LSD team seminar : Yao Yao

Seminar of

Yao Yao

(Hong Kong Polytechnic University)

Tuesday May 30, 2023, 12.00-12.45

Laboratoire Parole et Langage, conference room B011, Aix-en-Provence

Phonetic accommodation of tone: Reversing a tone merger-in-progress via imitation

Previous literature has documented phonetic accommodation for various segmental and suprasegmental features, but the accommodation of tone remains under-explored. The current study contributes to the literature by investigating the accommodation of two merging tones in Hong Kong Cantonese, mid-level Tone 3 (T3) and low-level Tone 6 (T6), in a speech shadowing experiment. Specifically, we ask whether shadowers will reverse the merging trend after exposure to a model talker with distinct T3–T6 productions and if so, what factors will modulate the accommodative behaviors. Evidence for phonetic imitation is observed, but the effect varies by shadower’s baseline production of T3–T6 distinction and across different model talkers. Shadowers with less baseline tonal distinction exhibit greater degrees of imitation, suggesting that greater linguistic distance facilitates imitation. More robust imitation is observed in the young model talker condition, but the effect is most likely driven by the talker’s idiosyncratic production. Shadowers’ impression of the model talker and attitudes towards ongoing changes in the language did not show substantial impact on imitation. We discuss the implications of these findings for theories of speech perception and production.

Language and Speech Dynamics (LSD) team seminar

Language and Speech Dynamics (LSD) - Team seminar 2023

Friday April 7th, 2023, 10.30 Conference room B011, LPL



10.30-11.30: Experimental investigations into turn taking by Martin Pickering (University of Edinburgh), Director of Research for the School of Philosophy, Psychology and Language Sciences.

Abstract: Dialogue is remarkably fluent, with the gaps between interlocutors’ contributions being remarkably short.  In this talk, I describe a series of question-answering experiments in which participants respond to predictable (vs. unpredictable) questions, speeded (vs. slow) questions, and questions in which the critical information occurs early (vs. late).  I then provide a theoretical explanation of these findings, based on the claims that listeners use their own production system to predict the speaker’s utterance and plan their own responses early.


11.45-12.15: Phonological Networks in Speech Perception and Production by Xenia Dmitrieva (LPL, AMU), PhD student.

Abstract: Phonological processing in word perception vs. production is the focus of this study. Two groups of models make distinct predictions on the brain regions involved across the language modalities: (1) Partial Separation Models (PSM) assume both frontal and temporal regions are recruited in production, but only temporal regions – in perception; (2) Integration Models (IM) assume the same fronto-temporal network to be recruited across modalities. In the current fMRI study we are contrasting these models by comparing minimal phonological pairs (e.g., bilabial: “ballon” vs. alveolar: “talon”) within the same participants performing picture naming and passive listening tasks. We’ve performed a region of Interest (ROI) analysis in the motor and auditory cortex to assess the presence (PSM) or absence (IM) of an interaction between phoneme and modality. Bilabial-initial words, both in production and perception, elicited more activation in the lip-associated ROI compared to the tongue-associated one, and vice versa for the alveolar-initial words. This result may indicate that phonological networks are shared across the language modalities.


12.15-12.45: When the going gets tough, be flexy and keep going. The reorganization of speech motor control in response to different sources of speech production complexity by Leonardo Lancia (LPL, CNRS/AMU), CNRS researcher

Abstract: The control of speech production relies on the balance between two fundamental but complementary tendencies of the sensorimotor system. A tendency toward particulation favours the emergence of task-specific functional modules and permits the production of complex behaviour.  At the same time, a tendency toward integration favours the emergence of globally coherent behaviour, as rhythmic patterns of activity, and permits a reduction of the degrees of freedom actively controlled by the speaker. To understand better the interplay between these two tendencies, in our work we aimed at studying how the organization of the sensorimotor system varies across utterances and speech conditions. To this aim, we employed a number of speech repetition tasks in which we manipulated one or more among the following factors: the complexity of the syllabic structure; the heterogeneity of the syllables composing the utterances to repeat; the simultaneous presence of repeated and alternating patterns (leading to speech errors); the presence of delays in the auditory feedback and/or of shifts of the perceived f0; the presence of an artificial speaker simultaneously executing the same speech repetition task at different degree of coordination with the participants’ speech signal. Produced utterances were modeled through a prosodically grounded hierarchy of modulatory patterns corresponding to different components of speech rhythm and behaving as low dimensional dynamical systems. The picture that emerges from these studies is one in which, in order to achieve the required linguistic goals, the coherency of the different rhythmic components as well as their temporal and spatial predictability (respectively related to the variability of syllabic nuclei durations and amplitude values) are modulated in a task-dependent fashion. These results suggest that the global strategies underlying the organization of speech motor control can vary quite drastically as a function of the phonetic content of the utterances to produce, of the conditions of enunciation, and potentially of factors affecting the experienced sense of agency (which is the feeling of control of one’s own actions).

