Winifred Strange - US grants

The overall objective of this research is the specification of the dynamic (time-varying) acoustic information that is used by listeners to identify vowels in their normal context, that is, when they are coarticulated with consonants in continuous speech. The research speaks to a very broad questions of continuing importance in speech and hearing research: What acoustic information in the speech signal is used by listeners to identify the phonetic message?" Traditional theories and many current models of vowel perception emphasize the role of static acoustic "targets" as the major source of information about vowel identity. However, much current research has shown that listeners rely critically on dynamic information that is available in the changing acoustic pattern of syllables (format trajectories) when they try to identify coarticulated vowels. Four types of studies will be performed: 1) Perceptual studies with electronically-modified natural speech materials will be conducted to investigate the perception of vowels in different phonetic contexts and at different speech rates. 2) Alternative auditory-perceptual representations of format trajectories will be studied to examine their relative efficacy in accounting for the perception of coarticulated vowels. 3) Format trajectories in synthetic consonant-vowel-consonant syllables will be manipulated to test alternative hypotheses about the perceptual relevane of spectral and temporal parameters characterizing "style of change". 4) Studies of the perception of coarticulated German vowels (by native and non-native speakers) will be conducted to assess the generality across languages of the findings about the role of dynamic acoustic information in the perception of vowels.

The overall objective of this research is the specification of the dynamic (time-varying) acoustic information that is used by listeners to identify vowels in their normal context, that is, when they are coarticulated with consonants in continuous speech. The research speaks to a very broad questions of continuing importance in speech and hearing research: What acoustic information in the speech signal is used by listeners to identify the phonetic message?" Traditional theories and many current models of vowel perception emphasize the role of static acoustic "targets" as the major source of information about vowel identity. However, much current research has shown that listeners rely critically on dynamic information that is available in the changing acoustic pattern of syllables (format trajectories) when they try to identify coarticulated vowels. Four types of studies will be performed: 1) Perceptual studies with electronically-modified natural speech materials will be conducted to investigate the perception of vowels in different phonetic contexts and at different speech rates. 2) Alternative auditory-perceptual representations of format trajectories will be studied to examine their relative efficacy in accounting for the perception of coarticulated vowels. 3) Format trajectories in synthetic consonant-vowel-consonant syllables will be manipulated to test alternative hypotheses about the perceptual relevane of spectral and temporal parameters characterizing "style of change". 4) Studies of the perception of coarticulated German vowels (by native and non-native speakers) will be conducted to assess the generality across languages of the findings about the role of dynamic acoustic information in the perception of vowels.

The long-term objectives of this basic research are to specify the distinctive acoustic patterns in speech signals that perceptually differentiate the vowels and consonants of different languages and to explore the nature of both language-universal and language-specific perceptual processes by which adult listeners recover the linguistic message from the complex acoustic signal. Three sets of experiments are proposed. A) Acoustic and perceptual studies of the role of dynamic (time-varying) acoustic information in the specification of German vowels produced in continuous speech will investigate similarities and differences in the acoustics and perception of North German (NG) and American English (AE)A vowels by native speakers. B) Studies of the perceptual similarities of native-language and non-native vowels and consonants will examine how subjects judge the goodness-of-fit of non- active phonetic segments to native categories (cross-language similarities) and how non-native listeners differ from native listeners in their perceptual similarity judgments of selected sets of vowels and consonants (within-language similarities). The vowel inventories of NG, Japanese (J), and AE will be compared. In addition, AE and J listeners' judgments of AE approximant consonants/r, 1, w, j/and AE listeners' judgment of Hindi stop consonants will be examined. C) Perceptual training studies will attempt to improve adult listeners' perception of non-active vowels and consonants. English speakers will be trained on NG vowels and Hindi stop consonants; Japanese speakers will be trained on AE vowels and approximant consonants. Results of these studies will provide a better understanding of the role that linguistic experience plays in the modification of perceptual processes. In addition, they contribute information which may be valuable in improving the teaching of foreign languages to adults and in the amelioration of perceptual problems in children and adults with speech disorders.

