Listener Ratings Research Articles

Gender typicality in children's speech, reconsidered: Effects of stimulus composition on listener judgments

Listeners rate the speech of boys and girls as young as four years old as sounding gendered: boys are rated as sounding boy-like and girls as girl-like (Perry et al., J. Acoust. Soc. Am. [2001]). Recent research found that the extent to which boys’ speech sounds boy-like is correlated with measures of their gender identity and expression (Li et al., J. Phonetics [2016], Munson et al., J. Acoust. Soc. Am. [2015]). Munson et al. found that boys with a diagnosis of gender identity disorder [GID] were rated as sounding less boy-like than boys without GID. Munson et al.’s experiment used only a small number of girls’ productions as filler items. The current study examined listeners’ ratings of the gender typicality of speech of boys with GID and both boys and girls without GID. Significant differences in sex-typicality ratings were found between the two groups of boys. Boys with GID elicited ratings intermediate to those for boys and girls without GID. However, the differences between boys with and without GID were much smaller than those in Munson et al., suggesting that the sex distribution in the stimulus set can affect ratings of the sex typicality of children’s voices.

Abstract WP420: Lesions Associated With Impaired Expression of Emotion Right Hemisphere Stroke

Introduction: Impaired expression of emotion through tone, loudness, rate, and rhythm of speech (affective prosody) is common after right hemisphere (RH) stroke and impedes all social interactions. Hypothesis: Abnormal patterns of acoustic features correlate with listener judgment of affective prosody and are associated with infarcts of specific RH gray and white matter regions. Methods: Forty acute ischemic RH stroke patients had MRI DWI and described a picture within 48 hours of onset. Affective prosody of picture descriptions was rated by 21 healthy volunteers (1-7 scale). We identified acoustic features that correlated with listener ratings with Spearman correlations and with linear regression to identify independent predictors. A neurologist identified percent damage to each of 9 regions after co-registration of DWI to an atlas. We identified lesions associated with abnormalities in important acoustic features using logistic regression. Results: Listener ratings of prosody correlated with duration of voiced segments divided by total duration of speech segments(DurV/Dur) (rho=0.63; p<0.00001) and Spectral Flatness (SF) (rho= -0.55; p=0.0002). Variability in listener rating was best accounted for by DurV/Dur, SF, pitch range, and fundamental frequency coefficient of variation (CoV) in both women (F (4,12) =6.58; p=0.0048) and men (F(5,18)=5.13; p=0.0056). The only feature independently associated with prosody rating after adjusting for other variables was DurV/Dur (p<0.0001) for women, and SF (p=0.007) for men. Low SF was predicted by sex, age, lesion volume, and percent damage to the 9 RH regions (X 2 =27; p=0.0081). Sex, lesion volume, damage to inferior frontal gyrus pars opercularis, inferior fronto-occipital fasciculus, superior longitudinal fasciculus (SLF), and uncinate were the only independent predictors. The same variables predicted CoV (X 2 =33; p=0.0005); age and damage to supramarginal gyrus and SLF were the only independent predictors. The same variables plus education predicted low DurV/Dur (X 2 =25; p=0.02). Conclusions: Impaired emotional expression after RH stroke is due to changes in specific acoustic features caused by infarct in right pars opercularis or supramarginal gyrus, or associated white matter tracts.

Deriving loudness growth functions from categorical loudness scaling data.

The goal of this study was to reconcile the differences between measures of loudness obtained with continuous, unbounded scaling procedures, such as magnitude estimation and production, and those obtained using a limited number of discrete categories, such as categorical loudness scaling (CLS). The former procedures yield data with ratio properties, but some listeners find it difficult to generate numbers proportional to loudness and the numbers cannot be compared across listeners to explore individual differences. CLS, where listeners rate loudness on a verbal scale, is an easier task, but the numerical values or categorical units (CUs) assigned to the points on the scale are not proportional to loudness. Sufficient CLS data are now available to assign values in sones, a scale proportional to loudness, to the loudness categories. As a demonstration of this approach, data from Heeren, Hohmann, Appell, and Verhey [J. Acoust. Soc. Am. 133, EL314-EL319 (2013)] were used to develop a CUsone metric, whose values were then substituted for the original CU values in reanalysis of a large set of CLS data obtained by Rasetshwane, Trevino, Gombert, Liebig-Trehearn, Kopun, Jesteadt, Neely, and Gorga [J. Acoust. Soc. Am. 137, 1899-1913 (2015)]. The resulting data are well fitted by power functions and are in general agreement with previously published results obtained with magnitude estimation, magnitude production, and cross modality matching.

