Most learning is the result of encounters with a limited number of environmental circumstances. Numerous experimenters have compared the ability of human observers to distinguish between various simple stimuli with the limits set by absolute discriminability. Under most conditions man's performance falls short of what could be achieved by appropriate ideal detecting devices. Stimulus-sampling theory has led to the development of numerous special models. The model on which the best-developed treatment of generalization is based is called the component model. Many of these experiments have been modeled on those done with animals, but, whereas some have used conditioning procedures, most have been concerned with "voluntary" behavior involving verbal instructions to the subjects. The procedures used in experiments on generalization are essentially similar to those used in psychophysics for the measurement of thresholds. The adequacy of any theory of generalization may be judged in part by its ability to describe results obtained in psychophysics and perception.