PhD supervisor : Marie-Hélène Grosbras
Labotary : CRPN Centre de Recherche en Psychologie et Neurosciences
Summary of the pre-proposal : Research in our group concerns the developmental and neural underpinnings of non-verbal communication. We are particularly interested in the interpersonal alignment of actions and emotions, and how this may change at the time of adolescence in relation to changes in the body and in self-perception. We welcome applications for a post-doc position to develop a project along this line, with possible collaboration with other members of ILCB. This project could build on an ongoing fMRI study investigating intergenerational emotion processing between adults and adolescents, with Valérie Chanoine from the Engineering support group. Interested candidates should send an email and CV to marie-helene.grosbras@univ-amu.fr.
PhD supervisor : Fenna Poletiek,
Labotary : Institute of Psychology, Leiden University; MPI of Psycholinguistics, Nijmegen; ILCB-Iméra Chair holder, Aix-Marseille Chair holder, 2023-2024.
Summary of the pre-proposal :
There is wide agreement that language processing is prediction based. We pre-activate a or several predicted items while processing incoming information. Brain activation is observed and processing slows down proportionally as the predicted item mismatches the actual item coming in. These mismatches indicate ‘error detection’ and processing ‘costs’.
Another view based on information theory, is that, in language, new information positively contributes to ‘inform’ our brain about the environment. Hence as information is more accurately predicted, its information value decreases. In this view, the goal of language processing is to detect new unpredicted information that is relevant for a future or current task.
A task might be predicting upcoming events, for example when crossing a street, we might predict a car coming from left or right in the street. In contrast to predicting in language processing (e.g., an upcoming word), mismatches between predicted and actual events, can come at high costs. By predicting about the environment, we control the efficacy of our actions. Thus; though prediction is a ubiquitous process in cognition, the stakes of prediction can differ with the context and the materials we are predicting. The goal of language processing is getting (new) information. The goal of perception is control the effect of our actions.
Question: Does our preference /search for predictable versus unpredictable targets depend on the context in which we predict? I.e. the goal of the prediction process. It can either be 1) getting information (Language domain) about an event or 2) predicting the occurrence of a future event.
Hypothesis: Language comprehension is information processing about relevant meaning. Language with targets that are less predictable, give higher information. Hence, agents prefer unpredictable sequences of language, generating high amplitude N400 ERP’s. Perception for action is processing information about the environment to control the outcome of actions. As targets are better predicted, action effectiveness is higher. Hence agents prefer predictable sequences. Hence, agents prefer sequences of events that are well predictable, and hence, generate low amplitude N400 ERP’s.
Experiment.
We design an experimental task involving processing structured sequences, that is carried out in two task goal conditions -within subjects-: 1) under ‘language processing condition’ and under ‘action condition’. We test preference for task trials. Participants choose the first element of a sequence, and observe the sequence of elements that ‘follow’ up to the one-before-last. Then they predict the last one. We observe what type of trials agents prefer (predictable or unpredictable) in each condition.
The language condition is induced by rewarding choosing any trial; not for making correct predictions. The action condition is induced by rewarding a correct prediction.
The postdoc carries out a thorough literature review on predictive coding under different conditions. And (s)he carries out a series of experiments exploring differences in markers of predictive coding before the target and at target onset (ERP) under conditions of language and action.