The neuronal bases of voice information processing studied using machine learning and primate electrophysiology

PhD supervisor : Pascal Belin, Thierry Artières

Labotary : Institut de Neurosciences de la Timone, Marseille and Laboratoire d’Informatique des Systèmes, Marseille

We invite applications from candidates to a PhD project to be presented in the annual PhD call by the Institute of Language, Communication and the brain ( in Aix-Marseille University, in which 3 PhD grants will be awarded on a competitive basis.

The successful candidate will be co-supervised by Prof Pascal Belin (Institut de Neurosciences de la Timone, Marseille ) and Prof Thierry Artières (Laboratoire d’Informatique des Systèmes, Marseille ). The project aims to measure, via multi-electrode arrays implanted chronically in the auditory cortex of monkeys, the neuronal activity evoked by a large array of complex sounds, and analyze that neuronal activity using machine learning tools, particularly deep-learning, to better understand the neurocognitive architecture underlying cerebral voice processing in primates. This project is part of a larger ERC-funded project entitled “Comparative Studies of Voice Perception in Primates” (COVOPRIM).

The successful candidate will have a Master’s degree or equivalent, strong interest and/or background in machine learning and deep learning, strong interest in Neuroscience and strong motivation to work with primates. He/she will be registered at Ecole Doctorale 62 at Aix-Marseille University (, embedded in the everyday activity of the two co-supervising teams, as well as in the PhD program of ILCB ( ).

Please send your Expression of Interest and CV to and thierry.artiè before June 1st, 2022


Apply here:

Formal syntactic dependencies syntax and syntax/semantics interface, the structure and computation

PhD supervisor : Viviane Deprez

Labotary : Laboratoire Parole et Langage

Keywords : Syntax, semantics, negation, questions, language deficiencies and acquisition, motor system

Expected competences of the candidate : experimental design, statistical analysis, basic linguistic knowledge

Apply here:

Towards a quantitative theory of child conversational development

PhD supervisorAbdellah Fourtassi

Labotary : Laboratoire d'Informatique & Systèmes

Keywords : child conversational development, deep learning modeling, unmoderated data collection, cross-cultural comparison, cognitive mechanisms 

Expected competences of the candidate : The project is highly interdisciplinary and we accept applications from researchers with an experimental background to supervise data collection and analysis and researchers with a background in computational modeling (especially in deep learning techniques) to help with modeling child conversational coordination from multimodal data 

Summary of the pre-proposal : How do children become able to use language to engage in coordinated conversations? The answer to this question has a far-reaching societal impact. Indeed, this development allows us to achieve crucial — socially mediated — goals such as learning from more knowledgeable people, expressing thoughts and needs, convincing others, collaborating with peers. If impaired, it can have negative consequences from the risk of developing mental health issues to the quality of academic attainment and employability (Murphy et al., 2004). Research has shown that this development takes several years to mature, spanning most of middle childhood (7 to 12 years). Yet, while there is a large body of research investigating children’s acquisition of linguistic structures such as phonology and syntax, in comparison, little is known about how children learn to translate this knowledge into conversational skills such as turn-taking management, negotiating shared understanding with the interlocutor, and the ability for a coherent exchange. This slow scientific progress can be attributed largely to methodological limitations in traditional research methods typically used to study this question. Our team ( proposes a research approach that goes far beyond the limitations of existing methods, allowing breakthroughs in our understanding of this phenomenon. Our approach combines two highly innovative methods in the field: 1) Unmoderated designs allowing large-scale cross-cultural data acquisition from children and 2) Deep learning modeling techniques, which will allow us to bridge across several dimensions of conversational complexity and to identify their cognitive mechanisms. This research approach will allow us to lay the groundwork for the first quantitative cognitive theory of conversational development. This theory will provide a formal framework that will allow us (and other researchers in the field) to articulate, contrast, and adjudicate between several hypotheses, answering crucial, lingering scientific questions about the universals of conversational development and its cognitive mechanisms. We coordinate data collection across 20 countries with members of our international community of Multimodal Child Conversation (MCC) ( 


Apply here:

Neural bases of vocal communication

PhD supervisor : Pascal Belin 

Labotary : INT 

Keywordsconspecific vocalizations auditory cortex comparative approach neurophysiology non-human primates 

Expected competences of the candidate :  experience and competence in electrophysiological recordings using Utah Arrays in the macaque; - experience and competence in training rhesus macaque to perform tasks; - competence with analysis of electrophysiological data; - competence in programming.  

Summary of the pre-proposal To exploit the highly valuable electrophysiological recordings in macaque temporal voice areas performed in Belin's team to investigate the macaque's higher-level representations using deep learning approaches, 


Apply here:

Neurophysiology of Social Interactions with Natural and Artificial Agents

PhD supervisorThierry Chaminade

Labotary : Institut de Neurosciences de la Timone (INT, UMR 7289)

Keywords : Human neurophysiology, Human-Robot interactions, Functional Magentic Resonance Imaging, Natural Conversation, , Artificial Conversational Agents.

Expected competences of the candidate : The candidate should have a borad interest in the field of natural conversational interactions and skills in the use of python tools for the analysis of behavioral and neurophysiological data.

Summary of a pre-proposal : This project provides an unique experimental environment, including a paradigm to record human-human and human-robot conversations in the MRI environment synchronously with brain activity and a number of other behaviours. Participants (n=25 in the corpus) have been scanned using fMRI while they interact with a human or a robot (the conversational head Furhat) through unconstrained language. A cover story hides to participants the real objective of the experiment, namely investigating social interactions. Multiple behaviours are recorded (participant and interlocutor speech, interlocutor head movements fed live to the participant, participant eye movements, brain activity recorded with functional MRI, respiration and peripheral blood flow) and curated to form a synchronized multidimensional corpus. This corpus, shared through various dedicated repository, has since been used to address multiple questions pertaining to the cognitive and (neuro)physiological foundations of natural social interactions, in particular through the comparisons of human-human and human-robot interactions. The project covers the development of new robots' behaviours to improve significantly its social competence, recording of a new corpus of 25 participants interacting with this new conversational robot, preparing the new corpus for analysis according to well-established pipelines and analysis of the multimodal corpus to investigate the physiological correlates of natural human-human conversations compared to the same interactions with a non-human agent.


Apply here:

Language and associative learning mechanisms

PhD supervisor : Arnaud Rey

Labotary : Laboratoire de Psychologie Cogntive

Keywords : Language perception (visual or oral) statistical learning associative learning chunking mechanisms computational modeling

Expected competences of the candidate : The candidate should have a background in psycholinguistics and/or statistical learning, reasonable computer programming skills, good english writing skills and a delicious sense of humor.

Summary of a pre-proposal : Our language system builds up through repetitive exposure and practice with linguistic material. Sequences of information (e.g., sequence of letters or phonemes) are encoded and grouped as units of processing through elementary chunking mechanisms. The goal of the research project should be to provide new insights about either the computational or brain mechanisms supporting the development of our language cognitive system through associative/hebbian learning. References: • Tosatto, L., Fagot, J., Nemeth, D., & Rey, A. (2022). The evolution of chunks in sequence learning. Cognitive Science. • Rey, A., Bogaerts, L., Tosatto, L., Bonafos, G., Franco, A., & Favre, B. (2020). Detection of regularities in a random environment. Quarterly Journal of Experimental Psychology. Doi: 10.1177/1747021820941356 • Rey, A., Minier, L., Malassis, R., Bogaerts, L., & Fagot, J. (2018). Regularity extraction across species: associative learning mechanisms shared by human and non-human primates. Topics in Cognitive Science. Doi: 10.1111/tops.12343


Apply here: