RELATIONS

Human mental representations are fundamentally compositional, built upon the ability to represent and reason about relations between things. A structured, coherent representation of the world depends on how we represent things but, more importantly, their relations. Relations between people and objects define how we perceive and interact with them: when we see a group of people, we readily infer in which way they interact with each other, and thus decide to approach or avoid them. Or, when we see multiple objects, we can readily infer how to combine them for efficient usage. We propose that the roots and earliest stages of relational representations are grounded in perception: perceptual representations are inherently structured by meaningful relations among people and objects. RELATIONS addresses the visual representation of relations in the social and non-social domain. We will measure performance and brain responses during perception of visual scenes structured around meaningful relations between people (a person fighting with another) or objects (a hammer hitting a nail). We will thereby delineate differences and similarities in how social and non-social relations shape perception and brain responses, are acquired across development, and altered in disorders that affect social (and possibly also non-social) perception such as autism. RELATIONS involves a unique combination of complementary approaches. First, behavioral experiments will reveal how meaningful relations facilitate visual perception and working memory for people and objects. Second, spatially and temporally resolved neuroimaging will reveal to what extent relations between people and objects are implemented via the same or different brain mechanisms. Third, we will trace the development of relational representations in the social and non-social domains by measuring brain responses in infants, children, and adults and concurrently studying the emergence of relational representations in artificial neural networks. Finally, inspired by current theories of autism –and our preliminary research findings–, we will study whether perceptual alterations in autism can be characterized by poor processing of visual relational information. The results will shed light on how relations define visual representations for social and non-social visual contents, which may be precursors of the structured representations in language, memory, and thought. By charting the emergence of relation perception during development and its alteration in autism, RELATIONS will also provide new theoretical impulses for understanding differences and commonalities between social and non-social vision. By bringing together the complementary expertise and resources of the two applicants who are among the authorities in this emerging field, RELATIONS represents the foundation of a new concerted effort towards an understanding of relations in perception and cognition.