S2S Research team seminar with Carolina Pletti

Wednesday April 5th 2023

S2S Research team seminar

Presentation by Carolina Pletti 

(Universität Wien)

Laughing Together – the influence of laughter on interpersonal neural synchrony, bonding and prosociality

10.30-12.00 LPL

In this talk, I will present preliminary results on the effect of laughter on neural synchrony, bonding and prosocial intentions towards an interaction partner. Laughter is a rhythmic, audiovisual expression of emotion, often occurring during social interactions, and has been shown to increase bonding in a social group  through endorphine release. Being a multimodal, rhythmic social stimulus, laughter could be hypothesized to facilitate interpersonal synchrony. This, in turn, could contribute to increased affiliation and prosociality between interaction partners. The present study aimed at testing these hypotheses using functional near-infrared spectroscopy (fNIRS) hyperscanning. Results from a pilot dataset show that laughter increases neural synchrony in the inferior frontal gyrus, but such effect is not long lasting. In this talk, I will discuss implications and challenges for future research on laughter and synchrony.

Carolina Pletti is Marie Skłodowska-Curie Fellow at the Institute for Developmental Psychology of the University of Vienna. Her research focusses on the development and neural mechanisms of social cognition, in particular moral cognition, prosociality, laughter, and interpersonal synchrony.

Seminar of Kate Watkins


Kate Watkins

(ILCB / IMERA Univ. Oxford)

Imaging and stimulating the brain in people who stutter

Friday March 3rd 2023, at 10.30 LPL, conference room B011 & online by Zoom

Zoom link:
ID: 835 9892 5819
Code: 862601

Developmental stuttering affects about 5% of children and 1% of adults.  Fluency can be enhanced in people who stutter by altering sensory feedback during speech production and by altering production e.g. by whispering, changing pitch or accent, or by external cueing, speaking in unison or singing.  Abnormal sensorimotor integration and impairment in cortico-striatal-thalamo-cortical circuits might explain the occurrence and persistence of speech disfluency.  Brain imaging studies show some support for these theories.  I will present our early structural and functional brain imaging studies that revealed abnormal white matter microstructure underlying cortical differences in brain activity during speech.  In recent work, we found increased amounts of iron in the striatum in a large sample of people who stutter.  We have also used brain stimulation to enhance speech fluency in people who stutter.  Five days of stimulation paired with fluency training increased fluency and brain activity in the striatum compared with fluency training alone.  Finally, I will present our findings from imaging the vocal tract during speech production that indicate instability in the speech motor system in people who stutter during perceptibly fluent speech.


Seminar of Julien Lefèvre


Environmental impacts of Information Technology and AI

10.30-12.00 LPL, conference room B011

The environmental impacts of information technology have been exposed for about 15 years. They have been recently updated through the shift project report [1] and a recent meta-analysis discussing trends and controversies [2], in particular regarding rebound effects. I will present more particularly quantitative tools in the field of AI (machine/deep learning) and a serious lack of evaluations for so-called AI solutions for environment [3], reflecting a general techno-solutionist state of mind. Those observations imply several conclusions, from a more critic approach in how to raise awareness for environmental issues and associated solutions [4] to a more global redirection of scientific activities taking into account planetary boundaries [5].

