حواس پرتی راننده، تداخل و استدلال فضایی

Abstract The hypothesis is explored that the precise influence of a secondary, unrelated, spatial reasoning task on driving performance also depends on the specific spatial cues used in this task, compared to those currently emphasized by the primary driving task. In a laboratory experiment, participants were presented with questions about spoken (familiar) city names while driving. The questions either required them to reason spatially about the cities or to process the same city names only acoustically (i.e., remembering and repeating one of the names). Amount of driver distraction was measured by means of a standardized tool called the Lane-Change Task (LCT) using a PC-based driving simulator. Results of the experiment showed that the spatial reasoning secondary task was more distracting than the acoustic one. In addition, participants performed worse on the LCT when switching to a right lane than when switching to a left lane. It is concluded that the results confirm an interpretation in terms of (in)compatible spatial cues emphasized simultaneously by primary and secondary task, but that alternative interpretations are also possible. The moderating influences of two cognitive ability variables on, and potential practical applications of, these findings are also addressed.

1. Introduction and problem statement An increasing number of automobile accidents is attributed to distracted driving. The percentage of accidents with such origin varies from 5 to over 25%, depending on the type of study (traditional crash studies or naturalistic driving studies) and the definition of distraction utilized by the study (Gordon, 2009 and Neale et al., 2005). Naturalistic studies arrive at percentages as high as 23% for crashes due to the performance of non-driving related secondary activities a few moments before the crash. These activities include personal grooming, and reaching for some object in the car. Moreover, there is evidence this percentage would be even higher if inattention-based crashes were included in the crash statistics (i.e., paying insufficient attention to the forward roadway due to daydreaming or other internal, but invisible activities). Obviously, when driving, not all sources of distraction can be avoided. However, knowledge about the human driver can help reduce the size of the distraction problem, for example by appropriately designing car-based equipment and procedures for its usage, but also by training and reinforcing drivers to behave prudently and responsibly with respect to the use of distracting equipment or activities. Though driver distraction can usefully be attributed to some kind of interference between a secondary task (driving-related or not) and the primary driving task, there are multiple sources of such dual-task interference in the driving context. Moreover, many psychological mechanisms have been proposed for explaining the nature and size of the interference. In an attempt to clarify these issues, we start by distinguishing structural interference from cognitive interference. Structural interference is based in the physiology of the sense organs. It refers to the influence of physiological limits on the ability of people to perform two or more tasks simultaneously ( Pashler & Johnston, 1998). Cognitive interference, on the other hand, refers to those types of interference that are observed when some or all subtasks of the time-shared tasks require information processing ( Kujala, 2010). Cognitive interference, in turn, can be divided into task-independent interference and task-dependent interference. The latter type of interference usually refers to the interference caused by task similarity. One particular type of similarity-based interference is also known as crosstalk ( Navon & Miller, 1987): performing two tasks simultaneously becomes difficult if they contain stimuli which generate conflicting response tendencies (e.g., a right-pointing arrow appears in one task, and an unrelated left-pointing arrow appears in the other ( Hommel, 1998, Lien and Proctor, 2002 and Pashler et al., 2001). This is similar to the Stroop task ( Stroop, 1935) and the Simon effect ( Simon, 1990) in single-task conditions. Though the empirical evidence for crosstalk is rather robust, disagreement exists in the literature with regard to the generality of, and the theoretical explanation for, this phenomenon. Explanations are based in theories varying from limited capacity models ( Pashler et al., 2001), to multiple-resources theory ( Wickens, 2002), to the concept of attention sharing ( Navon, 1984 and Navon, 1985), and to the theory of multimodal spatial attention. According to the latter account, shifts of spatial attention in one sensory modality (e.g., vision) tend to be accompanied by corresponding covert shifts in other modalities (e.g., audition) ( Driver and Spence, 1998 and Spence and Driver, 2004). These attentional shifts may be triggered by external events (exogenous attention shifts) or by internal events (e.g., intentions or thoughts: endogenous attention shifts). Therefore, this theory may explain crosstalk to the extent that the stimulus attributes causing crosstalk are spatial in nature. Regarding the empirical conditions necessary for crosstalk to be observed, there is evidence that time-shared tasks are less vulnerable to crosstalk with certain types of display design and task configuration (Carlson and Sohn, 2000 and Elio, 1986). Under some circumstances, task similarity also results in improved, rather than reduced, dual-task performance. Specifically, time-sharing efficiency may improve if tasks share some common display property, processing routine, mental set, or timing mechanism ( Duncan, 1979 and Fracker and Wickens, 1989). These findings probably reflect the fact that task reconfiguration becomes simpler, going from task to task, when the tasks are structurally identical or similar. This article describes a laboratory experiment in which one particular type of crosstalk was studied in the context of driver distraction. Specifically, the effect on driving performance was studied of having simultaneous activation in working memory of semantically related spatial codes, these codes either being invoked by a spatial reasoning auditory secondary task or a concurrently performed primary (driving) task. In addition to a spatial reasoning version, an acoustic version of an otherwise identical secondary task was employed. This acoustic version served as a baseline against which to assess the distracting effect of the spatial reasoning part of the first version, as the two versions only differed with respect to the amount of spatial reasoning they imposed on the participants. Driver distraction was measured by means of a standardized tool, called the Lane-Change Task (LCT). The rationale for studying this particular spatial reasoning task as a secondary task was that, though cognitive interference of a spatial nature has already been studied before in a driving context (Patrick & Elias, 2009), the distracting effect of semantically related spatial memory codes on driving performance has not. Moreover, there was an interest to test the generality of the observation that not only physically related items, but also items belonging to the same semantic category may cause crosstalk (Hirst & Kalmar, 1987). Specifically, in this experiment the hypothesis was tested that cardinal spatial cues such as “east” and “west” belong to the same semantic category as (and, therefore, may interfere with) egocentric spatial cues, such as “left” and “right”. The moderating influence of two cognitive ability variables (i.e., useful field of view, or UFOV, and task-switching ability, or TSA) was also investigated in this study. From a practical point of view, spatial reasoning can be considered an ecologically valid secondary task, because it can be linked to the use of in-car GPS-devices, to navigational conversations, or to reading a map while driving. Therefore, the results of this experiment may also have practical applications.

