Priority-1 (P1; lights and sirens) runs each 10-hour shift, and P1 arrival time is required to be within three minutes of receiving the call. Full driver training occurs when a new recruit into the service and if an officer changes jurisdictions. After training officers are then mentored under all driving conditions for approximately 3 months. The non-expert participants were recruited at a local public science museum in Canberra, Australia. All participants had current Australian driver’s licenses and reported normal or corrected to normal vision. All participants gave consent to participate and the research had institutional ethics approval (ANU Protocol 2013/445). Refer to Table 1 for demographic information. 2.2. Stimuli The task was adapted from Pammer et al. (2015). The IB stimuli were presented on either a laptop or projected onto a larger screen. The critical stimuli presented on the laptop subtended an approximate visual angle (VA) of 3°  1.5° and were 9° VA from the centre of the screen. The laptop screen had a refresh rate of 60 Hz and a screen resolution of 1920  1080

393

K. Pammer et al. / Transportation Research Part F 59 (2018) 389–400 Table 1 Summary of group demographics. Child Run

Child Stand

Adult Stand

Stroller

(N)

E (29)

NE (36)

E (34)

NE (56)

E (31)

NE (39)

E (25)

NE (22)

E (32)

NE (37)

Age in years

43.7 (7.9) 25.9 (8.4)

42.1 (10.1) 24.1 (11.1)

40.5 (9.9) 23.5 (7.9)

41.9 (10.7) 22.5 (9.9)

40.5 (8.3) 23.2 (7.9)

43.5 (11.5) 23.5 (12.2)

41.7 (9.5) 24.9 (10.1)

42.8 (11.1) 24.6 (12.7)

43.5 (11) 25.1 (10.7)

4.1 (4.2) 31

–

3.1 (2.9) 23

1.9 (1.4) 38

3.6 (2.9) 33

3.5 (3.6) 23

4.7 (5.5) 38

–

3.0 (4.2) 35

Driving Experience (years) Collisions over 10 yearsa Genderb %male

35

Bin

21

Main effect group Sig.

Main effect cond Sig.

Interaction Sig.

41.4 (10.7) 23.2 (10.7)

0.76

0.88

0.54

0.45

0.76

0.97

–

0.32

0.39

0.31

26

–

–

0.003

E = Expert participants, NE = Non Expert participants. a Sample size for non-experts was 28 as this data was not collected for most of the non-experts. b Chi square test used for between-groups comparison.

pixels. The projected stimuli were on average 2.5  0.9° VA and approximately 6.5° VA from the centre of the screen. The projected screen had a resolution of 1600  900 pixels. Each trial involved displaying of a high-resolution image of a driving scene taken from the driver’s perspective, and lasted 3000 ms: a black fixation cross on a grey screen was presented for 500 ms, followed by the driving image presented for 1500 ms, then a pattern mask of randomly crossed short black lines on a grey background presented for 1000 ms (see Fig. 1). There were three practice trials followed by five non-IB trials, then the critical IB trial, and a full attention trial. In the critical IB trial an unexpected object was added into the image on the median strip (for example a stroller or a child) that was not included in any of the previous trials. All of the images for all trials were of relatively normal driving situations such that no one image was any more hazardous than any other image. Five different unexpected objects constituted the five possible conditions to which participants were randomly assigned in a between-subjects design. The possible unexpected objects were a child standing (A), a child running (B), an adult standing (C), a bin (D), or a stroller (also known as a ‘buggy’ or ‘pram’) (E). Each possible object was located on the median strip to the right (refer to Fig. 1). The data for the child standing (A) and adult standing (C) conditions were ultimately collapsed in the results as IB rates were equivalent. We originally anticipated that participants would identify each as ‘child’ and ‘adult’ with different levels of risk; however, participants commonly classified both simply as a ‘‘pedestrian”, ‘‘person on side of road” or ‘‘someone on side of the road”. The unexpected object stimulus dimensions and colours were matched as closely as possible for all five conditions, and all stimuli have been used elsewhere (Pammer et al., 2015). Although Pammer et al. (2015) demonstrated small differences in the visual saliency between these different unexpected objects, this is inevitable when using naturalistic stimuli and physical saliency is not correlated with noticing the unexpected object (Humphrey & Underwood, 2009; Pammer et al., 2015). The unexpected objects have also been shown to vary in terms of danger to the driving situation (Pammer et al., 2015) with the child running being the most dangerous, followed by the stroller, the child standing and adult standing, and lastly, the garbage bin. Similar effects have been demonstrated elsewhere using different stimuli (e.g., Borowsky & Oron-Gilad, 2013). 2.3. Procedure For the expert drivers, testing was available to all paramedics in the ACT and was conducted during their in-service training in a classroom setting with 5–9 people. They moved through the experiment, paced by a research assistant so that everyone answered each question at the same time. For non-experts, the task was completed and was self-paced such that the participant moved through trials by hitting the space bar. We have run this task over a number of years in group, classroom settings and in individual settings with no difference in the results.2 Participants completed a demographic questionnaire prior to the IB task. They were told that we were interested in what people pay attention to when driving. Specifically, participants were told that they would see a series of driving images taken from the driver’s perspective and they would need to make judgements about the safety of the scene they saw, both for themselves as ’drivers’ and for other road users. After the presentation of each trial the participant was asked to indicate whether it was a safe or unsafe driving situation, when considered from the driver’s perspective, and to provide reasoning for their answer. It was stressed that there were no right or wrong answers and the experimenter was just interested in their opinion. A standard response might be ‘‘it was unsafe as there was a car ahead crossing my path”. It is not a task designed to simulate driving, but rather simulate some of the same decisions that are made

2 For example, here the ‘pedestrian’ stimulus condition gives a rate of noticing of 50% when presented individually. When we ran the same condition in a classroom situation as a pilot study, the rate of noticing was 51%.

394

K. Pammer et al. / Transportation Research Part F 59 (2018) 389–400

Fig. 1. Sequence of images in the static IB task. The images on the left side show an example of a non-IB trial (labelled ‘one trial’). The aim of the eight precritical trials (including 3 practice trials) is to engage the participant in the task. The critical trial, which contains the unexpected object, appears on the right side, and the critical trial occurs only once (adapted from Pammer et al, 2015, p. 784). The unexpected object varied between-subjects and could be a child standing (A), a child running (B), an adult standing (C), a bin (D), or a stroller (E).

under temporal limitations when driving. Although the safe/unsafe decision could also be considered priming for a hazard, we have demonstrated that other decisions like ‘‘what is the approximate time of day” gives the same results as the safety judgement (Pammer et al., 2015). Following the critical trial, which included the unexpected object, the participant was asked the safe-unsafe question as well as if they had noticed anything additional in this trial, other than trees, buildings or streetlights which had not been seen in the previous trials, and to describe it if possible. A forced choice selection sheet, consisting of eight objects that could plausibly be by the side of the road (household waste bin, commercial waste bin, stroller, adult standing, child standing, child running, motorcycle, dog) was shown and the participant was required to pick the image that matched what they had seen. If a participant reported not seeing anything additional in the critical trial, they were asked to guess which object might have appeared from the forced choice sheet. None of the objects had been presented during the pre-critical trials. The last trial was a full attention trial, which was a repeat of the critical trial. Participants were instructed to view the image without making a safety judgement afterwards, and to simply watch the next trial. The participant was again asked whether anything additional had been included in the trial, and this was followed by the forced-choice sheet. Full attention trials are used to determine that the unexpected object is clearly detectable to a participant when their attention is not engaged elsewhere (Mack & Rock, 1998). If a participant fails to detect the additional object on the side of the road in the full attention trial, then we cannot be sure that a failure to detect an object in an IB trial is in fact due to IB. All participants detected the additional object on the full attention trial. Participants only experience one critical trial, because after that the task becomes a divided attention task; that is, participants are primed to look out for an additional object. Thus, participants all saw the same pre-critical trials, but the unexpected object that appeared in the critical trial of each participant (either the child running, child standing, adult standing, stroller or garbage bin) was a between-subject variable, presented only once. The five different category conditions were grouped into clusters according to threat level based on previous research (Pammer et al., 2015). The cluster groups were; high (child running), medium-high (stroller), medium-low (pedestrian standing) and low (garbage bin) threat.

395

K. Pammer et al. / Transportation Research Part F 59 (2018) 389–400

3. Results 3.1. IB analyses Participants were coded as noticers if they correctly described the additional object and they could correctly identify it from the forced-choice sheet. Participants were coded as non-noticers if they claimed to see something but could not identify the correct object, or if they did not report an unexpected object (even if they subsequently guessed the correct item using the forced-choice sheet). To explore whether IB may be associated with drivers having greater risk perception (i.e., a stronger tendency to rate scenarios as ‘unsafe’), the proportion of times participants classified a trial as ‘‘safe” was calculated over 7 trials. This calculation included the critical trial; however, data was also calculated excluding the critical trial and the results were the same. This score comprised the safety judgement score, which was used as a predictor in the logistic regression. Expert drivers were significantly more likely to indicate that a situation was unsafe, compared with non-expert drivers t(338) = 7.48, p < .001 Binary logistic regression was used to assess which variables predicted the dependent variable (DV) of noticing the unexpected object. Initially we planned to run a two-step regression, with main effects for Condition (i.e., type of unexpected object) and Group (expert vs. non-expert drivers) entered as categorical predictors and Safety Judgement as a continuous predictor at the first step, and interaction terms entered at the second step, in accordance with Hosmer, Lemeshow and Sturdivant’s (2013) guidelines for structuring logistic regression. However, the Condition  Group interaction terms could not be reliably assessed, as one cell had zero observations for the expert drivers (refer to Fig. 2). As such, the Condition  Group interaction was explored using a series of chi-square tests, which were assessed using a Bonferroni-adjusted alpha level of 0.0125. As shown in Table 2, Condition predicted noticing. Note that odds ratios < 1 indicate observers were less likely to detect the unexpected object, relative to the medium-threat reference condition, and odds ratios >1 indicate observers were more likely to notice it. Thus in general, noticing the unexpected object was commensurate with the level of assumed threat that it represented. There is also a borderline significant effect of group, where an odds ratio of <1 indicates that non-expert drivers were less likely to detect the additional object relative to the expert drivers.

