How do we recognize other people’s actions? What does it mean to open something (e.g. a box, your mouth, or your mind)?
Our goal is to elucidate the neural pathway of action recognition and understanding – from the perception of single entities (e.g. body parts, objects) and their movements in space to more general, conceptual representations that capture, for example, what open a box and open your mind have in common.
A key aspect of this research focuses on how different aspects of action knowledge are topographically organized on the cortical surface and what determines this topography. Thereby, we seek to improve our understanding of the principles of knowledge organization in the brain.
We mainly use neuroimaging techniques (fMRI, MEG), in typical and special populations, to address these questions.
Identifying predictive representations using dynamic representational similarity models
How is dynamic visual input, such as the body movements of acting persons, represented in the brain? Do we represent dynamic stimuli in a lagged manner, or does the brain generate predictions of unfolding dynamics? We developed a novel dynamic extension to MEG-based RSA that uses temporally variable models, i.e. a unique model for each time point, to capture the representational nature of dynamic stimuli. This approach allows testing at which time point different aspects of dynamic stimuli - from low-level visual to body posture and kinematic features - are represented in the brain. Thereby, we aim at identifying and characterizing predictive representations that capture future states of the dynamic world.
Disentangling key components of action recognition
To recognize physical actions (e.g. "giving"), we have to process body movements (e.g. reaching and grasping) and how these movements causally impact on objects in the world (e.g. a change of object location). Which parts of the brain are involved in processing these distinct stages? Using fMRI-based multivariate pattern decoding across different stimulus formats, we aim at isolating different stages of action recognition in the brain.
The neural basis of high-level action-object integration
To date, the neural systems for action recognition and object recognition have been studied largely separately from each other. However, to understand what other people are doing and why, it is not sufficient to recognize actions (e.g. "opening") and objects ("present box") - at a higher level, conceptual knowledge about actions and objects needs to be integrated ("opening present"). This fMRI project aims at identifying neural representations that accomplish this integration. Thereby, our goal is to elucidate the bridge between action and object pathways and higher-level systems necessary for action interpretation and mental state inference (e.g. that the acting person is curious and happy).
- Moritz Wurm, Principal Investigator
- Nadezhda Murziakova, Master Student
- Paula Faimann, Master Student
- Philipp Flieger, Master Student
For a complete list see Moritz Wurm personal page
- Alfonso Caramazza (Harvard University, USA)
- Olivier Collignon (Université catholique de Louvain, Belgium)
- Floris De Lange (Donders/Radboud University, Netherlands)
- Ingmar de Vries (Donders/Radboud University, Netherlands; CIMeC)
- Angelika Lingnau (University of Regensburg, Germany)
- Ricarda I. Schubotz (University of Münster, Germany)
- Gilles Vannuscorps (Université catholique de Louvain, Belgium)