Kelly Berkson

Is linguistic injustice a myth? A response to Hyland (2016)

a Faculty of Linguistics, Philology and Phonetics, University of Oxford, Centre for Linguistics, Walton Street, Oxford, OX1 2HG, United Kingdom NYUAD Institute, New York University Abu Dhabi, P.O. Box 129188, Abu Dhabi, United Arab Emirates Department of Korean Language and Literature, Korea University, 145 Anam-ro, Seongbuk-gu, Seoul 02841, South Korea Child Language Doctoral Program, University of Kansas, 1000 Sunnyside Avenue, 3031 Dole, Lawrence, KS 66045, United States Department of German Linguistics, University of Cologne, Albertus Magnus Platz, D-50923 Köln, Germany Department of Linguistics, Indiana University, 322 Memorial Hall, Bloomington, IN 47405, United States

Real-time three-dimensional ultrasound imaging of pre- and post-vocalic liquid consonants in American English

Speech sound articulation is typically characterized in the mid-sagittal plane. However, lateral segments, such as the /l/ category in most varieties of English, are usually identified in introductory phonetics courses as exceptions to this rule. On the other hand, many productions of post-vocalic /l/ in varieties spoken in the lower Mid-west U.S. are noted as being non-lateral, involving only a dorsal articulation very different from the canonical coronal occlusion. Furthermore, a large body of literature indicates multi-constriction articulations, which vary by syllable position, for liquids like American English /l/ and /r/. This research presents results from a study of constriction location and laterality in pre- and post-vocalic /l/ and /r/ using real-time 3D ultrasound images of tongue motion, digitized impressions of the palate, and time-aligned acoustic signals.

Building a multilingual ultrasound corpus

This paper presents a corpus of three dimensional ultrasound data focused on tongue shape during speech sound articulation. Our ultimate goal is to collect data pertinent to phonetic structures from as many languages as possible (at present upwards of 20 languages are represented) and to curate these data in an open access corpus that is freely available for use by other researchers. In this presentation, we review the structure of the corpus and present a series of case studies illustrating the ways in which three-dimensional data are being used to address questions of phonetic interest. Examples of such areas include: examining the articulation of laterals; determining the point of articulation for dorsal consonants; analyzing consonant and vowel coarticulation patterns; and elucidating how coupling of the tongue body with the tongue blade and root affect place and manner of articulation. Like the work of Allard Jongman and Joan Sereno, this project is inherently collaborative and includes ample opportunity for using guided investigation of targeted research questions to ease novice scholars into the research process.This paper presents a corpus of three dimensional ultrasound data focused on tongue shape during speech sound articulation. Our ultimate goal is to collect data pertinent to phonetic structures from as many languages as possible (at present upwards of 20 languages are represented) and to curate these data in an open access corpus that is freely available for use by other researchers. In this presentation, we review the structure of the corpus and present a series of case studies illustrating the ways in which three-dimensional data are being used to address questions of phonetic interest. Examples of such areas include: examining the articulation of laterals; determining the point of articulation for dorsal consonants; analyzing consonant and vowel coarticulation patterns; and elucidating how coupling of the tongue body with the tongue blade and root affect place and manner of articulation. Like the work of Allard Jongman and Joan Sereno, this project is inherently collaborative and includes ample opportu...

Indic and Dravidian retroflexes: An ultrasound investigation

The languages of South Asia somewhat famously contain what might be deemed canonical retroflexes. These are often described as involving an extreme articulation with sub-apical post-alveolar or palatal contact, and imaging studies of varying types (electropalatography, MRI, ultrasound, etc.) suggest that this is an accurate characterization for Dravidian retroflexes (Kochetov et al. 2014, Krull and Lindblom 1996, Narayanan et al. 1999, Narayanan et al. 2004, Scobbie et al. 2013). Indic retroflexes reportedly involve less extreme articulations (Ladefoged and Bhaskararao 1983), but imaging data for Indic languages remains surprisingly rare. This raises an interesting question: how do Indic and Dravidian retroflexes compare articulatorily? We explore this question by presenting ultrasound imaging of retroflexes in several Indic (Marathi, Hindi) and several Dravidian (Kannada, Telugu) languages. Results indicate that all languages contain at least some of the extreme articulations often associated with South ...

Articulatory and acoustic investigation of coronals in Hakha Chin

Coronal obstruents are profoundly common typologically, occurring in most or all of the world’s known languages. Very few languages, however, contrast coronal obstruents at the dental and alveolar places of articulation. Hakha Chin—also known as Lai—is a Tibeto-Burman language spoken in western Myanmar that is reported to do so. Very little phonetic research on Chin exists, but Maddieson and Van Bik (2004) provide acoustic and static articulatory data (palatography and linguography) suggesting that the contrast is between a lamino-dental series and an apico-alveolar series. We follow up on that research using dynamic, volumetric articulatory imaging in the form of 3D/4D ultrasound. Articulatory and acoustic data from two native speakers of Chin are presented in order to contribute to a more thorough understanding of this typologically uncommon contrast.Coronal obstruents are profoundly common typologically, occurring in most or all of the world’s known languages. Very few languages, however, contrast coronal obstruents at the dental and alveolar places of articulation. Hakha Chin—also known as Lai—is a Tibeto-Burman language spoken in western Myanmar that is reported to do so. Very little phonetic research on Chin exists, but Maddieson and Van Bik (2004) provide acoustic and static articulatory data (palatography and linguography) suggesting that the contrast is between a lamino-dental series and an apico-alveolar series. We follow up on that research using dynamic, volumetric articulatory imaging in the form of 3D/4D ultrasound. Articulatory and acoustic data from two native speakers of Chin are presented in order to contribute to a more thorough understanding of this typologically uncommon contrast.

