No abstract

Multiple measures of operator workload may dissoci,ate, or fail to agree, for a given task. The goal of this study was to determine how workload as indexed by processing resource demand could explain the attendant variance in a second measure of workload-subjective ratings.A multiple structure model of processing resources (Wickens, 1984) guided construction of tasks that were both performed by subjects and rated on workload similarity. Multidimensional scaling analyses of the similarity ratings produced three dimensions of workload that were explained by parameters of the resource model. Performance data, effort ratings, and heart period variability scores used to support the scaling analyses revealed expected dissoci-, ations. These dissociations were examined within the structure of the resource model.

The processing resources enabling dual task performance are modeled as a vector quantity, with dimensions defined by processing stages, processing codes, and processing modalities. This multidimensional conception is employed to model operator workload and to predict task interference patterns, employing a Sternberg task methodology. The implications to time-sharing performance are addressed as they pertain to the stage dimensions (Experiment 1: Manual Tracking), the code dimension (Experiment 2: Failure detection), and the code and modality dimension (Experiment 3: Multi-element monitoring).

The present paper outlines three major assumptions often implicitly assumed in dual task experiments conducted to assess operator workload. These assumptions are shown to be incorrect. Three criteria which should be met in dual task experiments that draw inferences from secondary task decrements are discussed. An experiment, meeting the proposed criteria, was conducted which demonstrated that when the criteria are met secondary task performance can be predictive of primary task difficulty. However, the data also indicate that a simple assessment of efPort alone will not predict total task performance.

A multiple structure model of processing resources (Wickens, 1980) guided construction of tasks of differential resource demand that were both performed by subjects and rated according to workload similarity. Analysis of performance data generally supported model predictions. Multidimensional scaling analysis of the similarity data produced subjective dimensions of workload that were explained in terms of resource demand and task structure.