Structural Representation of Perceptual Similarity and Preference

of several stimulus processing techniques will produce the best result,
and of all the criteria that may be used for judging success,
human preference judgments provide the most valid criterion variable
for determining which technique is best. The reasons one technique is
preferred are often sought through comparison of experimental stimuli
in terms of identifiable attributes, but sometimes the stimulus dimensions
most relevant to human perception and preference are not well known.
This tutorial presentation examines methods from psychological science
that are typically used to find significant multidimensional differences
that exist between those stimulus samples. Examples are given of
experiments using naive subjects who make both dissimilarity
ratings and direct attribute ratings, and are also required to select which
stimuli are preferred with respect to particular parameters such as
naturalness, suitability, or pleasantness. The multidimensional perceptual
space for the selected stimuli, termed Stimulus Space, is derived from
dissimilarity judgments using INDSCAL analysis. In subsequent analysis,
termed multidimensional perceptual unfolding, the direct attribute ratings
and/or preferences for the stimuli provide a means for interpretation
of the dimensions of the Stimulus Space in terms of verbal descriptions
and/or ideal points.