Remembering the sound of a new word when it is first encountered is an important skill which plays a critical role in the development of vocabulary (Gathercole & Baddeley, 1989), yet the mechanisms underlying this form of verbal short-term memory are not well understood. Errors in the repetition and serial recall of nonwords indicate that structural properties of the syllable are represented in short-term memory, but existing accounts of serial learning and recall do not incorporate any representation of linguistic structure. Models of speech production implicate syllable structure in the representation of phonological form, but do not explain how such representations are acquired. This thesis draws together theories of speech production and serial memory to develop a computational model of nonword repetition based on the novel idea that short-term memory for the serial order of a sequence of speech sounds is constrained by a syllabic template. The results of simulations using the model are presented and compared with experimental findings concerning short-term memory for nonwords. The interaction of short- and long-term phonological memory systems and the aquisition of vocabulary are discussed in terms of the model. The model is evaluated in comparison with other contemporary theories.

The ability for English Language Learners (ELL) to communicate in a native-like manner can be vital when seeking to improve job opportunities and social interactions. Native Japanese ELL speakers commonly struggle with identifying and producing the English consonants /r/ and /l/ as separate phonemes, even for experienced Japanese ELLs. Traditional means of second language (L2) learning has primarily used auditory feedback. Electropalatography (EPG) is a visual biofeedback system designed to visualize lingua-palatal contact in real time. This study explores the impact of using EPG technology in L2 Speech Learning with Japanese ELLs. Research has shown that EPG technology is an effective method of treating speech sound disorders in children, however there is sparse research available concerning the impact of EPG use in L2 learning for ELLs. This study analyzed four native Japanese speakers' ability to produce the /r/-/l/ contrast before and after L2 learning treatment utilizing EPG technology. Target stimuli containing the /r/ and /l/ phonemes were produced by each participant in three different tasks: nonsense words, words, and spontaneous speech. The /r/ and /l/ phonemes were produced in word-initial and word-final position. The EPG technology and computer software were used to record lingua-palatal contact patterns in order to collect data on each participants' productions of the /r/-/l/ phonemic contrast. In general, all four participants demonstrated increased center of gravity (COG) difference of /r/ and /l/ in 50%-100% of testing conditions from baseline to post treatment and in 33%-100% of testing conditions from post treatment to follow-up. In combination with future research, results from this study will help deepen the knowledge of L2 learning in ELLs specifically related to EPG technology use in treatment.

Cette étude acoustico-perceptive concerne les limites de la perception et de la production des voyelles orales du français [i, e, ɛ, a, u, o, ɔ, y, ø, œ], en isolation et en contextes consonantiques divers, chez dix tchécophones, futures enseignantes de Français Langue Étrangère (FLE). Les résultats montrent que (1) La maîtrise phonétique des voyelles dépend de leurs graphies et de l’entourage consonantique. (2) Les voyelles fermées [i, y, u] et le [a] sont globalement maîtrisées avec authenticité. (3) Les capacités de perception des contrastes entre les voyelles moyennes e/ɛ, ø/œ et o/ɔ ainsi que leur production sont limitées. Ces résultats ne sont que partiellement en accord avec les prédictions établies à partir du Speech Learning Model (SLM) de Flege (1995), basé sur la notion de similarité phonétique qui existe entre la langue maternelle (LM) et la langue étrangère (LE). / This acoustic-perceptual study concerns the limits of perception and production of French oral vowels [i, e, ɛ, a, u, o, ɔ, y, ø, œ], in isolation and in different consonantal contexts, in ten pre-service Czech teachers of French as a Foreign Language (FFL). The results show that (1) Phonetic proficiency in vowels depends on their spellings and consonantal context. (2) Vowels [i, y, u] and [a] are generally mastered with authenticity. (3) The ability to hear contrasts between the vowels e/ɛ, ø/œ and o/ɔ and pronounce them is limited. These results are only partially consistent with the predictions established in the Speech Learning Model (SLM) by Flege (1995), based on the notion of phonetic similarity between the mother tongue (MT) and the foreign language (FL).