Efficacy of Electropalatography for Treating Misarticulation of /r/.

The purpose of the present study was to document the efficacy of electropalatography (EPG) for the treatment of rhotic errors in school-age children. Despite a growing body of literature using EPG for the treatment of speech sound errors, there is little systematic evidence about the relative efficacy of EPG for rhotic errors. Participants were 5 English-speaking children aged 6;10 to 9;10, who produced /r/ at the word level with < 30% accuracy but otherwise showed typical speech, language, and hearing abilities. Therapy was delivered in twice-weekly 30-min sessions for 8 weeks. Four out of 5 participants were successful in achieving perceptually and acoustically accurate /r/ productions during within-treatment trials. Two participants demonstrated generalization of /r/ productions to nontreated targets, per blinded listener ratings. The present findings support the hypothesis that EPG can improve production accuracy in some children with rhotic errors. However, the utility of EPG is likely to remain variable across individuals. For rhotics, EPG training emphasizes one possible tongue configuration consistent with accurate rhotic production (lateral tongue contact). Although some speakers respond well to this cue, the narrow focus may limit lingual exploration of other acceptable tongue shapes known to facilitate rhotic productions.

Discrete ratings and dimensional judgments of emotional speech: A preliminary look at gender differences

Two models of emotion identification and discrimination are not often compared and may show differing trends in how listeners rate emotional stimuli. The discrete or “basic” emotion model posits that emotions are categorical, and that listeners develop auditory prototypes of “basic” emotions (e.g., anger) based on their acoustic profile and past experiences. More complex emotions (e.g., frustration) fall under a “basic” category. The dimensional model examines emotions along continua of different emotional dimensions, such as activation/arousal and pleasantness/valence. The present study introduces the Morgan Emotional Speech Set and examines listener judgments of the stimuli in the corpus. The database consists of 2160 emotional speech sentences (90 sentences x 4 emotions x 6 talkers) produced by three male and three female actors for use in future studies. Each sentence was rated by 10 listeners (5 male, 5 female), who assigned an emotion category to each sentence and also rated each sentence by its activation (low to high) and pleasantness (very unpleasant to very pleasant). Discrete ratings will be compared with dimensional judgments made by listeners to examine intended and perceived emotional content in the database. A preliminary look at gender differences in the ratings and judgments will also be discussed, as male and female listeners may rate emotions differently for same- or different-sex talkers.

Combined Electric and Acoustic Stimulation With Hearing Preservation: Effect of Cochlear Implant Low-Frequency Cutoff on Speech Understanding and Perceived Listening Difficulty.

The primary objective of this study was to assess the effect of electric and acoustic overlap for speech understanding in typical listening conditions using semidiffuse noise. This study used a within-subjects, repeated measures design including 11 experienced adult implant recipients (13 ears) with functional residual hearing in the implanted and nonimplanted ear. The aided acoustic bandwidth was fixed and the low-frequency cutoff for the cochlear implant (CI) was varied systematically. Assessments were completed in the R-SPACE sound-simulation system which includes a semidiffuse restaurant noise originating from eight loudspeakers placed circumferentially about the subject's head. AzBio sentences were presented at 67 dBA with signal to noise ratio varying between +10 and 0 dB determined individually to yield approximately 50 to 60% correct for the CI-alone condition with full CI bandwidth. Listening conditions for all subjects included CI alone, bimodal (CI + contralateral hearing aid), and bilateral-aided electric and acoustic stimulation (EAS; CI + bilateral hearing aid). Low-frequency cutoffs both below and above the original "clinical software recommendation" frequency were tested for all patients, in all conditions. Subjects estimated listening difficulty for all conditions using listener ratings based on a visual analog scale. Three primary findings were that (1) there was statistically significant benefit of preserved acoustic hearing in the implanted ear for most overlap conditions, (2) the default clinical software recommendation rarely yielded the highest level of speech recognition (1 of 13 ears), and (3) greater EAS overlap than that provided by the clinical recommendation yielded significant improvements in speech understanding. For standard-electrode CI recipients with preserved hearing, spectral overlap of acoustic and electric stimuli yielded significantly better speech understanding and less listening effort in a laboratory-based, restaurant-noise simulation. In conclusion, EAS patients may derive more benefit from greater acoustic and electric overlap than given in current software fitting recommendations, which are based solely on audiometric threshold. These data have larger scientific implications, as previous studies may not have assessed outcomes with optimized EAS parameters, thereby underestimating the benefit afforded by hearing preservation.

