Revealing Multidimensional Representations of Actions

• Applicant: Professorin Dr. Angelika Lingnau
• Subject Area: General, Cognitive and Mathematical Psychology
• Term: since 2025
• Project identifier: Deutsche Forschungsgemeinschaft (DFG) - Project number 558213037

We constantly encounter actions performed by other people, e.g. when preparing a meal together or when performing team sports. Being able to understand these actions is not only important for the immediate planning of our own actions; it provides us with insights about what will happen next, it provides us with information about the person performing the action, and it lies at the core of our ability to learn new skills. Understanding the key principles underlying the ability to recognize and categorize actions is a fundamental question in Experimental Psychology and Cognitive Neuroscience. The current project is based on the hypothesis that actions – like objects – can be conceptualized as points in a multidimensional space, where dimensions correspond to psychologically meaningful distinctions between distinct action categories. Actions in close proximity within this space are expected to be perceived as more similar to one another than actions located farther apart. According to this hypothesis, action understanding involves the matching between an observed action and the storage of that action in the multidimensional space. This raises the question of how visual information and action knowledge are structured and represented, according to which rules perceptual evidence is matched with long-term representations, and how these representations and computations are implemented at the biological level. The current project aims to fill these gaps in the literature. This is made possible by several recent studies, including work from the applicant, that have developed behavioural and neuroimaging paradigms for examining these questions. The overall objective is to validate and evaluate the behavioral relevance of dimensions underlying a multidimensional action space, and to examine the neural representation of these dimensions in space and time. By transferring the concept of representational spaces to actions, the results of this project are expected to be of theoretical interest since they are likely going to contribute to ongoing debates regarding the algorithms, computations and neural implementations that form the basis of action understanding. Moreover, establishing an action space model makes it possible to quantify the similarity between different actions, which will be useful for the generation of hypotheses and for future experimental manipulations. Finally, though not the focus of the current project, potential future applications include the field of human-computer interactions, skill acquisition and training, physical therapy and automatic action recognition.

• DFG Programme: Research Grants