Fig. 2. Percent of observers in each condition who noticed the unexpected object. Significant differences in rates of noticing between Experts and nonexperts are starred.

Table 2 Summary of logistic regression analysis with noticing the unexpected object as the DV, over the categorical IB conditions. Predictor Experimental Condition Garbage bin (low threat) Stroller (med-high threat) Child running (high threat) Group Safety Judgement Group  Safety Judgement

B 2.23 0.84 2.31 2.12 0.29 0.36

SE

Wald

p

Odds ratio (OR)

95% confidence intervals OR

0.43 0.36 0.46 1.09 0.19 0.22

64.76 26.43 5.39 25.07 3.77 2.49 2.79

<.001 <.001 .020 <.001 .050 .114 .095

0.11 2.32 10.03 0.12 0.75 1.43

[0.05, [1.14, [4.06, [0.14, [0.52, [0.94,

Note. ‘‘Pedestrian” was used as the reference group for Condition, and ‘‘Expert” was used as the reference group for Group.

0.25] 4.72] 24.74] 1.02] 1.07] 2.20]

396

K. Pammer et al. / Transportation Research Part F 59 (2018) 389–400

The tendency to report a scenario as safe (over all the scenarios), did not predict noticing, and there was also no significant group by ‘safety judgement’ interaction. This suggests that although experts appeared to be overall more ‘hazard aware’ than non-expert drivers, this is unlikely to explain the differences between the two expert groups in terms of noticing the unexpected objects. Chi-square tests were conducted to compare expert and non-expert drivers in each condition. In the high-threat child running condition, there was no difference in rates of noticing between experts and non-experts; v2(1, N = 92) = 0.14, p = .83. Similarly, the results for the stroller condition were identical for experts and non-experts (statistical confirmation would be redundant). In the condition where there was either an adult or a child standing on the side of the road (Pedestrian condition), experts appeared to be more likely to notice the stimulus compared with non-experts v2(1, N = 160) = 5.2, p = .01. However, in low-threat conditions, experts were less likely to notice the additional object v2(1, N = 66) = 7.6, p = .006. Refer to Fig. 2. 3.2. Forced choice analysis To explore the possibility of pre-conscious processing of the additional object stimulus, we examined whether participants could correctly identify the additional object from a forced-choice array, even if they report not noticing it in the experimental trial. Among participants who did not notice the unexpected object, experts (23%) were significantly more likely than non-experts (10%) to accurately choose the correct option in the forced choice trial, v2 (1, N = 163) = 4.68, p = .031, irrespective of the stimulus condition. 3.3. Logistic regression for expert drivers Another regression analysis was conducted which included only expert drivers, where noticing was the DV. This was to explore whether the differences observed in noticing between the experts and non-experts is due to an intrinsic difference between them, or whether expert noticing could be attributed to any of the other recorded factors such as experience, gender or number of training courses. The IVs were: time since paramedic driver training (operationalised as the time since they did their first driver training as a paramedic or equivalent), gender, number of fines, number of hours spent driving each week, number of collisions, and driving experience in years. None of these variables, as listed in Table 3, predicted noticing the unexpected object for expert drivers. This implies it is the fact that experts are trained, not how much training or experience that they have, which accounts for the differences between experts and non-experts here (see Table 3). 4. Discussion The aim of the current study was to examine the differences between expert and non-expert drivers in terms of their attentional allocation to hazard detection, using a static driving IB task. Here, ‘experts’ were operationalized as emergency service responders – specifically paramedics, who had received explicit driver training that involved on-road vehicle control, attentional awareness, and continuous driver evaluation and mentoring. Non-experts were ‘normal’ drivers who had received no explicit driver training. The groups were equivalent in terms of age range, and general driving experience which allowed us to assess the contribution of expertise over experience. Although the number of hours driving did not predict noticing for the experts, we were not able to report this value for non-experts, therefore we cannot preclude the possibility that this might be an important factor for the non-expert drivers. We also assumed that although it is likely there are dispositional differences between the expert and non-expert groups, it is unlikely these trait personality differences would influence the results, i.e., paramedics are likely to have entered their profession due to medical or patient care reasons rather than for the chance to engage in high-level driver training. We hypothesised that experts would be more risk-averse, resulting in one of two possible outcomes: either experts would have a heightened general awareness resulting in higher rates of noticing all categories of objects, or experts may be more discriminating with a heightened ability to prioritise potentially risky objects in the scene. The results were consistent with the latter hypothesis.

