Les commentaires sont clos.

PhD offer (GIPSA-lab, Grenoble): modelling perceptual decision making in multistable vision

5 Juillet 2022

Catégorie : Doctorant

In the context of the recently accepted research proposal Vision-3E: Exploration, exploitation, and expectation in visual perception (ANR funded) as a collaborative project between Institut de Neurosciences de la Timone (INT) - Marseille, le laboratoire de neurosciences cognitives (LNC) - Marseille, Sciences Cognitives et Sciences Affectives (SCALAB) - Lille, le Laboratoire de Psychologie et de neurocognition (LPNC) - Grenoble et Grenoble Images Parole Signaux et Automatique (GIPSA-lab) – Grenoble four PhD scholarships are available.

With the Vision-3E project, we aim at gathering behavioural and physiological evidence about the three key functions in the closed-loop processing of human visual perception—expectation, exploration, and exploitation,—-in order to better understand and model their interaction. We will make use of multistable visual stimuli to access the temporal dynamics of ongoing perceptual decision processes and we address the causes of perceptual reversals in the broader context of perceptual decision making. Within this framework, we assume that eye movements play an important role in visual perceptual decision-making and aim at deeply investigating this role in the framework of perceptual rivalry. In addition, we will study the functional brain network involved in perceptual rivalry and aim at providing a comprehensive predictive model of percept reversals in time; both of these objectives will take eye movements into account as an active component of the perceptual decision loop.

Vision-3E proposes an original approach to this broad question, involving both correlational and interventional experiments. A series of behavioural experiments are planned to test the effect of ad-hoc selective manipulations of the oculomotor behaviour on the perceptual reversals. Our aim is to propose a novel and more comprehensive model of visual perceptual decision in the perceptual decision-making framework by combining behavioral experiments, advanced recording and stimulation techniques, data-analysis methods, and modelling.


PhD project: modelling perceptual decision making in multistable vision

Dynamic Bayesian sampling and evidence accumulation : role of eye movements and EEG frequency-band modulations in a comprehensive computational model of perceptual decision making

Keywords: drift-diffusion process, perceptual decision making, Bayesian sampling, multistable visual perception.

Whereas current models of multistable perception mainly use descriptive statistics (for instance, mean relative time spent in a perceptual state or estimated transition probabilities between different perceptual states; often combined into a semi-Markov chain), our aim is to move toward a predictive model, integrating eye movements and characteristic electro-encephalography (EEG) frequencies within the model as to predict the upcoming perceptual transition time. The underlying perceptual process is assumed to be composed of (i) a non-stationary noise process superimposed to (ii) an integration of constructive (in favour of a particular perceptual state) and destructive (opposed to a particular perceptual state) evidence. As such, the Bayesian sampling hypothesis is extended to an active vision framework, i.e., a drift-diffusion process with a time-varying diffusion rate, corresponding to a modulation of vigilance. This modulation is supposed to manifest itself as a modulation of the EEG amplitudes in the alpha and⋅or gamma band or higher variability in the viewing positions reported on the screen. A time-varying drift component will be inferred from both gaze positions and EEG markers that will be used for the integrated evidence over time.

Team expertise

The team has a vast experience with the moving plaid stimulus[1] that is used in multistable visual perception paradigms with implementation both in psychopy (python) and psychtoolbox (matlab)as well as with multivariate and multimodal data processing, specifically with EEG and Gaze. A semi-Markov model[2] has been implemented and used for hypothesis testing of a visual motion after effect. The team has a solid background in human visual perception[3], mastering the entire pipeline from experiment design to data analysis.

GIPSA-lab and LPNC offer an international and intercultural working environment. Both GIPSA-lab and LPNC possess their dedicated experimental platforms for running experiments on visual perception, with GIPSA-lab also offering the possibility of joint electro-encephalography and gaze recordings.

Sought for skill set and abilities

Mathematical modelling, machine learning, experiment design (prior experience with psychopy highly appreciated), electrophysiological (EEG) and behavioural (Gaze, response time) data processing and analysis, scientific programming, scientific communication (oral and written). Good coding practice will be particularly appreciated.


a master degree in computer science, signal and image processing, cognitive sciences, computational neuroscience, or a related field. English language proficiency, both oral and written communication (B2 level), is compulsary.

How to apply?

Contact the team members by email with your inquiries, an academic CV and a cover letter before August 31, 2022.

  • Anne Guérin-Dugué (PR, UGA, GIPSA-lab) :
  • Ronald Phlypo (MCF, Grenoble INP, GIPSA-lab) :
  • Alan Chauvin (MCF, UGA, LPNC) :


[1] Kevin Parisot. Multi-stability in visual perception and eye movements: does action control perception or vice-versa?. Cognitive science. Université Grenoble Alpes, 2020. English. ⟨tel-03109729⟩
[2] Juliette Lenouvel, Ronald Phlypo, Kevin Parisot, Emeline Lalisse, Alan Chauvin. Motion after effect is perception-based, not only signal-based. ECVP 2021 - European Conference on Visual Perception, Aug 2021, Online, France. ⟨hal-03506152⟩
[3] Kevin Parisot, Steeve Zozor, Anne Guérin-Dugué, Ronald Phlypo, Alan Chauvin. Micro-pursuit: A class of fixational eye movements correlating with smooth, predictable, small-scale target trajectories. Journal of Vision, Association for Research in Vision and Ophthalmology, 2021, 21 (1), pp.9. ⟨10.1167/jov.21.1.9⟩