This dissertation focuses on vowel harmony in Maasai, an Eastern Nilotic language spoken in Kenya and Tanzania. The main goal of this dissertation is to determine whether an adequate account of the Maasai pattern of Advanced Tongue Root (ATR) harmony can be formulated in Optimality Theory. Ultimately it is seen that it can, relying on directional Maximal Licensing constraints Walker (2011). Maasai is a language with dominant-recessive harmony. There are two sets of vowels--ATR and non-ATR. A word can only include members of one vowel set; if there is an AT R vowel anywhere in a word, all vowels will be ATR in the output. The only exception to this is the non-ATR low vowel, which lacks an ATR counterpart. It is opaque--it does not harmonize and it blocks the spread of harmony if it is followed by an ATR vowel, but it harmonizes to [o] when preceded by an ATR vowel. All earlier analyses have been based on mainly one source, Tucker and Mpaayei (1955). To avoid using inaccurate or inaccurate data, the data analyzed in this thesis were collected from native speakers in Arusha, Tanzania. Earlier accounts have been based on impressionistic transcriptions. Acoustic analysis of the data were performed to explore the properties of the vowels. The height of the first formant was found to be the most robust acoustical cue to differentiate ATR and non-ATR vowels, though the height of the second formant has some use as a secondary cue. Like many previous studies of languages with an ATR contrast, in this study, it was found that the ATR vowels in Maasai have lower F1s than their non-ATR counterparts (Ladefoged 1964, Lindau et al. 1972, Lindau 1976, Jacobson 1980, Hess 1992, Maddieson and Gordon 1996, Fulop et al. 1998, Anderson 1999, 2007, Przezdziecki 2005, Gick et al. 2006, Starwalk 2008, Kang and Ko 2012). Guion (2004)'s acoustic analysis of Maasai, which showed that ATR and non-ATR vowels in minimal pairs or near minimal pairs differ in F1 was confirmed. Unlike previous research, vowels that have undergone harmony were also investigated. It was observed that not only does Maasai show an ATR/ non-ATR distinction, but that the harmony process is neutralizing. An ATR suffix will force a non-ATR root to harmonize, and an ATR root will force a non-ATR prefix to harmonize. The vowel that has undergone harmony to become ATR is not distinguishable from one is always ATR. It was also found that distance from the trigger (the ATR vowel that causes harmony) does not affect the harmony process. Maasai has been described as having one lexically ATR prefix which causes only less peripheral prefixes to harmonize (Tucker and Mpaayei 1955, Mol 1995, 1996). This claim was investigated, but no acoustic evidence was found to support the claim that there is an ATR prefix. Instead, it is suggested that the perception of the prefix as ATR arises from coarticulatory effects that are the result of the unique environment of the prefix. Acoustic analysis of prefixes preceding the putative ATR prefix were found to be non-ATR. Although previous OT analyses of Maasai have been either unduly complex, incapable of accounting for all the data or have dismissed elements of the harmony as morphological, the harmony system can be accounted for rather simply with two directional harmony constraints. Walker (2011) suggests that languages which appear to demonstrate one bidirectional harmony process might actually be the result of two unidirectional harmony processes. The analysis of Maasai presented her supports this suggestion. There are two directional Maximal Licensing constraints which are high-ranked there is another constraint that must be ranked between them to account for the asymmetric behavior of the low vowel.

