Trained Listeners Research Articles

Different learning patterns for tone detection in three simultaneous-masking conditions with different temporal characteristics

Can the processes underlying performance on basic masking tasks be modified with practice? To address this question, 17 listeners completed pre- and posttests in which detection thresholds for a 20-ms 1-kHz tonal signal were measured in three notched-noise simultaneous-masking conditions that differed in the masker duration and relative onset times of the signal and masker, and three forward-masking conditions that differed in the tonal frequencies of the maskers. For the ∼8 days between tests, eight of these listeners practiced 240 trials/day on each simultaneous-masking condition. Learning differed across conditions. When the signal and a 20-ms masker began and ended together, both trained and untrained listeners improved, but trained listeners improved more, learning gradually. When the signal began 400 ms after a 500-ms masker, only the trained group improved, learning quickly. When the signal and the 500-ms masker began together, both groups improved equally. Both groups also improved equally on each of the three forward-masking conditions. These results suggest that processes underlying performance on simultaneous- and forward-masking tasks are modifiable, training affects different mechanisms in simultaneous-masking conditions with different temporal characteristics, and the simultaneous-masking mechanisms affected by multi-hour training are not involved in forward masking. [Work supported by NIH.]

The effects of auditory–visual vowel and consonant training on speechreading performance

Recent work examined the effects of a novel approach to speechreading training using vowels, for normal-hearing listeners tested in masking noise [C. Richie and D. Kewley-Port, J. Acoust. Soc. Am. 114, 2337 (2003)]. That study showed significant improvements in sentence-level speechreading abilities for trained listeners compared to untrained listeners. The purpose of the present study was to determine the effects of combining vowel training with consonant training on speechreading abilities. Normal-hearing adults were tested in auditory–visual conditions in noise designed to simulate a mild-to-moderate sloping sensorineural hearing loss. One group of listeners received training on consonants in monosyllable context, and another group received training on both consonants and vowels in monosyllable context. A control group was tested but did not receive any training. All listeners performed speechreading pre- and post-tests, on words and sentences. Results are discussed in terms of differences between groups, dependent upon which type of training was administered; vowel training, consonant training, or vowel and consonant training combined. Comparison is made between these and other speechreading training methods. Finally, the potential benefit of these vowel- and consonant-based speechreading training methods for rehabilitation of hearing-impaired listeners is discussed. [Work supported by NIHDCD02229.]

Correlations among within-channel and between-channel auditory gap-detection thresholds in normal listeners.

We obtained data on within-channel and between-channel auditory temporal gap-detection acuity in the normal population. Ninety-five normal listeners were tested for gap-detection thresholds, for conditions in which the gap was bounded by spectrally identical, and by spectrally different, acoustic markers. Separate thresholds were obtained with the use of an adaptive tracking method, for gaps delimited by narrowband noise bursts centred on 1.0 kHz, noise bursts centred on 4.0 kHz, and for gaps bounded by a leading marker of 4.0 kHz noise and a trailing marker of 1.0 kHz noise. Gap thresholds were lowest for silent periods bounded by identical markers--'within-channel' stimuli. Gap thresholds were significantly longer for the between-channel stimulus--silent periods bounded by unidentical markers (p < 0.0001). Thresholds for the two within-channel tasks were highly correlated (R = 0.76). Thresholds for the between-channel stimulus were weakly correlated with thresholds for the within-channel stimuli (1.0 kHz, R = 0.39; and 4.0 kHz, R = 0.46). The relatively poor predictability of between-channel thresholds from the within-channel thresholds is new evidence on the separability of the mechanisms that mediate performance of the two tasks. The data confirm that the acuity difference for the tasks, which has previously been demonstrated in only small numbers of highly trained listeners, extends to a population of untrained listeners. The acuity of the between-channel mechanism may be relevant to the formation of voice-onset time-category boundaries in speech perception.

Learning foreign vowels.

