ABSTRACT
The inherent complexity of psychological processes and phenomena requires theories and methodologies that recognize and address this complexity. Decades of overreliance on univariate analysis of variance, unidimensional item analysis techniques, and unidimensional scaling methods has, in our opinion, led to oversimplified and incomplete understandings of social, cognitive, developmental, and behavioral processes and phenomena. Although conventional multivariate data analysis methods, for example factor analysis, have aided the definition and delineation of constructs measured via explicit item domains (e.g., personality and attitude), these methods have been of limited use in the measurement and modeling of domains where subjects’ perceptions of stimuli and stimulus relationships have been the focus. For this class of problems, multidimensional scaling and related models are potentially very useful. They can be used to reveal and quantify the structure of complex stimulus domains, isolate and identify individual differences in perception, cognition, and preference, and measure changes in perceived structure over time, across subject populations, and experimental interventions.
