Burton S. Rosner

Loudness predicts prominence: Fundamental frequency lends little

We explored a database covering seven dialects of British and Irish English and three different styles of speech to find acoustic correlates of prominence. We built classifiers, trained the classifiers on human prominence/nonprominence judgments, and then evaluated how well they behaved. The classifiers operate on 452 ms windows centered on syllables, using different acoustic measures. By comparing the performance of classifiers based on different measures, we can learn how prominence is expressed in speech. Contrary to textbooks and common assumption, fundamental frequency (f0) played a minor role in distinguishing prominent syllables from the rest of the utterance. Instead, speakers primarily marked prominence with patterns of loudness and duration. Two other acoustic measures that we examined also played a minor role, comparable to f0. All dialects and speaking styles studied here share a common definition of prominence. The result is robust to differences in labeling practice and the dialect of the labeler.

Auditory frequency discrimination in adult developmental dyslexics

Developmental dyslexics reportedly discriminate auditory frequency poorly. A recent study found no such deficit. Unlike its predecessors, however, it employed multiple exposures per trial to the standard stimulus. To investigate whether this affects frequency discrimination in dyslexics, a traditional twointerval same-different paradigm (2I_1A_X) and a variant with six A-stimuli per trial (2I_6A_X) were used here. Frequency varied around 500 Hz; interstimulus interval (ISI) ranged between 0 and 1,000 msec. Under 2I_1A_X, dyslexics always had larger just noticeable differences (JNDs) than did controls. Dyslexic and control JNDs were equal at shorter ISIs under 2I_6A_X, but dyslexics became worse than controls at longer ISIs. Signal detection analysis suggests that both sensory variance and trace variance are larger in dyslexics than in controls.

Rhythm measures and dimensions of durational variation in speech.

Patterns of durational variation were examined by applying 15 previously published rhythm measures to a large corpus of speech from five languages. In order to achieve consistent segmentation across all languages, an automatic speech recognition system was developed to divide the waveforms into consonantal and vocalic regions. The resulting duration measurements rest strictly on acoustic criteria. Machine classification showed that rhythm measures could separate languages at rates above chance. Within-language variability in rhythm measures, however, was large and comparable to that between languages. Therefore, different languages could not be identified reliably from single paragraphs. In experiments separating pairs of languages, a rhythm measure that was relatively successful at separating one pair often performed very poorly on another pair: there was no broadly successful rhythm measure. Separation of all five languages at once required a combination of three rhythm measures. Many triplets were about equally effective, but the confusion patterns between languages varied with the choice of rhythm measures.

The Perceptual Roles of Melodic Process, Contour, and Form

A melodic process traces a melody9s principal motions from its main beginning to its closure. Western tonal music seems built around a handful of different melodic processes. In each of three experiments, we took melodically complete excerpts from fully instrumented recordings of music of the Classical and Romantic periods. The stimuli instantiated either of two different melodic processes. The processes varied across experiments. Subjects judged the dissimilarity of pairs of stimuli. Multidimensional scaling and hierarchical clustering showed that melodic processes and overall properties of contour were important in determining the subjects9 perceptions. In a fourth experiment, two-part forms proved perceptually distinguishable from three-part forms. A final experiment which systematically varied both process and form indicated that process and contour can completely mask any perceptual effects of form. On balance, melodic process, contour, and form influence the perception of music in descending order of importance.

Voice-onset times for Castilian Spanish initial stops

Abstract A previous experiment on the Castilian Spanish perceptual voice-onset time (VOT) boundary indicated possible differences between Spanish dialects. We therefore measured VOTs for the six Castilian stops in initial position. Significant main effects of voicing and place on VOT reproduced previous findings on Latin American Spanish dialects. Voicing also interacted with place and postconsonantal vowel. Voice-onset times for stops in the same voicing category but not in different categories correlated across speakers. Some previously published Latin American VOTs differ significantly from our measurements, and the majority of the Latin American VOTs fall outside the 99% confidence limits of our means. Spanish articulatory VOT values seem to vary with dialect, possibly influencing Spanish VOT perceptual boundaries.

Perception of English Intonation by English, Spanish, and Chinese Listeners

Native language affects the perception of segmental phonetic structure, of stress, and of semantic and pragmatic effects of intonation. Similarly, native language might influence the perception of similarities and differences among intonation contours. To test this hypothesis, a cross-language experiment was conducted. An English utterance was resynthesized with seven falling and four rising intonation contours. English, Iberian Spanish, and Chinese listeners then rated each pair of nonidentical stimuli for degree of difference. Multidimensionals caling of the results supported the hypothesis. The three groups of listeners produced statistically different perceptual configurations for the falling contours. All groups, however, perceptually separated the falling from the rising contours. This result suggested that the perception of intonation begins with the activation of universal auditory mechanisms that process the direction of relatively slow frequency modulationd A second experiment there fore employed frequency-modulated sine waves that duplicated the fundamental frequency contours of the speech stimuli. New groups of English, Spanish, and Chinese subjects yielded no cross-language differences between the perceptual configurations for these nonspeech stimuli. The perception of similarities and differences among intonation contours calls upon universal auditory mechanisms whose output is molded by experience with ones native language.