Cette étude acoustico-perceptive concerne les limites de la perception et de la production des voyelles orales du français [i, e, ɛ, a, u, o, ɔ, y, ø, œ], en isolation et en contextes consonantiques divers, chez dix tchécophones, futures enseignantes de Français Langue Étrangère (FLE). Les résultats montrent que (1) La maîtrise phonétique des voyelles dépend de leurs graphies et de l’entourage consonantique. (2) Les voyelles fermées [i, y, u] et le [a] sont globalement maîtrisées avec authenticité. (3) Les capacités de perception des contrastes entre les voyelles moyennes e/ɛ, ø/œ et o/ɔ ainsi que leur production sont limitées. Ces résultats ne sont que partiellement en accord avec les prédictions établies à partir du Speech Learning Model (SLM) de Flege (1995), basé sur la notion de similarité phonétique qui existe entre la langue maternelle (LM) et la langue étrangère (LE). / This acoustic-perceptual study concerns the limits of perception and production of French oral vowels [i, e, ɛ, a, u, o, ɔ, y, ø, œ], in isolation and in different consonantal contexts, in ten pre-service Czech teachers of French as a Foreign Language (FFL). The results show that (1) Phonetic proficiency in vowels depends on their spellings and consonantal context. (2) Vowels [i, y, u] and [a] are generally mastered with authenticity. (3) The ability to hear contrasts between the vowels e/ɛ, ø/œ and o/ɔ and pronounce them is limited. These results are only partially consistent with the predictions established in the Speech Learning Model (SLM) by Flege (1995), based on the notion of phonetic similarity between the mother tongue (MT) and the foreign language (FL).

This dissertation explores the potential differences in the acquisition of the sound system of a second language (L2) versus a third language (L3), building on recent research on adult L2 speech learning and testing the hypothesis that new category acquisition is available across the lifespan. On the one hand, recent influential theories of L2 speech learning predict that new sound categories will be difficult to acquire due to complex interactions among the phonetic categories residing in one same perceptual space. On the other, there exists the common assumption that the more sound categories one’s native language contains, the less difficult it will be to acquire new ones in a native-like fashion. My work provides additional evidence to this discussion, while providing speech data from the following five different speaker groups: (1) native speakers of American English; (2) native speakers of Mexican Spanish; (3) native speakers of Brazilian Portuguese; (4) native speakers of English learning Portuguese as their L2, and (5) native speakers of English learning Portuguese as their L3, who speak Spanish as their L2.The dissertation consists of three content chapters. The first of the content chapters describes the vowel systems of Mexican Spanish, American English and Brazilian Portuguese as informed by production data of the three native speaker groups above (1, 2 and 3). The second and third chapters describe and contrast the vowel system(s) of the two learner groups (4 and 5) as informed by their production and perception of Portuguese vowels respectively. The results from the analysis of the data from the various production and perception experiments performed as part of this dissertation provide evidence for (a) phonetic category assimilation and dissimilation processes in post-L1 speech learning and phonetic category interactions in general; (b) the relevance of quality and quantity of input in language learning; (c) the discussion on the relation between perception and production in post-L1 speech learning; and (d) the need for an extension of current models of L2 speech learning and cross-linguistic speech perception in order for these to address post-L2 speech learning processes, among other topics.

Unsupervised learning of speech is concerned with automatically finding patterns such as words or speech sounds, without supervision in the form of orthographical transcriptions or a priori knowledge of the language. However, a fundamental problem is that unsupervised speech learning methods tend to discover highly speaker-specific and context-dependent representations of speech. We propose a method for improving the quality of posteriorgrams generated from an unsupervised model through partitioning of the latent classes discovered by the model. We do this by training a sparse siamese model to find a linear transformation of input posteriorgrams, extracted from the unsupervised model, to lower-dimensional posteriorgrams. The siamese model makes use of same-category and different-category speech fragment pairs obtained through unsupervised term discovery. After training, the model is converted into an exact partitioning of the posteriorgrams. We evaluate the model on the minimal-pair ABX task in the context of the Zero Resource Speech Challenge. We are able to demonstrate that our method significantly reduces the dimensionality of standard Gaussian mixture model posteriorgrams, while also making them more speaker invariant. This suggests that the model may be viable as a general post-processing step to improve probabilistic acoustic features obtained by unsupervised learning. / Obevakad inlärning av tal innebär att automatiskt hitta mönster i tal, t ex ord eller talljud, utan bevakning i form av ortografiska transkriptioner eller tidigare kunskap om språket. Ett grundläggande problem är dock att obevakad talinlärning tenderar att hitta väldigt talar- och kontextspecifika representationer av tal. Vi föreslår en metod för att förbättra kvaliteten av posteriorgram genererade med en obevakad modell, genom att partitionera de latenta klasserna funna av modellen. Vi gör detta genom att träna en gles siamesisk modell för att hitta en linjär transformering av de givna posteriorgrammen, extraherade från den obevakade modellen, till lågdimensionella posteriorgram. Den siamesiska modellen använder sig av talfragmentpar funna med obevakad ordupptäckning, där varje par består av fragment som antingen tillhör samma eller olika klasser. Den färdigtränade modellen görs sedan om till en exakt partitionering av posteriorgrammen. Vi följer Zero Resource Speech Challenge, och evaluerar modellen med hjälp av minimala ordpar-ABX-uppgiften. Vi demonstrerar att vår metod avsevärt minskar posteriorgrammens dimensionalitet, samtidigt som posteriorgrammen blir mer talarinvarianta. Detta antyder att modellen kan vara användbar som ett generellt extra steg för att förbättra probabilistiska akustiska särdrag från obevakade modeller.

