William F. Katz

Time-varying spectral change in the vowels of children and adults

Recent studies have shown that time-varying changes in formant pattern contribute to the phonetic specification of vowels. This variation could be especially important in childrens vowels, because children have higher fundamental frequencies (f0s) than adults, and formant-frequency estimation is generally less reliable when f0 is high. To investigate the contribution of time-varying changes in formant pattern to the identification of childrens vowels, three experiments were carried out with natural and synthesized versions of 12 American English vowels spoken by children (ages 7, 5, and 3 years) as well as adult males and females. Experiment 1 showed that (i) vowels generated with a cascade formant synthesizer (with hand-tracked formants) were less accurately identified than natural versions; and (ii) vowels synthesized with steady-state formant frequencies were harder to identify than those which preserved the natural variation in formant pattern over time. The decline in intelligibility was similar across talker groups, and there was no evidence that formant movement plays a greater role in childrens vowels compared to adults. Experiment 2 replicated these findings using a semi-automatic formant-tracking algorithm. Experiment 3 showed that the effects of formant movement were the same for vowels synthesized with noise excitation (as in whispered speech) and pulsed excitation (as in voiced speech), although, on average, the whispered vowels were less accurately identified than their voiced counterparts. Taken together, the results indicate that the cues provided by changes in the formant frequencies over time contribute materially to the intelligibility of vowels produced by children and adults, but these time-varying formant frequency cues do not interact with properties of the voicing source.

Measures of the sentence intonation of read and spontaneous speech in American English

The visual abstraction procedure used in previous studies of declination was tested using 12 subjects who each fit the F0 contours of 19 spoken short simple sentences with baselines. These baselines were found to be poorly replicated the fitters. An objective all-points least-squares best-fit procedure was tested on this corpus and on a set of sentences that had been produced in both spontaneous and read speech by six speakers. The all-points linear regression line was a better descriptor of the F0 contours than either baselines or toplines. Declination did not always occur in these simple declarative sentences; there was more variation present in the F0 contours of sentences that had been uttered during spontaneous speech; 35% of the spontaneous sentences did not show declination; 45% of these sentences better fit the breath-group model. Their F0 contours could be described by a level all-points linear regression line followed by a falling terminal segment.

Rapid automatized naming and gesture by normal and language-impaired children.

This study investigates whether language-impaired (LI) children show deficits in rapid automatized naming and whether RAN performance is specific to verbal output (or to rapid motor output in general). A total of 67 LI and 54 age-matched control children were tested with the Rapid Automatized Naming (RAN) test (Denckla & Rudel, 1976) and with a manual version of the RAN (RAN-manual) in which subjects were required to provide a nonverbal, pantomime response. Subjects also completed tests of rapid oral and manual sequencing skills and standardized tests of reading ability. Each subject was tested at 4, 6, and 8 years old. The results showed that LI children perform significantly poorer on both versions of the RAN than age-matched controls. Correlations between RAN scores and tests of reading ability were significant for normal and LI subjects and were particularly high for 8-year-old LI children. RAN-manual scores also correlated with 8-year-old LI childrens reading scores. Further, RAN and RAN-manual scores for the LI children correlated significantly with these childrens manual sequencing abilities, whereas this was not the case for the control subjects. These findings suggest that LI childrens rapid sequential processing deficits are not limited to verbal output, but also generalize to other motoric domains.

Synthesis fidelity and time-varying spectral change in vowels

Recent studies have shown that synthesized versions of American English vowels are less accurately identified when the natural time-varying spectral changes are eliminated by holding the formant frequencies constant over the duration of the vowel. A limitation of these experiments has been that vowels produced by formant synthesis are generally less accurately identified than the natural vowels after which they are modeled. To overcome this limitation, a high-quality speech analysis-synthesis system (STRAIGHT) was used to synthesize versions of 12 American English vowels spoken by adults and children. Vowels synthesized with STRAIGHT were identified as accurately as the natural versions, in contrast with previous results from our laboratory showing identification rates 9%-12% lower for the same vowels synthesized using the cascade formant model. Consistent with earlier studies, identification accuracy was not reduced when the fundamental frequency was held constant across the vowel. However, elimination of time-varying changes in the spectral envelope using STRAIGHT led to a greater reduction in accuracy (23%) than was previously found with cascade formant synthesis (11%). A statistical pattern recognition model, applied to acoustic measurements of the natural and synthesized vowels, predicted both the higher identification accuracy for vowels synthesized using STRAIGHT compared to formant synthesis, and the greater effects of holding the formant frequencies constant over time with STRAIGHT synthesis. Taken together, the experiment and modeling results suggest that formant estimation errors and incorrect rendering of spectral and temporal cues by cascade formant synthesis contribute to lower identification accuracy and underestimation of the role of time-varying spectral change in vowels.

