Testing and training geometrical knowledge in baboons through a conceptual categorization task

Symbolic match-to-sample training enabled baboons to learn and generalize geometric concepts for line perpendicularity and parallelism, but not length equality, without producing human-like quadrilateral regularity effects.

• Why it matters: The work addresses whether baboons’ lack of advantage for geometric regularity reflects limited geometry learning or a failure to build human-like compositional shape perception.
• What they did: Baboons and human controls were trained on a symbolic categorization task (A vs not-A) across two non-geometric domains and three geometric domains, with frequent new generalization stimuli.
• The result: Baboons performed above chance and generalized to novel exemplars for all categories except length equality, yet explicit geometry training did not make the quadrilateral intruder task show a geometric regularity effect.

AI summary of the abstract · may contain errors · not medical advice · read the paper.