Optimum characterization of individual information-processing skills requires isolation of assessable components. In matching-to-sample, three elementary processes might be separated if a proposed model of serial, selfterminating search were supported: registration of the stimulus in short-term memory, comparison of the two registered stimuli, and execution of the identifying response. Support for the model was obtained when response latencies were examined as a function of left-to-right position of the target stimulus in a display, but quantitative estimates of components could not be made because of interactions between tasks and the linearity of the scanning process. A second experiment, which varied the number of comparison stimuli, yielded highly linear functions whose slopes and intercepts violated certain aspects of the model. Data on eye movements obtained in a third experiment again supported the basic model but indicated that median response latencies represent a confounding of optimal and suboptimal performances.