An investigation of lexical ambiguity in Broca's aphasics using an auditory lexical priming technique

Brocas aphasics and normal control subjects performed an auditory lexical decision task with stimulus pairs containing ambiguous (semantically-related) or unambiguous (unrelated) words as primes. Brocas aphasics, though having considerably longer latencies, produced a pattern of results similar to normal subjects; namely, faster reaction times for target words preceded by semantically related than unrelated words (i.e., semantic priming). These results do not support the hypothesis of BLUMSTEIN and colleagues that Brocas aphasics show a selective impairment in automatic processing associated with lexical access.

Treating apraxia of speech (AOS) with EMA-supplied visual augmented feedback

Background: Previous studies have suggested that visual augmented feedback provided by electromagnetic articulography (EMA) helps persons with apraxia of speech (AOS) recover speech motor control following stroke (e.g., Katz et al., 2007). However, the data are few, both in terms of the variety of participants and the speech motor targets investigated. Aims: This study was designed to determine whether EMA supplied feedback improves articulatory accuracy in an adult with acquired AOS. We also examined whether reduced feedback frequency results in (1) decreased performance during acquisition and (2) enhanced maintenance and generalisation of the targeted behaviours. Methods & Procedures: A multiple-baseline across-behaviours design was used to assess the efficacy of this treatment for an individual with AOS. Over a 27-week period, the participant received visual feedback provided by an EMA system for treatment of three groups of speech motor targets (SMTs): /j/, /θ/, and /t∫  / with various following VCs. The consonant clusters /br/ and /sw/ served as untreated controls. Frequency of feedback scheduling was 100% for /j/ and /t∫ /, and 50% for /θ/. Outcomes & Results: For the first group of SMTs treated, /j/, there was acquisition for 4/5 trained words. These were maintained post-treatment and at the long-term probe. Improved performance and maintenance were also noted for 5/8 untreated stimuli, with maintenance shown for most of these words by 1 month post-treatment. The next treated SMT, /θ/, showed acquisition for all five treated items. Two of these five targets were maintained one month post-treatment. All three untreated /θ/ probes showed generalisation, with two of these showing maintenance post-treatment. The third treated group of SMTs, /t∫/, showed improved performance for all of the five treated words. However, these gains could only be attributed to /t∫/ treatment for three of the five words. Two treated items appeared well maintained at 1 month post-treatment. Generalisation and maintenance were also noted for all six untreated /t∫ / words. However, generalisation from previously treated /j/ and /t∫/ targets was involved in their improved performance. The untrained (control) word data suggested that the gains noted for treated items did not result from across-the-board improvement or unassisted recovery. There were no consistent differences corresponding with low- versus high-frequency feedback conditions. Conclusions: Augmented kinematic feedback provided by an EMA system improved production for some, but not all, treated targets. Generalisation to untreated probes was also evident. Predictions concerning the effects of feedback frequency on the acquisition, maintenance, and transfer of trained behaviours were not supported.

Identification of children's and adults' vowels: Intrinsic fundamental frequency, fundamental frequency dynamics, and presence of voicing

