Elsevier

Accident Analysis & Prevention

Volume 50, January 2013, Pages 955-963
Accident Analysis & Prevention

Attentional differences in driving judgments for country and city scenes: Semantic congruency in inattentional blindness

https://doi.org/10.1016/j.aap.2012.07.026Get rights and content

Abstract

‘Looked-but-failed-to-see’ vehicle collisions occur when a driver gives all indications of having responsibly evaluated the driving situation yet still fails to see a hazard that is clearly in view. The experience maps well onto the psychological phenomenon called inattentional blindness (IB). IB occurs when a viewer fails to see an unexpected object that is clearly visible, particularly if they are concentrating on an additional primary task. In this study, a driving-related IB task was used to explore whether an unexpected stimulus (US) such as a pedestrian or animal, is more likely to be seen in country or city-related driving scenarios if it is congruent or incongruent with the semantic context of the scenes, and thus congruent or incongruent with the attentional set of the viewer. Overall, participants were more likely to see the US in the City scenarios, which also demonstrated a borderline effect of congruency, with incongruent stimuli less likely to be seen than congruent stimuli. Analyses suggested that driver experience was related to detection of the US in City scenarios but not Country scenarios. However, analyses also revealed that participants generally tended to drive in city rather than country environments, thus prompting speculation that the results may reflect attentional requirements for familiar and unfamiliar driving scenarios. Thus we suggest that the analysis of the driving situation, and the attentional set that we develop to filter information, change when the driving situation is more familiar.

Highlights

IB is used to determine the detection of an unexpected object in a driving scenario. ► The attentional-set of the driving scenario is a city or country driving situation. ► The unexpected object is consistent or inconsistent with the attentional set. ► Results are dependent on experience, familiarity and evaluation of driving scenario.

Introduction

The driving situation is one which requires constant monitoring of the visual environment, and continuous filtering of visual information in order to attend to the most important cues. In many cases however, we fail to see something of critical importance while driving. Referred to as ‘looked-but-failed-to-see’ accidents (Hills, 1980, Treat, 1980), they have been implicated as the third most frequent type of driver error (Brown, 2005). Such accidents have been attributed to a number of different factors, for example insufficient visual search strategies, where the driver fails to systematically move their eyes around the visual environment to capture relevant visual information (Summala et al., 1996), or attentional load (Hancock et al., 1990, White and Caird, 2010). Another explanation for looked-but-failed-to-see-accidents is that the ‘unseen’ item does not fit the driver's current attentional expectations. Such a situation could occur when a cyclist comes from an unexpected direction (Summala et al., 1996), or when a highly visible vehicle such as a police car, is sitting in an unexpected location (Langham et al., 2000).

Cognitively, the looked-but-failed-to-see experience appears to map onto a known psychological phenomenon called inattentional blindness (IB). In IB, an observer fails to see an unexpected stimuli or event while attending to another, primary task (Mack and Rock, 1998). In one of the earliest versions of IB, Neisser and Becklen (1975) participants were required to follow one of two superimposed videos. In one video, participants were required to follow an informal basketball play, and ignore a ‘hand-slapping’ game similar to the child's ‘patty-cake’ game. In the other participants followed the hand-slapping game, and ignored the basketball game. In each of the unattended situations, an unusual event occurred, such as swapping the gender of one of the basketball players, or having the ‘slapping hands’, stop, shake, and resume play. Very few of these unusual events were detected by the participants, suggesting that while attending to one situation, unrelated events in the same visual space go unnoticed by the viewer.

In one common version of IB, participants are presented with a primary attention task in the form of a briefly presented, masked cross (Mack and Rock, 1998). The cross appears approximately five times, and in each trial, one arm of the cross is longer than the others. At each trial the participants are required to indicate which arm of the cross was longer. On about the fifth trial, an additional unexpected stimulus (US) appears such as a small square, and IB is determined by whether the participant notices the US. This is the static version of IB. In the dynamic version, the primary task consists of objects such as circles and triangles moving randomly around the computer screen, In this case, the participants’ task is to count how many times the tracked items bounce off the edges of the computer screen, and the US is an additional object moving a set course through the centre of the display. In general, approximately 25% of people fail to notice the additional US in these tasks, however this number varies depending on the nature and location of the US. For example, the US is noticed far more frequently if it is salient to the participant such as their name or a ‘smiley face’ (Mack and Rock, 1998), but failure to notice it increases dramatically when the primary task is made more difficult (e.g., Cartwright-Finch and Lavie, 2007).

