Modeling Spatial Abstraction during Mental Rotation

Andrew LovettUniversität Bremen
Holger SchultheisUniversität Bremen


We present a computational model of mental rotation and shape comparison. The model posits that spatial abstraction, in which spatial details are removed to simplify a representation, is a key skill underlying spatial ability. Shapes are represented as collections of two-dimensional parts, and abstraction is applied by merging parts or ignoring certain part features. Using the model, we simulate a classic mental rotation experiment, demonstrating how abstraction explains the study’s key finding. Finally, we compare the part-based approach to a previous edge-based approach, demonstrating that the current approach better explains human shape comparisons.


Modeling Spatial Abstraction during Mental Rotation (246 KB)

Back to Table of Contents