[1] Lean ICT, Shift Project, 2018
[2] Freitag et al, 2021, The real climate and transformative impact of ICT: A critique of estimates, trends, and regulations, Patterns, 2021
[3] Ligozat, Lefevre, Bugeau, Combar, Unraveling the Hidden Environmental Impacts of AI Solutions for Environment Life Cycle Assessment of AI Solutions, Sustainability, 2022 et la version grand public
[4]  Ligozat, Marquet, Bugeau, Lefevre, Boulet, Bouveret, Marquet, Ridoux Michel. 2022. How to Integrate Environmental Challenges in Computing Curricula?. In SIGCSE’22
[5] Monnin, Bonnet, Landivar, Héritage et Fermeture, 2021

Julien Lefèvre is part of the Aix-Marseille Political Ecology Workshop which “participates in the construction of a multidisciplinary community of the academic world working or reflecting on the multiple dimensions linked to ecological upheavals. »

Seminar by Mortaza Taheri-Ardali


Mortaza Taheri-Ardali

(Shahrekord University, Iran & Universität zu Köln, Germany)

Analysing the prosody of native speakers of Persian in L1 and L2 (English) using the OMProDat corpus

Friday December 9, 2022 at 10.30 – 12.00

LPL, conference room B011

After a brief presentation of our current knowledge of the prosody of the Persian language, this talk focuses on the findings from a bilingual Persian-English prosodic database read by native Persian speakers as part of the OMProDat project. This corpus contains 40 continuous and thematically-connected passages each with five sentences. The collection was read by 5 male and 5 female Persian speakers. Moreover, the ASCII phonetic alphabet SAMPA was used to romanize and phonetically transcribe the Persian sentences. The database includes TextGrid annotations obtained semi-automatically for two speakers (1 male and 1 female) from the sound and the orthographic transcription using the SPPAS alignment software. Then, this part of the corpus was annotated prosodically using the Momel and INTSINT algorithms. Preliminary results of this analysis will be reported in this talk. Using this substantial amount of data, we intend to conduct a cross-linguistic comparison with other languages in OMProDat.

Boersma, P., & Weenink, D. (2021). Praat: doing phonetics by computer [Computer program]. Version 6.1.55, retrieved October 2021 from
Bigi, B., & Hirst, D. J. (2012, May 22-25). SPeech Phonetization Alignment and Syllabification (SPPAS): a tool for the automatic analysis of speech prosody. Speech Prosody 2012, Shanghai, China.
Bigi, B. (2015). SPPAS – Multi-lingual Approaches to the Automatic Annotation of Speech. The Phonetician – International Society of Phonetic Sciences. 111-112, 54-69.
Ding, H., Hirst, D. J. & Hoffmann, R. (2015, Oct 28-30). Cross-linguistic prosodic comparison with OMProDat database. 18th Oriental COCOSDA & Conference on Asian Spoken Language Research and Evaluation (O-COCOSDA/CASLRE), Shanghai, China.
Hirst, D. J., Bigi, B., Cho, H.-S., Ding, H., Herment, S., Wang, T. (2013) Building OMProDat, an open multilingual prosodic database. TRASP, Tools and Resources for the Analysis of Speech Prosody [satellite workshop of Interspeech], Aix-en-Provence, France. Hirst, D. J. (2007, Aug 6-10). A Praat plugin for Momel and INTSINT with improved algorithms for modelling and coding intonation. 16th International Conference of Phonetic Sciences, Saarbrücken, Germany.
Hirst, D. J. (2013, Aug 25-29). Melody metrics for prosodic typology: comparing English, French and Chinese. Interspeech 2013, Lyon, France. Hirst, D. J. (2016, May 31 – June 3). On the automatic comparison and cloning of native and non-native speech prosody. Speech Prosody 2016, Boston, USA. Hirst, D. J. (2021, Oct 20-21). Automatic visual and auditory feedback for second language (L2) speech prosody [Keynote]. The 2nd International Conference on Laboratory Phonology and Phonetics (ICLPP 2), Tehran, Iran.
Keshtiari, N., Kuhlmann, M., Eslami, M., & Klann-Delius, G. (2015). Recognizing emotional speech in Persian: A validated database of Persian emotional speech (Persian ESD). Behav Res 47, 275-294.
Mohamad Nezami, O., Jamshid Lou, P. & Karami, M. (2019). ShEMO: a large-scale validated database for Persian speech emotion detection. Lang Resources & Evaluation 53, 1–16. Taheri-Ardali, M., Hirst, D. (2022). Building a Persian-English OMProDat database read by Persian speakers. Speech Prosody 2022, Lisbon, Portugal.
Taheri-Ardali, M., Khorram, S., Assi, M., Sameti, H., & Bijankhan, M. (2016). Designing and recording a speech database for Persian TTS systems. Iranian Journal of Comparative Linguistic Research, 6, 69-84, 2016.

