On the Accuracy and Reliability of Predictions by Perceptual Control Theory: Five Years Later

Abstract

This is a demonstration of an idiographic procedure that makes very accurate quantitative predictions of one person’s specific actions during a task. In June, 1988, I completed Condition 1 for each of eight different runs of a two-condition pursuit tracking task. Using data from Condition 1, when only the control handle affected the position of the cursor, I estimated the constants in the model of behavior from perceptual control theory, then I used the model to predict the results for Condition 2, when a random disturbance would also affect the cursor. Every five years I will do two of the eight predicted runs; I did two of them in July, 1993. Correlations between 1800 pairs of predicted and actual handle positions were .998 for the first run, and .997 for the second. Data from the model and me were nearly indistinguishable. I briefly discuss some misconceptions about data like these.