PROSOPAGNOSIA MARILYN C. KAY, M.D., AND HARVEY S. LEVIN, PH.D. Galveston, Texas Three patients with prosopagnosia, an acquired inability to recognize familiar faces usually resulting from cerebrovascular insufficiency, had left-sided visual field defects and color vision abnormalities of central origin. Prosopagnosia, although clinically associated with posterior right hemispheric disease in most cases, represents a disconnection of both occipital poles from the final processing center for facial recognition in the right temporal lobe. This problem can profoundly affect everyday activities requiring visual recognition. Prosopagnosia, the inability to recognize familiar faces, is a rare subjective visual complaint. It usually accompanies more generalized visual agnosia secondary to cerebral vascular insufficiency. When it occurs as an isolated phenomenon, the patient cannot recognize previously familiar faces and uses the voice or visual cues such as a piece of clothing worn by the familiar person to recognize the visage presented. Because other localizing or recognizable neurologic signs and symptoms may be absent, these patients are sometimes thought to have functional complaints or are referred to an ophthalmologist for routine examination. We have examined three such patients. CASE REPORTS Case i-This 60-year-old right-handed man with insulin-dependent diabetes experienced bilateral visual blurring while driving in December 1979. When he arrived home, his wife noted that he was ataxic and that his speech was slurred. When he was Accepted for publication March 22, 1982. From the Department of Ophthalmology (Dr. Kay) and the Division of Neurosurgery (Dr. Levin), University of Texas Medical Branch, Galveston, Texas. Reprint requests to Marilyn C. Kay, M.D., The Eye Institute, Medical College of Wisconsin, 8700 W. Wisconsin Ave., Milwaukee, WI 53226. examined by his ophthalmologist two weeks later, his visual acuity was correctable to 6/9 (20/30) in both eyes. His pupillary light responses and ocular motility were normal. A slit-lamp examination disclosed mild nuclear cataracts in both eyes but no other abnormalities. Fieldmaster perimetry showed an almost complete left homonymous hemianopia to the 425apostilb stimulus. Ophthalmoscopy showed no evidence of diabetic retinopathy. A more detailed history disclosed that since the episode of visual blurring the patient could not recognize familiar people by their faces. If a neighbor arrived at his house, he did not know who it was until the person spoke. This facial recognition problem extended even to his wife. Colors appeared to be "washed out." He felt disoriented when his wife drove him through his neighborhood, although he was able to find his way around his own house. Because of continuing visual complaints and concern that he might have a progressive lesion, the patient was hospitalized for further examination two months later. At that time his visual acuity was R.E.: 6/9 (20/30)and L.E.: 617.5 (20/25). Pupillary response to light and near stimuli were normal, as was ocular motility. Goldmann perimetry (Fig. 1) showed a left homonymous hemianopia with some sparing superiorly. Optokinetic responses were symmetric horizontally and vertically. Ophthalmoscopic findings were normal. The general neurologic examination disclosed no abnormalities and the results of routine blood studies, including an erythrocyte sedimentation rate, were normal. Computed tomography showed bilateral hypodense lesions of the occipital lobes, with further extension on the right into the parietal lobe and some involvement of both temporal lobes as well. Cerebral arteriography showed a normal left posterior cerebral artery, but the right posterior cerebral artery was narrowed with decreased branching in the region of the calcarine cortex. When undergoing a neuropsychological evaluation ©AMERICAN JOURNAL OF OPHTHALMOLOGY 94:75-80, 1982 75 76 JOURNAL OF OPHTHALMOLOGY A~IERICAN LEFT JULY, 1982 RIGHT Fig. 1 (Kay and Levin). Case 1. Visual fields. at this time, the patient reported that his facial recognition problem had improved for relatives and acquaintances, although he still had trouble recognizing television performers. He was oriented to person, place, and time. He was able to name objects presented visually or tactilely; writing and reading comprehension were all average or better. Remote memory was well preserved. Severe impairment of his visual memory was demonstrated by testing his ability to draw geometric designs