abstract: Much evidence has shown that first language (L1) plays an important role in the formation of L2 phonological system during second language (L2) learning process. This combines with the fact that different L1s have distinct phonological patterns to indicate the diverse L2 speech learning outcomes for speakers from different L1 backgrounds. This dissertation hypothesizes that phonological distances between accented speech and speakers' L1 speech are also correlated with perceived accentedness, and the correlations are negative for some phonological properties. Moreover, contrastive phonological distinctions between L1s and L2 will manifest themselves in the accented speech produced by speaker from these L1s. To test the hypotheses, this study comes up with a computational model to analyze the accented speech properties in both segmental (short-term speech measurements on short-segment or phoneme level) and suprasegmental (long-term speech measurements on word, long-segment, or sentence level) feature space. The benefit of using a computational model is that it enables quantitative analysis of L1's effect on accent in terms of different phonological properties. The core parts of this computational model are feature extraction schemes to extract pronunciation and prosody representation of accented speech based on existing techniques in speech processing field. Correlation analysis on both segmental and suprasegmental feature space is conducted to look into the relationship between acoustic measurements related to L1s and perceived accentedness across several L1s. Multiple regression analysis is employed to investigate how the L1's effect impacts the perception of foreign accent, and how accented speech produced by speakers from different L1s behaves distinctly on segmental and suprasegmental feature spaces. Results unveil the potential application of the methodology in this study to provide quantitative analysis of accented speech, and extend current studies in L2 speech learning theory to large scale. Practically, this study further shows that the computational model proposed in this study can benefit automatic accentedness evaluation system by adding features related to speakers' L1s. / Dissertation/Thesis / Doctoral Dissertation Speech and Hearing Science 2018

The inability to differentiate the English vowels /o/ and / O/ has become a stigmatized marker of a lower prestige and widespread dialect of Sri Lankan English. This lower prestige (LP) dialect is often referred to with the derogative phrase “Not pot English”. This study aims to investigate the production of the vowel contrast by native Sinhala speakers of English. To this end, speech samples of three adult learners were analyzed. The findings of the study are discussed according to hypotheses of the Speech Learning Model, which suggests that the existent L1 specific phonetic categories hinder the formation of new L2 sound categories. Here, sounds that are similar, but not identical to L1 sounds are considered to be the most difficult to acquire. Also, the percentage of L1 use and the age of second language acquisition seem to have influenced the production of the vowels. Finally, in order to address this pronunciation issue, an instructional framework to teach pronunciation is proposed.

This study examines the influence of formal L2 English instruction (FEI) onpronunciation accuracy of similar and new vowels in the framework of Flege’s SpeechLearning Model (Flege, 1995a, 1999, 2002), focusing on selected segmentalpronunciation features produced by native German speakers. A spectral vowelproduction experiment was carried out with 24 German students who attend sixth, ninthand twelfth grade of a south German secondary school by the time this study wasconducted. Spectral characteristics of the similar vowel /ɛ/ and the new vowel /æ/,produced by each subject of the three groups (G6, G9, and G12), were compared tovowel data of native English speakers. In order to test the influence of formalinstruction on selected phonetic segments, which are known to cause difficulties fornative German learners of English, three native English listeners judged on a three pointscale the intelligibility of the segments produced. The results of the vowel experimentshowed significant influence of FEI on pronunciation accuracy of some characteristicsof the similar vowel /ɛ/ but not of the new vowel /æ/. The listener rating experimentshowed that FEI had on average no influence on pronunciation accuracy of the three L2English learner groups.

This study examines the influence of formal L2 English instruction (FEI) on pronunciation accuracy of similar and new vowels in the framework of Flege’s Speech Learning Model (Flege, 1995a, 1999, 2002), focusing on selected segmental pronunciation features produced by native German speakers. A spectral vowel production experiment was carried out with 24 German students who attend sixth, ninth and twelfth grade of a south German secondary school by the time this study was conducted. Spectral characteristics of the similar vowel /ɛ/ and the new vowel /æ/,produced by each subject of the three groups (G6, G9, and G12), were compared to vowel data of native English speakers. In order to test the influence of formal instruction on selected phonetic segments, which are known to cause difficulties for native German learners of English, three native English listeners judged on a three pointscale the intelligibility of the segments produced. The results of the vowel experiment showed significant influence of FEI on pronunciation accuracy of some characteristics of the similar vowel /ɛ/ but not of the new vowel /æ/. The listener rating experiment showed that FEI had on average no influence on pronunciation accuracy of the three L2 English learner groups.