An acoustic study of Sardinian and Algherese Catalan vowels

Abstract

The present dissertation sets out to instrumentally investigate the quality of the vowels of Algherese Catalan and of four Sardinian varieties in order to pursue three different aims. The first aim concerns the evaluation of the possible role played by Sardinian in modifying the quality of Algherese Catalan vowels. To fulfil this goal, a formant analysis of the seven Algherese vowels /i, e, ɛ, a, ɔ, o, u/ in stressed position has been carried out for 525 realisations (7 vowels x 3 consonantal contexts x 5 repetitions x 5 native speakers), and the corresponding formant frequency values have been compared to those obtained for 2,105 Sardinian realisations (7 vowels x 3 consonantal contexts x 5 repetitions x 5 native speakers x 4 varieties + 5 extra tokens). As a framework of reference for Catalan, the formant data reported in Recasens & Espinosa (2006) for the stressed vowels of four Catalan varieties have also been included in the present work. The analysis results show that the acoustic quality of Algherese vowels is more similar to that of the Sardinian vowels than to that of the vowels of other Catalan varieties. In terms of individual dialects, as expected, the vowels most resembling those of Algherese were found to belong to the Northern Logudorese variety. These tentative conclusions were also strongly supported by statistical analyses (ANOVA) of the formant frequency data. According to this scenario and given the fact that Algherese still possesses the same seven vowel phonemes as Catalan, its vowel system was defined as phonologically Catalan and phonetically Sardinian. The second aim of the present work is to evaluate two predictions of the Adaptive Dispersion Theory (ADT) against instrumental data from Algherese, Sardinian and other Catalan varieties. Specifically, the first prediction suggests that larger vowel inventories should cause the vowel space to expand, while the second suggests that vowels in more crowded inventories should vary less than vowels in less crowded ones. These predictions were evaluated with the same formant frequency database which was used in order to fulfil the first aim of this dissertation. The first prediction was tested by calculating the point-vowel area for each dialectal variety, and the second by calculating the standard deviations of the contextual formant frequency values for each vowel in each dialect and across dialects. In general terms, the results challenged the validity of the two ADT predictions, since, in comparison to less crowded inventories, those which were more crowded did not always exhibit larger vowel spaces and less context-dependent variability. The third aim of this dissertation is to gather information on the acoustic reduction of Sardinian unstressed mid vowels, and was investigated by analysing a further set of 1,215 unstressed vowels of four Sardinian varieties (12 vowels x 5 repetitions x 5 informants x 4 varieties). In this respect, some very interesting data published by Contini (1987:449) indicate that in three Sardinian varieties unstressed /ɛ, ɔ/ do not present, on average, visible signs of rising with respect to their stressed counterparts. This finding is in contrast with two models normally used to describe acoustic vowel reduction, i.e. the undershoot-based model, which predicts an overall elevation of the vowel floor for reduced vowels, and the centralisation-based model, according to which similar contextual and prosodic conditions are expected to trigger a shift of the target vowels towards a central or ‘schwa’ region. Our results confirmed Contini’s finding in that the Sardinian unstressed open-mid vowels analysed in this study lacked visible rising. These data led us to conclude that, when evaluating acoustic vowel reduction for Sardinian mid vowels, language-dependent or systemic variables should also be taken into account

Date of Award	28 Feb 2014
Original language	English
Supervisor	Daniel Francisco Recasens Vives (Director)