The long-term objectives of this basic research are to specify the distinctive acoustic patterns in speech signals that differentiate the vowels in different languages and to understand the processes by which listeners recover the linguistically-relevant message from the complex acoustic signals. The goal is to investigate the acoustics and perception of vowels as they occur in continuous speech so that results may be generalized to actual situations. A specific focus of the proposed research is to explore how vowels vary acoustically with local and global situations. (immediate phonetic and phonotactic context, speaking style, speaking rate, sentence prosody) and to investigate how these acoustic variations affect the perception of non-native vowels by second language learners. Three sets of studies are proposed: (l) cross-language comparisons of acoustic variation in vowels of American English (AE), North German, Danish, German French, and Japanese; (2) studies of cross-language perceptual similarity of vowels by AE listeners judging Danish, German and French vowels and by Japanese listeners judging AE and German vowels; and (3) perceptual training studies which investigate the effects of perception training on the production and perception of difficult non-native vowels by adult learners of a foreign language (AE speakers learning German and French, Japanese speakers learning English). These studies will contribute to our understanding of how experience with a particular language shapes inborn capabilities. Results will allow us to make better predictions about problems facing adult foreign language learners and to develop techniques for foreign language instruction and accent reduction. Results will also provide evidence relevant to current theories about native-language phonological representation. Development of better theories of speech perception and production in normal children and adults form the basis for advances in research into problems of atypically developing children with phonological disorders.

This dissertation project attempts to identify the acoustic features of Spanish-accented speech that give it the perception of being 'accented'. While considerable information is available on phonological aspects of Spanish-accented English, only a limited amount of research has been conducted on the objective acoustic features. In this study, Spanish speakers of English and native speakers of American English will be recorded, and speech samples will be compared in terms of temporal acoustic measurements, e.g. vowel duration, consonant duration, and stop voice onset time. These measurements will provide descriptive data for the timing of Spanish-accented English. The second part of the study is experimental. Having identified the temporal differences, the researchers will select those features which differ most consistently across the two groups of speakers. These features will be acoustically modified by waveform editing for the Spanish-accented speakers to bring them into accord with the measurements of the native (unaccented) speakers. The modified and unmodified speech of the accented speakers will then be subjected to listener ratings for relative accent judgments. The aim is to determine the extent to which temporal modification makes accented speech less accented and more native-like.

This study will examine how American listeners perceive differences in duration of vowels (e.g., kiro vs. kiiro) and consonants (e.g., kite vs. kitte) in Japanese words using brain and behavioral measures. Specifically, the study will examine how well American English speakers, with no previous exposure to Japanese, can distinguish differences in vowel and consonant lengthening in Japanese. American English and Japanese speakers are likely to focus on different speech sound properties to recognize Japanese words because, unlike English, Japanese relies heavily on sound duration to distinguish meaning at the word level. This study will examine the question from a novel direction by considering the implications of first-language (L1) perception as an "over-learned" process. The notion of "highly over-learned processes" interfering with second-language (L2) speech perception relates to the claim that in first language learning, infants learn to automatically focus attention on important properties of speech sounds needed for word recognition (Jusczyk, 1997). The brain electrical measure, Mismatch Negativity (MMN), will be used to compare pre-attentive, automatic discrimination of the target sounds to discrimination with focused attention in the two groups of listeners. Research indicates that MMN serves as an index of pre-attentive, automatic discrimination. Subjects whose L1 is American English may not be able to discriminate between words with short and long sounds in Japanese without careful attention. In contrast, subjects whose L1 is Japanese will discriminate between these words without attention as shown by MMN. Behavioral discrimination of the words (pressing a button to a sound change) will provide further information regarding the relationship between preattentive brain responses and behavioral discrimination, which requires attention.