Table 3 Summary of logistic regression analysis for the expert drivers, with noticing the unexpected object as the DV. Independent variable

B

SE

Wald

p

OR

95% CI OR

Time since paramedic driver training Gender Fines Hours driving (per week) Collisions Driving experience Number of driving courses

0.01 0.94 1.09 0.03 0.09 0.11 0.09

0.03 0.06 0.12 0.03 0.05 0.03 0.11

0.207 0.201 0.812 1.064 2.803 0.156 0.736

.649 .811 .368 .302 .090 .693 .391

0.99 0.98 0.88 1.03 1.09 0.99 0.91

[0.93, [0.42, [0.71, [0.97, [0.99, [0.94, [0.74,

1.05] 1.97] 1.14] 1.09] 1.21] 1.05] 1.13]

Note. ‘Time since ambulance qualification’ was taken as the time in years since their first driver training as an ambulance officer. In a small number of cases participants had undergone previous specialist training such as in the police force – in this case, time was taken from this initial specialist training. Driving experience was included rather than Age, as Age and Driving experience were highly correlated, r(151) = 0.98, p < .001.

K. Pammer et al. / Transportation Research Part F 59 (2018) 389–400

397

When comparing the different threat conditions, expert drivers were no different from non-expert drivers in high threat conditions such as a child running on the side of the road. This is perhaps not that surprising, one would expect all drivers to consistently detect high risk stimuli in the driving situation. Our data demonstrate more than 80% of non-experts and almost all experts noticed the child running on the side of the road. The real point of difference emerged in the low-medium threat condition, with pedestrians on the side of the road. Here the experts were significantly more likely to detect the pedestrian, compared with non-experts. Finally, in the low threat condition where a garbage bin was on the side of the road, not one expert reported noticing the object, whereas almost 20% of non-experts noticed it. This result is important because it reveals that experts are not just better than non-experts in detecting medium-range threats in driving scenarios, they are also better at filtering out low-threat objects. Moreover, when considering only the expert drivers, none of the demographic variables predicted noticing. Thus, the differences observed are unlikely to be due to the fact that expert drivers simply do more driving as part of their job, because if this were the case we would expect work-related driving years would predict noticing the unexpected object, and this was not the case. Instead, our data suggest that there is something qualitatively different about how expert drivers attend to the scene compared with non-expert drivers, irrespective of their age, experience or qualifications. This difference in attentional filtering seemingly allows them to not only detect what is important in the driving scene, but also reject and thus ignore what is unimportant. In driving research and driver training, much of the emphasis is on detecting hazards, and we tend not to focus on the flipside of this equation which is that it is not only important to learn to identify potential hazards, but also to quickly identify what is not important so as to most effectively deploy cognitive resources. Certainly, it has been demonstrated elsewhere that efficient cognitive processing is based on being able to filter out irrelevant information. For example, recent research has demonstrated that selective attention training improves decision making (Schmicker, Müller, Schwefel, & Müller, 2017). The explanation is typically that selective attention training frees up cognitive resources to be deployed elsewhere. Consistent with this, Lansdale, Underwood, and Davies (2010) compared ordinance survey experts to non-experts in their processing of aerial photographs. The most striking difference between the experts and non-experts was in their respective eye movements; non-experts were far more likely to direct their attention to physically salient parts of the image, whereas non-experts appeared to be able to shift their attention away from such salient attentional capture. Moreover, although experts and non-experts processed the same number of stimuli within the first few seconds, the non-experts continued to process objects over the remaining 12 s, whereas experts reduced the number of objects they processed. These results suggest that experts engage in a faster and more efficient processing of the search space. Indeed, in driving, the noticeable shift in the transition to novice driving in terms of spread of fixations (Mourant & Rockwell, 1972), the adaptability of fixations to the environment (Underwood, Chapman, Brocklehurst, Underwood, & Crundall, 2003), and the time it takes to fixate on hazards (Crundall et al., 2012), speaks to a commensurate shift in the ability to effectively deploy attentional resources as drivers become more experienced. It seems likely that expert drivers have developed a more sophisticated and nuanced driving schema allowing them to predict outcomes and thus be more discriminating in terms of their attentional allocation on the basis of potential consequences that objects represent. Given finite cognitive resources, it is cognitively more efficient to direct resources to those objects that have a higher probability of potential impact. Crundall et al. (2005) demonstrated that expert drivers such as police pursuit drivers are more likely to direct their attention to areas in a visual scene that may be more likely to represent a potential hazard. Our findings complement Crundall et al.’s study by demonstrating that expert drivers are not only likely to direct attentional resources to locations of interest, but also to explicitly prioritise objects of interest when driving. We tentatively suggest that attentional filtering occurs at a pre-attentive level, that is, at a level preceding conscious awareness. This hypothesis was based on an interpretation from Pammer et al.’s (2015) paper showing that the coding required to assign an ‘importance’ rating to the unexpected objects might be occurring at a pre-attentive level. More support for the pre-attentive coding theory comes from earlier studies by Mack and Rock (1998) who suggested that priming effects might be indicative of unattended stimuli being subconsciously perceived and encoded. Mack and Rock (1998) conducted priming experiments, using word stem completion tasks, and successfully demonstrated that supposedly undetected stimuli were in fact positively influencing the participant’s performance in the word stem task. There is also evidence from more recent research to suggest that even without conscious awareness a substantial amount of processing still occurs (Lathrop, Bridgeman, & Tseng, 2011). This ‘pre-conscious’ processing appears to provide enough information to the participant to allow them to correctly identify the unexpected object in a forced choice trial, even if the unexpected object never reaches conscious awareness (Lathrop et al., 2011). This proposal is consistent with the contextual guidance model of attentional selection (Torralba, Oliva, Castelhano, & Henderson, 2006) which posits a fast, pre-saccadic, parallel processing of contextual information in a scene. In this, it is proposed that a global representation of scene features provides a cognitive shortcut for subsequent attentional allocation to scene objects. It is possible that top-down control in the form of driving-specific schemas distinguish between expert and non-expert drivers by modifying global context features. This model has been applied to other areas of expertise such as medical imaging (see Sheridan & Reingold, 2017 for a review), and lends itself nicely to the driving context. In doing so, it offers a number of exciting opportunities for further research to explore the cognitive processes underlying attentional allocation and selection in driving. The forced choice component of the current research offered a unique opportunity to test the prediction that stimuli may be pre-attentively coded and filtered. The results showed that experts were significantly more likely to correctly identify the image (that is, the image of the unexpected object they failed to detect in the critical trial) from the forced choice selection sheet, despite reporting they noticed nothing. This pattern did not occur with the non-experts, who showed no significant

