This thesis investigates the role of voice quality in expression of politeness under conditions of varying relative social status among Japanese male speakers. The thesis also sheds light on four important methodological issues: 1) experimental control of sociolinguistic aspects; 2) eliciting semi-natural spontaneous speech which satisfies naturalness; 3) recording quality suitable for voice quality analysis; and 4) the use of direct waveform and spectrum measurement as a non-invasive method for measuring glottal characteristics related to perceived voice quality. Japanese has been believed to rely on what has been called “negative politeness” (formality and deference). Since explicitly expressing deference under the Keigo (Japanese system of honorifics) requires mastery of a highly complex system, in daily conversation, this function may be taken over by vocal paralinguistics. Also, as the Keigo system is not supposed to convey “positive politeness” (friendliness and solidarity), vocal paralinguistics may contribute to this aspect of politeness. The use of high fundamental frequency (FO), which has been considered a universal cue of politeness, is more associated with femininity in Japanese, and this usage of FO is not observed in male speakers, who possibility employ other vocal cues to express politeness. Therefore, this study focuses on voice quality of male speakers in expressing politeness. To obtain natural, unscripted utterances, the speech data were collected with the Map Task. This task also allows us to study the effect of manipulating relative social status differences among participants in the same community. For voice quality analysis, a direct waveform and spectrum measurement (Hanson, 1995) was employed. An experiment was conducted to observe the alignment between acoustic measures and the perceived politeness from both written and spoken versions of the utterances obtained from the Map Task. The results suggest two principal findings. First, Keigo does not play a role in conveying politeness in everyday conversation, but speakers showed politeness through voice quality variations. Second, in judging the politeness of test utterances, listeners reacted to the irregularity of the waveform and spectral characteristics in the third formant region.