Visual selection: Scanning vs filtering

Ss were required to report whether two successively presented lines of Os and Ds were same or different. In a pretest, failure to detect differences was a function of the position of the substitute letter, increasing with the distance of that letter from the left of the line. However, after training, it is the utility of a position, the probability of its containing a substitute letter, and not position per se, which determines accuracy. This result, and the conditions of training which produced it, cast some doubt on theories which regard visual selection as a scanning process derived from overt eye-movements. A filter model seems more appropriate to the data.