We report evidence from three sets of experiments dealing with spatially pa
rallel grouping of parts in single objects. A first set of experiments demo
nstrates that parts can be encoded in a spatially parallel manner in three-
dimensional (3-D) objects, while there is a serial selection of parts acros
s objects. A second set of experiments further shows that grouping in 3-D i
s less affected by eliminating collinearity between the parts of objects th
an grouping in two dimensions, suggesting that 3-D constraints operate dire
ctly on visual grouping. A final pair of experiments demonstrates that rota
ting the elements in the plane, to make a physically unstable 3-D object, d
isrupts the benefit found with 3-D stimuli when collinearity is eliminated.
The evidence indicates that there is rapid and spatially parallel encoding
of 3-D object descriptions in vision.