To catch a thief with a recognition test: the model and some empirical results

The purpose of the present paper is to describe a new technique and a mathematical model--called the "Catch model"--for identifying a face previously seen (i.e., the target face). Both the technique and the model were developed on the basis of the general approach of information processing used with respect to human memory. Subjects were presented with a pair of test faces on each trial. Neither of the test faces was the target face. Their task was to choose from the two test faces the one most similar to the target face. The data furnished by the subjects were used to reconstruct the target face in the following way: At each trial the differentiating values, such as a long nose and blue eyes, of the test face chosen by the subject were recorded. These values were the ones that accounted for the difference between the two test faces. Over the whole run of the test trials, the differentiating values were associated with various frequencies of occurrence. The target face was reconstructed by selecting the differentiating values having the highest frequency of occurrence. Only one differentiating value per facial dimension such as a nose and eyes could be selected. Thus, given that the facial dimension of the nose has three different values consisting of the long, short, and wide varieties of nose, the value chosen would be the one associated with the highest frequency of occurrence. Mathematical derivations show that, given different variations of the proposed technique, the target face will be detected. These derivations were supported by the results of three experiments.