Journal ofNeurology, Neurosurgery, and Psychiatry 1997;62:261-264 261

Impairment of facial recognition in patients with J Neurol Neurosurg Psychiatry: first published as 10.1136/jnnp.62.3.261 on 1 March 1997. Downloaded from right cerebral infarcts quantified by computer aided "morphing"

A Rosler, S Lanquillon, 0 Dippel, H J Braune

Abstract It is unlikely that we are able to distinguish any Objective-To investigate where facial other object to such a degree as we can human recognition is located anatomically and to faces. The idea that recognition of a human establish whether there is a graded transi- face could be a function which is independent tion from unimpaired recognition offaces of other recognition modalities of the , to complete after infarc- and therefore could also be independently dis- tions in the territory of the middle cere- turbed, was finally established by Bodamer in bral artery. 1947.' Previous to this reports of disturbances Methods-A computerised morphing in facial recognition had been made by program was developed which shows 30 Willebrand2 and Charcot.3 If we assume an frames gradually changing from portrait isolated function of facial recognition, two photographs of unfamiliar persons to substantial issues arise: (1) Where is it located those of well known persons. With a stan- anatomically? (2) Is there a graded transition dardised protocol, 31 patients with right from of faces to com- Department of unimpaired recognition and left sided infarctions in the territory plete prosopagnosia? A Rosler of the and an age H J Braune and sex matched control group were com- Department of pared by non-parametric tests. Patients and methods Psychiatry Results and conclusion-Facial recogni- To elucidate these questions, 31 right handed S Lanquillon tion in patients with right sided lesions patients with ischaemic infarcts in the Institute of Computer supply Science, Philipps was significantly impaired compared with area of the middle cerebral artery (15 left sided University, Marburg, controls and with patients with left sided and 16 right sided infarcts) were studied and Germany lesions. A graded impairment in facial compared with 21 age and sex matched con- 0 Dippel recognition in patients with right sided trols. The distribution of Barthel indices was Correspondence to: Dr Alexander Rosler, ischaemic infarcts in the territory of the the same for left and right sided infarcts (mean University Hospital middle cerebral artery seems to exist. = 84 and median = 90 for left and mean = Marburg, Rudolf-Bultmann Strasse 8, 35033 Marburg, 84 and median = 95 for right sided infarcts). Germany. (7 Neurol Neurosurg Psychiatry 1997;62:261-264) The following inclusion criteria were applied: Received 4 March 1996 cranial CT confirmed a single infarction in the and in revised form supply area of the middle cerebral 19 September 1996 artery (table http://jnnp.bmj.com/ Accepted 30 October 1996 Keywords: prosopagnosia; facial recognition; 1 shows the topographical distribution); no

:#:.. _.: w__ r.>s #._ _.:E ^:

wr: oF .: S' '-° At i *': ' S,9tt

'. .g; r on September 24, 2021 by guest. Protected copyright. *;l' tdjle;j3 [' ;b Ctfe [.: .e i;;'0 '* _E

o S..tzl-tlllrx...... ' ....tR a1llL 4 - tI-;ll-pt:1110to-raph .tC .:fi,^., wm0- i nE... aSijrtilnc,...... frln.li : i Figure 1 Selection offrames illustrating the morphing process. 262 Rosler, Lanquillon, Braune, Dippel

Table I Distribution of infarcts by CT J Neurol Neurosurg Psychiatry: first published as 10.1136/jnnp.62.3.261 on 1 March 1997. Downloaded from Regio7n Right inifarcts Left infarcts Cortical: Temporo-occipital 4 3 Temporoparieto-occipital 5 4 Frontotemporoparietal 1 2 Subcortical: and internal capsule 4 3 2 3

Table 2 Results with frame Diana Attenmpts s< 25 26-30 Failed Controls** 16 (76 2) 2 (9-5) 3 (14 3) Left infarcts 6 (40) 4 (26-7) 5 (33-3) Right infarcts** 3 (18-8) 6 (37-5) 7 (43-8) Percentages are in parentheses. **P = 0-002. Figure 2 Group comparisons of attempts to identify Diana. Table 3 Results with frame Kohl Attemipts previous or multiple infarcts; tests were done < 24 25-30 Failed in the second week after the stroke; exclusion Controls** 16 (76 2) 5 (23-8) 0 (0) of patients with moderate or severe dementia Left infarctst 9 (60) 6 (40) 0 (0) Right infarcts**t 5 (31-2) 7 (43-8) 4 (25) by mini mental status examination4; and exclusion of patients with receptive by Percentages are in parentheses. tP = 0 034; **p = 0-004. the Aachener aphasia test.5 Each patient was shown two series of 30 pictures on a video screen (fig 1). In the first Table 4 Effect of cortical and subcortical lesions on results with frame Diana series a photo of Jil Sander turned into a por- trait of Lady Diana; in the second Michael Attempts Douglas was transformed into Helmut Kohl. < 25 26-30 Failed Patients were given 30 seconds with each to Diana and Helmut Cortical: frame recognise Lady Left infarcts* 3 2 4 Kohl. The number of frames needed to iden- Right infarcts* 0 5 5 either name or social function Subcortical: tify the target by Left infarcts 3 2 1 was counted. Characteristics which could have Right infarcts 3 1 2 aided in the recognition of the prominent per- *P = 0-06 (NS). son-for example, hair, glasses, or a tie were not shown in the photo. Based on the results of the controls for both Table 5 Effect of cortical and subcortical lesions on results with frame Kohl series of pictures, categories were developed http://jnnp.bmj.com/ Attempts for normal, late, and non-identification. Seventy five per cent of the controls recog- < 24 25-30 Failed nised Diana within 25 and Kohl within 24 Cortical: attempts; therefore these numbers were set as Left infarcts** 5 4 0 for "normal" Right infarcts** 0 6 4 arbitrary upper boundaries Subcortical: recognition. More attempts up to the end Left infarcts 4 2 0 "late" and no on September 24, 2021 by guest. Protected copyright. Right infarcts 5 1 0 point of 30 were considered identification at all as "failed". Patients or **P = 0-001. controls who did not know the target persons even after being told their names, were excluded from analysis. Results were analysed Table 6 Effect ofcortical brain region involved in right sided infarcts with frame Diana by maximum likelihood X2 tests.67 Attemipts < 25 26-30 Failed Results Temporoparieto-occipital 0 1 4 Temporo-occipital 0 3 1 Healthy controls recognised Helmut Kohl early in 16 out of 21 cases and late in the remaining five. Diana was recognised early in 16, late in two, and not at all in three cases. Patients with left sided infarcts recognised Diana and Kohl later than controls but this Table 7 Effect ofcortical brain region involved in right sided infarcts with frame Kohl was not significant (P = 0-086 for Diana; P = Attempts 0-58 for Kohl). Patients with right sided infarcts identified s< 24 25-30 Failed Diana and Kohl significantly later than the Temporoparieto-occipital 0 1 4 controls (P = 0-002 for Diana and P = 0 004 Temporo-occipital 0 4 0 for Kohl). Impairment offacial recognition in patients with right cerebral infarcts quantified by computer aided "morphing" 263

