Theoretical aspects of sensory discrimination

Publisher Website
Google Scholar
Abstract
Discrimination methods are commonly used in research on the chemical senses. Chemosensory stimulation often leads to multivariate sensations. A multivariate theory for a class of discrimination techniques is described here along with a discussion of the practical implications of the theory to experimenters. One of the most important findings is that the probability of a correct response for the discrimination tasks modelled is not monotonically related to the Euclidean distance between the means of the populations from which the stimuli are drawn. There are two important consequences which flow from this finding. First, the power of these methods cannot be determined without specifying values for the multivariate parameters. Second, the traditional assumption in multidimensional scaling that proximity measures and multivariate distances are monotonically related is invalid. The theory presented here is the basis for a new approach to multidimensional scaling using these discrimination methods without invoking the assumption of monotonicity. The conceptual basis is established to extend the model to include stimulus variation and an explicit assumption, based on Stevens' power function, about the nature of the relationship between the stimulus continuum and the sensation continuum.