Search

Approximately 140 students take Introduction to Phonetics each year at the University of New Mexico. They are diverse in several ways: in age (many are returning to school after years away), in ethnicity (UNM is majority-minority), in major (the course serves Linguistics and Speech & Hearing), and in preparation (there are no prerequisites, but many students have some related coursework). This paper reports on design considerations and experience with the first online offering of this course. Studying online requires self-motivation and the ability to work independently, which are difficult for the large number of students with poor study skills and time management. This course attempts to counter these challenges with a high degree of structure, and activities in a variety of formats for learning and assessment. Although some in distance education view structure as a “constraint” for learner and teacher (Farquhar 2013 Eur. J. e-learning), this instructor believes that the flexibility and isolation of onl...

Listeners’ perception of prosodic structure may differ depending on whether they are instructed to attend to the meaning of a spoken passage, or to the acoustics. Real-time perceptions of prominence and phrasal boundaries were obtained from Rapid Prosody Transcription [Cole et al. (2010)]. Twenty naive French listeners were divided into two groups that were given either meaning- or acoustically based instructions for listening to passages of spontaneous speech. While listening, they read an orthographic unpunctuated transcript of the speech. Half of each group first labeled words they perceived as prominent in five passages, then phrasal boundaries in five different passages; the other half performed the tasks in the opposite order. Consistent with previous results, listener agreement (measured by kappa) was higher for labeling boundaries than for prominence. The mean kappa was 0.80 for both meaning- and acoustically based responses, but the difference between prominence and boundaries was greater in meaning-based responses. Listeners labeled more words as prominent than they labeled boundaries, with a greater difference in meaning-based responses than acoustically based, although the frequency of labeling did not differ overall between acoustic- and meaning-based instructions. The divergence between prominence and boundary labeling challenges the assumption that prominence in French derives from pre-boundary position.

Prosodic structure in speech derives from both linguistic structure, particularly syntax, and performance factors such as speech rate. The combination of influences means that predicting the structure of a spontaneous utterance remains a major challenge. One approach has been to survey listeners and take their perceptions as the basis for analysis [e.g., Cole et al., Lab.Phonol. (2010)]. A study of this type in French compared perceptions of two types of spontaneous speech: informal conversation and more formal journalistic discussion. 25 naive native speakers of French listened to recordings of both types and indicated where they perceived boundaries between groups of words. Their responses were then compared to acoustic measurements and analyzes of the phrasal structure of the conversations. The presence of a pause and a rise in F0 were the most reliable acoustic correlates of a perceived boundary. Preliminary analyzes of phonological phrase structure suggest that the thresholded boundary weight hypothesis of Breen et al. [2010 Lang. Cog. Proc.] is a promising metric for explaining the pattern of boundary perception in these data. Locations were more often perceived as boundaries when the sum of the sizes of just‐completed and upcoming constituents (measured in numbers of phonological phrases) is larger.

Although a speech style known as foreigner talk has been described, speech addressed to a reasonably fluent non‐native interlocutor may show little or no accommodation by native speakers. This issue is of interest in understanding whether on‐line modifications to speech are an important contributor to phonetic variation. Ten native speakers of Parisian French were recorded individually as they performed map tasks, one with another native speaker of French, and another with an American who identified himself as such at the beginning of the conversation, but who spoke French during the task. Previous work has found that French listeners who heard brief extracts from these conversations could not reliably identify which interlocutor was involved. In an attempt to identify other possible differences in speech to natives and non‐natives, two prosodic characteristics of the conversations were analyzed, vowel devoicing and the epenthesis of word‐final schwas. The hypothesis was that vowel devoicing, which is sty...

Cross‐language prosodic differences are well‐known at the phrasal level and below, and a few studies have shown language‐dependent differences in speakers’ prosody over a discourse or text [Fon (2002), Smith and Hogan (2003)]. For comparison with the patterns found in production, this study examines listeners’ preferences in text‐level prosody. Previously‐analyzed recordings of one speaker of American English reading aloud an English text, and one speaker of Parisian French reading a comparable French text, were used as a basis. The discourse organization of each text was analyzed by several native speakers who categorized each sentence‐to‐sentence transition as a Topic Shift, Continuation or Elaboration. The original recordings were manipulated in several ways to alter rate, sentence‐final lengthening and pause duration, and the manipulated versions were presented to listeners. English listeners liked best those versions where values for rate and lengthening matched the speakers’ means for each separate ...

Previous work [Smith and Hogan (2001), (2002)] has shown that the organization of topics in a text affects durational patterning when the text is read aloud. Comparable texts in English, French, and Japanese were selected, and native speakers of each language were recorded reading them aloud. For all three texts, the transition in topic from each sentence to the next was labeled as a Shift, Continuation, or Elaboration. Acoustic measurements were made of lengthening of the sentence‐final word, the duration of pauses between sentences and changes in speaking rate from one sentence to the next, and in all languages, these measures were significantly different for different topic transition types. The direction and magnitude of the effects differed considerably among the single speakers of the three languages examined so far. However, with just one speaker per language, it is impossible to distinguish language‐specific from speaker‐specific behavior. Additional data from more speakers will be reported, suggesting that differences among languages can be attributed to known prosodic differences, whereas differences among speakers of the same language are less systematic. [Work supported by NSF Grant No. BCS‐9983106.]