LINGUISTIC DIMENSIONS OF L2 ACCENTEDNESS AND COMPREHENSIBILITY VARY ACROSS SPEAKING TASKS

AbstractThis study critically examined the previously reported partial independence between second language (L2) accentedness (degree to which L2 speech differs from the target variety) and comprehensibility (ease of understanding). In prior work, comprehensibility was linked to multiple linguistic dimensions of L2 speech (phonology, fluency, lexis, grammar) whereas accentedness was narrowly associated with L2 phonology. However, these findings stemmed from a single task (picture narrative), suggesting that task type could affect the particular linguistic measures distinguishing comprehensibility from accentedness. To address this limitation, speech ratings of 10 native listeners assessing 60 speakers of L2 English in three tasks (picture narrative, IELTS, TOEFL) were analyzed, targeting two global ratings (accentedness, comprehensibility) and 10 linguistic measures (segmental and word stress accuracy, intonation, rhythm, speech rate, grammatical accuracy and complexity, lexical richness and complexity, discourse richness). Linguistic distinctions between accentedness and comprehensibility were less pronounced in the cognitively complex task (TOEFL), with overlapping sets of phonology, lexis, and grammar variables contributing to listener ratings of accentedness and comprehensibility. This finding points to multifaceted, task-specific relationships between these two constructs.

Linguistic Dimensions of Accentedness and Comprehensibility: Exploring Task and Listener Effects in Second Language French

AbstractThis study explored the effects of task and listeners’ linguistic background on judgments of accentedness and comprehensibility in second language (L2) French. Forty Spanish speakers of L2 French recorded a picture narrative and answered an interview question. These audio samples were assessed by native French listeners with (n = 10) and without (n = 10) knowledge of L2 Spanish using 1,000‐point sliding scales to evaluate accentedness and comprehensibility, as well as nine linguistic variables targeting pronunciation, fluency, lexis, grammar, and discourse. In the picture narrative task (cognitively less complex), accentedness could be distinguished from comprehensibility at the level of the linguistic dimensions associated with each construct. While accentedness was related to pronunciation and fluency (consonant and vowel errors, intonation, speech rate), comprehensibility was additionally linked to lexis (accuracy, richness), grammar (accuracy, complexity), and discourse richness. However, in the interview task (cognitively more complex), both accentedness and comprehensibility had similar linguistic profiles, associated with all nine linguistic variables. Listeners’ experience with L2 Spanish had little impact on listener ratings of accentedness or comprehensibility in both tasks. Findings are discussed in relation to L2 instruction and the role of task complexity in targeting various dimensions of L2 accentedness and comprehensibility.

Perception of isolated chords: Examining frequency of occurrence, instrumental timbre, acoustic descriptors and musical training

This study investigated the perception of isolated chords using a combination of experimental manipulation and exploratory analysis. Twelve types of chord (five triads and seven tetrads) were presented in two instrumental timbres (piano and organ) to listeners who rated the chords for consonance, pleasantness, stability and relaxation. Listener ratings varied by chord, by timbre, and according to musical expertise, and revealed that musicians distinguished consonance from the other variables in a way that other listeners did not. To further explain the data, a principal component analysis and linear regression examined three potential predictors of the listener ratings. First, each chord’s frequency of occurrence was obtained by counting its appearances in selected works of music. Second, listeners rated their familiarity with the instrumental timbre in which the chord was played. Third, chords were described using a set of acoustic features derived using the Timbre Toolbox and MIR Toolbox. Results of the study indicated that listeners’ ratings of both consonance and stability were influenced by the degree of musical training and knowledge of tonal hierarchy. Listeners’ ratings of pleasantness and relaxation, on the other hand, depended more on the instrumental timbre and other acoustic descriptions of the chord.

