Mental rotation depends on the number of objects rather than on the number of image fragments.

It is often intuitively assumed that disconnected image fragments result in a representation of separate objects. When objects are partly occluded, disconnected image fragments can still result in a representation of a single object, based on visual completion. In a simultaneous matching task, displays showing one object, partly occluded objects, or two objects were compared with each other. When only a translation was required to match pairs of displays, one-object displays were matched faster than both occluded-object and two-object displays, which did not differ significantly from each other. When mental rotation and translation were required, the one-object displays were again matched the fastest. In addition, an advantage for occluded-object displays compared with two-object displays was found. We conclude that when the generation of a mental representation is likely, object-based connectedness determines object matching. Mental rotation then seems to depend on the number of objects rather than on the number of image fragments.