|
Article Excerpt
INTRODUCTION
Currently more than 180 million Americans use wireless phones (Cellular Telecommunications and Internet Association, 2005). It is further estimated that 3.9% of passenger car drivers are using their wireless phones while driving at any point in time (Utter, 2001). The rapid growth of wireless phone use while driving has fostered interest among safety professionals and, more recently, among the public at large in the associated problem of driver distraction. Despite the significant limitations of crash data for determining the incidence of distraction as a contributing factor in crashes (e.g., Goodman, Tijerina, Bents, & Wierwille, 1999) and the absence of matched exposure data necessary for assessing the crash risk associated with phone use (Stutts, Reinfurt, Staplin, & Rodgman, 2001), the scientific evidence continues to accumulate in support of the conclusion that wireless phone use while driving can compromise safety (e.g., Laberge-Nadeau et al., 2001; Redelmeier & Tibshirani, 1997). It is also becoming clear that the problem of amassing sufficient scientific evidence to rigorously demonstrate the role of distraction as a causal factor in crashes is a significant and daunting problem, requiring convergent evidence from numerous studies (e.g., Tijerina, 2000).
Indeed, the real-world incidence of distraction depends on numerous motivational factors that contribute to drivers' willingness to engage in secondary tasks. Furthermore, the resulting impact on safety depends on the timing of distracting activities in relation to the rapidly changing dynamics of the immediate driving situation. In this context, experimental studies are best suited for evaluating the relative distraction potential associated with various in-vehicle tasks, defined as the decrement in driving performance associated with a particular secondary task under a given set of driving conditions.
The emergence of mobile technologies that permit message retrieval and Internet access allows for the integration of a variety of complex task capabilities into moving vehicles. The safe operation of these technologies is predicated on the assumption that voice-activated and speech-based interfaces are sufficient for preventing significant distraction because they allow drivers to keep their hands on the wheel and their eyes on the road. This assumption is the basis for what Strayer and Johnston (2001) referred to as the peripheral-interference hypothesis, according to which distraction is attributable primarily to direct conflicts between the visual and manual demands of (secondary task) interface manipulation and the corresponding demands of driving. The alternative to this, the attentional hypothesis, attributes a significant component of distraction to the "attention-demanding generative components of processing" (Strayer & Johnston, p. 466).
Wickens's multiple resource model (Wickens, 1984, 1999; Wickens & Hollands, 2000) provides a framework for predicting how secondary tasks may interfere with driving. Strayer and Johnston's (2001) distinction between peripheral and attentional interference is similar to Wickens's stages of processing dimension, which is one of three structural dichotomies that constitute the model. Accordingly, driving and in-vehicle secondary tasks are likely to interfere with each other because both tasks involve activity at the perceptual-cognitive (attentional) stage as well as the responding (peripheral) stage. The multiple resource model also supports predictions concerning the relative level of interference expected when one performs secondary tasks using different interfaces. Specifically, the model predicts that the amount of interference will be modulated by the mode of secondary task performance--that is, the auditory mode will involve less interference than will the visual mode because driving inputs are mainly visual. The amount of interference will also be modulated by the secondary task response mode--that is, responding via voice will interfere less than will responding manually, given that driving control responses are primarily manual.
These predictions form the basis of the present study, which had the objective of determining whether using a speech-based interface would reduce the interference associated with the performance of secondary tasks during driving. The experimental approach was to control the primary (driving) task demands and measure the relative level of interference associated with secondary tasks with different task characteristics and performed with different interfaces. To accomplish this, we developed a set of in-vehicle secondary tasks that could be performed using either a visual/manual or a voice-based interface. The number of steps and associated cognitive demands were roughly equivalent.
As PC-based technologies expand the types of tasks that can be conducted while driving, it is important to determine whether increases in task complexity will provide a significantly greater potential for distraction, relative to the more simple tasks that are commonplace among drivers today. In particular, tasks that require navigation of hierarchical menu systems or searching a database of messages or phone numbers may impose a significant burden on the driver's working memory, which could increase distraction. Jacko and Salvendy (1996) argued that menu depth is related directly to task complexity because "increased depth involves additional visual search, decision-making, response selection, and greater uncertainty as to the location of the target" (p. 1195). In the present study, we defined complexity in terms of the number of steps required to complete a task and the working memory load associated with remembering the task sequence and selecting a response. Because our complex tasks involved additional effort at both of Wickens's (Wickens, 1984, 1999; Wickens & Hollands, 2000) stages of processing, we predicted that performing complex tasks would impair driving performance more than performing simple tasks would and that this impairment would be evident at both the perceptual-cognitive and response stages.
A common experimental paradigm used to assess the distraction potential of in-vehicle tasks utilizes a car-following task, in which the lead vehicle decelerates unexpectedly while the driver is engaged in a secondary task (e.g., Lee, McGehee, & Brown, 2000). Concerns have been raised that drivers' behavior may change as a result of the use of repeated surprise trials (Muto & Wierwille, 1982). To address this problem, we selected a methodology that allows assessment of distraction potential in noncritical situations. The method, based on the work of Brookhuis, de Waard, and Mulder (1994), utilized a car-following task in which the speed of the lead vehicle was varied systematically and the speeds of the lead and following vehicles were used to compute coherence and the associated measures of phase shift and modulus. Coherence is a measure of squared correlation, reflecting the extent to which the respective speed traces have the same periodicity. When coherence is relatively high (e.g., [greater than or equal to] .80), the driver is adequately following the lead vehicle's speed changes, which implies that the associated measures are meaningful.
Brookhuis et al. (1994) found that wireless phone use while driving increased the phase shift of the two speed signals, reflecting an increase in the lag or car-following response time, which they referred to as delay. They argued that delay incorporates both perceptual and cognitive factors, in contrast to measures that focus primarily on operational-level behaviors, which are more automatic and thus less susceptible to effects of cognitive distraction. Like response-time measures, increased delay reflects slower response, which could directly increase the likelihood of crash involvement in the event of a sudden stop by the lead vehicle. Modulus (gain) reflects the following driver's responses at the extreme values of the lead vehicle speed. Specifically, modulus values near 1.0 indicate that the following driver is closely matching the extreme speed values of the lead...
|