Comparison of Two Multiparameter Acoustic Indices of Dysphonia Severity: The Acoustic Voice Quality Index and Cepstral Spectral Index of Dysphonia

The Acoustic Voice Quality Index (AVQI) and the Cepstral Spectral Index of Dysphonia (CSID) are two multiparameter acoustic indices designed to objectively estimate dysphonia severity and track treatment outcomes. This study compared the performance of these two indices using a common corpus of dysphonic speakers. Pre- and posttreatment samples of sustained vowel and connected speech were elicited from 112 patients across six diagnostic categories: unilateral vocal fold paralysis (n = 20), adductor spasmodic dysphonia (n = 20), primary muscle tension dysphonia (n = 20), benign vocal fold lesions (n = 20), presbylaryngis (n = 20), and mutational falsetto (n = 12). Listener ratings of dysphonia severity were compared to acoustic estimates of severity derived from two iterations of the AVQI (versions 2.02 and 3.01) as well as the CSID. The AVQI- and CSID-estimated severity for sustained vowels, connected speech, and a combined context were strongly correlated and significantly associated with listener ratings pretreatment, posttreatment, and change observed pre- to posttreatment. However, multiple regression analysis (adjusted for age, sex, and diagnostic category) revealed that the CSID generally accounted for more variance in listener-perceived severity ratings, and the contribution of the AVQI was small and statistically insignificant when the CSID was already in a combined model. The AVQI and the CSID were strongly correlated and both provided valid estimates of dysphonia severity. However, associations observed between the CSID- and listener-estimated dysphonia were almost uniformly stronger than either version of the AVQI, suggesting that the CSID outperformed the AVQI.

Perceptual Anchors and the Dispersion of Nasality Ratings.

This investigation studied the effects of perceptual anchors on the dispersion and reliability of listener ratings of nasality. Listeners (N = 129) were assigned to one of six listening groups. Each group rated nasality independently for 100 speech samples on a seven-point scale that ranged from 1 = normal nasality to 7 = severe hypernasality. The anchors used were examples of a 1, 3, 4, 5, and/or 7 on the rating scale. These anchors were played selectively to group 2 (4), group 3 (1 and 7), group 4 (3 and 5), group 5 (1, 4, 7), and group 6 (7). Group 1 had no anchor. Of the speakers, 95 were children followed by a craniofacial team and five were children without histories of speech disorders. The outcome measures were 12,900 ratings of nasality on a seven-point scale. Q values showed that group 5, which was the only group to receive three anchors, had the lowest, or best, Q value (0.78), and group 1 (no anchor) had the highest, or worst, Q value (0.99). Across groups, the most reliable ratings were those at scale values 1 (Q = 0.46) and 7 (Q = 0.56). The least reliable ratings were at scale values 3 (Q = 1.01), 4 (Q = 1.03), and 5 (Q = 1.06). Nasality rating reliability/dispersion was influenced by the presence and location of anchor stimuli. Consistent with absolute judgment theory, nasality ratings showed a strong end effect.

Speech Rate Normalization and Phonemic Boundary Perception in Cochlear-Implant Users.

Normal-hearing (NH) listeners rate normalize, temporarily remapping phonemic category boundaries to account for a talker's speech rate. It is unknown if adults who use auditory prostheses called cochlear implants (CI) can rate normalize, as CIs transmit degraded speech signals to the auditory nerve. Ineffective adjustment to rate information could explain some of the variability in this population's speech perception outcomes. Phonemes with manipulated voice-onset-time (VOT) durations were embedded in sentences with different speech rates. Twenty-three CI and 29 NH participants performed a phoneme identification task. NH participants heard the same unprocessed stimuli as the CI participants or stimuli degraded by a sine vocoder, simulating aspects of CI processing. CI participants showed larger rate normalization effects (6.6 ms) than the NH participants (3.7 ms) and had shallower (less reliable) category boundary slopes. NH participants showed similarly shallow slopes when presented acoustically degraded vocoded signals, but an equal or smaller rate effect in response to reductions in available spectral and temporal information. CI participants can rate normalize, despite their degraded speech input, and show a larger rate effect compared to NH participants. CI participants may particularly rely on rate normalization to better maintain perceptual constancy of the speech signal.

Can Hearing Aids Affect Emotional Response to Sound?

Can Hearing Aids Affect Emotional Response to Sound?

