Stochastic Prediction in Pursuit Tracking: an Experimental Test of Adaptive Model Theory

In this paper we test the proposition that in pursuit tracking, subjects compute stochastic (statistical) models of the temporal variations in position of the target and use these models to forecast target position for at least a response time interval into the future. A computer simulation of a human operator employing stochastic model prediction of target position is used to generate a synthetic pursuit tracking response signal. Actual pursuit tracking response signals are measured from 10 normal subjects using the same stimulus signal. Cross correlation and spectral analysis are employed to compute gain and phase frequency response characteristics for both synthetic and actual tracking data. The similarity of the gain and phase curves for synthetic and actual data provides compelling evidence in support of the proposition.