from memory. When the faces of famous statesmen and entertainers were presented, he showed good recognition. However, matching of unfamiliar faces was severely impaired. A Farnsworth-Munsell 100-hue test, performed two months after the hospitalization, showed a pattern of errors most pronounced in the greenyellow to blue sector, a pattern unlike that associated with congenital color blindness or diabetic retinopathy.l Case 2-On Feb. 4, 1981, this 26-year-old righthanded man underwent ligation of a pseudoaneurysm of the left vertebral artery secondary to a self-inflicted gunshot wound suffered eight months previously. He had had no neurologic complaints before surgery, but when he awoke in the recovery room he complained of being blind. Optokinetic responses were elicited at this time. When he was examined the next day he was distraught and tried to keep his eyes closed. Pupillary light responses were intact and ocular versions were full. Ophthalmoscopic findings were normal. The patient reported retrospectively that during this postoperative period his vision seemed "dream-like" and was "in black and white." Computed tomography done at this time showed large lucent areas in both occipital lobes corresponding to both posterior cerebral arteries. Throughout his two-week hospital stay, the patient did not recognize close relatives until they spoke. Computed tomography performed three weeks after surgery showed bilateral hemorrhagic occipital lobe infarctions (Fig. 2). At this time neuropsychological testing showed intact reading and oral comprehension as well as normal naming of objects presented visually or described orally. Visual memory for line designs was impaired. Recognition of photographs of famous people was very poor, yet Fig. 2 (Kay and Levin). Case 2. Hemorrhagic infarction in both occipital lobes. VOL. 94, NO. 1 77 PROSOPAGNOSIA matching of unfamiliar faces was within normal limits. Six weeks after surgery his uncorrected visual acuity was 6/5 (20/15) in both eyes. There was I mm of anisocoria but pupillary light and near responses were normal. Ocular versions were full and optokinetic responses were symmetric. Goldmann perimetry showed a dense right homonymous hemianopia as well as involvement of the left quadrant superiorly in both eyes (Fig. 3). Ophthalmoscopic findings were normal. Although the patient reported that his facial recognition problems and color vision were improved at this time, he still performed poorly on matching famous faces and testing with Farnsworth-Munsell lOO-hue test showed a pattern of errors most pronounced in the green to purple-blue sector. Case 3-This 50-year-old right-handed man underwent a technically difficult right carotid endarterectomy because of transient left-sided visual field loss and arm weakness in April 1981. Because he complained to his surgeon that "faces were blank" and that he could not recognize his wife until she spoke, he was referred one week postoperatively for a neuroophthalmic examination. He also complained that he had trouble seeing to the left, although he believed that his facial recognition problem had improved by the time of the examination. He denied having any color vision problems. The patient was depressed and had a short attention span. His uncorrected visual acuity was 6/6 (20/20) in both eyes. Pupillary responses and ocular versions were normal, and optokinetic testing showed a slightly decreased response to a right-going target. Aimark perimetry showed an incomplete left homonymous hemianopia. Ophthalmoscopy showed many cotton-wool patches LEFT around the right optic disk; the fundus of the left eye was normal. Two weeks later, Goldmann perimetry showed a left inferior quadrantanopia in each eye; it was denser in the right eye (Fig. 4). Neuropsychological testing at this time showed that recognition of famous faces was normal, and that discrimination of unfamiliar faces was borderline normal. The FarnsworthMunsell lOO-hue test showed response errors primarily in the purple-blue sector. A second color test done ten weeks after the first test showed errors not only in the purple-blue sector but also in the greenyellow sector. He was able to name objects presented tactilely and visually in a normal fashion. Computed tomography done at this time showed a small right parietal lobe infarct. DISCUSSION Prosopagnosia is a disturbing subjective complaint that is rarely reported as an isolated neurologic symptom. It usually occurs as part of the spectrum of generalized visual agnosia associated with diffuse cerebrovascular disease. When the patient's only problem is difficulty in recognizing familiar faces, the condition may be mistakenly diagnosed as functional because of the patient's apparent ability to function well visually otherwise. The' recognition problem may even extend to the patient's own face, so that looking RIGHT Fig. 3 (Kay and Levin). Case 2. Visual fields. 