Web page of M. Taheri-Ardali

Seminar Snizhana Holyk


Snizhana Holyk

(Uzhhorod National University)

Conceptualisation of old age in English : Interdisciplinary perspective

Friday November 4th, 2022, 10.30-12.00

LPL, conference room B011



In the present study, I aim to analyse cultural and social constructions of old age and the aged in English. Drawing on the thesis about the interrelation between language and culture, I hypothesize that the concept old age in English is verbalized by a set of socially and culturally loaded representations. This research adopts a cognitive-semantic approach when meaning is defined as conceptualization. It aims at revealing the notional content of the concept old age in English with the focus on the definitions of the lexeme ‘old age’ in present-day lexicographical sources (both printed and on-line). Firstly, the concept of old age as a mental unit and complex knowledge cluster is analysed, its nominative field is defined. It appears to be a multicomponent structure representing the period or time in life when one is old, the person or the group of persons who are old and the state or condition of being old. Secondly, I will highlight the definitions of old age in English and compare different discursive models of ageing conceptualisation. Old age as a social phenomenon is usually defined through the prism of its medico-biological and socio-philosophical characteristics as a final stage of human life with considerable physiological and psychological changes leading to little involvement of a person in social activities. However, the demographic processes of the recent decades have resulted in radical changes in old age delimitation, when it gets measured not only within chronometric framework, but also by sociocultural criteria. Old age as a cultural phenomenon is different from the chronological age, as it is characterized by vague boundaries, the interaction between different age stages, and deviations from their logical order. Furthermore, adopting the interdisciplinary research perspective, the study reveals discursive models of old age conceptualization, providing the description of the new reality of old age not only as a biological phenomenon, but also as a social construct with a complex cognitive structure represented in discourse with normative, axiological, stereotypical, figurative and symbolic components. Moreover, the paper gives insight into the metaphorical conceptualization of old age in the literary discourse and authors’ creative associations about this stage of life.

Key words: old age concept, ageing, conceptualisation, conceptual metaphor, discourse analysis, interdisciplinary approach.

Seminar Leonor Becerra & Philippe Blache


10.30 - 11.15 Leonor Becerra (LPL/AMU)

Multimodal learning: bringing together natural language acquisition and machine learning

I will briefly present my main research activities, which are in the intersection of linguistics and computer science. My goal has been to develop computational systems that learn to understand and speak natural languages from multimodal data, inspired in the way children learn their native language. I will also present applications of my work to solve Natural Language Processing tasks and ongoing research projects that involve the integration of a larger set of multimodal cues than the one I used so far in my research.

11.15 - 12.00 Philippe Blache (LPL/CNRS)

Prédiction, retardement, chunking: les mécanismes permettant d'expliquer la compréhension en conversation

Comment accède-t-on au sens ? La plupart des expériences abordant cette question, que ce soit en linguistique, en neurosciences ou en psycholinguistique, portent sur des mots ou des phrases isolés. Ces études permettent de décrire avec précision les mécanismes à l'œuvre, y compris au niveau du cerveau. Cependant, elles reposent toutes sur une hypothèse forte : celle de la compositionnalité, stipulant que le sens d'une phrase est une fonction du sens de ses constituants et se construit donc progressivement  en partant de mots, agrégés en constituants supérieurs avant de composer la structure d'une phrase. Cette hypothèse se double de plus d'une méthode de construction, reposant sur un mapping direct entre structure syntaxique et structure sémantique (Montague, 1973; Dowty, 2006). Il existe cependant un grand nombre de cas où la compositionnalité ne permet pas d'expliquer l'accès au sens: de nombreuses constructions (ditransitivité, idiomes, etc.) montrent qu'il existe également une possibilité d'accès direct au sens, non pas en partant des mots, mais en reconnaissant globalement un "pattern" (Construction Grammars : Goldberg, 2015). De plus, de nombreuses expériences montrent également que nous sommes capables d'accéder au sens à l'aide d'un traitement  superficiel (Good-enough Theory : Ferreira et Patson., 2007), voire de reconstituer le sens à partir d'une entrée syntaxiquement ou sémantiquement mal construite (Semantic attraction : Kim & Osterhout, 2005). A ce jour, il n'existe pas de modèle neuro-cognitif permettant de décrire la coexistence de ces phénomènes et d'une façon plus générale permettant d'expliquer l'accès au sens au delà de la phrase, dans des situations naturelles (typiquement des conversations).