Voice-onset Time and Tone-onset Time: The Role of Criterion-setting Mechanisms in Categorical Perception

Problems in modelling categorical perception (CP) and attempts to apply signal detection theory (SDT) to CP are reviewed. An approach based on SDT supplemented by a theory of criterion setting is presented. Criterion setting theory (CST) postulates mechanisms that reset the response criterion on each trial, and it accounts for sequential dependencies. A criterion setting model for discrimination is shown to fit data from the literature. The hypothesis that “sharp” category boundaries may arise from the suppression of noise caused by intertrial dependencies was examined in an experiment on the identification of [ba] and [pa] syllables, and tone combinations of varying tone-onset time. However, it was shown that both positive and negative intertrial dependencies were present. They could be fitted by the criterion-setting model; in this respect, CP resembles standard psychophysical judgements. Examination of the psychometric functions from the two CP tasks shows that they are not normal ogives, as in standard psychophysical tasks: these curves are steeper centrally and flatter at the extremes than a Gaussian ogive; we describe them as “hypersigmoid”. The description of CP identification functions as hypersigmoid provides a new, qualitative characterization of the “sharp” category boundaries traditionally claimed for CP. Their causation remains to be determined.

Harmonic Closure: Music Theory and Perception

Music theorists have often disagreed about the material variables that determine the perception of harmonic closure. To investigate this controversial topic, we presented subjects with pairs of selected two-chord progressions. The subjects judged which member of each pair seemed more closed. Preferences varied across pairs of cadences and generally obeyed transitivity. Quantitative reformulation of theoretical harmonic variables permitted correlational analysis of the results. Three or four variables, including one or two that reflect learned stylistic structures, best explained our findings. Conventional harmonic factors of scale step, soprano position, and root position demonstrated surprisingly little explanatory power.

Perception of voice‐onset‐time continua: A signal detection analysis

An absolute identification task and various discrimination tasks were performed on synthetic syllables initiated by alveolar stops varying in voice-onset-time (VOT). Signal detection analyses of the data yielded three main findings. First, discrimination ds exceeded those for absolute identification. Greater instability of judgment in the latter than in the former task explained this difference. Second, stimuli within a phonetic category were discriminable, even in a delayed 2IAX paradigm with 4s between stimuli. Third, a category boundary effect appeared: identification and discrimination ds were largest around the voiced/voiceless boundary for alveolar stops. The relationship between discrimination and identification ds varies with response constraints, the number of just noticeable differences in the stimulus array, and stability of judgments in the various tasks. Nearly equal performance in discrimination and identification tasks is not a sufficient condition for a category boundary effect to occur. Several arguments are advanced against a dual-coding model. A continuous model which invokes only a single decision process may account more gracefully for the relationship between identification and discrimination of both consonants and vowels.

Identificación del orden temporal en sonidos de habla y de no-habla

espanolUn grupo de monolingues espanoles y otro de monolingues ingleses fueron sometidos a experimentos de identificacion respecto a tres continuos acusticos: un continuo de sonidos sintetizado o [ba]-[pa] que variaba en el tiempo de inicio de la sonoridad (tis), un continuo de tiempo de inicio del ruido (tir) compuesto por un ruido y un zumbido, y un continuo de dos tonos puros (tit) en el que el inicio de uno de ellos (500 hz) se retrasaba o adelantaba con relacion al otro (1500 hz) se retrasaba o adelantaba con relacion al otro (1500 hz). Las asincronias temporales para todos los continuos fueron desde -35 a 55 ms. Como se esperaba, la frontera de tis fue inferior en el grupo de sujetos espanoles. El mismo resultado se obtuvo el continuo de tit, pero no asi con el de tir. La forma de las funciones de identificacion diferia en cada continuo acustico. Segun los resultados obtenidos, la percepcion del tis en adultos no parece estar relacionada causalmente con la percepcion de asincronias temporales por el sistema auditivo. EnglishNative English and Spanish speakers provided identification functions for three acoustic continua: synthetic [ba]-[pa] stimuli varying in voice-onset-time (VOT), a noise-lead-time (NLT) continuum where a noise and a buzz started at different times but terminated together, and a tone-onset-time (TOT) series with a variable onset time between 500 Hz and 1500 Hz coterminous sinusoids. Onset-time differences for each continuum ranged from-35 to 55 ms. As expected, the Spanish VOT boundary was shorter than the English one. The TOT, but not the NLT, boundaries also differed between groups. Identification functions changed systimatically in shape across continua. The results show that perception of VOT by adults seems not to be causally related to the auditory temporal lag perception