398

K. Pammer et al. / Transportation Research Part F 59 (2018) 389–400

preference for selecting one stimulus over another. This finding supports our argument that attentional filtering is occurring at a pre-attentive level, particularly in the expert drivers who are expected to have more nuanced and specific attentional sets for both driving, and driving related stimuli. Further research would be required to substantiate this ‘pre-attentive’ claim; such a design would need to be able to elicit the function by which experts are selectively filtering out unimportant information yet are still able to correctly identify unimportant stimuli in a later trial. Although the experts here are experienced in varied driving situations that no doubt contribute to a highly nuanced and sophisticated driving schema, the findings nevertheless have implications for driver training programs. The range and value of driver training in general, has been considered elsewhere (e.g., Beanland, Goode, Salmon, & Lenné, 2013), however for the sake of the current study it is clear that some models of training may be able to help shift the balance of attentional allocation when driving, which appears to be the critical cognitive factor demonstrated in the current study. For example, Walker, Stanton, Kazi, Salmon, and Jenkins (2009) compared a group of drivers who had undergone an 8-week advanced driver training course, to drivers who had not undergone training. They demonstrated that from Time 1 to Time 2 (the start and end of the 8-weeks), the ‘trained’ group of drivers did not differ so much in the number of environmental driving elements they reported, but they did differ in complexity with which the elements were linked conceptually, suggesting an increase in network connectivity and higher situational awareness. However – crucially for the current study – they demonstrated that those information items that changed from T1 to T2 in the ‘trained’ group, were information items that were more strongly linked to other information items. In other words, the trained drivers re-prioritised the elements that they attended to, based on the element’s importance. The study also provides indirect support for the value of reinforcing training paradigms. Expert drivers such emergency first responders such as police and paramedics, regularly maintain their skill level with peer monitoring and high-demand driving. The message here is that regular ‘refresher’ courses may be beneficial to the average driver in order to consolidate good driving practices. Even if non-expert drivers participate in post-license driver training courses, there is a tendency for those skills to decline over time presumably as drivers fall back into well-worn repertoire of driving behaviours. 5. Summary Overall, the results from the current study suggest that expert drivers differ from non-expert drivers in terms of their attentional allocation in driving-relevant scenarios. When the threat value of an unexpected object in the driving scene is high, such as a child running toward the road, both expert and non-expert drivers are highly likely to detect it. In contrast, when the threat value is more moderate, such as a pedestrian standing by the road, experts are significantly more likely to detect it. And finally, when there was a benign, non-threatening object on the side of the road, such as a garbage bin, not one expert noticed it, whereas almost 20% of non-experts noticed it. Moreover, when all participants who failed to notice the unexpected objects were given the forced choice option to select an object that might have been there, experts were significantly more likely than non-experts to correctly choose the item that they had previously failed to notice. These results are consistent with the notion that expert drivers learn to prioritise information in a way that allows them to selectively attend to objects that are most important, but filter out objects that are not important. These results have implications for the continued development of driver training programs, by advocating for a focus on developing more nuanced models of object identification when driving rather than simply encouraging a broad visual scanning of the driving scene. Given that visual information is filtered differently by experts compared with non-experts, driver training could benefit from explicitly training drivers in their ability to discriminate the type of hazards. Task irrelevant information is more readily disregarded at a pre-attentional level by experts when compared with non-experts, suggesting differing driving situation awareness and schemata. These appear to develop through focussed driving training rather than generic driving experience. Thus conscious rehearsal of the process of locating and targeting those scene elements that pose a hazard risk could hasten the development of driving schema that efficiently filter information, so that safety-relevant elements are consistently targeted. Acknowledgements This research was funded in part by grants from the NRMA-ACT Road Safety Trust, Australia and the Australian Research Council LP13010081. The authors would like to thank the Australian Capital Territory Ambulance Service and Questacon National Science and Technology Centre for their support in this project. References ACEM (Association des Constructeurs Européens de Motocycles). (2009). In-depth investigations of accidents involving powered two-wheelers (MAIDS). Retrieved from http://www.maids-study.eu/pdf/MAIDS2.pdf. ACT Government, Justice and Community Services (2017). 2016 ACT Road Crash Report. Canberra, Australia: ACT Government. Retrieved from: http://cdn. justice.act.gov.au/resources/uploads/JACS/PDF/ACT_Road_Crash_Report_2016_web_accessable.pdf. Beanland, V., Filtness, A. J., & Jeans, R. (2017). Change detection in urban and rural driving scenes: Effects of target type and safety relevance on change blindness. Accident Analysis and Prevention, 100, 111–122. https://doi.org/10.1016/j.aap.2017.01.011. Beanland, V., Fitzharris, M., Young, K. L., & Lenné, M. G. (2013). Driver inattention and driver distraction in serious casualty crashes: Data from the Australian National Crash In-depth Study. Accident Analysis and Prevention, 54, 99–107. https://doi.org/10.1016/j.aap.2012.12.043. Beanland, V., Goode, N., Salmon, P. M., & Lenné, M. G. (2013). Is there a case for driver training? A review of the efficacy of pre- and post-licence driver training. Safety Science, 51(1), 127–137. https://doi.org/10.1016/j.ssci.2012.06.021.

