Eliciting and analysing perceptions of prosodic prominence: a Maori case study

Abstract

This project outlines a methodology developed for analysing perception of prosodic prominence based on Māori, the indigenous language of New Zealand. Development of the study began with a pilot, which employed a perception test involving a recorded tapping exercise. The second phase implemented a participant-friendly, purpose-built web survey, increasing efficiency and accessibility. The third phase increased the scale and developed a statistical processing method for determining significantly perceptually prominent syllables (PPMs) based on the perception test responses. Ninety-two New Zealand and overseas listeners responded to 30 short recordings of continuous Māori speech from the MAONZE database (King et al., 2011a). Since native language cues affect prominence perception in any language, participants were further divided into three self-rated Māori proficiency groups. Results indicated a pattern of difference between high proficiency and lower or zero proficiency participants, both in response to the perception task, the effects of which were handled by the result-processing, and in PPM identification, though the latter difference was not always significant. Analysis of PPMs from all groups indicated strong alignment with syllables where existing expected phrase and word stress overlap, and with word stress (but not phrase stress) alone; a sensitivity to heavy (2+ morae) syllable weight and whole syllable duration; and influence from F0 movement around the phrase F0 peak or from the F0 peak itself. There were indications that the strongest prominence effect is cumulative, but that word stress cues are, or have become, stronger than phrase stress cues, and that this may be increasing with time.