5. Conclusions and discussion This article reported a laboratory experiment in which a particular type of crosstalk was studied in the context of driver distraction. Specifically, the effect on driving performance was studied of having simultaneous activation in working memory of semantically related spatial codes, these codes being invoked by either a spatial reasoning secondary task or a concurrently performed primary (driving) task. In addition to a spatial reasoning version, an acoustic version of an otherwise identical secondary task was also employed. Driver distraction was measured by means of a standardized tool, called the Lane-Change task. 5.1. Hypothesis 1: distracting effect of type of secondary task The finding regarding the distracting effect of spatial reasoning about spoken city names is consistent with a previous study showing the distracting effect of performing an irrelevant mental navigation task while driving (Patrick & Elias, 2009). Both findings lend support to multiple-resources theory (Wickens, 2002) and to functional distance theory (Kinsbourne & Hicks, 1978). According to these theories, the more dissimilar two concurrent activities are in a cognitive sense and from the point of view of brain anatomy, the easier it is to time-share these activities. In Section 4.3Hypothesis 2 and Hypothesis 3, it was seen that the expected main effect of type of secondary task on amount of distraction was not observed when only those parts of the LCT-tracks were analyzed at which switching maneuvers were actually performed. However, additional data analysis revealed that this was only true for individuals with a higher-than-average task-switching ability (as measured by the TSA). In contrast, for participants with a lower-than-average task-switching ability, the expected inferiority of the spatial reasoning secondary task was observed. This suggests that high-ability individuals are somehow able to resist distraction by the spatial secondary task (at least, in comparison to the acoustic secondary task), but only when actually performing switching maneuvers. (The interactions with UFOV-standing did not reveal additional insights, in this regard.) The finding that the acoustic processing of spoken city names was not distracting vis-à-vis a control condition without secondary task is also consistent with multiple-resources theory: an acoustically oriented task and a (spatially-oriented) driving task can be assumed to require different mental resources, with the result that these tasks can be time-shared relatively easily. In summary, the present study confirms that a spatial reasoning task, representative of the secondary tasks performed during real driving, may degrade performance on the primary driving task. 5.2. Hypothesis 2: amount of distraction as a function of the (in)compatibility between spatial directions currently emphasized in secondary and primary task Though the expected effect of switching direction on LCT-performance was confirmed in this study for the spatial reasoning version of the secondary task, the same effect was also observed for the acoustic version of this task (significant main effect of direction and non-significant task * direction interaction). This suggests that the overall effect of switching direction has little to do with the hypothesized (in)compatibility between spatial directions that are emphasized simultaneously in the primary and secondary task, but must have another explanation. In a first attempt to identify the conditions under which the overall effect of switching direction is observed, the data of this study belonging to the control conditions C1 and C2 were analyzed. This analysis revealed no significant effect of switching direction, p > 0.10. Second, the LCT-data collected in a previous experiment (Hurts & Sjardin, 2009) were re-analyzed, using a visual (rather than auditory) secondary task. This analysis also failed to show evidence for an effect of switching direction. Third, it is possible that the secondary task instructions had not been understood properly by some participants, with the result that both types of secondary task were performed more or less in the same way. However, this possibility is not very likely, because all participants had received three practice trials on each type of secondary task before they started to perform this task concurrently with the LCT. The experimenter used these trials to make sure the participants had understood his instructions correctly. As another manipulation check, Table 1 shows that secondary task questions under condition SR were not always easy to answer correctly (average accuracy percentage of nearly 80%). In contrast, the questions for condition AC were invariably answered without errors by any participant (no accuracy percentage is shown for condition AC in Table 1). Fourth, additional data analysis revealed that the expected task * direction interaction was observed for participants with a higher-than-average task-switching ability, as measured by the TSA. In contrast, for participants with a lower-than-average task-switching ability, the effect of switching direction was stronger for the acoustic version than for the spatial reasoning version of the secondary task. This may (partly) explain the absence of an overall task ∗ direction interaction. (The interactions with UFOV-standing did not reveal additional insights, in this regard.) Fifth, in future research the present experiment should be replicated, but with the following simple modification to the spatial secondary task: participants should be asked to state which of the two cities has the most eastern (rather than western) orientation. If, as a result of this modification, the effect of switching direction would become significantly smaller than the one observed in the present study (or even becomes the opposite of it), we would have more confidence that the effect is due to spatial cues in concurrently performed tasks being (in)compatible with each other. In summary, it must be conducted that it has not yet been shown convincingly that cardinal spatial cues such as “east” and “west”, appearing