Abstract
Object-based affordances, as defined by the Gibsonian account, refer to the possibilities of interaction that an object offers to an agent. Within this conventional framework, affordances are operationalized as the agent’s automatic motor-system responses on objects, such as reaching to grasp. However, our capacity to perceive interactions extends beyond such *simple agent-object relationships*, to also include how objects might interact with other objects. For instance, we almost instinctively see a hook as a place for a raincoat to hang or a flat tabletop as a surface for objects to sit on. To explore these *agentive inferences of object-object interactions*, we conducted several experiments involving a distance reproduction task. Participants were asked to recreate the distance between two objects seen in a target image: (1) a cupholder and (2) either a cup or a box. Critically, the hook on the cupholder was oriented either upward (hook-up) or downward (hook-down). This allowed us to modulate the presence or absence of possible interactions between objects. In the hook-up condition, the possible interactions are present because the cup can be hung on the hook in a stable manner. In the hook-down condition, no such interaction is possible since the cup can no longer be hung. Meanwhile, regardless of the hook orientation, no such interactions are possible in the box condition. The results were clear and striking: The effect of the hook orientation on the accuracy of the reproduced distances was significantly different between the cup and the box conditions, even when all other variables were equated for. These results held across several direct replications. This phenomenon establishes a new domain of affordances that are as automatic as simple agent-object affordances. We suggest the existence of ‘object interaction fields’ in the mind that identify stable configurations of multi-object systems under gravity and goal-directed action.