Search

We present a sociophonetic, acoustic, and articulatory analysis of coda rhoticity in East Lancashire, North-West England. We analysed data from 24 participants aged 8–73 recorded at a public engagement event in Blackburn Market (598 tokens). Auditory analysis shows coda rhoticity is declining across generations, with speakers born after 1990 being mostly non-rhotic. Audible rhoticity is realised by lowered F3 and raised F2. GAMMs fitted across the vowel(+rhoticity) interval show that audibly rhotic tokens have a significantly smaller distance between F3 and F2 than audibly non-rhotic tokens in all vowel contexts. Our ultrasound analysis compares minimal pairs e.g. ‘core’ and ‘caw’. Principal Component Analysis of tongue splines shows that speakers use different tongue shapes in auditorily rhotic tokens.

It is widely accepted in forensic speech science that formants have speaker discriminatory power, and therefore formant measurements commonly feature in forensic speech casework. Courts in Northern Ireland even go as far as to insist that formant analysis has to be carried out as part of forensic voice comparison analysis. However, work in the broader phonetics field has started to challenge the weight that has been traditionally attached to formants in a range of related subdisciplines – this justifies a review of the use and performance of formants in forensic speech science research studies. This paper therefore presents part of a meta-analysis of forensic speech science studies which test formants as a parameter for speaker discrimination. The results are highly variable across the 277 results from the 37 papers included. Some performance trends that might be expected are not evident from this meta-analysis.

Dynamic tongue movements of intervocalic English and Japanese liquids are analysed based on ultrasound data obtained from 17 L1 Japanese and 12 L1 North American English speakers. Principal Component Analysis (PCA) identified three key articulatory properties, including variation in tongue retraction and height as well as tongue tip movement. The time-varying changes showed different magnitude and timing for tongue retraction for Japanese speakers compared to English speakers. Substitution of English liquids with Japanese /r/ was not clearly observed in the tongue movement pattern. The findings highlight the complexity involved in the articulation of English liquids in L2 speech and the usefulness of the finer-grained articulatory analysis in understanding the particular challenges L2 learners face in producing English liquids in L2 speech.

We outline a method for collecting simultaneous high-speed ultrasound and electromagnetic articulography (EMA) data using widely-available commercial hardware and software. Previous research demonstrates the utility of a simultaneous EMA-ultrasound set-up and reports the magnitude of probe rotation and displacement over an experimental session. We build upon this research by (1) combining systems with higher temporal and spatial accuracy than those previously used; (2) using a more principled method for temporal synchronisation; (3) reporting the effects of ultrasound on the accuracy of EMA tracking; (4) tracking probe movement using an alternative stabilisation method. Our system shows a high degree of temporal and spatial accuracy and can be easily implemented by other researchers.

This study examines jaw and lip aperture characteristics of syllables. High correlations were found between velocity and displacement of opening/closing of lips and jaw, similar to previous studies. However, lip aperture opening velocity changed as a function of manner and voicing of syllable onset while jaw opening velocity, except for a plosive onset, did not change. In addition, differences were found in timing between jaw opening and lip aperture opening, and the amount of difference varied as a function of syllable type (open/closed) and voicing/manner of the consonant. Our interpretation of the asynchronous patterns of lip and jaw articulation is that the jaw is the syllable articulator and the lips are the syllable onset/coda articulators. The findings of this study have application to language acquisition as well as to stuttering research.

This paper reports a cross linguistic study that compares the coarticulation patterns between consonant and vowel (CV) in Mandarin Chinese and Southern Swedish. Kinematic data were collected using the Electromagnetic Articulography (EMA) for both languages and were subjected to three types of CV time lag measurement, based on more or less equivalent landmarks on lips and tongue, and partially adopted in previous studies [1, 2, 3]. We found rather consistent CV coordination patterns in these two typologically different languages with both the velocity-based and the acceleration-based measurements on the lips and the tongue body. The most striking result to emerge from the data is the same effect of gender on the variation of CV coarticulation in both languages, which has not been reported previously. In addition, only when gender was added as a factor, did we find the language differences on the CV time lags.

We investigated the auditory-perceptual space across a substantial portion of the human vocal range (2201046 Hz) using multidimensional scaling analysis of cochlea-scaled spectra from 250-ms vowel segments. Three native German speakers produced vowels /i y e ø ɛ a o u/ (N=240) spanning a wide range of their respective voice frequency ranges, which were subsequently identified in a closed-set task by five phonetically trained listeners (>80% correct response). Our findings revealed systematic spectral alterations associated with vowel height and frontness as fundamental frequency increased and a distinct clustering around /i a u/ above 523 Hz. These results emphasize the importance of the overall spectral shape in vowel perception and demonstrate listeners' utilization of acoustic anchors at higher pitches. Moreover, this study sheds light on the quantal nature of these vowels and their potential impact on language evolution, offering a possible explanation for their widespread presence in the world’s languages.

Human listeners have a remarkable ability to recognize speakers by their voice, but within-speaker voice variability through different speaking styles, for example, can reduce recognition performance. In this study, we investigated voice discrimination across speaking styles in Persian. One hundred and fortythree naïve Persian listeners were asked to decide whether pairs of style-matched utterances in childdirected, spontaneous, read and clear speech originated from the same or different speakers. Listeners’ performance across speaking styles was examined using the bias-free sensitivity measure A’, and the bias measure b’’D. Results showed that listeners performed accurately across all speaking styles but more poorly in child-directed speech. They had a bias toward responding ‘different’ regardless of the speaking style, thus suggesting a more general difficulty telling people together than apart

This paper presents an acoustic analysis of the variability in stress group patterns, i.e. the fundamental frequency contour associated with stressed and following unstressed syllables, in Copenhagen Danish with a particular focus on the register known as the Copenhagen multiethnolect. This is a register of Danish spoken by multilingual adolescents. It is prosodically distinct from the register used by monolingual peers. Previous work has shown a difference in vowel durations between the two registers with a near-neutralisation of the vowel length contrast in the multiethnolect. Recordings with 4 speakers from a larger corpus of map tasks containing trisyllabic words with initial stress as place names have been analyzed. All speakers show an expected rise in pitch from the stressed to the first posttonic syllable, but the rise is significantly less steep for speakers of the multiethnolect, which may contribute to the impression of a more syllable-timed rhythm.

In this study we propose a new tool for the perceptual assessment of voice quality. For its design, we have adapted the Simplified Vocal Profile Analysis so that the new tool features two main characteristics: (1) the ordinal scalar degrees of the original protocol are turned into a visual analog scale; and (2) the original paper-based version of the protocol is now a computer-based implementation. In order to assess the reliability of the new tool, five phoneticians listened to 12 different speakers and evaluated their voice quality using the proposed tool. Inter-rater agreement was then calculated using the Intraclass Correlation Coefficient. The results show that high agreement was reached for most of the perceptual settings of the protocol. Yet more investigations seem necessary into the continuous nature of the perceptual dimensions making up the voice quality of a speaker. As a preliminary approach to the graphical possibilities that the visual analog scale offers to the Simplified Vocal Profile Analysis, we explore the usefulness of multiple dot plots and propose an adaptation of the Bland-Altman plot to be used in pairwise comparisons. In this study, these visualization techniques are tested on two pairs of identical twins.