Multiple Resource Time Sharing Models

The multiple-resource model predicts the degree of interference between two time-shared tasks, i.e., it predicts the loss in performance of one or both tasks carried out concurrently, relative to their single-task baseline measures. As such, it is a model of workload effects on performance, in which the sources of workload are multiple task demands. The model we describe here is based upon multiple-resource theory (Navon and Gopher, 1979; Wickens, 1980, 2002), which posits that three factors are important in predicting how well (or poorly) a task will be performed when time-shared with another:1. The difficulty, or demand for resources, of each single task component (e.g., driving in traffic is more resource demanding than driving on an open road)2. The allocation of those limited resources between two time-shared task (e.g., whether driving is emphasized at the expense of using in-vehicle technology, or the converse)3. The extent to which the two tasks demand common or separate attentional resources (e.g., an invehicle visual display will demand more common resources with driving than will an in-vehicle auditory display)As the above example suggests, separate resources are defined by auditory versus visual processing. In addition, synthesis of dual-task research (Wickens and Hollands, 2000) also suggests that separate resources are defined by spatial (analogue) versus verbal (linguistic) processing; by perception and working memory versus responding; and by focal versus ambient vision. Thus any task can be represented by a set of demands on either dichotomous level of one or more of these four dimensions. Any pair of tasks can then be represented by the degree to which they share common levels on each dimension (e.g., both auditory) and by their combined demand for resources. The amount of shared resources and the combined demand predict the total interference between tasks. Then the resource-allocation policy between tasks (the extent to which one is favored and the other neglected) determines how this interference (dual-task decrement) is apportioned between them. 0-415-28700-6/05/$0.00+$1.50 © 2005 by CRC Press LLC40.2 ProcedureThe computational version of the multiple-resource model (Wickens, 2002b) involves the following steps:40.21. Step 1: Code Time-Shared TasksEach task that will be time-shared is coded by the extent to which it depends on separate resources defined by the four dichotomous dimensions mentioned above, as shown in Table 40.1. Demand levels (including 0) within each resource can take on simple integer values, with greater demand implying greater value. As an example, a very simple conversational task would be coded as: Perception: auditionverbal (=1), Working memory: verbal (=1), Response: verbal (=1). A very demanding vigilance task (detecting weapons in x-rayed luggage) would be coded as: Perception: vision-focal-spatial (=3), Working memory (=0), Response: vocal or manual (=1). Thus each task spawns what is called a demand vector. This vector has two important properties for computing interference:1. The average level of demand across all resources involved. For the simple conversation task, this is 1.0; for the vigilance task it is 3.0.