Attention Allocation in Dynamic Environments.

Three policies of attention resource allocation between tasks of dynamically varying difficulty are described. These policies--optimal allocation, optimal resource expansion, and non-optimal allocation are distinguished analytically by the gain of the transfer function between task difficulty and primary and secondary task performance. Eight subjects time-shared two compensatory tracking tasks in which the control dynamics of the primary task fluctuated continuously between first and second order. Linear control analysis of the difficulty and filtered RMS error performance measures indicated that subjects were initially non-optimal in their allocation policy, failing to guard the primary task in the face of fluctuations in its difficulty. With practice, a trend toward more optimal performance was observed. This appeared to be related to greater automation of performance at the most difficult level. However, close anlaysis and comparison of the variable difficulty data with performance in constant difficulty dual task conditions indicated a persisting limitation in subjects ability to reallocate resources from the secondary task when required by demand changes of the primary.