It is conventional to assess the status of the somatosensory system in such a way that patients cannot see whether or not they are being touched: either the part tested is shielded from view or the eyes are closed. The main clinical reason for this well-established constraint on method is to control for any expectations that the visual stimulus may generate in the patient. In this study we compare the assessment of light touch with and without visual feedback in 20 well-oriented and cooperative patients after stroke, and 20 age-matched controls. The results show that in 18 of the patients, visual feedback produced only small (or no) differences in reports of sensation. Controls showed a similar result to these patients; visual feedback produced only small effects upon reported touch. In two patients, by contrast, sensation was confidently and reliably reported only when they viewed the affected limb during assessment. In one of these patients, we demonstrate that "suggestibility" is not an appropriate description of the phenomenon. The conventional clinical method of somatosensory assessment would not have elicited this informative discrepancy between feeling in the two conditions of testing in these two patients. We accordingly suggest that it is valuable to test tactile perception with and without vision when diagnosing somatosensory loss after brain damage.