Two hypotheses have recently been put forward to account for listeners' ability to distinguish and learn contrasts between speech sounds in foreign languages. First, Best's Perceptual Assimilation Model and Flege's Speech Learning Model both predict that the ease with which a listener can tell one non-native phoneme from another varies directly with the extent to which these sounds assimilate to different native phonemes (Best, 1994; also Best, McRoberts, & Goodell, 2001; Flege, 1991). Second, Logan, Lively, & Pisoni (1991) have argued that training listeners to identify non-native phonemes teaches them sets of exemplars rather than more abstract distinctive feature values. I report here the results of three sets of experiments designed to test these hypotheses, in which American English listeners were trained to categorize German nonlow vowels. The first set of experiments show that some instances of the same contrast between German vowels are more easily discriminated than others, a result incompatible with the predictions of either Best's or Flege's models, but compatible with the alternative category recognition interpretation. The second set of experiments reveals effects of contextual and speaker variation on listeners' ability to learn [tense] but not [high] contrasts between foreign vowels, and are thus at least partly compatible with an exemplar model of foreign category learning (Pisoni, Lively, & Logan, 1994; also Nosofsky, 1986). The third set of experiments compares the predictions of Nosofsky's (1986) selective attention exemplar model of category learning with those of a feature learning model in tests of listeners' learning the natural classes to which the German vowels belong. The results are mixed: listeners learned the features that define the natural classes of [+/- high] and [+/- back] vowels, but could have learned either the feature that defines the natural classes of [+/- tense] vowels or sets of [+/- tense] exemplars. Natural classes defined by abstract distinctive feature values are thus learnable, even if their membership is phonetically polymorphous.

Speech habilitation of hard of hearing adolescents using electropalatography and ultrasound as evaluated by trained listeners

Four adolescents with moderate to severe sensorineural hearing losses and moderately unintelligible speech participated in a 14-week speech therapy study using two dynamic visual feedback technologies, electropalatography and ultrasound imaging. Electropalatography provides information about tongue-hard palate contact points. Ultrasound displays images of tongue shape and movement in two dimensions from the tip to the root. Treatment targets for all participants included a sibilant place contrast (/s/ versus / Z /), liquids /l/ and /r/, and the tense-lax vowel contrast with the high vowels. Trained listener evaluations of pre- and post-treatment transcripts are reported in this paper. Significant improvements in speech production were noted across students and targets. Treatment targets improved significantly more than non-treatment test targets overall. Students showed greatest gains on consonants that were absent or marginal in their speech pre-treatment. No particular advantage of one technology over the other was evident in this sample.

On learning to recognize spectrally reduced speech: I. The effects of multiple repetitions

In a previous study [Chiu et al., J. Acoust. Soc. Am. 109, 2501 (2001)], normal listerers learned open-set recognition of words in Harvard IEEE sentences that were spectrally reduced to 4 channels using the CIS processing strategy [Shannon et al., J. Acoust. Soc. Am. 104, 2467–2476 (1998)]. Half of the listeners listened to a male talker during training, the other half a female talker. At different points in the training listeners were tested with blocks of IEEE sentences spoken by the familiar talker, blocks of IEEE sentences by the novel talker, blocks of IEEE sentences of the 2 talkers randomly intermixed, and blocks of HINT sentences. In that study, listeners could listen to each sentence during early training for an unrestricted number of times before they were given feedback. In the current study, listeners were allowed to listen to each sentence only once during training before feedback was given. Preliminary data obtained from the current study were largely comparable to those obtained in the previous study. The results suggest that massed repetitions and drills, at least in the context of our experimental conditions, wa have minimal effect on learning to recognize spectrally reduced speech. [Work supported by URC, Univ. of Cincinnati.]

The perception of speed-modified recordings

The overall quality of audio tapes may be affected by such factors as background noise and the intensity of the signal, or whether the recording was made at the correct speed. The last would affect the speaking fundamental frequency (SFF), and therefore perceived pitch, and phonetic characteristics of the speech sample. Listeners’ skills regarding the recognition of recording speeds that had been manipulated are investigated here. It was found that a majority of correct identifications of speed manipulation is only achieved when the speed is –6% or +6% off the correct speed in an experimental procedure that compared deviations of 3, 6 and 9%. Judging samples that are played at higher speeds is easier than those played at lower speeds. Expert listeners did not perform any better than naive or trained listeners. However, they were more cautious and were more often correct when very confident about their judgments.

The Relationship Between Cepstral Peak Prominence and Selected Parameters of Dysphonia

Traditional measures of dysphonia vary in their reliability and in their correlations with perceptions of grade. Measurements of cepstral peak prominence (CPP) have been shown to correlate well with perceptions of breathiness. Because it is a measure of periodicity, CPP should also predict roughness. The ability of CPP and other acoustic measures to predict overall dysphonia and the subcategories of breathiness and roughness in pathological voice samples is explored. Preoperative and postoperative speech samples from 19 patients with unilateral recurrent laryngeal nerve paralysis who underwent operative intervention were analyzed by trained listeners and by measures of smoothed CPP (CPPS), noise-to-harmonic ratio (NHR), amplitude perturbation quotient (APQ), relative average perturbation (RAP), and smoothed pitch perturbation quotient (sPPQ). The data were analyzed with bivariate Pearson correlation statistics. Grade of dysphonia and breathiness ratings correlated better with measurements of CPPS than with the other measures. CPPS from samples of connected speech (CPPS-s) best predicted overall dysphonia. None of the measures were useful in predicting roughness.

