Selective Visual Attention during Multiple-Process Control

During multiple-task control, eye movements are central for the performance of aimed actions and the development of working strategies. In order to investigate visual and informational workload the behavioural pattern has been analysed. Information presented on a screen has to be visually analysed in order for a decision to be reached on the action to be performed. Normally peripheral vision and the knowledge of the displayed optical structure are used to select the next position of fixation. The resulting scan path is determined by the context and the dynamics of the processes to be controlled. Therefore, the question arises to what extent characteristics of the eye movements are affected by the amount of workload.

In order to test this an experimental setup was designed which makes it possible to record eye and head movements simultaneously. From these data the gaze position on the screen could be computed. During the experimental sessions trained subjects were asked to control a complex process simulation by means of cursor movements. This simulation displayed on a high-resolution colour screen consisted of four process units similar to Crossman's water bath paradigm. Temperature and flow rate represent the control variables for each process unit and have to be adjusted to and kept within a certain tolerance range. The visual scan path and all actions on the graphic tablet were recorded continuously during the experimental sessions.

An attempt is made here to describe the allocation of attention during multiple-task performance. The results provide evidence that the contextual structure of the task determines the uptake of specific visual information. The amount of visual workload affecting the subject during the control of all four simulated processes is determined by two independent variables: the flow rate (kinetics) and the number of activated processes. On the one hand fixation duration is affected by the kinetics and on the other hand the stability of the activated processes is very sensitive to both independent variables. These effects are due to the limited capacity of the subject to deal with more than one process state within a certain time interval. He or she is forced to divide his or her attention between different processes. There seems to exist only a poor mental representation of the dynamics of several processes running in parallel.