It is well known that voiced stops in English tend not to be fully voiced. For many American speakers, voiced fricatives may also have little or no voicing. In what phonological contexts do speakers devoice underlying /z/? Previous research [T. Veatch, Ling. Soc. Am. mtg., 69 (1989)] emphasized the influence of the following segment. The present study confirms this, and investigates the likelihood of devoicing in different positions in syllable, word, and sentence. Five speakers of American English recorded multiple repetitions of sentences in which /s/ and /z/ occurred in matched environments. Measures of acoustic durations, airflow, and vocal fold vibration as evidenced by EGG were used to compare the production of /z/ and /s/. Preliminary data from one speaker show a variety of /z/’s, with the duration of vocal fold vibrations as a percentage of fricative duration ranging from 0% to 100%. However, the acoustic vowel length difference preceding /s/ and /z/ is maintained, and even those tokens of /z/ in ...

In a model of articulatory movement, based on a dynamical systems approach, parameter values derived from actual articulatory data are being used. One way of obtaining these values relies on the program NEWPAR, which has been developed to analyze articulatory movement in order to extract dynamic parameters such as frequency that will serve as coefficients in the model equation [McGowan et el., J. Acoust. Soc. Am. Suppl. 1 83, S113 (1988)]. Alternative hypotheses for the application of NEWPAR have been tested, including starting the half‐cycle windows at different points in the trajectory and fixing either the initial displacement and velocity or the initial and final displacement as a means of reducing the number of parameters to be fitted. Tests have been made on simulated data, consisting of sinusoidal curves with damping varied from completely undamped to critically damped. Patterns in the results of these tests and some preliminary analyses of actual lower lip movement trajectories will be shown. [Work supported by NSF and NIH.]In a model of articulatory movement, based on a dynamical systems approach, parameter values derived from actual articulatory data are being used. One way of obtaining these values relies on the program NEWPAR, which has been developed to analyze articulatory movement in order to extract dynamic parameters such as frequency that will serve as coefficients in the model equation [McGowan et el., J. Acoust. Soc. Am. Suppl. 1 83, S113 (1988)]. Alternative hypotheses for the application of NEWPAR have been tested, including starting the half‐cycle windows at different points in the trajectory and fixing either the initial displacement and velocity or the initial and final displacement as a means of reducing the number of parameters to be fitted. Tests have been made on simulated data, consisting of sinusoidal curves with damping varied from completely undamped to critically damped. Patterns in the results of these tests and some preliminary analyses of actual lower lip movement trajectories will be shown. [Wor...

Vowel‐to‐vowel coarticulation across consonants can be seen as evidence of overlap between the articulatory gestures of the vowels. Further evidence for the organization of the gestures can be found in contextual variation in the durations of the vowels. This acoustic study of Japanese examines the effects of geminate consonants on the durations and the F2 values of the surrounding vowels in VCV utterances. Because of the syllable structure of the language, geminate consonants in Japanese might be expected not to close the preceding syllable, eliminating this potential source of vowel shortening. In agreement with this hypothesis, vowels did not shorten before geminate consonants (in fact they lengthened), and the second vowel affected the formants of the first vowel less across geminate than across single consonants. [Work supported by NSF.]

Two opposing explanations of the timing of consonants and vowels are that (1) they are timed relative to each other, or (2) vowels are produced in a continuous rhythm with consonants superimposed on them [Fowler, J. Exp. Psychol. 112, 386–412 (1983)]. Languages with different temporal properties, such as Japanese, a mora‐timed language, and Italian, a syllable‐timed language, might be expected to exemplify these different relationships. Speakers of each of these languages were recorded producing pairs of disyllabic words with medial consonants of contrasting length (single or geminate). Evidence from measures of acoustic duration and vowel‐to‐vowel coarticulation, shown by contextual changes in F2, suggest that Fowler's model (2) predicts the timing patterns found in Italian, while the other model (1) predicts the behavior of Japanese. [Work supported by NSF.]

In order to model the movements of the speech articulators using a dynamical systems approach, it is necessary to specify (1) the specific form of the dynamical equation, (2) the portion of the articulatory trajectory being described/generated by the equation—the “window,” and (3) the values of the coefficients to be used in the equation. In this work, a damped mass‐spring dynamical system is assumed, as used in previous task dynamic modeling [Saltzman et al., J. Acoust. Soc. Am. Suppl. 1 82, S15 (1987)]. A program was developed to determine the values of the coefficients for the equations, using various definitions of the articulatory window. The program uses a nonlinear least‐squares algorithm to determine the best fit between the observed trajectories and the trajectories generated using a range of values for the coefficients. This program is currently being tested on simulated data with known parameters in order to establish the accuracy and limitations of the procedure. There will be a report on the reliability of the analyses performed using this approach, with an emphasis on numerical considerations. [Work supported by Grant BNS‐8520709 from NSF and Grants HD‐01994 and NS‐13617 from NIH.]