The study involves tests on the perception by speakers of Finnish of vowel sounds in natural speech in four important varieties of standard Swedish and as produced by an OVE III synthesizer.The contrastive analysis takes the form of experiments to reveal the divisions of the vowel space in Swedish and Finnish. Perceptually optimal areas and the diffuse zones in the vowel space are contrasted. On the basis of these tests predictions are made concerning second language perceptual problems.The contrastive relations of the vowel sounds in Swedish and Finnish are further illuminated by perceptual tests to study the inherent tendency of speakers of Finnish to confuse certain Swedish vowels. The distribution of the phonemic reactions of speakers of Finnish not knowing Swedish are taken as material for an error analysis relevant to the time of starting to learn Swedish. A number of phonemically long vowel sounds were discovered to cause problems of identification for Finnish listeners. These are / y:/, /in:/, /e:/, / e :/, /#:/, and /o:/. A comparison of the predictions and actual confusions shows a high degree of correlation.Finnish vowel harmony was found to be a factor influencing the responses of Finns to Swedish vowels. The responses to combinations of vowels not complying with the vowel harmony rules contained an increased proportion of confusions leading to harmony-compatible or neutral vowels. Also the vowel qualities which were in an intermediate position in the vowel space in relation to the Finnish vowels tended to be influenced by assimilation or sequential contrast.The perceptual interference properties of the mother tongue in the perception of the vowel qualities as illuminated by the results of the perceptual test are discussed.Certain didactic conclusions are drawn concerning the problems experienced by speakers of Finnish in the identification and also the production of the Swedish vowels. Gestures of lip position and its contrastive perceptual effects are noted to be of especial importance in the learning of the vowel qualities of Swedish by Finns. / digitalisering@umu

This study hypothesizes that the Epenthetic Vowel (EV) that occurs in Spanish consonant clusters, although produced unconsciously, is part of the articulatory plan of the speaker. As part of the plan, the epenthetic vowel occurs more often in the least perceptually recoverable contexts in order to enhance them. To achieve a better understanding of the role of the epenthetic vowel, this study shows that the linguistic and phonotactic contexts condition the occurrence of these vowels. Specifically, it argues that linguistic and phonotactic contexts that are perceptually weak compel a significantly higher occurrence of EVs. The EV was analyzed from both production and perceptual standpoints. The results show that from the production standpoint, the occurrence of the EV is affected by the type of liquid that forms the clusters: in clusters with /r/ the variables that made a statistical contribution were post-tonic position (odds ratio, 4.46), and voiceless consonants (odds ratio, 1.42). In the case of clusters with /l/ an EV has a higher probability of occurring in the context of bilabial consonants (odds ratio, 4.19), and voiceless consonants (odds ratio, 1.3). As for the effects of speech rate on the duration of EVs, the results show that speech rate accounts for 14% of the variation in an EV’s length. From the standpoint of perception, listening was divided into the tasks of perceptual identification and perceptual discrimination. The results show that the strongest predictor is the interaction voiceless x post-tonic position (odds ratio, 4.8). For the identification of the Cr clusters, the strongest predictor is the context of voiceless consonants (odds ratio, 4.42). Regarding identification of the Cl clusters, the strongest predictors are the tonic position (odds ratio, 1.54) and the labial place of articulation (odds ratio, 1.39). With regard to the discrimination of the Cr clusters, the strongest predictors for perceptual recoverability are the interaction voiceless x post-tonic position (odds ratio, 2.22), and the labial place of articulation (odds ratio, 1.37), while for the Cl cluster, the strongest predictors are the tonic position (odds ratio, 5.83) and voiceless consonants (odds ratio, 3).

