Eliminating Stimulus Bias in Category Scaling

Abstract

Category scaling provides a listener with a fixed set of numbers, or ratings, for assignment to sounds under test. Unfortunately, category scales are sensitive to stimulus bias—that is, to the specific distribution of sounds arbitrarily selected for testing. Stimulus bias may, however, be reduced by an iterative procedure suggested by Stevens. A series of experiments is successively performed in which the results of the first experiment provide the stimuli for the second, etc. It is demonstrated that the procedure eliminates stimulus bias over a wide range of conditions for auditory frequency and intensity, as well as for several nonauditory variables. (Research performed at the Applied Psychology Research Unit, Cambridge, England, under a National Science Foundation Fellowship.)