Developing a methodology for eliciting and analyzing perceived prosodic prominences : a m ā ori case study

This paper outlines a methodology developed for eliciting and analyzing perceived prosodic prominences, based on a study of Māori, the indigenous language of New Zealand. Māori has some existing description of expected stress patterns and pitch contours, including which syllables may be expected to be prominent (e.g. Biggs 1969, Bauer 1993). Owing partly to the influence of English, there exists the perception that the sound of the language has changed over time. This has been observed by listeners of all proficiency levels, from fluent speakers to first-timers, and is found in the recordings from different time periods collected in the MAONZE database (King et al. 2011). The principal aim of the study described here was to determine where listeners hear prosodic prominences in Māori, and examine what it is acoustically that makes them hear certain units (in this case, syllables) as prominent. In the study, a set of 30 utterances from three different time periods were heard by 92 listeners. The utterances were continuous speech, taken from interviews in the MAONZE database: 10 each from male Historical Elders (b. 1880s, rec. 1940s), male Present Day Elders (b. 1920-30, rec. 2000s), and male Young L1 speakers (b 1970-90, rec. 2000s). The study participants were divided into three roughly equal-sized groups according to their self-rated Māori proficiency: 'High' (28) meant conversational ability from fluent to basic; 'Exposed' (30) meant word-and-phrase or parsing level ability and was the category for most New Zealanders. 'Zero' (34) meant no or low exposure, and applied to most of the overseas participants. All participants had high English proficiency. They were sourced via advertising, networks, and social media applications. As highly proficient speakers of some languages (including Māori) are difficult to come by, sourcing participants is a key problem in a study of this type; the others being selection of stimuli, appropriateness of delivery method, and result-wrangling. First, the stimuli. Here, they had to be chosen based on clarity, completeness, length and content. Though 'perfect' delivery is very hard to obtain, interruptions in the source recordings will affect an inherently relative concept such as prominence. To find a balance between participant stamina and the opportunity to test enough tokens, study size was set at 32 utterances. They are a mixture of simple and complex sentences. In hindsight, most participants could probably have handled twice the number of stimuli. The delivery method was a web form interface. The utterances …