Relationships between acoustic and perceptual measures of stress in nonwords produced by children with developmental apraxia of speech and children with phonological disorders

This study investigated relationships between acoustic measures of stress in nonwords produced by children with suspected developmental apraxia of speech (sDAS) and children with phonological disorders (PD), and perceptual judgments made by trained listeners. Five children with sDAS and five children with PD produced multiple tokens of eight disyllabic nonwords with iambic and trochaic stress patterns. Measures of vowel duration, f0, and intensity were made. Children from both groups produced acoustic differences between stressed and stressless syllables. In a perception experiment, 10 listeners judged whether tokens were produced with stress on the initial or final syllable. Listeners judged that children with sDAS accurately matched target stress contours less often than children with PD. The proportion of listeners judging that a word was produced with final syllable stress was used as the dependant variable in a regression, in which the ratio of vowel duration, f0, and intensity between the two syllables served as the independent variables. All three variables predicted a significant proportion of variance in listener judgments, although the overall proportion of variance accounted for was low (R2=0.212). Thus, children with sDAS are able to produce contrasts between stressed and stressless syllables, but these are often imperceptible to listeners.

Training /v/–/w/ perception with native speakers of Indian languages

Experiments in training listeners to perceive non-native categories are an important aspect of cross-language perception research. In this study, ten native speakers of various Indian languages participated in training designed to improve perception of the English /v/–/w/ contrast using a fading technique (Jamieson and Morosan, 1986). First the subjects labeled English /p, b, f, v, w/ using native language (L1) orthography to ensure that other L1 consonants were not evoked by acoustic cues to English /v/, /w/. They also rated the closeness of the English sound they heard to the sound in their L1. In the five-day training, subjects heard natural stimuli produced by four (two males and two females) English speakers. The steady-state portion of the initial consonant was reiterated by varying amounts with subjects hearing progressively shorter consonants as the training progressed. Both discrimination and identification training with feedback were used. Pre- and post-training testing with stimuli from new speakers as well as recording of production of Indian subjects was completed. Results and implications will be discussed.

Specificity of learning in an auditory temporal-order task

The ability to determine the onset order of two tones at different frequencies is a measure of temporal acuity and an analog of voice-onset time. Learning in this temporal-order task and its generalization to untrained frequencies and temporal-acuity tasks was explored. Temporal-order thresholds of six listeners trained 1 h per day for 6 days with tones at 0.25 and 4 kHz improved from 58 ms before to 12 ms after training, yielding a proportional improvement greater than on all untrained conditions. Comparisons between proportional improvements of trained and control listeners revealed that trained listeners learned considerably more than controls on the trained condition, and only slightly or no more than controls on: (1) the trained task at two sets of untrained frequencies (0.5 and 1.5 kHz, 0.75 and 1.25 kHz), and (2) three untrained temporal-acuity tasks (offset temporal order, onset asynchrony, offset asynchrony) at the trained frequencies. The lack of robust generalization to untrained frequencies and tasks suggests learning in onset temporal-order tasks is mediated by a frequency-dependent mechanism different from the mechanism(s) underlying performance in other temporal-acuity tasks. By increasing our understanding of plasticity in temporal-acuity mechanisms, these data may guide the treatment of related disorders. [Work supported by NIDCD.]

