This thesis investigates the perceptual categories associated with contrasting pitch accents in Tokyo Japanese. Various aspects of pitch movements throughout the words are systematically varied in order to determine what aspects of the pitch contour affect categorization of words on the basis of accent. In addition, this thesis also investigates the effect on categorization of loss of pitch due to lack of voicing in various parts of the word. The model determined by these studies reveals two more-or-less orthogonal perceptual dimensions; pitch alignment with speech segments determines accent location and the amount of pitch drop determines accent presence. This study also investigates how to quantify the distinctive function of pitch accent in a way which incorporates the frequency of the contrasting items, as well as the peculiar category structure of accents. This model was applied in the analysis of a large-scale lexical database, revealing many irregularities in the distribution and use of accents. Comparing this quantification of the lexical use of accent with the perceptual experiments shows that accent-location detection is functionally more fundamental than accent-presence detection in short, 2-mora words. (PsycINFO Database Record (c) 2016 APA, all rights reserved)