تشخیص محیطی به عنوان یک اندازه گیری از حواس پرتی راننده. مطالعه مبتنی بر حافظه در مقابل ناوبری مبتنی بر سیستم در یک منطقه مملو از ساختمانها

Abstract The effect of in-vehicle information systems (IVIS) on traffic safety is currently under debate and suitable methods for measuring and comparing the impact of such devices on driver behaviour are urgently required. The secondary-task technique may be a good tool for objective measurement of driver distraction caused by IVIS. The present study summarises previous results of secondary-task studies in traffic contexts and investigates the suitability of one secondary-task method, the peripheral detection task (PDT)-method, as a standard procedure for safety testing and evaluation of IVIS. The study was concerned with the effect of navigation messages on PDT-performance (reaction time and hit rate) taking into account also behavioural variables. Professional drivers served as subjects. They had extensive prior local-knowledge and experience of driving in the built-up area in which the experiment took place. They were required to drive two different routes, one after memory and the other in accordance with navigation messages a standard navigation system installed in the car. In the navigation system condition subjects were subdivided into three groups, receiving either verbal, visual or both visual and verbal (full) navigation messages. Driving behaviour was virtually uninfluenced by navigation condition (memory versus navigation system) and message modality (full, visual or verbal) whereas PDT-performance, showed some effects of navigation condition on subjects’ reaction times and hit rates. Pairwise comparison of message modality within each three groups showed a prolongation in reaction time and a marginally significant decrease in hit rate with full navigation messages (combined visual and verbal ones). Visual navigation messages affected only hit rate and no significant differences between navigation conditions were observed for the group presented with verbal messages. The pattern of results suggests that the PDT-method is biased toward visual sources of information from IVIS. As visual information processing is an important component in safe driving, the PDT-method is suitable as a predominant method in a test battery, but for unbiased measurement of distraction, methods less dependent on mode of presentation would be more appropriate.

Introduction New in-vehicle technologies such as navigation systems, vision enhancement systems and onboard Internet connections are presumed to increase in popularity and number in the not-so-far-away future vehicle. A subdivision has been made between advanced driver assistance systems (ADAS) with driving support functions and in-vehicle information systems (IVIS) that have other functions than those related to driving. Both car drivers and passengers may benefit from new technology, but some in-vehicle systems may not be suitable or appropriate for use in moving vehicles in the current road transport system. It is generally acknowledged that IVIS can cause distraction by diverting the driver’s attention from the driving task. In fact, even some driver assistance systems may occasionally call for attention. However, being an integrated part of the driving task, their impact is usually not classified as distraction but as cognitive load. Regardless of type, IVIS and ADAS require drivers to sometimes divide their attention between in-vehicle information and information in the driving environment. The impact of these devices on attention may depend both on the design and on the function of such devices. Therefore it is possible that minor physical differences between devices with the same general functions affect driver attention, and methods sensitive for such differences would produce valuable information for safer IVIS design useful to authorities, users and producers of ADAS and IVIS. The secondary-task method is a frequently used tool for the measurement of human capacity limitation. Although the theoretical status of dual-task and secondary-task methods has been under debate, many secondary tasks have been used (see Ogden, Levine, & Eisner, 1979; Wierwille & Gutman, 1978; Wierwille, Rahimi, & Casali, 1985) in the attempt to objectively measure cognitive load or “spare capacity” in applied contexts including driving in real traffic. The present paper outlines two main lines of research with secondary-task methods particularly in traffic-related contexts. Moreover, the sensitivity of the recently developed peripheral detection task (PDT)-method to the presence and to the modality of navigation messages during driving in a built-up area is tested. 1.1. Factors affecting processing efficiency under dual-task conditions: Automaticity, multiple resources and task priority Time-sharing between concurrent tasks or components of a complex task is usually associated with a cost of concurrence assumed to reflect the limited capacity of the human information system (Broadbent, 1958; see also Broadbent, 1982). Exceptions are the concurrent performance of very easy tasks i.e. tasks with a small demand on processing resources or tasks that can be processed in parallel due to automatic processing, which is known to develop from considerable amounts of consistent training (Shiffrin & Schneider, 1977). The degree of interference between concurrent tasks can be manipulated by task structure and task priority. Results from dual-task studies (Brooks, 1968; see also Baddeley, 1976; Wickens & Liu, 1988) strongly suggest that verbal and spatial processing use different resource pools. In accordance with