Integration of Novel Shape Templates During Human Spatial Navigation Leads to Prototype Extraction

Abstract

The ability to rapidly form stable representations of novel environments is crucial for successful navigation, yet the majority of studies involve environments composed of overlearned geometrical configurations, like rectangles. To address this gap in knowledge, participants underwent extensive pre-training with a non-regular shape and then navigated environments organized around this shape and other parametrically related novel non-regular shapes. Concurrent high-resolution fMRI indicated that, consistent with past studies on regular geometries, hippocampal neural activation patterns differentiated these environments, although with a bias toward the pre-exposed shape. Surprisingly, we found that hippocampal representations were most stable for the average geometric configuration of the novel environments, with subsequent map-drawing accuracy predictive of the stability of this average. Our findings challenge existing models of memory and navigation by suggesting unique mechanisms underlie the formation of novel spatial contexts, which may only become apparent under conditions that directly manipulate the influence of prior knowledge.