in a spatial reasoning task, are subject to crosstalk when egocentric cues such as “left” and “right” appear in a concurrently performed, but unrelated, driving task. However, it was shown that several trivial, alternative explanations for the main effect of switching direction could be ruled out. 5.3. Hypothesis 3: influence on effect described by Hypothesis 2 of secondary task phase typicality Though the expected two-way interaction between direction and phase typicality and the expected three-way interaction between direction, task, and phase typicality were confirmed, other findings cast doubt on the validity of Hypothesis 3. Specifically, and contrary to expectations, it turned out that right-going maneuvers were not only performed worse than left-going maneuvers during the typical phases of the spatial reasoning secondary task, but also during the non-typical phases of the acoustic secondary task. This may reflect a phenomenon quite different from the type of cognitive interference and crosstalk discussed so far. It is known from neurological studies that, when performing spatial tasks, people perform more accurately when processing stimuli originating in the left-hand visual field (Kogure & Hatta, 1999) due to cerebral lateralization. In our study, such a perceptual bias may have caused participants to perform right-going switching maneuvers less accurately than left-going ones. (The interactions with UFOV-standing or TSA-standing did not reveal additional insights, in this regard.) Furthermore, it can be seen that such a perceptual bias may also explain why the overall effect of switching direction (regardless of phase typicality), that was predicted by Hypothesis 2, turned out to be equally strong for both versions of the secondary task. As was mentioned in Section 5.2Hypothesis 2, it may be that participants with a lower-than-average task-switching ability are particularly vulnerable to this bias. However, as yet it is unclear why such a bias would particularly affect the non-typical phases of the acoustic secondary task. This also represents an issue for future research. In summary, this experiment provided evidence for an effect of switching direction (larger LCT-deviations when steering to the right than when steering to the left). Also, this effect seemed to be larger when participants were actually involved in spatial reasoning during a secondary task. Nonetheless, the precise theoretical interpretation of these effects is not clear yet, as a similar effect of switching direction was also observed (but not expected) for those parts of the acoustic secondary task (i.e., all second trial phases) where (probably) no acoustic processing (or any other subtask) was going on. 5.4. Potential applications and recommendations for future research First, the finding that a realistic spatial reasoning task performed concurrently with the primary driving task may degrade primary task performance also has potential practical implications. Specifically, this finding highlights the importance of carefully designing in-car navigation systems to make sure they impose the least amount of spatial reasoning on the driver as possible. For example, the travel directions that are automatically generated by these systems should not be too complex or require the driver to engage too much in navigational planning while driving, as this may degrade driving performance. Second, the findings concerning the effects on driver distraction of the (in)compatibility among spatial cues, emphasized by a secondary and a concurrently performed driving task, may also have practical applications, if confirmed in future studies. Obviously, drivers cannot always be prevented from thinking about spatial directions that are irrelevant to the concurrently performed driving task and that may conflict with the direction of specific maneuvers performed as part of this task. Nonetheless, these findings suggest the importance of the precise type and timing of spatial reasoning drivers engage in while driving. This knowledge can be used, for example, by designers of in-car navigation systems: when automatically generated travel directions are presented to the driver too early, they may interfere (and cause crosstalk) with the currently performed driving maneuver. The importance of having compatible spatial directional cues can also be seen in the design of display maps in navigation systems: egocentric and (3-D) head-up display maps are generally easier to use for navigation purposes than exocentric (north-up) maps. In other words, physical and mental maps should have similar reference frames ( Bryant & Tversky, 1999). Finally, the following limitations of this study should be noted. First, though participants were instructed to pay an equal amount of attention to the primary and secondary task, it was not verified whether they actually succeeded in doing so. Nor were the effects on driver distraction tested of alternative attention allocation policies. Second, in order to test Hypothesis 2 and Hypothesis 3, the overall LCT-track length was broken up in smaller parts corresponding to the precise time periods when participants actually performed switching maneuvers. Though we have not done so, we could also have zoomed in on overall secondary task performance, i.e., measured the accuracy of spatial reasoning performance for only those periods where switching maneuvers were actually performed. In this way, additional insights could have been obtained into possible tradeoffs between secondary and primary task performance. Third and last, though the LCT is a convenient tool for measuring driver distraction in the laboratory, it is not (yet) precisely known to what extent results of behavioral experiments conducted using this tool can safely be generalized to more realistic driving scenarios. Each of these limitations represents a reason for conducting future studies validating and exploring more deeply the scope and precise interpretation of the present findings. Other recommendations for future research were given earlier in this section, at the end of the discussion of Hypothesis 2 and Hypothesis 3.