This study hypothesizes that the Epenthetic Vowel (EV) that occurs in Spanish consonant clusters, although produced unconsciously, is part of the articulatory plan of the speaker. As part of the plan, the epenthetic vowel occurs more often in the least perceptually recoverable contexts in order to enhance them. To achieve a better understanding of the role of the epenthetic vowel, this study shows that the linguistic and phonotactic contexts condition the occurrence of these vowels. Specifically, it argues that linguistic and phonotactic contexts that are perceptually weak compel a significantly higher occurrence of EVs. The EV was analyzed from both production and perceptual standpoints. The results show that from the production standpoint, the occurrence of the EV is affected by the type of liquid that forms the clusters: in clusters with /r/ the variables that made a statistical contribution were post-tonic position (odds ratio, 4.46), and voiceless consonants (odds ratio, 1.42). In the case of clusters with /l/ an EV has a higher probability of occurring in the context of bilabial consonants (odds ratio, 4.19), and voiceless consonants (odds ratio, 1.3). As for the effects of speech rate on the duration of EVs, the results show that speech rate accounts for 14% of the variation in an EV’s length. From the standpoint of perception, listening was divided into the tasks of perceptual identification and perceptual discrimination. The results show that the strongest predictor is the interaction voiceless x post-tonic position (odds ratio, 4.8). For the identification of the Cr clusters, the strongest predictor is the context of voiceless consonants (odds ratio, 4.42). Regarding identification of the Cl clusters, the strongest predictors are the tonic position (odds ratio, 1.54) and the labial place of articulation (odds ratio, 1.39). With regard to the discrimination of the Cr clusters, the strongest predictors for perceptual recoverability are the interaction voiceless x post-tonic position (odds ratio, 2.22), and the labial place of articulation (odds ratio, 1.37), while for the Cl cluster, the strongest predictors are the tonic position (odds ratio, 5.83) and voiceless consonants (odds ratio, 3).

This dissertation is a proposal for foot structure in Québécois that uniformly accounts for high vowel distribution with respect to tenseness, devoicing and deletion within a single prosodic framework. The complementary distribution of tenseness in the final syllable and the variable realizations in the non-final syllable are shown to be regulated by the proposed foot structure. A grammatical, sonority-based surface weight distinction is assumed for vowels: tense high vowels are associated to a full mora µ, along with non-high vowels, while lax high vowels are associated to a hypomora λ, a weight value less than µ. This grammatical weight is regulated at the level of the minimally monomoraic foot. The final, Head Foot is necessarily monosyllabic. Thus, a final hypomoraic rime is quantitatively insufficient to host a foot projection, resulting in a monomoraic, tense vowel in an open syllable. The foot expands to include an adjacent syllable in words consisting of more than two syllables, following the Trochaic Markedness Hierarchy, based on the following three principles, in the order of priority: 1) quantitative minimum: light and heavy rimes are preferred to superlight (λ) rimes, 2) quantitative evenness: even trochees are preferred to uneven trochees, and 3) quantitative dominance: the left branch that is heavier than the right branch is preferred to the left branch that is lighter. A form like /kamizᴐl/ surfaces with a monomoraic, tense vowel in the left branch of the trochee, (ska. wmi)(szᴐl), given that an even foot (L L) is preferred to an uneven foot with a hypomoraic branch, (L SL). The trochaic instantiation (H) is also better-formed than (L SL), preferring deletion to a hypomoraic rime: (kam)(zᴐl). In the Optimality-theoretic analysis, variation is modeled via the mechanism of a Floating Constraint (Reynolds 1994): a constraint whose ranking status can be varied with respect to a set range of a fixed ranking of constraints, within a single grammar. The variation in question is shown to be largely a function of the floating status of the constraint regulating the grammatical weight association of vowels, (Son-Weight), and its relative ranking with respect to the Trochaic Markedness constraints. / text

This dissertation presents the results of a study conducted on the speech of male and female students training to become Drill Instructors (hereafter DIs) in the United States Marine Corps. Both high amplitude and low pitch are reported to be important characteristics of the DI Command Voice; these characteristics are also strongly associated with masculinity (Hicks 1997). However, previous research argues that female DIs do not view these qualities as the most important characteristics of the Command Voice. They focus instead on being "direct" (Hicks 1997, Hicks Kennard 1999). The question I address is whether or not the Command Voice taught in DI School is used differently by female and male DI students in authoritative speech.The data were recorded at the DI School in Parris Island, South Carolina. Six subjects participated in the study. The two factors considered were sex and speech style, which included: 1) teachback: high-amplitude recitations of training procedures, 2) locker-box discussion: a speech style used in academic settings, 3) interviews, and 4) a reading sample. Both vowel duration and peak pitch measurements were done in each speech style; measurements on larger thirty-second "chunks" of discourse were taken for mean pitch, standard deviation of pitch, range of pitch, and speaking rate for each speaker in each speech style. There was a significant main effect on vowel duration for all subjects except for one male and a significant main effect on peak pitch for all subjects. For discourse measurements of pitch, there was a significant main effect for mean pitch, standard deviation of pitch, and pitch range.Pair-wise comparisons resulted in significant differences in peak pitch for all subjects in all speech styles. Both females and males exhibited the same pattern for both vowel and discourse pitch measurements, from highest to lowest pitch: teachbacks, lockerbox discussion, reading sample, and interview. However, females' vowels were significantly longer in the most authoritative speech style--precisely where males had their shortest vowel duration. This difference suggests that in authoritative speech, females use vowel duration as part of the "directness" in authority, where males do not.

