- You wouldn’t drink from a glass if a fish is IN the glass, or would not grab a chair if one is already sitting ON it. How do we represent relations between things? And how does it relate to object and action recognition? New paper out osf.io/preprints/ps... @mmellon.bsky.social
- We presented scenes showing two people (or two objects). We designed a behavioral paradigm to compare how much it takes to recognize that: • There are people (vs other objects) • They are facing vs facing away from each other • They are dancing vs fightingMay 21, 2025 12:00
- We found that it takes no longer to access the category of objects (people vs other) in a scene, than to access the relation between them (facing vs non-facing), while recognizing their (inter)action (fight or dance) takes significantly longer.
- Two main observations: 1. A striking parallel between object and relation categorization suggests that relations between objects are extracted as rapidly as categorical object information; 2. Representations of objects and relations are the building blocks of scene recognition
- The paper includes an internal replication of the effects and a set of interesting analysis in the supplementary material, which nicely support the conclusions of the study.
