Bimodal compensatory tracking and attention

Abstract

Subjects performed visual and auditory compensatory tracking separately as well as in dual task combination. In a third condition Ss tracked one signal and copied with his other hand the control movement, thus emitting two identical responses to one input. Normalized mean squared error (NMSE) was lowest in single task, highest in bimodal dual task, with copying/tracking NMSE falling somewhere in between. NMSE was also found to vary as a function of input bandwidth, order of control, and input/plant bandwidth and order of control combinations. The error signal in each loop was sampled once every 100 msec and the samples were then compared. The absolute magnitude of the error in one channel at any given instance was found to be independent of the magnitude of the error in the other loop. This relationship held even when the information processing rate was increased by increasing the input bandwidth. The results do not support single-channel, information processing theories of attention. Moreover, they indicate a revision of current undifferentiated capacity models of attention (i.e., Kahneman, 1973; Moray, 1967). It was found that the performance of either one of two concurrently performed tracking tasks was a function of the informational content of the other. This result implies that attentional resources are not allocated freely to the various tasks, but rather, it suggests that the amount of attention allocated to a given task depends and is a weighted function of the concurrent informational content of other, unrelated and distracting, tasks and events. Although, it was fairly evident from the results that subjects could perform continuous tracking tasks simultaneously the observed task interference effects do indicate that the human controller of two, otherwise parallel, multiple single-loop systems does not behave either as a true parallel, or as a serial, information processor. It is suggested that one of the key functions of man's attentional process is the modulation of stimulus information to enable more or less optimal input to more or less permanent structures in the brain. Thus it is argued, a general undifferentiated capacity theory of attention may not be incompatible with current multiprocessor theories of attention, (Allport, Antonis and Reynolds, 1972). Moreover, in view of supportive existing evidence, it is also suggested that one other functional role played by attention is to enable the establishment in memory of an internal representation of task invariant descriptions than can actively be drawn upon (resources) and implemented to reduce the magnitude of task-related information. Hence, the amount of information, or uncertainty, associated with the performance of a given task is conceived to be the magnitude of discrepancy, or mismatch, between expected and actual task dependent events. Within such an undifferentiated capacity framework, it is possible, therefore, to account for time-sharing decrements of performance arising directly from either (1) changes in processing linearity, or (2) response delay, or both of these factors without having to appeal to discrete, serial, human information processing models of attention.