K. Pammer et al. / Transportation Research Part F 59 (2018) 389–400

399

Beanland, V., Lenné, M. G., & Underwood, G. (2014). Safety in numbers: Target prevalence affects the detection of vehicles during simulated driving. Attention, Perception, and Psychophysics, 76(3), 805–813. https://doi.org/10.3758/s13414-013-0603-1. Borowsky, A., Shinar, D., & Oron-Gilad, T. (2010). Age, skill, and hazard perception in driving. Accident Analysis and Prevention, 42, 1240–1249. https://doi. org/10.1016/j.aap.2010.02.001. Borowsky, A., & Oron-Gilad, T. (2013). Exploring the effects of driving experience on hazard awareness and risk perception via real-time hazard identification, hazard classification, and rating tasks. Accident Analysis and Prevention, 59, 548–565. https://doi.org/10.1016/j.aap.2013.07.008. Borowsky, A., Shinar, D., & Parmet, Y. (2008). Sign location, sign recognition, and driver expectancies. Transportation Research Part F: Traffic Psychology and Behaviour, 11(6), 459–465. https://doi.org/10.1016/j.trf.2008.06.003. Brown, I. (2002). A review of the ‘looked but failed to see’ accident causation factor. London, UK: Department for Transport. Chapman, P. R., & Underwood, G. (1998). Visual search of driving situations: Danger and experience. Perception, 27(8), 951–964. https://doi.org/10.1068/ p270951. Charlton, S. G. (2006). Conspicuity, memorability, comprehension, and priming in road hazard warning signs. Accident Analysis and Prevention, 38(3), 496–506. https://doi.org/10.1016/j.aap.2005.11.007. Charlton, S. G., & Starkey, N. J. (2013). Driving on familiar roads: Automaticity and inattention blindness. Transportation Research Part F: Traffic Psychology and Behaviour, 19, 121–133. https://doi.org/10.1016/j.trf.2013.03.008. Cicchino, J. B., & McCartt, A. T. (2015). Critical older driver errors in a national sample of serious U.S. crashes. Accident Analysis and Prevention, 80, 211–219. https://doi.org/10.1016/j.aap.2015.04.015. Clabaux, N., Brenac, T., Perrin, C., Magnin, J., Canu, B., & Van Elslande, P. (2012). Motorcyclists’ speed and ‘‘looked-but-failed-to-see” accidents. Accident Analysis and Prevention, 49, 73–77. https://doi.org/10.1016/j.aap.2011.07.013. Crundall, D., Chapman, P., France, E., Underwood, G., & Phelps, N. (2005). What attracts attention during police pursuit driving? Applied Cognitive Psychology, 19(4), 409–420. https://doi.org/10.1002/acp.1067. Crundall, D., Chapman, P., Phelps, N., & Underwood, G. (2003). Eye movements and hazard perception in police pursuit and emergency response driving. Journal of Experimental Psychology: Applied, 9(3), 163–174. https://doi.org/10.1037/1076-898X.9.3.163. Crundall, D., Chapman, P., Trawley, S., Collins, L., Van Loon, E., Andrews, B., & Underwood, G. (2012). Some hazards are more attractive than others: Drivers of varying experience respond differently to different types of hazard. Accident Analysis and Prevention, 45, 600–609. https://doi.org/10.1016/j. aap.2011.09.049. Crundall, D., Clarke, D., Ward, P. & Bartle, C. (2008). Car Driver’s skills and attitudes to motorcycle safety: A review (Road Safety Research Report No. 85). London, UK: Department for Transport. Crundall, D., Stedmon, A. W., Crundall, E., & Saikayasit, R. (2014). The role of experience and advanced training on performance in a motorcycle simulator. Accident Analysis and Prevention, 73, 81–90. https://doi.org/10.1016/j.aap.2014.08.009. Fisher, D., Pollatsek, A., & Pradhad, A. (2006). Can novice drivers be trained to scan for information that will reduce their likelihood of a crash? Injury Prevention, 12, 25–29. https://doi.org/10.1136/ip.2006.012021. Furley, P., Memmert, D., & Heller, C. (2010). The dark side of visual awareness in sport: Inattentional blindness in a real-world basketball task. Attention, Perception, & Psychophysics, 72(5), 1327–1337. https://doi.org/10.3758/app.72.5.1327. Galpin, A., Underwood, G., & Crundall, D. (2009). Change blindness in driving scenes. Transportation Research Part F: Psychology and Behaviour, 12(2), 179–185. https://doi.org/10.1016/j.trf.2008.11.002. Garrison, T. M., & Williams, C. C. (2013). Impact of relevance and distraction on driving performance and visual attention in a simulated driving environment. Applied Cognitive Psychology, 27(3), 396–405. https://doi.org/10.1002/acp.2917. Harms, I. M., & Brookhuis, K. A. (2016). Dynamic traffic management on a familiar road: Failing to detect changes in variable