Acquiring and visualizing 3D/4D ultrasound recordings of tongue motion

Abstract Ultrasound is increasingly common in speech and phonetics research as technology continues to improve. The first digital 3D/4D ultrasound system was utilized for speech research nearly a decade ago, but data access, processing, and visualization were limited to (non-speech) clinical imaging applications. Access to the raw (pulse-echo or scan-converted) image data is a critical step toward making 3D/4D ultrasound an effective tool for speech research. In addition, there is a need for technical characterization of 3D/4D ultrasound systems together with a presentation of their strengths and limitations for research. This paper gives a general technical description of ultrasound systems, beginning with conventional 2D image acquisition and working up toward 3D/4D systems. Emphasis is given to one particular system – the Philips EPIQ 7G system with xMatrix X6-1 digital 3D/4D transducer – for which access to raw scan-converted data is now possible. For parameter settings typical of general abdominal imaging, frame rates around 20 fps are easily achieved (with frame rates possible up to 173 fps depending on image volume size, scan line density, and imaging depth), with spatial resolution comparable to research-grade 2D transducers. Using a modified Philips foot pedal, ultrasound recordings can be made with synchronous audio, and time-aligned with a minimum uncertainty equal to the ultrasound frame duration. In speech research, frame rates around 20 fps – with corresponding time-alignment uncertainty near 50 ms – provide good 3D coverage and enable many new phonetics questions to be posed and addressed.

Three dimensional ultrasound imaging of pre- and post-vocalic liquid consonants in American English: Preliminary observations

This study presents articulation data of American English laterals and rhotics captured by combining real-time 3D ultrasound, digitized 3D palate impressions, and time-aligned audio recordings. While articulation of laterals and rhotics has long been of interest, traditional two-dimensional imaging techniques are subject to limitations because the vocal tract is three-dimensional. The technological advances rendering 3D imaging possible therefore enable more sophisticated understanding of these complex articulations. Results from two speakers are presented, one from New England and one from Indiana. Lateral productions are different for each speaker, but both speakers exhibit mirrored temporal phasing of coronal and dorsal occlusions in /l/ in onset vs. coda position, with dorsal articulations occurring temporally closer to the neighboring vowel. Lingual shape for rhotic productions is complex but consistent across speaker, syllable position, and vowel context.

Articulatory and acoustic correlates of tongue root contrasts in Gua

Vowel systems in West African languages are often noted for using the position of the tongue root (TR) to contrast vowels throughout the vowel space. E.g. X-ray studies of Igbo show that pairs of vowels such as /i/ and /ɪ/ contrast with regards to tongue root position. A similar study of Akan shows that vowel height also gets incorporated into the contrast (Ladefoged and Maddieson 1990). While many languages are noted for having TR contrasts, imaging data are available for only a small subset. Gua, a Kwa language from the Niger Congo family spoken in coastal Ghana (Simons and Fennig 2017, Yeboah-Obiri 2013), is a critically under-documented language which contains TR contrasts in all high and mid vowels (Advanced TR: /i e o u/; Retracted TR: /ɪ ɛ a ɔ ʊ/). Acoustic analysis and articulatory data from 3D ultrasound recordings reveal that RTR vowels show a variety of deformations of the tongue surface, depending on the vowel. However, these deformations are linked by the mechanics of tongue root retraction. ...

Rhoticity in Cajun French

Previous studies of Cajun French (CF) report a shift in pre-rhotic vowel quality (Conwell & Juilland 1963, Blainey 2015, Dubois & Noetzel 2005, Salmon 2007, Lyche, Meisenburg & Gess 2012). This is unsurprising, in that such behavior is reported for numerous Francophone varieties, but prior work on rhotics in CF has been largely impressionistic. As such, the current study presents acoustic analysis of pre-rhotic vowels in CF. Formant measurements were taken from data collected from three native speakers. The question of whether rhotic deletion is present in CF is also addressed: /r/-dropping is widely attested cross-linguistically, and has been reported in the closely related Francophone variety of Acadian French (Cichocki 2012, 2006). Given the previous literature on /r/-dropping and rhoticity in CF and other Francophone varieties, it was expected that these data would show variable /r/ production, compensatory lengthening of vowels preceding a dropped /r/, and a general trend of /r/-induced vowel lowerin...

Articulatory data for a five-way liquid contrast: 3D ultrasound of Marathi

Lateral and rhotic consonants show great crosslinguistic variation, and are traditionally described as articulatorily complex (Ladefoged & Maddieson 1996; Proctor 2011; Wiese 2001, 2011). A good body of work has investigated the characteristics of liquids in languages like English (Dellatre & Freeman 1968; Guenther et al. 1998; Sproat & Fujimura 1993; Westbury, Hashi, & Lindstrom 1998; many others), which contains a two-way contrast. What of South Asian languages, however, which often contain a greater number of liquids? Tamil liquids have been imaged using palatography and electropalatography (McDonough & Johnson 1997) as well as MRI (Narayanan et al. 1999), and Malayalam liquids have been imaged using mid-sagittal ultrasound (Scobbie, Punnoose, & Khattab 2013). Little has been done with Marathi, though. Like Tamil and Malayalam, Marathi—an Indic language spoken in the Indian state of Maharashtra—contains a five-way liquid contrast. This work utilizes recent advances in 3D ultrasonography to provide deta...