The general finding – that attending to a primary task decreases a viewer's tendency to detect an additional unexpected stimulus – is a consistent result, and many physical parameters of the phenomenon have been explored, such as the nature of the primary task (Beanland et al., 2011, Koivisto and Revonsuo, 2008, Most et al., 2000, Most et al., 2001), the colour/shape/location and size of the US (e.g., Mack and Rock, 1998), the role of eye movements (Beanland and Pammer, 2010a), experience (Beanland and Pammer, 2010b), and the difficulty of the primary task (e.g., Cartwright-Finch and Lavie, 2007, Lavie, 1995).

Outside the IB literature, it is also known that the ‘attentional set’ of an observer is important for the detection of a target in experimental settings. Here, attentional set refers to the recognition that we generate a mental template for objects, situations or behaviours that have consistent semantic features and feature relationships. Attentional set can be simple, such as looking for a red circle amongst red and blue shape distractors, where the attentional set would be “red” + “circle”. A complex attentional set is more akin to Bartlett's concept of schema (e.g., Bartlett, 1932) where the attentional set for driving for example, would involve the objects and procedures involved in the driving situation. Folk et al. (1992) demonstrated that participants detecting a target take advantage of a cue that is consistent with a specific attentional set, for example when participants know that a target is defined by a particular colour, a cue defined by that colour is more effective than a cue defined by some other feature. Such expectation is similarly important in the detection of the US in IB. Most et al., 2001, Most et al., 2005 employed a dynamic IB task in which the primary task was to monitor black and white objects – such as circles and squares – moving randomly around the computer screen. IB was determined by whether the participant noticed an additional US, such as a “+” sign traverse the display through the moving circles and squares. In one experiment, participants were required to track the black items and ignore the white items or the reverse – track the white items, ignoring the black. Rates of IB were much higher when the US was congruent with the colour of the items to be tracked, i.e., if participants monitored the black items, they were far more likely to see the US when it was also black, than if the US was white. The suggestion here was that participants were more likely to see the US when it shared a common feature with the tracked items, because the attentional set in this case was “black” and black items gained priority access to higher order recognition systems.

Most et al. (2005) further demonstrated an IB effect for faces, in order to determine if the same attentional set mechanism was applicable for complex features, as well as simple features such as shape and colour. In a dynamic IB display consisting of four Caucasian faces and four African American faces moving randomly around the computer screen, participants were required to track either the Caucasian or African American faces. Participants were more likely to see an US when it was consistent with the race of the face that they had been asked to monitor.

Koivisto and Revonsuo (2008) through another series of dynamic IB experiments systematically manipulated the features of the distractor and attended items while keeping the US constant. Their question was whether participants established an attentional set for the features of the attended items and ignored the unattended items, or conversely, coded the unattended items in order to selectively ignore them and concentrate on the attended items. They concluded that participants developed an attentional set consistent with the features of the attended items, not the unattended items.

Overall, these results suggest that observers are able to establish an attentional set that makes salient the features of the objects that are explicitly attended. In this then, an additional unexpected item in a visual scene is more likely to be seen when it shares one or more features of the attended items. However in general this research has been in the context of setting an attentional set for stimulus features such as colour and/or shape. One question that arises as a consequence of this research is whether realistic settings engender a similar attentional set. For example, given a city scene, are we more likely to see an US if it is congruent with the attentional set – such as a man in a business suit, compared to an US that is incongruent with a city schema, such as a deer or tractor?

Set theory is the theoretical framework for the concept of attentional set. Such a model derives from a number of different studies that identify the cognitive advantage incurred by semantic relationships between objects. For example, in early research, Biederman et al. (1982) demonstrated automatic activation of the semantic relationships between objects that appeared to occur in parallel to identification. Such results illustrate the power of contextual information in object identification. Set theory, would therefore predict that a ‘city’ attentional set would include items such as buses, buildings, cars, people, but would be unlikely to include tractors, sheep or hay silos, which would be more likely to be included in a ‘country’ attentional set. Thus we would predict that if set theory works for semantic schemas in the same way as for features, then IB would be higher when the US is a sheep in a city scene, compared to when the US is a man in a business suit in a city scene because the sheep does not fit our attentional set for the city.

It is well known that we can extract the salient features of a scene extremely quickly and accurately; Potter (1975), presented participants with 16 scenes presented as an RSVP (rapid serial visual presentation) task. Here participants were able to correctly identify a target scene even at the fastest presentation rate of 125 ms if they knew beforehand what would be in the scene (e.g., a child). This rate of performance was consistent with subjects who were previously shown the actual target scene. Thus participants were equally accurate in detecting a scene if they knew exactly what they were looking for, compared to whether they simply knew what would be in the scene. This result shows the speed with which our attentional system can extract relevant information from a scene at a glance.