speed limits. Transportation Research Part F: Traffic Psychology and Behaviour, 38, 37–46. https://doi.org/10.1016/j.trf.2016.01.005. Herslund, M., & Jørgensen, N. O. (2003). Looked-but-failed-to-see-errors in traffic. Accident Analysis and Prevention, 35(6), 885–891. https://doi.org/10.1016/ S0001-4575(02)00095-7. Horswill, M. S., Taylor, K., Newnam, S., Wetton, M., & Hill, A. (2013). Even highly experienced drivers benefit from a brief hazard perception training intervention. Accident Analysis and Prevention, 52, 100–110. https://doi.org/10.1016/j.aap.2012.12.014. Horwood, S., & Beanland, V. (2016). Inattentional blindness in older adults: Effects of attentional set and to-be-ignored distractors. Attention, Perception, & Psychophysics, 818–828. https://doi.org/10.3758/s13414-015-1057-4. Hosmer, D. W., Lemeshow, S., & Sturdivant, R. X. (2013). Applied logistic regression (3rd ed.). New York, NY: Wiley. Humphrey, K., & Underwood, G. (2009). Domain knowledge moderates the influence of visual saliency in scene recognition. British Journal of Psychology, 100 (2), 377–398. https://doi.org/10.1348/000712608X344780. Johnston, K. A., & Scialfa, C. T. (2016). Hazard perception in emergency medical service responders. Accident Analysis and Prevention, 95, 91–96. https://doi. org/10.1016/j.aap.2016.06.021. Lansdale, M., Underwood, G., & Davies, C. (2010). Something overlooked? How experts in change detection use visual saliency. Applied Cognitive Psychology, 24, 213–225. https://doi.org/10.1002/acp.1552. Lathrop, W. B., Bridgeman, B., & Tseng, P. (2011). Perception in the absence of attention: Perceptual processing in the Roelofs effect during inattentional blindness. Perception, 40(9), 1104–1119. 10.1068%2Fp6859. Mack, A., & Rock, I. (1998). Inattentional blindness. Cambridge: MIT Press. McCarley, J. S., Steelman, K. S., & Horrey, W. J. (2014). The view from the driver’s seat: What good is salience? Applied Cognitive Psychology, 28(1), 47–54. https://doi.org/10.1002/acp.2955. Memmert, D. (2006). The effects of eye movements, age, and expertise on inattentional blindness. Consciousness and Cognition, 15(3), 620–627. https://doi. org/10.1016/j.concog.2006.01.001. Most, S. B., Scholl, B. J., Clifford, E. R., & Simons, D. J. (2005). What you see is what you set: Sustained inattentional blindness and the capture of awareness. Psychological Review, 112(1), 217–242. https://doi.org/10.1037/0033-295X.112.1.217. Most, S. B., Simons, D. J., Scholl, B. J., Jimenez, R., Clifford, E., & Chabris, C. F. (2001). How not to be seen: The contribution of similarity and selective ignoring to sustained inattentional blindness. Psychological Science, 12(1), 9–17. https://doi.org/10.1111/1467-9280.00303. Mourant, R., & Rockwell, T. (1972). Strategies of visual search by novice and experienced drivers. Human Factors, 14(4), 325–335. https://doi.org/10.1177/ 001872087201400405. Neisser, U. (1976). Cognition and reality: Principles and implications of cognitive psychology. San Francisco: W.H. Freeman. Pammer, K., Bairnsfather, J., & Burns, J. (2015). Not all hazards are created equal: The significance of hazards in inattentional blindness for static driving scenes. Applied Cognitive Psychology, 29(5), 782–788. https://doi.org/10.1002/acp.3153. Pammer, K., & Blink, C. (2013). Attentional differences in driving judgments for country and city scenes: Semantic congruency in inattentional blindness. Accident Analysis and Prevention, 50, 955–963. https://doi.org/10.1016/j.aap.2012.07.026. Pammer, K., & Blink, C. (2018). Visual processing in expert drivers: What makes expert drivers expert? Transportation Research Part F: Traffic Psychology and Behaviour, 55, 353–364. https://doi.org/10.1016/j.trf.2018.03.009. Pammer, K., Sabadas, S., & Lentern, S. (2017). Allocating attention to detect motorcycles: The role of inattentional blindness. Human Factors, 60(1), 5–19. https://doi.org/10.1177/0018720817733901. Pollatsek, A., Narayanaan, V., & Fisher, D. (2006). Using eye movements to evaluate a PC-based risk awareness and perception training program on a driving simulator. Human Factors, 48, 447–464. https://doi.org/10.1518/001872006778606787. Pradhan, A., Pollatsek, A., Knodler, M., & Fisher, D. (2009). Can young drivers be trained to scan for information that will reduce their risk in roadway traffic scenarios that are hard to identify as hazardous? Ergonomics, 52, 657–673. https://doi.org/10.1080/00140130802550232.