The Effect of Uni- and Bilateral Thalamic Deep Brain Stimulation on Speech in Patients With Essential Tremor: Acoustics and Intelligibility

Deep brain stimulation (DBS) of the ventral intermediate nucleus (VIM) is performed to suppress medically-resistant essential tremor (ET). However, stimulation induced dysarthria (SID) is a common side effect, limiting the extent to which tremor can be suppressed. To date, the exact pathogenesis of SID in VIM-DBS treated ET patients is unknown. We investigate the effect of inactivated, uni- and bilateral VIM-DBS on speech production in patients with ET. We employ acoustic measures, tempo, and intelligibility ratings and patient's self-estimated speech to quantify SID, with a focus on comparing bilateral to unilateral stimulation effects and the effect of electrode position on speech. Sixteen German ET patients participated in this study. Each patient was acoustically recorded with DBS-off, unilateral-right-hemispheric-DBS-on, unilateral-left-hemispheric-DBS-on, and bilateral-DBS-on during an oral diadochokinesis task and a read German standard text. To capture the extent of speech impairment, we measured syllable duration and intensity ratio during the DDK task. Naïve listeners rated speech tempo and speech intelligibility of the read text on a 5-point-scale. Patients had to rate their "ability to speak". We found an effect of bilateral compared to unilateral and inactivated stimulation on syllable durations and intensity ratio, as well as on external intelligibility ratings and patients' VAS scores. Additionally, VAS scores are associated with more laterally located active contacts. For speech ratings, we found an effect of syllable duration such that tempo and intelligibility was rated worse for speakers exhibiting greater syllable durations. Our data confirms that SID is more pronounced under bilateral compared to unilateral stimulation. Laterally located electrodes are associated with more severe SID according to patient's self-ratings. We can confirm the relation between diadochokinetic rate and SID in that listener's tempo and intelligibility ratings can be predicted by measured syllable durations from DDK tasks.

音の時系列データに対する聞き手の印象評価結果に基づく音生成システム

We propose a method of generating a sound which is expected to give a listener a good impression. We aim to use the system to help a robot give a person a good impression in interaction between the robot and the person. In the proposed method, an impression of a sound is estimated based on results of a listener's evaluation to a sound. A spectrogram of a sound is used to learn the relationship between a sound and an evaluation. The evaluation of a sound is estimated based on the spectrogram and not based on the criteria of a sound. Therefore, the proposed method is expected to generate unpredictabile sound. The relationship between the value of an evaluated impression and the data of a sound is estimated by using RNN. We conducted some experiments. It was composed of gathering data of evaluation to some sounds, estimating a listerner's evaluation to sound and generating sounds which was expected to give a listener a good impression.

Listeners' and Performers' Shared Understanding of Jazz Improvisations.

This study explores the extent to which a large set of musically experienced listeners share understanding with a performing saxophone-piano duo, and with each other, of what happened in three improvisations on a jazz standard. In an online survey, 239 participants listened to audio recordings of three improvisations and rated their agreement with 24 specific statements that the performers and a jazz-expert commenting listener had made about them. Listeners endorsed statements that the performers had agreed upon significantly more than they endorsed statements that the performers had disagreed upon, even though the statements gave no indication of performers' levels of agreement. The findings show some support for a more-experienced-listeners-understand-more-like-performers hypothesis: Listeners with more jazz experience and with experience playing the performers' instruments endorsed the performers' statements more than did listeners with less jazz experience and experience on different instruments. The findings also strongly support a listeners-as-outsiders hypothesis: Listeners' ratings of the 24 statements were far more likely to cluster with the commenting listener's ratings than with either performer's. But the pattern was not universal; particular listeners even with similar musical backgrounds could interpret the same improvisations radically differently. The evidence demonstrates that it is possible for performers' interpretations to be shared with very few listeners, and that listeners' interpretations about what happened in a musical performance can be far more different from performers' interpretations than performers or other listeners might assume.