multiple-resource theory, performance decrements are less severe during concurrent performance of cross-modal tasks compared with intra-modal ones (Wickens, Sandry, & Vidulich, 1983). Instructions about prioritisation of concurrent tasks have been found to affect performance accordingly (Schneider & Fisk, 1982; see also Wickens & Hollands, 2000) and in fact a most popular example of spontaneous task prioritisation is a car driver interrupting an ongoing conversation due to increased traffic task demand. Apparently, during driving, the driving task has a natural first priority while tasks unrelated to driving have a lower priority. However, spontaneous prioritisation of traffic-related information over other sources of information, should not be overestimated. Recent research indicates that in-car activities such as the use of cellular phones are associated with an increase in accident risk (Sagberg, 2001; see also Stevens & Paulo, 1997). Numerous results from basic research also indicate vulnerability of visual search to automatic capture of attention by salient stimulus features (Theeuwes, 1994). A most intrusive stimulus feature relevant to the secondary-task technique is abrupt onset and offset of stimuli (see Yantis & Jonides, 1990). 1.2. Studies with cross-modal and intra-modal secondary tasks during driving Secondary tasks have been used in real traffic for measuring variation in drivers’ cognitive load while driving. Brown and Poulton (1961), Wiegand (1974) and Harms (1991) used mental arithmetic i.e. calculation tasks as secondary tasks during driving in different traffic environments. The purpose of these studies was to investigate the relationship between “cognitive load” or “spare capacity” and variation in the driving environment. Both Brown and Poulton (1961) and Wiegand (1974) reported that subjects completed fewer calculation tasks in more complex driving environments than in simpler ones. The former study also included analysis of speed variations and a lower speed was found in complex driving environments (city traffic) than in simpler ones (residential areas). In the studies reported by Harms (1991) prolonged reaction time to calculation tasks was consistently found, in combination with a lower driving speed in more complex driving environments––i.e. different village areas and close to rural junctions––than on the observed adjoining road segments. The general interpretation of the results in the above-mentioned studies in terms of “cognitive load” or “spare capacity” is reasonable since the calculation tasks were presented verbally, were presented on a regular basis and required verbal responses. Regular presentation supports task prioritisation and reduces intrusiveness of secondary-task stimuli. Moreover, separation of the auditory-verbal secondary task from the task of driving (visual-motor) by sense modality should have reduced structural interference to a minimum in all these studies. Another line of research on cognitive load in traffic is that based on visual secondary tasks. Visual secondary tasks are primarily indicators of visual demand, which is most prevalent in driving. Some studies have suggested that detection of simple visual stimuli is also sensitive to variation in cognitive load. In an early study, Lee and Triggs (1976) found that increased environmental complexity during driving affected the detection of peripherally presented stimuli. They also reported a stronger effect in the left visual hemi-field than in the right one. Later, Miura (1986) demonstrated that task demand rather than visual complexity affected eye movement patterns as well as sensitivity in the driver’s visual periphery. Results obtained in other contexts than traffic support the assumption that peripherally presented stimuli are less likely to be detected at high levels of perceptual load in the foveal field of view (see Rinalducci & Rose, 1986; Williams, 1988). Chan and Courtney (1993) demonstrated that variance in cognitive demand also affected sensitivity in the visual periphery at a constant level of perceptual load in the foveal field. Their subjects were required to either name or add digits, presented in the foveal field and these instructions were found to affect the detection of peripheral stimuli although these were identical in both conditions. Recently, a study of visual search for hazards in videotaped traffic scenes (Crundall et al., 1999a and Crundall et al., 1999b) has demonstrated that the presence of traffic hazards on videotapes resulted in lower detection rate for peripherally presented stimuli. Apparently, detection of simple stimuli presented in the visual periphery is sensitive both to perceptual load in the foveal field and to task load. 1.3. Previous work with the PDT-method The PDT-method is based on the above-mentioned findings. The method involves frequent presentation of simple visual stimuli at random positions with eccentricities ranging between 5° and 25° left of the driver’s normal line of sight, and 2–5° over the horizon (in a driving simulator) or over the car console during driving (in a real car) (see Fig. 1). Subjects are instructed to detect as many stimuli as fast as possible without, at any moment of driving, withdrawing attention from the road scene. The method was first used in a simulator study in which different designs and functions of an ADAS was evaluated (Martens & Van Winsum, 1999). In the first part of this study, PDT-performance on baseline sections on motorways and rural roads was compared with PDT-performance on the same road segments where subjects were exposed to critical traffic scenarios. Compared to road sections without prespecified scenarios a decrease in hit rate and an increase in mean reaction time was found on road segments with unexpected traffic events and road scenarios requiring immediate action. The subjects were subdivided into three groups, a control group, and two experimental groups, equipped with an ADAS. Experimental groups were provided with visual warnings in case of critical traffic scenarios. In addition to visual warnings, one group received tactile warnings while the other group received verbal warnings. PDT-performance was analysed from the onset of a warning until 10 s after (corresponding to the duration of the visual message). Analysis of reaction times and hit rates in those intervals showed that tactile warnings were less damaging to PDT-performance than were verbal warnings. Olsson (2000) successfully transferred the PDT-method to field conditions. She used the same road types as the two previous studies and related PDT-performance to subjects’ monitoring of in-car equipment during driving. Subjects were presented with three different tasks, one (the radio task) was to report the preset car-radio frequency and find a specific radio station, the other (the CD-task) was to turn on the CD, play a specific track on a specific CD and then to return to the radio mode. Subjects in this experiment were also presented with a third task, mental calculation, which demand neither visual nor motor activity. Compared to a baseline, both reaction time and hit rate were generally affected by the presence of in-car tasks. Some differences between tasks were reported: Counting-backwards increased PDT reaction time dramatically and more than any other task but this effect was only found on the rural road. A higher hit rate was observed for the radio task than for the CD task. Some difference in PDT-performance between in-car tasks and different road types remained unexplained in this study. However, the finding that the counting-backwards task affected PDT-performance essentially the same as the tasks that demanded motor and visual activity, strongly suggested that PDT-performance is also sensitive to cognitive load. The figure shows position of stimuli in the instrumented car. The white arrow ... Fig. 1. The figure shows position of stimuli in the instrumented car. The white arrow marked with A points to the reflection of the LED in the windscreen. The arrow marked with B points to the navigation system. The telephone was not used in this experiment. Figure options 1.4. Another test of the PDT-method: Is it sensitive to the modality of equivalent IVIS messages? The results of the previous studies make the PDT-method a promising candidate for a standard procedure for the estimation of distraction from IVIS and ADAS, but it requires some further testing. In fact, the PDT-method is based on a visual secondary task and although previous studies suggested that this method is also sensitive to cognitive load, it might be sensitive to the sense modality of equivalent messages as well. This assumption is consistent with findings reported by Verwey (1993) that variance in traffic scenes affected a visual secondary task––regardless of its demand on processing (naming versus adding visually presented digits)––more than an equivalent verbal secondary task. Prior to the use of the PDT-method for comparative measurements of the impact of IVIS on driver attention, it is important to test the method with equivalent visual and verbal information primary sources of information. This would clarify the importance of stimulus modality to PDT-performance. However, sensitivity to task demand variation in the driving environment might easily overshadow the occasional captures of attention by IVIS messages. This would even be more likely in built-up areas with greater variation in traffic scenarios and road user categories than on the rural roads or motorways as used by Olsson (2000). In the present study, all subjects were exposed to the two navigation conditions (memory-based versus navigation system based). However, in the navigation system condition the subjects were randomly assigned to one out of three groups with different message modalities: Navigation messages were presented either visually, verbally or both visually and verbally (full instruction). Assuming that variance in traffic task demand would not overshadow the occasional captures of attention caused by navigation messages, the effect of the modality of the navigation messages on PDT-performance can be analysed and the importance of sense modality of the messages to PDT-performance can be clarified.

5. Conclusions In line with results from previous studies, PDT-performance showed a remarkable sensitivity to calls for attention due to navigation messages in the present experiment. By presenting subjects the navigation messages with the same content and on identical locations during driving, only the mode of presentation was varied in the present experiment. It was demonstrated that PDT-performance was not unrelated to the mode of presentation. On the basis of this finding it may be questioned whether the PDT-task is suitable as the only standard method for measuring distraction from IVIS. Apparently, a method or combination of methods, being less dependent on mode of presentation, would be more useful for measuring distraction and cognitive load more generally. On the other hand, the driving task requires continuous visual information processing and visual distraction is a very important component in safety evaluation of IVIS, therefore the method should have a prominent status in a test battery for safety evaluation of IVIS and ADAS.