Je propose de présenter dans cet exposé les bases d'un tel modèle. Il repose sur l'hypothèse que la mémoire déclarative contient non seulement les informations relatives aux mots, mais également à ces "patterns" (ou constructions). La compréhension repose donc principalement non pas sur un mécanisme séquentiel, incrémental et cumulatif (un traitement mot par mot,) mais de façon plus globale grâce à un double mécanisme de retardement et de segmentation en chunks et associés à la prédictibilité du langage. Je présenterai de façon plus précise comment ces hypothèses permettent de généraliser le modèle "Memory, Unification and Control" (Hagoort, 2013), rendant compte des aspects linguistiques, cognitifs et cérébraux de l'accès au sens en situation naturelle.

Dowty, D. (2006) "Compositionality as an Empirical Problem", in Direct Compositionality, Barker C. & Jacobson P. (eds), Oxford University Press
Ferreira, F. & Patson N. (2007) "The ‘Good Enough’ Approach to Language Comprehension", Language and Linguistics Compass 1, 1‑2
Goldberg, A. (2015) "Compositionality". In The Routledge Handbook of Semantics, Taylor and Francis Inc..
Hagoort, P. (2013) "MUC (Memory, Unification, Control) and beyond", Frontiers in Psychology, 4
Kim, A. & Osterhout, L. (2005) "The independence of combinatory semantic processing: Evidence from event-related potentials", Journal of Memory and Language, 52(2)

Seminar Francesca Di Garbo


Francesca Di Garbo


La diversité linguistique et ses corrélats socio-culturels : études de cas, méthodes et perspectives de recherche

October 21st 2022, at 10.30 in B011 and online via Zoom

(Please contact us for the Zoom access)


Abstract (Seminar in French):

Ce séminaire a comme objectif de présenter mes activités de recherche et les replacer dans le contexte général des recherches en cours au sein de l’équipe SystUs et du laboratoire.

Mes recherches portent sur la diversité linguistique, sa distribution dans l’espace et le temps et ses corrélats non-linguistiques (par exemple, sociaux, environnementaux). Les deux thématiques les plus centrales dans mes travaux récents sont la typologie morphosyntaxique et le développement des méthodes visant à intégrer la typologie linguistique et la sociolinguistique (typologie sociolinguistique).

Dans le cadre de mes recherches en typologie morphosyntaxique, je présenterai deux études de cas sur les systèmes d’accord en genre grammatical dans un échantillon de 179 langues bantoues du Nord-Ouest, et sur les facteurs morphosyntaxiques, sémantiques et sociohistoriques qui peuvent expliquer leur distribution et évolution dans cette région d’Afrique.

Dans le cadre de mes travaux en typologie sociolinguistique, je présenterai un nouveau programme de recherche pour l’étude comparative des changements langagiers dus au contact, qui a été développé à l’université d’Helsinki au sein du projet GramAdapt ( Une nouvelle technique d’échantillonnage typologique et un nouveau questionnaire sociolinguistique ont été développés au sein de ce projet afin de mener des études comparatives sur les dynamiques de contact entre communautés voisines et plurilingues, ainsi que sur l’impact potentiel de ces contacts sur les structures langagières. L’efficacité de ces nouveaux outils sera illustrée à l’aide de quelques résultats préliminaires des recherches encore en cours.


Seminar Markus Syring: Heterogeneity in Research and Practice: Perspectives of the School Pedagogy


Markus Syring

(Universität Tübingen)

Heterogeneity in Research and Practice: Perspectives of the School Pedagogy

LPL, conference room B011 at 10.30 a.m. and online via Zoom

(Please contact us for the Zoom access)

The topic of heterogeneity not only plays an important role in schools and teaching in German-speaking countries, it is also being researched worldwide. The special view that is taken in the presentation refers to cultural, social and performance-related diversity in the classroom. The school pedagogy makes a genuine contribution to pedagogical professionalism in dealing with heterogeneity for (future) teachers. Therefore, in the short presentation, some perspectives of school pedagogy on the topic will be presented and documented with own research.