Establishing the nature of context in speaker vowel space normalization
William F. Styler IV
Defended Spring 2008 at the University of Colorado and later featured on LolMyThesis
This study was designed to gain insight into the process through which humans are able to adjust to and understand the speech of unfamiliar speakers, referred to as "speaker normalization". Prior research has suggested that, in order for normalization to occur, a listener has to have some speech data (or "context") to process. The goal of this work was to further elucidate the role of this context, by searching for any effects that frequency of token occurrence and ordinal primacy may have on normalization.
Comparison of different stimuli in a forced-choice vowel identification task yielded no statistically significant differences, and even after an analysis of sources of error, neither a primacy effect nor a frequency effect was found to be supported by the data. This lack of support raises many interesting theoretical questions and suggests a variety of future avenues of exploration in the field of speaker normalization research.