Prosodic changes in Parkinson's disease

Parkinson’s disease often results in a hypokinetic dysarthria, causing well-known effects on speech and voice. Speech from individuals with Parkinson’s is typically studied through listening tasks (i.e., by family members and naïve listeners) and is known to affect intelligibility. However, the exact speech changes causing degradations in speech quality have not been defined. Previous investigations have focused on standard measures of mean fundamental frequency, intensity, and rate. We hypothesize that speech changes are present at the prosodic level in addition to at the acoustic level. To test this hypothesis, we collected spontaneous conversational speech samples from 15 individuals with Parkinson’s and 13 age-and-gender matched neurologically healthy speakers (Control group). Several parameters were extracted from the speech signal to describe rate, phrasing, and stress in addition to fundamental frequency and intensity. These measures were compared with listener ratings of intelligibility performed by 12 naïve listeners. Listeners rated the speech samples using a visual analog scale. Preliminary findings suggest that prosody is impaired in Parkinson’s disease. Statistical analysis of variance between the speaker groups (Parkinson’s vs. Controls) are reported in addition to linear regressions with listener ratings. Our findings can facilitate in early diagnosis and the development of targeted speech therapy for individuals with Parkinson’s.

Objective Measure of Nasal Air Emission Using Nasal Accelerometry.

This article describes the development and initial validation of an objective measure of nasal air emission (NAE) using nasal accelerometry. Nasal acceleration and nasal airflow signals were simultaneously recorded while an expert speech language pathologist modeled NAEs at a variety of severity levels. In addition, microphone and nasal accelerometer signals were collected during the production of /pɑpɑpɑpɑ/ speech utterances by 25 children with and without cleft palate. Fourteen inexperienced raters listened to the microphone signals from the pediatric speakers and rated the samples for the severity of NAE using direct magnitude estimation. Mean listener ratings were compared to a novel quantitative measurement of NAE derived from the nasal acceleration signals. Correlation between the nasal acceleration energy measure and the measured nasal airflow was high (r = .87). Correlation between the measure and auditory-perceptual ratings was moderate (r = .49). The measure presented here is quantitative and noninvasive, and the required hardware is inexpensive ($150). Future studies will include speakers with a wider range of NAE severity and etiology, including cleft palate, hearing impairment, or dysarthria. Further development will also involve validation of the measure against airflow measures across subjects.

Variation in the second formant: Its role in a creole's coarse-refined contrast

Baba Malay encodes an interesting coarse-refined contrast. Speakers overtly recognize words ending with -al, -aɾ, and -as (e.g., bakaɾ “burn") as kasar “coarse,” while their counterparts ending with -ɛ (e.g., bakɛ “burn") are halus “refined.” The contrast is neither phonetic nor is it constrained by phonology or morphology. The coarse-refined system may be based on sound symbolism. The frontness of -ɛ is associated with a smaller articulatory space in the oral cavity, and hence refinedness, while the more backwards al, -aɾ, and -as forms, are associated with a larger articulatory space and hence coarseness. A perception experiment, mainly a within-subjects matched guise task, is conducted. Coarse and refined pairs are elicited from speakers. The relevant F2 of both variants are adjusted twice upwards and twice downwards in steps of 100 Hz. Listeners rate these forms on a scale of 1 to 5, 5 being most associated with values that are “refined.” Results show that the lower F2 is, the more likely listeners are to associate the coarse guise with “coarse” values, and the higher F2 is, the more likely listeners are to associate the refined guise with “refined” values. The kasar-haluscontrast is thus mitigated by sound symbolism operationalized by F2.

Human recognition of familiar voices.

Recognizing familiar voices is something we do every day. In quiet environments, it is usually easy to recognize a familiar voice. In noisier environments, this can become a difficult task. This paper examines how robust listeners are at identifying familiar voices in noisy, changing environments and what factors may affect their recognition rates. While there is previous research addressing familiar speaker recognition, the research is limited due to the difficulty in obtaining appropriate data that eliminates speaker dependent traits, such as word choice, along with having corresponding listeners who are familiar with the speakers. The data used in this study were collected in such a fashion to mimic conversational, free-flow dialogue, but in a way to eliminate many variables such as word choice, intonation, or non-verbal cues. These data provide some of the most realistic test scenarios to-date for familiar speaker identification. A pure-tone hearing test was used to separate listeners into normal hearing and hearing impaired groups. It is hypothesized that the results of the Normal Hearing Group will be statistically better. Additionally, the aspect of familiar speaker recognition is addressed by having each listener rate his or her familiarity with each speaker. Two statistical approaches showed that the more familiar a listener is with a speaker, the more likely the listener will recognize the speaker.