Title:EFFECTS OF THE PROMINENCE OF FIRST HARMONIC ON THE PERCEPTION OF BREATHINESS AND VOWEL IDENTITY Authors: Emily Lin, Samuel Sloane,and Donal Sinex Background: Human communication relies on adequate speech intelligibility to enable the comprehension of verbal messages. Dysphonia (i.e., aberrant voice) may not only result in distraction during communication but also interfere with speech intelligibility leading to a communication barrier. One voice quality commonly found in dysphonia is breathiness, which is related to the presence of excessive airflow during phonation due to incomplete glottal closure. Breathiness has been associated with the prominence of the first harmonic (H1) in the acoustic analysis of voice. Objectives: This study aimed to determine whether excessiveness in the first harmonic (H1) dominance, which has been associated with breathy voice, may result in the perception of breathiness and compromise vowel intelligibility. Methods: Participants included 10 female and 10 male normal-hearing adults, aged between 19 to 40 years. Participant’s tasks included a “breathiness rating” and a “vowel identification” task. For the “breathiness rating” task, a direct magnitude method was employed for the participant to rate a 500-ms long vowel (/i/ and /a/) segmented from sustained vowel phonation. For the “vowel identification” task, the vowel stimuli were segmented out from running speech (“Rainbow passage”) and the participants were asked to listen to one vowel stimulus (/i/, /a/, or /o/; duration: 60 ms) at a time and indicate which vowel (i.e., /i/, /e/, /a/, /o/, or /u/) they perceived the stimulus to be. The vowel stimuli included processed and unprocessed voice recordings of individuals with and without voice disorders. Voices showing the lowest, median, and highest amplitude differences between the first two harmonics (H1-H2) were chosen from a voice database for female and male voices respectively. The 18 selected vowel signals (3 vowels X 3 H1-H2 levels X 2 speaker genders) were processed through 12 signal manipulation conditions. The 12 signal conditions involved increasing or decreasing the H1 amplitude of the original signals in six 2-dB interval steps in both directions. Results: For the “breathiness rating” task, the five-way (3 vowels X 2 speaker genders X 3 H1-H2 levels X 13 signal conditions X 2 listener genders) Mixed Model Analysis of variance (ANOVA) conducted on the breathiness scores for normal speakers and voice patients separately showed significant findings for various main and interaction effects, such as a significant speaker gender by signal condition by vowel interaction effect on the perception of breathiness [F(12, 96) = 1.95, p = 0.038] for normal voice. An increase of H1-H2 through signal manipulation led to an increase of perceived breathiness only when performed on the vowel /i/ produced by female normal speakers. As for the “vowel identification” task, a relationship between H1-H2 increment and vowel intelligibility was found but the relationship was affected by vowel type, speaker gender, and H1-H2 level. With all vowel types, speaker genders, and H1-H2 levels combined, a significant signal condition effect on the number of incorrect vowel identification was found (2 = 188.585, df = 10, p < 0.001). Generally, it appeared that an increase of H1-H2 would worsen the identification of /i/ but enhance that of /o/. Conclusion: The relationship between H1 dominance and perceived breathiness was non-linear. Factors found to disrupt the linear relationship included speaker gender, vowel type, and the extent of H1 dominance. In addition, there was evidence that acoustic manipulation of the H1 amplitude would affect vowel intelligibility and the relationship between vowel intelligibility and H1-H2 values also vary by speaker genders and vowel types.

