How we see relations: The case of containment and support

To navigate and act in the environment, it is necessary to recognize objects and the relationships between them. Recognition of objects in visual scenes is rapid and automatic. Are relations between objects in a scene also extracted rapidly and automatically during perception? Here, we considered containment and support, two relations that are fundamental to our understanding of the physical world. Using frequency-tagging electroencephalography on human subjects, we measured stimulus-driven neural responses during rapid visual presentation of images (4 Hz), with a periodic oddball structure to isolate the selective response to object pairs presented side-by-side (e.g., a book and a table) versus two objects in support or containment relations (e.g., a book on a table). We found enhanced responses to objects in support/containment relations (vs. objects side-by-side) when a relation was instantiated by different exemplars of the same scene (e.g., a book on a table), but not by different scenes involving distinct object categories (e.g., a book on a table, a lamp on a stand, a laptop on a desk). These findings indicate that, during scene perception, relations are extracted in a format that generalizes within, but not across, types of visual scenes. We propose that, during perception, relations are not represented as discrete, abstract tokens ("in" or "on") but rather as objects' properties, such that changes in the objects entail changes in the relational representation. These results provide new insight into the perceptual foundations of relational processing underlying physical reasoning. (PsycInfo Database Record (c) 2026 APA, all rights reserved).