400

K. Pammer et al. / Transportation Research Part F 59 (2018) 389–400

Rudin-Brown, C., George, M., & Stuart, J. (2014). Human factors issues of accidents at passively controlled rural level crossings. Transportation Research Board, 2458, 96–103. https://doi.org/10.3141/2458-12. Salmon, P. M., Read, G. J. M., Stanton, N. A., & Lenné, M. G. (2013). The crash at Kerang: Investigating systemic and psychological factors leading to unintentional non-compliance at rail level crossings. Accident Analysis and Prevention, 50, 1278–1288. https://doi.org/10.1016/j.aap.2012.09.029. Schmicker, M., Müller, P., Schwefel, M., & Müller, N. G. (2017). Attentional filter training but not memory training improves decision-making. Frontiers in Human Neuroscience, 11(138), 1–7. https://doi.org/10.3389/fnhum.2017.00138. Shanteau, J., Weiss, D. J., Thomas, R. P., & Pounds, J. C. (2002). Performance-based assessment of expertise: How to decide if someone is an expert or not. European Journal of Operational Research, 136(2), 253–263. https://doi.org/10.1016/S0377-2217(01)00113-8. Sheridan, H., & Reingold, E. M. (2017). The holistic processing account of visual expertise in medical image perception: A review. Frontiers in Psychology, 8 (1620), 1–11. https://doi.org/10.3389/fpsyg.2017.01620. Stutts, J., Reinfurt, D., Staplin, L., & Rodgman, E. (2001). The role of driver distraction in traffic crashes. Washington, DC: AAA Foundation for Traffic Safety. Symmons, M., Haworth, N., & Mulvihill, C. (2005). Characteristics of police and other emergency vehicle crashes in New South Wales. In Australasian road safety research policing education conference (pp. 121–126). Wellington, NZ. http://acrs.org.au/publications/conference-papers/database/, http://acrs. org.au/files/arsrpe/RS050098.pdf. Torralba, A., Oliva, A., Castelhano, M. S., & Henderson, J. M. (2006). Contextual guidance of eye movements and attention in real-world scenes: The role of global features in object search. Psychological Review, 113(4), 766–786. https://doi.org/10.1037/0033-295X.113.4.766. Underwood, G., Chapman, P., Bowden, K., & Crundall, D. (2002). Visual search while driving: Skill and awareness during inspection of the scene. Transportation Research Part F: Traffic Psychology and Behaviour, 5(2), 87–97. https://doi.org/10.1016/S1369-8478(02)00008-6. Underwood, G., Chapman, P., Brocklehurst, N., Underwood, J., & Crundall, D. (2003). Visual attention while driving: Sequences of eye fixations made by experienced and novice drivers. Ergonomics, 46(6), 629–646. https://doi.org/10.1080/0014013031000090116. Underwood, G., Crundall, D., & Chapman, P. (2002). Selective searching while driving: The role of experience in hazard detection and general surveillance. Ergonomics, 45(1), 1–12. https://doi.org/10.1080/00140130110110610. Underwood, G., Ngi, A., & Underwood, J. (2013). Driving experience and situation awareness in hazard detection. Safety Science, 56, 29–35. https://doi.org/ 10.1016/j.ssci.2012.05.0125. Vlakveld, W., Romoser, M., Mehranian, H., Diete, F., Pollatsek, A., & Fisher, D. (2011). Do crashes and near crashes in simulator-based training enhance novice drivers’ visual search for latent hazards? Transportation Research Record, 2265(1), 153–160. 10.3141%2F2265-17. Walker, G. H., Stanton, N. A., Kazi, T. A., Salmon, P. M., & Jenkins, D. P. (2009). Does advanced driver training improve situational awareness? Applied Ergonomics, 40(4), 678–687. https://doi.org/10.1016/j.apergo.2008.06.002.