Cette thèse se propose d'étudier deux processus phonologiques bien distincts qui sont la réduction des voyelles atones à schwa que l'on trouve dans certains dialectes italiens et le redoublement phono-syntaxique induit par l'accent (Raddoppiamento Fonosintattico), désormais RF. Il semblerait que l'existence de ces deux phénomènes (le RF induit par l'accent et la réduction vocalique) soit en distribution complémentaire puisque l'analyse du processus de (non-)réduction vocalique en coratin permettrait de mettre en lumière la motivation structurelle de cette incompatibilité. Cette thèse va mettre à jour et vérifier cette prédiction avec des données diatopiques. En effet, le coratin dispose d'une contrainte pour que les voyelles soient réalisées non réduites rendant le RF déclenché par l'accent incompatible avec les dialectes à réduction localisés situé dans la partie centre-méridionale de la péninsule. / This thesis proposes to study two distinct phonological processes that are reducing unstressed vowels to schwa found in some Italian dialects and Raddoppiamento Fonosintattico induced stress RF now. It seems that the existence of these two phenomena (RF-induced by stress and vowel reduction) is in complementary distribution since the analysis process (no-) vowel reduction coratin would highlight the structural motivation of the inconsistency. This thesis will update and verify the prediction with diatopic data. Indeed, coratin has a constraint that the vowels are carried unreduced making RF triggered by stress incompatible with reducing localized dialects located in the central-southern part of the peninsula.

abstract: Distorted vowel production is a hallmark characteristic of dysarthric speech, irrespective of the underlying neurological condition or dysarthria diagnosis. A variety of acoustic metrics have been used to study the nature of vowel production deficits in dysarthria; however, not all demonstrate sensitivity to the exhibited deficits. Less attention has been paid to quantifying the vowel production deficits associated with the specific dysarthrias. Attempts to characterize the relationship between naturally degraded vowel production in dysarthria with overall intelligibility have met with mixed results, leading some to question the nature of this relationship. It has been suggested that aberrant vowel acoustics may be an index of overall severity of the impairment and not an "integral component" of the intelligibility deficit. A limitation of previous work detailing perceptual consequences of disordered vowel acoustics is that overall intelligibility, not vowel identification accuracy, has been the perceptual measure of interest. A series of three experiments were conducted to address the problems outlined herein. The goals of the first experiment were to identify subsets of vowel metrics that reliably distinguish speakers with dysarthria from non-disordered speakers and differentiate the dysarthria subtypes. Vowel metrics that capture vowel centralization and reduced spectral distinctiveness among vowels differentiated dysarthric from non-disordered speakers. Vowel metrics generally failed to differentiate speakers according to their dysarthria diagnosis. The second and third experiments were conducted to evaluate the relationship between degraded vowel acoustics and the resulting percept. In the second experiment, correlation and regression analyses revealed vowel metrics that capture vowel centralization and distinctiveness and movement of the second formant frequency were most predictive of vowel identification accuracy and overall intelligibility. The third experiment was conducted to evaluate the extent to which the nature of the acoustic degradation predicts the resulting percept. Results suggest distinctive vowel tokens are better identified and, likewise, better-identified tokens are more distinctive. Further, an above-chance level agreement between nature of vowel misclassification and misidentification errors was demonstrated for all vowels, suggesting degraded vowel acoustics are not merely an index of severity in dysarthria, but rather are an integral component of the resultant intelligibility disorder. / Dissertation/Thesis / Ph.D. Speech and Hearing Science 2012

