Methods & Meta-science

Modeling and collecting distance data

Distance is a fundamental construct in theories of cognitive processes (e.g., categorization). Not surprisingly, several decades of research have focused on the design of mathematical models of distance data (e.g., multidimensional scaling, MDS), and on the assessment of methods for the collection of behavioral distances (e.g., dissimilarity ratings). I will dedicate the initial part of this talk to presenting the three major classes of distance models: spatial models (e.g., MDS), set-theoretic models (e.g., Tversky's contrast model), and graph-theoretic models (e.g., additive trees). In the second part of this talk I will introduce a number of methods for collecting behavioral distances, and describe how multiple distance models of the same behavioral data map onto each other. I will finally describe how changes in the behavioral method affect the fit of various distance models, and show that, at least in part, such changes in model fit can be explained by the influence of the method on the distributional properties of the behavioral data. A paper describing this research is available at: