Spatial proximity and scene grammar: shaping spatial representations for memory-guided actions in naturalistic environments

Spatial object representations are necessary for effective goal-directed actions. These representations can be produced with respect to the self (i.e., egocentrically) and/or to surrounding objects (i.e., allocentrically). For the latter, one (spatial) factor of influence is landmark proximity to target. This factor shapes allocentric representations and informs future memory-guided actions, in both impoverished and rich scenes. More recent work has suggested that cognitive factors (e.g., object semantics) can also influence allocentric coding. Here, we investigated how both spatial and cognitive factors affect allocentric coding in naturalistic environments. In a memory-guided virtual reality (VR) experiment, semantically-related targets (i.e., local objects, which are movable and manipulable, e.g. a cup) were presented on congruent or incongruent anchors (i.e., large, immovable objects, predictive of local object locations). After an imperceptible anchor shift (or not), a target had to be placed in its remembered position. Placement performance results indicated that anchors were, indeed, used to build up spatial target representations. Additionally, these representations showed strong effects of proximity-to-anchor, with no differential effects of phrase-level semantic congruency. Taken together, in our memory-guided task, we found that spatial proximity plays a dominant role in allocentric coding in naturalistic (virtual) scenes.