Intelligibility/Severity Measurements of Prekindergarten Children’s Speech

Intelligibility/severity measurements were obtained for 48 prekindergarten children with varying levels of phonological proficiency/ deficiency. The measure used as the “standard” was percentage of words understood (i.e., orthographically transcribed correctly) in continuous speech in a known context by unfamiliar trained listeners. The children were divided into four groups based on the percentage of words understood from their continuous speech samples. The ranges of intelligibility for each group were: (a) 91–100% for children with “adult-like” speech; (b) 83–90% for children in the “mild” category; (c) 68–81% for children with moderate intelligibility/speech involvement; and (d) 16–63% for the 12 children in the “severe” (i.e., least intelligible) category. When the percentages of the children in the severe group were excluded, the range of the top three groups combined was 68–100% and the mean was 85%. For a child 4 years of age or older, any percentage of words understood in connected speech that falls below 66% (2 standard deviations below the mean) may be a potential indicator of speech difficulty. In addition, data were obtained from the 48 children to determine the correlations between the standard measure and the following intelligibility/severity measures: (a) imitated sentences, (b) imitated words, (c) listener ratings of intelligibility, and (d) phonological deviation averages. All five measures, including the standard measure, investigated in this study were strongly intercorrelated. Multiple regression analysis results yielded a prediction model that included listener ratings and imitated sentences measures. Results of multivariate analysis of variance (MANOVA), univariate analysis, and post-hoc Bonferroni tests indicated that differences between all pairs of groups were significant for the listener rating measure based on the continuous speech sample. For the percentage of words understood in continuous speech samples, the differences between all pairs of groups, except between the adult-like and mild groups, were also significant. The only group that differed significantly from the other three groups for all five measures was the severe group.

Diplophonia reappraised.

The acoustic structure of diplophonia was investigated spectrographically and in terms of perturbation measures, while the perception of diplophonia among other voice qualities was confirmed auditorily by trained listeners. Recordings of pathological voices were presented to listeners for systematic evaluation, and narrowband spectrography was used to quantify the subharmonics within each sample. Subharmonics strongly correlated with the perception of diplophonia, but occurrence was mostly intermittent, and structure was highly variable between samples. Uniquely among voice qualities identified perceptually, diplophonia was found to correlate positively with the number of subharmonics (irrespective of percentage of occurrence) and with perturbation parameters measured separately. Exceptions to these group results indicated that diplophonia was perceived sometimes in the absence of subharmonic structure, and subharmonic structure was observed without a commensurate perception of diplophonia. In light of these data, a less deterministic relation between diplophonia and subharmonic structure is proposed.

Learning to discriminate reduced-channel /bɑ/–/wɑ/ syllables in normal-hearing adults

Acoustic information delivered to individuals with cochlear implants differs in many ways from the same information delivered via a normally functioning cochlea. One difference is that the number of frequencies that can be distinguished is sharply reduced in implanted listeners. This results in poor frequency selectivity where only sounds that fall into different frequency bands are distinguishable (all other things being equal). Shannon [Science 270, 303–304 (1995)] used an electrical filtering and rectification system to simulate this aspect of cochlear implant function in normal-hearing listeners. In the present study software filtering was combined with envelope-shaped-noise (ESN) processing to accomplish the same end. The immediate goal was to examine the efficacy of a simple technique in training listeners to distinguish between reduced-channel versions of syllables. The stimuli tested were synthetically produced /bɑ/ and /wɑ/ syllables, identical in every aspect except for their onset duration. These syllables were filtered into eight frequency bands and ESN processed. Following this they were once-again filtered at the same frequencies to create reduced-channel versions of the original syllables. The resulting stimuli were presented to 16 normal-hearing adult subjects in behavioral and electrophysiological tasks. Results demonstrated that the training rapidly improved listeners’ abilities to identify reduced-channel /bɑ/ and /wɑ/.

Acoustic and perceptual evaluation of laryngeal reinnervation by ansa cervicalis transfer.

To evaluate the acoustic and perceptual results of laryngeal reinnervation with ansa cervicalis to recurrent laryngeal nerve anastomosis. Retrospective study of voice samples from 12 patients with unilateral recurrent laryngeal nerve paralysis, treated with ansa cervicalis to recurrent laryngeal nerve anastomosis. Samples were recorded before surgery and at least 8 months after surgery. The samples were subjected to several acoustic analyses sensitive to paralytic dysphonia, including cepstral peak prominence, noise-to-harmonics ratio, and measures of frequency and amplitude perturbation. The voice samples from the patients were randomized with age- and sex-matched samples from normal subjects and judged by trained listeners for overall dysphonia, roughness, breathiness, asthenia, and strain. The preoperative and postoperative results were compared statistically, and the postoperative results were compared with the matched normal subjects. As a group, the patients showed improvement (P < .05) in cepstral peak prominence, frequency perturbation, and perceptual judgments of overall dysphonia, breathiness, and asthenia. The best results occurred in patients with isolated vocal fold paralysis. The postoperative group as a whole did not improve to the level seen in matched normals. Suboptimal results were seen primarily in patients with untreated laryngeal or extralaryngeal pathology beyond the laryngeal paralysis. These data suggest that laryngeal reinnervation has the potential to bring about a return to normal or near-normal voice in patients with isolated unilateral vocal fold paralysis.