The project will provide insight into how participants discriminate sounds automatically, as indexed by brain measures. In addition, information concerning the relationship between brain responses and overt behavioral responses will reveal how underlying brain processes contribute to the behavioral response and provide insight into the intricacies of the Japanese sound system. The broader impact of this research include several areas. First, the project will allow for development of evidence-based methods of foreign language education that can lead to more native-like speech perception and production. Second, it will contribute to the bilingual education field (infants through adults) and help determine critical periods in which speech perception skills in L2 should be introduced. Third, because it will improve the current approach to foreign language instruction, it will benefit society by improving the communication abilities of L2 learners.

This research investigates the difficulties that Spanish, Russian, and Japanese speaking adult learners of English have in hearing and pronouncing differences among vowel sounds. English includes 11-12 distinct vowels sounds, whereas the three languages listed above have only 5-6 vowel sounds. Thus, for instance, where Spanish, Russian and Japanese have one vowel (similar to the [a] sound as in "hot"), English has three different vowels (the sounds in "hat", "hot" and "hut") that are very similar to each other and to the single vowel sound in Spanish, Russian and Japanese. Other confusing vowels include those that distinguish word pairs such as "luck-look", "mate-met", "feet-fit" and "pit-pet". To the English learner, these word pairs may sound like the same word (i.e. homonyms). A failure to pronounce these vowel sounds correctly leads to 'foreign-accented' speech that may interfere with communication. What is less obvious is that failures to perceive the differences between vowels may interfere with English language learners' ability to understand English speakers' utterances. As adult learners of English gain experience in listening to and speaking the new language, some of these perception and production problems are ameliorated; however, even after years of immersion in an English speaking environment, some vowel sounds continue to be confused, especially in difficult listening conditions (in noisy rooms, when the language learner is otherwise occupied, or is fatigued).

This research project examines the perceptual difficulties of Spanish, Russian and Japanese adult English learners using speech materials and listening conditions that more closely simulate 'real life' listening conditions. Both behavioral measures and brain measures of perceptual processing will be obtained on groups of listeners who vary across a wide range of English language experience. From the results, the underlying processing difficulties of these English learners and the effects of non-optimal listening conditions on their perception and production performance can be determined. These findings will contribute to basic theories of speech processing and will also lay an empirical foundation for improving English-language testing and instruction for immigrant populations. In an increasingly globalized economic and political world, understanding and ameliorating communication problems by speakers of different 'mother tongues' is of great importance. Spoken English proficiency is extremely important for the educational and economic advancement of immigrant populations in the USA.

People have difficulties communicating with non-native speakers, both understanding their speech and being understood. This study examines the difficulties that non-native English learners have in understanding continuous speech. English word-final consonants in connected speech are produced very differently from those produced in isolation; this phenomenon is referred to as coarticulation. The proposed study will examine the identification of words ending with /p/, /t/, /k/, /b/, /d/, /g/, /m/, /n/, and /ng/ (as in "king") followed by a word beginning with /p/, /t/ or /k/ in meaningful sentences by Japanese listeners. The task is to identify the target word in minimal triplets (e.g., sip, sit, sick) after listening to a sentence, such as "He said the word "sip" (or "sit" or "sick") positively" (or "tauntingly" or "cautiously"). The sentences will be spoken in both "clear speech" and in "fast conversational speech" styles. The former is used when native English speakers know that the listener may have difficulty understanding them; the latter is used when speaking with friends.

The proposed studies will investigate whether Japanese listeners have difficulty recognizing the final consonants in English words relative to native English listeners. whether their error patterns are different from those of English listeners, and whether they perform significantly better when the sentences are produced in "clear speech" style. Correctly identifying English word-final stops is expected to be challenging for Japanese listeners because Japanese words do not end in word-final consonants, with the exception of a nasal stop (/N/). Word-final nasal stops are also potentially confusable because the Japanese nasal is produced as /m/ /n/ or /ng/ depending on the following consonant. These data will have implications for how native speakers can modify their speech to improve communication with non-native speakers in English language classrooms and in work environments.