There is currently very little research looking at IB and the attentional set of the observer. Koivisto and Revonsuo (2007) presented participants with an array of pictures in which they were required to identify either the animals, or the furniture in the array. The US in this case was the presentation of printed word that was either congruent or incongruent with the semantic category that they were attending to. Their results demonstrated that detection for the US was higher when the attended items and the US were congruent, supporting the suggestion that detection of an unexpected stimulus is enhanced not only when its features match the attended items, as shown previously, but also when it fits an attended semantic category

We argue that there is a conceptual link between IB and looked-but-failed-to-see-accidents, such that IB could provide the theoretical framework for the experience of such accidents. However, again, there is little research exploring incidents of IB in the driving situation. The only study that we know of is by Most and Astur (2007). In this study participants were required to drive through a virtual environment, following indicator signs designated either blue or yellow. Participants followed the yellow signs, ignoring the blue, or followed the blue signs, ignoring the yellow. The assumption was that the participant would develop an attentional set for blue or yellow, depending on which sign they were following. The US was a motorcyclist that was either blue or yellow such that it was either congruent or incongruent with the attended signs. Drivers were slower to brake for the motorcycle when it was the incongruent colour, and were also more likely to ‘collide’ with it. Thus the results remain consistent with the idea that we are more likely to see an unexpected stimulus when it fits a predetermined attentional set. Although this study combines the notion of attentional set with IB and driving, the link between the attended stimuli and the US is again featural – in this case colour. It still remains to be seen whether a particular attentional set as defined by a real world scenario engenders the same response. Thus, if when driving we develop an attentional set for the environment in which we are driving, such as the city or country, then set theory would predict that this attentional set would mediate our detection of hazards when driving. Moreover, positioning the study within the context of IB allows us to extend the IB literature to include the possibility that set theory can extend from simple features to real-life contexts. In this sense then, the study is designed to address a point of theoretical significance in the cognition of IB, as well as the practical element of hazard detection when driving.

What then do we know within the driving literature about the degree to which different driving environments engender different attentional requirements? Again, the literature is limited. Borowsky et al. (2012) demonstrated that eye movements and hazard detection was different depending on whether participants were driving in urban compared to residential environments. Although these researchers were in interested in driver experience, they nevertheless demonstrated that pedestrian-related hazards were detected more frequently in residential environments, and eye fixations for the curb and the road varied depending on whether it was the urban environment or the residential environment. Similarly, Borowsky et al. (2008) presented experienced and inexperienced drivers with typical road signs (e.g., No Right Turn) either on the expected right side of the road or the unexpected left side of the road. Experienced drivers – who presumably have a well-developed attentional set for driving, were significantly better at detecting road signs in the expected location compared to the unexpected location. Moreover, the inexperienced drivers showed no difference in detecting signs presented on the ‘correct’ right side, or ‘incorrect’ left side of the road. This is consistent with the inexperienced drivers having a less well-defined attentional set for the driving environment compared to experienced drives. These findings suggest that it is possible to develop a driving attentional set that dictates different driving requirements. But again, we remain unenlightened as to whether a particular attentional set would facilitate or hinder detection of an unexpected stimulus. In the current study we present participants with a static IB task in which ‘city’ or ‘country’ driving scenes are varied between subjects, and the US is either congruent or incongruent with the scenes. Consistent with set theory we would predict that incongruent stimuli – in this case the presence of a kangaroo in the city – would be more likely to produce IB (less likely to be seen) as it does not fit the attentional set generated by the city scene. From this we could also conclude that set theory extends beyond the development of an attentional set for simple features, to the development of an attentional set for semantic information in the form of semantic schemas. Moreover, in the current study participants will make consecutive judgments about the relative safety of a driving scenario, allowing us to tentatively explore whether a tendency towards caution when driving induces its own attentional set. One possibility is that participants who are more cautious and thus more likely to describe driving situations as “unsafe”, may have a heightened sense of vigilance, and more likely to detect the US. Finally, driving experience was recorded. It is well known in the driving literature that increases in driving experience change visuo-attentional mechanisms such as increasing search speed, eye movement and fixations (Underwood et al., 2011a, Underwood et al., 2011b; refer to Konstantopolous et al. (2012) for a review), as well as increased hazard detection (e.g., Underwood et al., 2005). Thus it is not unreasonable to predict that detection of the US is mediated by driving experience. The current findings are an attempt to understand the generation and flexibility of attentional-set when driving in specific environments, with the practical significance in terms of the context in which drivers are vulnerable to lapses of attention.

