Spatial ability at two scales of representation: A meta-analysis