78 AMERICAN JOURNAL OF OPHTHALMOLOGY JULY, 1982 270 LEFT RIGHT Fig. 4 (Kay and Levin). Case 3. Visual fields. into the mirror can be unsettling.f Patients who are farmers or birdwatchers may be unable to recognize previously familiar animals." Why is facial recognition specifically impaired when object recognition is not? We know that facial recognition is an important feature of an infant's early development, one that is separate from other acquired visual abilities; the ability to recognize approximately 1,000 faces by visual cues alone during a lifetime suggests a unique faculty." The problem does not lie with discrimination of subtle facial characteristics, because, as Levin, Hamsher, and Benton! have shown, the ability to recognize unfamiliar faces may be preserved despite prosopagnosia, as happened in our Case 2. Our Patients 1 and 3 recovered the ability to recognize familiar faces although recognition of unfamiliar faces remained abnormal. This strongly suggests that there are different pathways for the coding of familiar and unfamiliar facial characteristics. 6,7 Prosopagnosia was once thought to result from impaired general mental func- tion, because in the early case reports the patients were demented and usually suffered from visual object agnosia as well. 7 However, many more recent case reports, including our three cases, have shown that prosopagnosia occurs in the presence of intact orientation, verbal function, and visual object recognition. 2,3,7-10 Some investigators have suggested that restricted visual fields or decreased visual acuity prevent adequate viewing of facial details. 9,11,12 This does not explain why the ability to recognize unfamiliar faces is intact in some patients. Of course, there are many patients with substantial visual field defects who do not complain of facial recognition problerns.P Patients whose visual fields are extremely constricted because of retinitis pigmentosa or glaucoma do not report specific facial recognition difficulties. Finally, prosopagnosia has been reported in the absence of any demonstrable visual field defect.v? Poor visual acuity is not the cause of prosopagnosia. Our patients, like others described previously, 7,10,14,15 had good to VOL. 94, NO. 1 PROSOPAGNOSIA excellent visual acuities. The inability to "put together" fine details to form a picture of the face being viewed is not the problem because these patients can retain the ability to discriminate subtle differences in unfamiliar faces. 7 Rather, it seems that the pathway from the occipital lobes to the right temporal lobe, the center for famous face recognition," is disturbed. Where is the lesion? Reviewing the previously reported cases suggested a clinical correlation between a left-sided visual field defect and prosopagnosia. Meadows," in 1974, in a review of 42 cases of prosopagnosia, found that 34 patients (19 with unilateral and 15 with bilateral visual field defects) had leftsided visual field defects. All involved the left upper quadrant. He suggested that a lesion involving the right occipital radiation, along its inferior aspect, might cause such visual field loss and also involve pathways from the occipital lobe to the right temporal lobe. Meadows also reviewed the reported autopsies of prosopagnosic patients and found that the area of the right fusiform gyrus, specifically the inferior longitudinal fasciculus along the medial inferior aspect of the occipital lobe just external to the optic radiation, was always involved. This finding was substantiated by other autopsy studies in which fusiform gyrus destruction was found to be associated with interruption of the occipitotemporal paths.v" Is it possible for a unilateral lesion to cause a global facial recognition problem? Published reports of autopsies and computed tomographic flndings" and our own Cases 1 and 2 indicate that a right-sided lesion causes the left-sided visual field defect but that a second, left-sided, lesion is required to disconnect the patient's visual pathways from the facial memory center in the right temporal lobe. Meadows' found that all