Two operations in character recognition: Some evidence from reaction-time measurements

Theories of the recognition of a visual character may be divided into three sets.defined by the way in which the stimulus is encoded before being compared to a memorized target character. A character-classification experiment was performed in which the test stimuli were characters that were either intact or degraded by a superimposed pattern. Analyses of reaction-times in the experiment lead to the rejection of two of the three sets of theories. There appear to be at least two separate operations in the recognition or classification of a character. The first encodes the visual stimulus as an abstracted representation of its physical properties. The second, which may occur more than once, compares such a stimulus representation to a memory representation, producing either a match or a mismatch. A theory of high-speed exhaustive scanning in memory underlies the experiment and is given new support. The method of reaction-time analysis that is introduced, an elaboration of the Uelmholtz-Donders subtraction method, may be applicable to the general problem of the invariance ofperceived form under certain transfomations of the stimulus.