On the contribution of “Fine Phonetic Detail” (FPD) to Gradience in Phonology: Acoustic, Articulatory, Perceptual and Automatic Methods

This HDR thesis summarises the various research activities I engaged in during my time at Lyon, Newcastle and Paris. The central idea of this thesis is that Fine-Phonetic-Detail (FPD) is an important aspect to consider for speech production, perception, articulation and in automated methods. FPD is defined as speaker-specific details, which are stored in the mental representation and used to identify specific categories and/or the speaker/dialect, following exemplar-based approaches to speech production and perception. Using FPD, I show its role in defining gradiency in Phonology.

My work spanned various areas. Starting from dynamic specification of vowels and the role of FPD (within the vowel and/or due to coarticulation) in identifying vowel categories, in production, perception and learning, in Arabic, French, English, and Mandarin. Then we move to the importance of the phonological feature [±Tense] in specifying geminate consonants in Lebanese Arabic, in both adult and child speech. We look at how adults systematically use this feature in their productions and how it influences the production of the first words in L1 acquisition. We look then at the Voicing contrast in Arabic in interaction with gemination, claiming that the phonological features [Long], [Tense] and [Voice] are all active, following a gradient privative features account. The specification of the Voicing contrast in other dialects shows variable behaviour for the Voiceless set, where it can either receive an [SG] or a [Tense] feature depending on the dialect. The next major research activity looks at the role of the epilarynx in the production of pharyngealisation and of gutturals, using acoustic, Ultrasound Tongue Imaging and ElectroGlottoGraphy. The combination of techniques allows to evaluate the articulatory to acoustic mapping to identify primary and secondary correlates of a combined supra-laryngeal and laryngeal gesture. Using extensions to this area, we look at dialectal and speaker identification in the context of gutturals and evaluating nasalisation in the production of pharyngeal consonants in Iraqi Arabic. We then travel to a new area of research I have engaged in on the use of automatic signal processing algorithms and machine learning to support diagnosis of cognitive disorders, followed by the development of the Arabic WebMAUS and WebMINNI services for forced-alignment. Other collaborative work is presented, such as, relationship between Adult-Directed-Speech (ADS) and Child-Directed-Speech (CDS), acoustics of the fifth liquid in Malayalam, impact of exposure on perceptual adaptations to various English accents, research on L2 phonology, the impact of native language on the preferred tongue postures, use of GAMMs (Generalised Additive Mixed-effects Models) on UTI data and comparison with SSANOVAs, and finally, our contribution to the Many Speech Analyses project, before finishing with the Open Source material that I put to the service of the research community.

We then move to a presentation of my significant activities as a PI, research management and co-supervision, and ending with my future research projects, where I aim to continue working in the same areas, but by expanding methodologies and performing cross-linguistic comparisons.