Section snippets

Participants and design

Ninety-two undergraduates participated voluntarily, either for course credit or $20AUS. Participants were excluded if they had participated in an IB study before (3 participants) or failed to see the US on the full attention trial (one participant). The remaining 88 participants (43% female; Mage = 22 years, SD = 4.5) were randomly assigned to the four experimental conditions. This experiment used a 2 (environment: city vs. country) × 2 (unexpected stimulus: kangaroo, businessman) between-subjects

Results

Unlike most IB tasks, accuracy judgments for the primary task were not analysed as there was no explicit correct or incorrect response for each trial. Participants were required to make a “safe” or “unsafe” judgment on each trial with an explanation for their decision that was recorded by the experimenter. A typical response would be; “safe, because the car had stopped at a red light”. The aim was to simply get participants to attend to the primary task. All participants made valid and logical

Discussion

When monitoring a visual stimulus, a viewer is more likely to notice an unexpected object appearing if it shares some features with the stimulus being monitored. It is suggested that when attending to a visual stimulus we develop an attentional set for the object of our attention (Most et al., 2005), and objects that share features of this attentional set are more likely to be seen (Folk et al., 1992). The aim of the current study was to look at IB in a realistic situation such as driving,

Acknowledgements

This research was supported by the NRMA-ACT Road Safety Trust, grant number: 2/09. The authors would like to thank Andy Thomson and Vanessa Beanland for programing assistance.

References (43)

  • M.H. Martens et al.

    Do familiarity and expectations change perception? Drivers glances and response to changes

    Transportation Research Part F

    (2007)
  • U. Neisser et al.

    Selective looking: attending to visually specified events

    Cognitive Psychology

    (1975)
  • H. Summala et al.

    Bicycle accidents and driver's visual search at left and right turns

    Accident Analysis and Prevention

    (1996)
  • G. Underwood et al.

    Driving simulator validation with hazard perception

    Transportation Research Part F – Traffic Psychology and behavior

    (2011)
  • G. Underwood et al.

    Decisions about objects in real-world scenes are influenced by visual saliency before and during their inspection

    Vision Research

    (2011)
  • C. White et al.

    The blind date: the effects of change blindness, passenger conversation, and gender on looked-but-failed-to-see (LBFTS) errors

    Accident Analysis and Prevention

    (2010)
  • F. Bartlett

    Remembering: A Study in Experimental and Social Psychology

    (1932)
  • V. Beanland et al.

    Gorilla watching: effects of exposure and expectations on inattentional blindness

  • H. Benda et al.

    Estimating hazards in traffic situations

    Accident Analysis and Prevention

    (1983)
  • A. Borowsky et al.

    The effect of hazard perception training on young-inexperienced drivers’ movies categorization

  • A. Borowsky et al.

    The relation between driving experience and recognition of road signs relative to their locations

    Human Factors

    (2008)
  • Cited by (28)

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

      2018, Transportation Research Part F: Traffic Psychology and Behaviour
      Citation Excerpt :

      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).

    • The impact of attentional set and situation awareness on dual tasking driving performance

      2018, Transportation Research Part F: Traffic Psychology and Behaviour
      Citation Excerpt :

      Most and Astur’s (2007) feature-based attentional set suggestion is supported by the findings of Experiment 1, where more driving-relevant items (road signs) were detected than driving-incongruent items. The suggestion that attentional sets can be semantic, rather than just feature-based (Pammer & Blink, 2012) is also supported by this investigation. The finding that dual tasking participants, in Experiment 2, failed to notice unexpected yet driving-relevant events supports the notion of an attentional set based on an understanding of what ‘normally’ occurs in these situations, rather than simple failure to detect specific objects (e.g. a green traffic light).

    • Visual processing in expert drivers: What makes expert drivers expert?

      2018, Transportation Research Part F: Traffic Psychology and Behaviour
      Citation Excerpt :

      The latter is an explanation for Inattentional Blindness (IB), and is the observation that we fail to see an object that is in full view, when our attention is directed elsewhere, and this is magnified when the unexpected object does not match our mental template (Most, Scholl, Clifford, & Simons, 2005). In the driving context, this equates to failing to see hazards such as police cars parked in an unexpected location (Langham et al., 2016), or rural animals in city driving scenes (Pammer & Blink, 2013). It is also the psychological framework for looked-but-failed-to-see-crashes (Pammer, Bairnsfather, & Burns, 2015).

    • Compensating for failed attention while driving

      2017, Transportation Research Part F: Traffic Psychology and Behaviour
      Citation Excerpt :

      While the complexity of the visual environment plays a role in the ability to detect information, the current experiences and situations of the driver may also play a factor in tracking ability and target detection. Differences have been found between urban and rural landscapes when making driving judgments (Pammer & Blink, 2013). Road familiarity may also play a pivotal role in driving ability (Charlton & Starkey, 2013).

    View all citing articles on Scopus
    View full text