|Abstract:||The phonological feature [±nasal] is used to describe sounds that are distinguished based on the relative state of the velopharyngeal opening, but does not distinguish oropharyngeal differences. The opening of the velopharyngeal port for [+nasal] vowels couples the oral and nasal cavities, which significantly alters the acoustic spectrum of nasal vowels. It is therefore difficult to discuss the articulatory processes that are responsible for acoustic differences compared to oral vowels. Furthermore, many studies now show that oropharyngeal articulations are not trivial in the differentiation of [±nasal] sounds. Rather, these oropharyngeal differences systematically enhance or compensate for the acoustic effects of nasal coupling. Additionally, the phonetic implementation of [–nasal] vowels in oral and nasal contexts is understudied, but is necessary to understand the relationship between [±nasal] vowel pairs.
In this dissertation, I present the results of an articulatory and acoustic study regarding the phonetic realizations of oral, nasal, and nasalized vowels in Brazilian Portuguese (BP), including data from thirteen speakers from southeastern Brazil. BP arguably has [+nasal] vowels (i.e. phonemic nasal vowels), and two classes of [–nasal] vowels (oral and phonetically nasalized vowels) in its inventory. The systematic comparison of these vowels is made to disambiguate the oropharyngeal mechanisms that differentiate these three types of vowels. Articulatory data includes real-time magnetic resonance imaging (rt-MRI) data collected in 1) a single-slice midsagittal orientation, and 2) four simultaneously-acquired slices in coronal, oblique and axial orientations, through the tongue tip, velopharyngeal opening and tongue dorsum, hyperpharynx, and hypopharynx, respectively. Acoustic data includes 1) recordings made simultaneously with the MR scans, and 2) high-fidelity recordings made in a noise-attenuating booth. Comparisons are made between phonemic oral and nasal vowel congeners, as well as phonetically nasalized vowels, for the vowel categories /a/, /i/, and /u/.
Results show articulatory differences in the oropharyngeal cavity between oral and nasal vowel congeners, specifically in the tongue blade and hyperpharyngeal regions. This is especially apparent for the vowel category /a/. Articulatory differences include a higher tongue blade and a less constricted hyperpharynx for /ã/ compared to /a/. The acoustic effects of nasalization are argued to be enhanced for /ã/. Based on phonetic evidence that this vowel has diverged from its oral counterpart, I agree with alternate transcriptions of the central low nasal vowel as /ɐ̃/ . For /ũ/, the tongue blade is lower and the hyperpharyngeal region is more constricted, compared to /u/. This is also argued to enhance the acoustic effects of nasalization. Two patterns emerged for /i/—for the majority of speakers, the tongue blade was lower and the hyperpharynx was more constricted for /ı̃/ compared to /i/. For other speakers, the opposite pattern occurs. In addition to articulatory distinctions made throughout the vowels’ durations, high nasal vowels in BP are undergoing nasal coda emergence, which is argued to be a strategy to further enhance the oral/nasal vowel distinction.
Results in the acoustic domain largely mirror those in the articulatory results. F1 is much higher for /a/ compared to /ã/, which is due to oral/nasal coupling as well as relative expansion in the hyperpharynx and a higher tongue blade position for /ã/. F1 of /ũ/ is higher than that of /u/. This is expected due to nasal coupling, increased constriction in the hyperpharynx, and lower tongue position. F1 of /i~ı̃/ shows more variation, though two patterns emerge—one in which F1 of the nasal vowel is higher than that of the oral vowel, due to oral/nasal coupling as well as oropharyngeal articulation (tongue blade lowering and pharyngeal constriction), and one in which F1 of the nasal vowel is lower than that of the oral vowel. This second pattern is seen in speakers who produce the nasal vowel with a relatively higher tongue blade position than the oral vowel.
Findings also include articulatory distinctions between phonemic and phonetically nasalized vowels. Nasalized vowels pattern differently vis-à-vis oral and nasal vowels, depending on vowel category. Nasalized /a/ shows articulatory and acoustic characteristics intermediate to those of the oral and nasal vowels. Nasalized /i/ is more similar to oral /i/ than nasal /ı̃/ in its articulatory and acoustic characteristics. Nasalized /u/ is more similar to nasal /ũ/ than oral /u/ in its articulatory and acoustic characteristics. Nasalized vowels crucially do not manifest nasal coda emergence. This indicates that multiple articulatory strategies are important in maintaining the nasal/nasalized distinction. The phonetic variability between nasalized and oral vowels is argued to occur because of underspecification of the nasalized vowels.
Based on the articulatory and acoustic characteristics of these vowels, I argue that nasal vowels in BP are underlyingly /Ṽ/, rather than /VN/ sequences, the latter distinction being the underlying form of nasalized vowels. This dissertation lends further evidence to the importance of oropharyngeal shape in the production of nasal vowels, and more generally, to a holistic study of the vocal tract when analyzing speech sounds. The articulatory divergence of [+nasal] and [–nasal] vowels has implications in perception, sound change, and the phonetic implementation of nasality. These findings lend further evidence to the claim that the goal of speech communication is acoustic, rather than articulatory, in nature.
The findings presented in this dissertation were possible because of recently-developed techniques in rt-MRI data acquisition that allow researchers to see and resolve articulatory configurations with the necessary spatiotemporal resolution. The results of this study further exemplify the complexities of the many-to-one problem of phonetics, and outline a way to deal with it using high-dimensional imaging data. I present findings that discuss potential solutions to the drawbacks of using rt-MRI to conduct speech research. Specifically, I show minimal positional effects on acoustic output between supine and upright positions. I also show promising results of acoustic denoising algorithms, which can be used to make direct comparisons between the articulatory data and simultaneously-acquired acoustic recordings.