O fenômeno de sândi vocálico externo apresenta três realizações possíveis: a elisão, a degeminação e a ditongação. Este trabalho tem como proposta analisar o fenômeno de ditongação como processo de resolução de hiato em fronteira de palavras (camisa usada ~ cami[zaw]sada), baseado na amostra da cidade de Lages, em Santa Catarina, incluída no banco de dados do projeto VARSUL (Variação linguística na região sul do Brasil). O embasamento teórico se apoia em Bisol (1996, 2002 e 2005) para a descrição do processo; em Labov (2008) para fundamentos de teoria da variação; além de teorias fonológicas, como Fonologia Prosódica e Fonologia Lexical. Os objetivos específicos são os que seguem: a) identificar, em nossa amostra, fatores linguísticos que possam favorecer ou bloquear a aplicação da ditongação como processo de sândi externo; b) a partir dos resultados obtidos, corroborar ou não resultados de pesquisas já realizadas a respeito deste processo; c) ampliar a compreensão do fenômeno e oferecer subsídios para uma descrição geral do processo do sândi externo e do português falado no sul do Brasil. As hipóteses que procuramos confirmar são as seguintes: a) quanto à tonicidade, o contexto ideal para a aplicação da ditongação é o de atonicidade máxima (casa escura ~ ca[zaj]scura) (conforme Bisol, 1996; Bisol, 2002); b) o contexto interno à frase fonológica é mais favorecedor na aplicação do fenômeno (velho exemplo ~ velh[we]zemplo) (conforme Bisol, 1996). A análise estatística dos dados foi realizada pelo pacote de programas VARBRUL/GoldvarbX. A amostra, constituída de 16 informantes, mostrou, dentre outros fatores, que a ditongação crescente e a ditongação decrescente têm diferentes contextos favorecedores em relação ao acento e à categoria das vogais e têm contextos favorecedores semelhantes em outras variáveis em comum. Para a ditongação, confirmamos nossas hipóteses: a atonicidade máxima e o contexto interior à frase fonológica se mostraram favorecedores. A seleção das variáveis relevantes não foi idêntica para as duas realizações. / The external vowel sandhi phenomenon presents three possible ways of realization: the elision, the degemination and the diphthongization. This work analyzes the diphthongization phenomenon as a gap resolution process in words boundary (camisa usada ~ cami[zaw]sada) based on the sample of the city of Lages, Santa Catarina, included in the project database VARSUL (linguistic variation in southern Brazil). The theoretical basis is Bisol (1996, 2002 and 2005) for the description of the process; Labov (2008) for the variation theory fundamentals; and Brescancini (2005) for variational research methodology. The specific objectives are the following: a) to identify, in our sample, linguistic factors that may favor or block the application of external vowel sandhi processes; b) from the results obtained, to prove or disprove results of previous studies regarding this process; c) to increase the understanding of the phenomenon and provide support for a general description of the process of external sandhi and Portuguese spoken in southern Brazil. The hypotheses we intend to confirm are the following: a) the ideal context for the three external sandhi processes is an unstressed vowel + an unstressed vowel (casa escura ~ ca[zaj]scura) (as Bisol, 1996; Bisol, 2002; Ludwig-Gayer, 2008; Vianna, 2009); b) the internal context to phonological phrase is more favorable to the application of the phenomenon (velho exemplo ~ velh[we]zemplo) (as Bisol, 1996). Statistical analysis of data was performed by VARBRUL / GoldvarbX software. The sample consisted of 16 informants showed that rising diphthongization and decreased diphthongization have different favorable contexts related to stress and vowel category, and they have similar favorable contexts in same variables. Related to diphthongization, we confirmed our hypotheses: the sequence of unstressed vowels and the phonological phrase are favorable to the process. The selection of the variable was not exactly the same for both processes.