Evidence for feature perturbations from character misidentifications

Evidence that features from characters in a briefly exposed display undergo perturbation has come from tasks requiring direct reports of feature locations. In the current experiments, a different approach was taken in which perturbations were inferred from incorrect name responses to letter and number stimuli. The relative positions of two segments (features), one of which was anchored so that its perturbation would be difficult, determined the name of the character that was to be reported. The procedure produced data which suggest that feature perturbations occur when characters are exposed briefly, and that these perturbations tend to move toward fixation.