A theory is proposed to account for unawareness of blindness, hemianopsia, and hemiplegia, and for phantom limb after amputation. It is assumed that interruption of a sensory pathway at any level--from peripheral nerve to primary sensory cortex--is not associated with any immediate sensory experience that uniquely specifies the defect. Instead the sensory loss must be discovered by a process of self-observation and inference. Discovery is easy for defects that create major functional disability, such as total blindness. Hence unawareness of total blindness occurs only in association with severe intellectual impairment, precluding the required self-observation and inference. In contrast, hemianopsia is difficult to discover because several mechanisms automatically compensate the defect effectively. Thus unawareness of hemianopsia is common, even in intellectually normal individuals. Insensate fields are often the source of suggested (false) percepts, because no information from such a field specifies the absence of a sensory stimulus. The most powerful source of suggestion is sensory activity in uninvolved portions of the affected sensory field. Thus hemianopsics may perceive complete geometric forms when only incomplete forms are shown and the missing portion falls in the hemianopsic fields. Such perceptual completion also occurs in hemianesthetic hemiplegics, creating the illusion that there are normally functioning limbs on the affected side. This perceptual completion increases the difficulty of discovery of hemianesthetic hemiplegia, but the disability is still sufficiently obvious that some additional cognitive impairment is invariably present in patients with lasting unawareness of hemiplegia. Phantom limb after amputation is the product of perceptual completion without associated cognitive impairment. The patient with phantom limb is thus aware of the illusory quality of his phantom. Some insight into the neural basis of perceptual completion and of unawareness of sensory loss may derive from considering sensory systems and associative cortex as parallel-distributed processing mechanisms.