seven of his autopsied 79 patients had clinically undetected lefthemisphere disease. The patient described by Benson, Segarra, and Albert" had bilateral fusiform gyrus destruction, although both calcarine cortices were intact, pointing out the importance of the occipitotemporal path in relaying raw visual information. One patient of Cohn, Neumann, and Wood 9 had bilateral fusiform gyrus destruction and loss of the right occipital pole. The left calcarine cortex was intact, as were the left areas 18 and 19. However, they suggested that destruction of the left fusiform gyrus made it impossible to transmit visual information from the left hemisphere for further encoding. Our Patient 3, whose only clinically evident disease was in the right hemisphere, might have had leftsided infarcts that were not detected by computed tomography, thus explaining his prosopagnosia. Two of our three patients complained of color vision abnormalities associated with the onset of prosopagnosia and all three had persistent abnormal responses when they were tested with the Farnsworth-Munsell 100-hue test. These color sorting problems are not an agnosia or a naming problem but rather one of hue discrimination in cerebral achromatopia." This clinical problem is often associated with prosopagnosia and may lead to perplexing subjective complaints. 10,15,17 It also appears to represent bilateral inferior occipital lesions, specifically those involving the junction of the fusiform and lingual gyrus just anterior to the striate cortex. 17 In a report of five cases of cerebral achromatopia, Green and Lessell" demonstrated the importance of bilateral lesions in achromatopia. Three of their patients had had normal color vision despite unilateral cerebral lesions but developed color vision difficulties when the other hemisphere became involved. Meadows" has also suggested that simple 80 AMERICAN JOURNAL OF OPHTHALMOLOGY color VISIOn screening tests, such as Ishihara plates, may not detect all cases of cerebral achromatopia. He suggested that more subtle hue discrimination difficulties may be detected only by more detailed tests such as the FarnsworthMunsell lOO-hue test. Although most cases of prosopagnosia are secondary to cerebral vascular insufficiency, as in our three cases, isolated cases have been reported secondary to encephalopathy, 7 closed head trauma.? tumor.v" temporal lobectomy," and carbon monoxide poisoning." This explains why most patients improve as new collateral vessels form and functioning of the cerebral tissue improves. REFERENCES 1. Ramsay, W. J., Ramsay, R. c., Purple, R. L., and Knobloch, W. H.: Involutional diabetic retinopathy. Am. J. Ophthalmol. 84:851, 1977. 2. MaCrae, D., and Trolle, E.: The defect of function in visual agnosia. Brain 79:94, 1956. 3. Bomstein, B., Sroka, H., and Munitz, H.: Prosopagnosia with animal face agnosia. Cortex 5:164, 1969. 4. Meadows, J. C.: The anatomical basis of prosopagnosia. J. Neurol. Neurosurg. Psychiatry 37:489, 1974. JULY, 1982 5. Levin, H. S., Hamsher, K. S., and Benton, A. L.: A short form of the test of facial recognition for clinical use. J. Psychol. 91:223, 1975. 6. Warrington, E. K., and James, M.: An experimental investigation of facial recognition in patients with unilateral cerebral lesions. Cortex 3:317, 1967. 7. Benton, A. L., and Van Allen, M. W.: Prosopagnosia and facial discrimination. J. Neurol. Sci. 15:167, 1972. 8. Benton, A. L.: The neuropsychology of facial recognition. Am. Psychol. 35:176, 1980. 9. Cohn, R., Neumann, M. A., and Wood, D. H.: Prosopagnosia. A clinico-pathological study. Ann. Neurol. 1:177, 1977. 10. Whiteley, A. M., and Warrington, E. K.: Prosopagnosia. A clinical, psychological, and anatomic study of three patients. J. Neurol. Neurosurg. Psychiatry 40:395, 1977. 11. Bay, E.: Disturbances of visual perception and their examination. Brain 76:515, 1953. 12. Bender, M. D., and Feldman, M.: The socalled visual agnosias. Brain 95:173, 1972. 13. Hamsher, K. de S., Levin, H. S., and Benton, A. L.: Facial recognition in patients with focal brain lesions. Arch. Neurol. 36:837, 1979. 14. Rubens, A. B., and Benson, D. F.: Associative visual agnosia. Arch. Neurol. 24:305, 1971. 15. Green, G., and Lessell, S.: Acquired cerebral dyschromatopsia. Arch. Ophthalmol. 95:121, 1977. 16. Benson, D. F., Segarra, J., and Albert, M. L.: Visual agnosia-prosopagnosia. Arch. Neurol. 30:307, 1974. 17. Meadows, J. C.: Disturbed perception of colours associated with localized cerebral lesions. Brain 97:615, 1974.