Abstract Three experiments investigated the role of fundamental frequency ( F 0 ) in the perception of vowels produced by children and adults. The stimuli were based on a set of American English vowels recorded in /hVd/ syllables by three groups of children (ages 3, 5, and 7 yr) and adult males and females. Experiment 1 presented natural and synthesized vowels to adult listeners for identification. To determine whether vowel-specific differences in F 0 (intrinsic F 0 ) contribute to the accuracy of vowel identification, listeners identified stimuli in which F 0 was held constant at the mean value across the entire vowel set for a given talker. These manipulations had no effect on identification accuracy, either when F 0 conditions varied from trial to trial (Experiment 1) or were blocked within a set of trials (Experiment 2). Holding F 0 constant over the duration of the vowel did not lower accuracy, suggesting that time-varying changes in F 0 make little contribution to vowel identification. Experiment 3 assessed whether the presence of F 0 per se provides benefits for vowel identification. This was achieved by replacing the pulsed excitation source of the synthesized voiced vowels with white noise to simulate whispered speech. Identification accuracy was lower for noise-excited than pulse-excited vowels for all talker groups. Differences between noise-excited and voiced vowels were larger for adult males than for children, suggesting that the benefits of F 0 are not due simply to improved specification of the spectrum envelope. Taken together, the results suggest that (1) F 0 plays a small but complex role in vowel identification in isolated words; (2) the periodicity or harmonicity contributed by F 0 carries more perceptual weight than either vowel-inherent F 0 dynamics or a vowels intrinsic F 0 , and (3) F 0 characteristics (dynamic patterns and intrinsic F 0 ) provide a relatively small contribution to the reduced intelligibility commonly noted for childrens vowels.

Anticipatory coarticulation and aphasia: implications for phonetic theories

Abstract Studies of clinical populations usually (1) track more variables than in studies of normal subjects and (2) encounter a high degree of behavioral variability, making generalization difficult. However, these apparent limitations may in the long run prove to be strengths. For example, in studies of normal talkers variability is often treated as noise against which the size of an effect may be measured, while in studies of disordered talkers the extent of variability itself may be an important goal of inquiry. Information obtained from clinical populations has proven valuable in addressing the neuroanatomical substrates and functional organization of speech. This review focuses on recent studies of anticipatory coarticulation in the speech of focally-lesioned aphasic talkers, examining how these data address the neural representation of language processes, the role of suprasegmental units (e.g., the syllable) in speech production, and the manner in which coarticulatory variability is perceived by normal listeners. The reviewed data indicate that the variable speech of clinical populations does not necessarily pose an impediment to research, but rather constitutes an important database from which further understanding of normal and disordered speech processes may be obtained.

The role of prosody in a case of foreign accent syndrome (FAS).

Foreign accent syndrome (FAS) is a rare disorder characterized by the emergence of a perceived foreign accent following brain damage. The symptomotology, functional bases, and neural substrates of this disorder are still being elucidated. In this case study, acoustic analyses were performed on the speech of a 46‐year old monolingual female who presented with FAS of unknown aetiology. The patient had a pseudo‐accent frequently described as ‘Swedish’ or ‘Eastern European’. Stop consonant VOT, consonant burst spectra and duration, vowel durations, formant frequencies, and trajectories were analysed, along with prosodic cues for lexical stress assignment and sentence‐level intonation. Results indicated VOT values were generally preserved, while there was a strong tendency to realize the English alveolar flap as a full stop, and to produce flaps that had greater‐than‐normal closure durations. The spectral properties of the patients vowels resembled those of normal talkers (with the possible exceptions of decreased F1 values for /i/ and slight differences in formant dynamics for /u/, /o/, /i/, and /ϵ/). However, vowel durations were relatively long, contributing to exaggerated tense/lax contrasts. Token‐to‐token variability in vowel production was slightly higher than normal for duration, but not for formant frequency values. Lexical stress assignment was inaccurate and highly variable (with similar problems noted for non‐speech materials), and sentence level intonation showed occasional deviations from typical American English patterns. For this patient, an underlying timing/rhythm difficulty appeared responsible for the range of segmental and suprasegmental changes leading to the impression of a foreign accent.

Visual Feedback of Tongue Movement for Novel Speech Sound Learning.

Pronunciation training studies have yielded important information concerning the processing of audiovisual (AV) information. Second language (L2) learners show increased reliance on bottom-up, multimodal input for speech perception (compared to monolingual individuals). However, little is known about the role of viewing one’s own speech articulation processes during speech training. The current study investigated whether real-time, visual feedback for tongue movement can improve a speaker’s learning of non-native speech sounds. An interactive 3D tongue visualization system based on electromagnetic articulography (EMA) was used in a speech training experiment. Native speakers of American English produced a novel speech sound (/ɖ/; a voiced, coronal, palatal stop) before, during, and after trials in which they viewed their own speech movements using the 3D model. Talkers’ productions were evaluated using kinematic (tongue-tip spatial positioning) and acoustic (burst spectra) measures. The results indicated a rapid gain in accuracy associated with visual feedback training. The findings are discussed with respect to neural models for multimodal speech processing.