Acento tonal pré-nuclear ascendente no português brasileiro: comparação com a fala disártica parkinsoniana

Abstract

This thesis investigates how the rising pre-nuclear accent is performed in Brazilian Portuguese. Based on the Metrical Autossegmental Theory and Prosodic Phonology, the literature presents two representations for this accent, L*+H and L+H*, indicating a tonal alignment variation. Thus, we aimed to elucidate the phonetic implementation of this tonal accent and, then, to discuss its phonological representation. In addition, we aimed to dialogue with the anchoring hypotheses, which propose a fixation of tonal targets to the segmental layer. For that, based on the concept of natural experiment, we analyzed the speech of patients with Parkinson's disease (PD) along with that of healthy individuals (CNT). A natural experiment consists in analyzing a phenomenon in a group that, by certain pathology, naturally shows modifications in the phenomenon. It was carried out in order to know more about the phenomen under studying in healthy people. Because of its phonoarticulatory constraints due to dysarthria, PD can shed light on the phonetic implementation of the prenuclear accent. We investigated the distance between the tonal targets, the positions of these targets, their heights, the amplitude of the melodic movement (f0 excursion) and the f0 variation rate. Besides that, we investigated the influence of the preceding intonation environment and the presence of boundaries of units above the word level - phrases. It was found that the valley lies in the regions of the initial boundary of the stressed syllable, moving to the right when the onset is voiced. The H, on the other hand, is in the final region of the stressed vowel, but the peak also moves to the right when the onset is voiced. L is more sensitive to the pressures of the preceding intonational environment: the higher the preceding peak, the more displaced to the right is the valley. H is less affected by these influences. Furthermore, the PD analysis showed vowels longer than in CNT and H aligned earlier. Thus, the PD analysis contributed to show that H is performed preferentially in the stressed vowel when there is more time (or syllabic space) available. This was confirmed in CNT in contexts where the pre-nuclear accent was followed by phrase boundaries. In that context, vowels were also longer and H was also performed earlier in the stressed syllable. In general, this study found a greater instability of L target and a quite stability of H. Then, we assert that the L+H* representation is the most adequate to represent rising pre-nuclear accents in Brazilian Portuguese. As for its phonetic implementation, we observed that the tonal targets were realized in specific regions (stressed syllable boundaries), but not with a stability as rigid as that predicted in the more conservative versions of the anchoring hypothesis. In addition, we found that the f0 variation rate is the most stable parameter among the others, and that the tonal targets height changes along with their displacements. From this, we interpreted that the Brazilian Portuguese speakers have to perform an f0 movement which is codified with a given f0 variation rate, but they adjust it according to the temporal constraints during the phonetic implementation. Those temporal pressures might cause the accent to shift to the right, and then, H may be performed in the post-stressed syllable.