Subjective evaluations of professional microphones

Attempts to evaluate microphone performance by listening to individual models via a reproduction chain often reveal subtle, and not-so-subtle, ‘‘sonic signatures’’ which in turn reveal preferences and dislikes among trained listeners. However, there are many factors contributing to the process of evaluation, even if only one microphone type, and only one listener, are involved: (i) monaural versus stereophonic versus ambisonic reproduction; (ii) headphone versus loudspeaker auditioning (again, mono or stereo?); (iii) choice of music, if indeed listening is limited to music; (iv) ‘‘direct feed’’ versus analog versus digital reproduction; (v) personal and environmental factors; (vi) the enhancing or detrimental aspects of microphone ‘‘coloration.’’ The foregoing partial list has prompted this listener to treat AB and ABX comparisons of microphones, by individuals or groups, with a cetain level of suspicion—in spite of many tests he has conducted involving others. Therefore this presentation will indeed offer subjective views: those of one person with a 35-year history of recording and listening, and with an even longer history of formulating a personal musical taste.

The effect of transition slope and transition duration on vowel reduction in V1V2 complexes

Digital V1V2 vowel pairs were generated using a cascade-parallel formant synthesizer. The duration of V1 was 100 ms and that of a completed V2 150 ms, while the duration of the transition was varied between 40 and 100 ms, depending on the condition. For each of six English V1V2 pairs (/iɑ/, /ɑi/, /iu/, /ui/, /uɑ/, and /ɑu/) 8–12 tokens were obtained by progressively cutting back the V2 portion and the vowel transition in 10-ms steps. Separately for each vowel pair, the cut-back tokens were presented to trained listeners who had to indicate their confidence level (using a 4-point scale) that the final portion of the token was, indeed, the vowel V2. Since high-confidence V2 responses were given not only to tokens that actually reached V2, we observed vowel reduction; the degree of vowel reduction (VR) was measured as the severity of the cut-back at which such a misperception occurred at a certain percent of the time. Using this measure, we determined the degree of VR for different durations and velocities of transition. In general, VR was larger for steeper transition slopes and for durations around 80–100 ms. [Work supported by NATO Office of Scientific Research, a grant from N.I.H., and the V.A. Medical Research.]

Mark DeBellis. Music and Conceptualization. Cambridge: Cambridge University Press, 1995. viii, 163 pp. ISBN 0-521-40331-6 (hardcover)

Mark DeBellis. Music and Conceptualization. Cambridge: Cambridge University Press, 1995. viii, 163 pp. ISBN 0-521-40331-6 (hardcover). Un article de la revue Canadian University Music Review / Revue de musique des universités canadiennes (Volume 18, numéro 2, 1998, p. 1-134) diffusée par la plateforme Érudit.

The ability of right- and left-hemisphere-damaged individuals to produce and interpret prosodic cues marking phrasal boundaries.

Two experiments were conducted with the purpose of investigating the ability of right- and left-hemisphere-damaged individuals to produce and perceive the acoustic correlates to phrase boundaries. In the production experiment, the utterance pink and black and green was elicited in three different conditions corresponding to different arrangements of colored squares. Acoustic analyses revealed that both left- and right-hemisphere-damaged patients exhibited fewer of the expected acoustic patterns in their productions than did normal control subjects. The reduction in acoustic cues to phrase boundaries in the utterances of both patient groups was perceptually salient to three trained listeners. The perception experiment demonstrated a significant impairment in the ability of both left-hemisphere-damaged and right-hemisphere-damaged individuals to perceive phrasal groupings. Results are discussed in relation to current hypotheses concerning the cerebral lateralization of speech prosody.

An Algorithm for Harmonic Analysis

An algorithm is proposed for performing harmonic analysis of tonal music. The algorithm begins with a representation of a piece as pitches and durations; it generates a representation in which the piece is divided into segments labeled with roots. This is a project of psychological interest, because much evidence exists that harmonic analysis is performed by trained and untrained listeners during listening; however, the perspective of the current project is computational rather than psychological, simply examining what has to be done computationally to produce "correct" analyses for pieces. One of the major innovations of the project is that pitches and chords are both represented on a spatial representation known as the "line of fifths"; this is similar to the circle of fifths except that distinctions are made between different spellings of the same pitch class. The algorithm uses preference rules to evaluate different possible interpretations, selecting the interpretation that most satisfies the preference rules. The algorithm has been computationally implemented; examples of the program's output are given and discussed.