UC Irvine

This dissertation applies various mathematical and computational methods to investigate properties of color naming systems and their evolution. An advantage of using quantitative analyses to study behavioral topics is the ability to concretely define complex processes and phenomena in a precise and controlled manner. In this manuscript, I present three studies which utilize computational models and mathematical analyses to examine features of both static and evolving color naming systems. The first two studies use evolutionary game theory to simulate the evolution of categorization systems modeled using data from the World Color Survey. The first study investigates the evolution of these naturally occurring systems under a simple communication mechanism and the second study explores how the introduction of non-salient regions in the color space affects the evolutionary process. The third study applies a Bayesian nonparametric mixture model to the World Color Survey data in order to reveal universal patterns in color naming across different languages. This dissertation argues for the effectiveness of applying quantitative methods to behavioral data sets, such as the World Color Survey, and explores the types of inferences which can be drawn from these results. Finally, preliminary results from a pilot study examining the correlation between minimized flicker settings and observer phenotype are also presented.