Figure 3 Group The disorder showed associated signs such J Neurol Neurosurg Psychiatry: first published as 10.1136/jnnp.62.3.261 on 1 March 1997. Downloaded from comparisons of attempts to as left visual field defects in 91% of the cases, identify Kohl. disturbances of spatial orientation, construc- 100 tive apraxia, and neglect. This enumeration alone indicates that right sided lesions are .-w 80 ull obviously more often related to impaired facial ull 60 Iz recognition. The high number of combina- 40 tions with left upper quadrantanopia led to the Z; assumption that the underlying lesion could be 2- 20 6S. S s^s_ in the right basal . Miller

.E ii ..--_ described delayed recognition of faces in <2 4 .. '_st patients who had undergone right temporal Infarcts leftt lobectomy.'4 In a more recent article, Evans et parietal lobe in right familiar faces, and the matching of unfamiliar sided lesions was associated with more recog- faces. 18 nition failures compared with temporo-occipital We know that object recognition is lesions alone (tables 6 and 7). No statistical processed by a ventral pathway constituted by analysis was carried out because of the few occipito-temporal cortices. On the other hand, patients in these subgroups. visuospatial processing of the visual scene is comprised of parieto-occipital pathways. Different modalities for the recognition of Discussion faces play a synergistic part. When healthy Our patients with right sided infarcts seemed persons have to discriminate elementary fea-

to have more difficulties in identifiying familiar tures like colour, the prestriate cortex and in http://jnnp.bmj.com/ faces than healthy controls and patients with particular the fusiform gyrus is activated. left sided infarctions. Perceptual integration takes place in the poste- The neuroanatomical localisation of facial rior part of the lingual gyrus and, in a third recognition and its disturbances still remain step, especially important for the recognition controversial. Kolb and Wishaw postulated of faces, mainly the anterior part of the lingual that the damage which leads to prosopagnosia gyrus and the parahippocampic gyrus are acti- must be bilateral. Brodman's areas 18 and 19, vated in attributing meaning to the "per- and additionally 20, 21, and 37 seem to play a cept". 19 on September 24, 2021 by guest. Protected copyright. part in facial recognition.8 Kandel too argued Despite this well established topographical that "lesions that cause prosopagnosia are association between the visual recognition always bilateral".9 In a comprehensive article process and the brain region involved, there is by Meadows'0 there were seven postmortem still no consensus on the question of lateralisa- results available from patients who had had tion of different facial recognition modalities. prosopagnosia. Five of these had bilateral For instance, results of analysing emotional occipitotemporal lesions, mostly symmetric, in facial expression are controversial: accordance with the study by Damasio et al.11 Young et al examined 32 servicemen 40 The two other patients had right sided occipi- years after they had sustained a penetrating totemporal lesions. missile injury to the brain in the second world Young et al reported a 51 year old woman war and found that expression analysis was who showed a deficit in recognising faces and selectively impaired in five patients with left matching photographs of facial emotional hemispheric lesions,20 although earlier this task expression after bilateral partial amygdalo- had mainly been thought to be located in the tomy.'2 right hemisphere.2' 22 On the other hand there is little doubt that Superiority of the right hemisphere was also the right hemisphere plays an important part claimed for the function of visual attention in facial recognition: Hecaen and Angeluerges according to Dimmond and Heilmann.2324 It made a list of neurological signs which were could therefore be argued that the impairment positively associated with prosopagnosia.'3 of facial recognition in our patients with right 264 Rosler, Lanquillon, Braune, Dippel

sided lesions might be partly due to failure of As Puschel and Zaidel showed, the Benton- J Neurol Neurosurg Psychiatry: first published as 10.1136/jnnp.62.3.261 on 1 March 1997. Downloaded from arousing visual interest in the emotional Van Allen test is not able to discriminate expression of the face shown. However, the between patients with damage in the left or lateralisation of visual attention remains con- right hemisphere.29 troversial: Posner showed that patients with Morphing programs with a variable number left or right parietal lesions were equally of features are widely available for personal impaired in directing their attention to stimuli computers and enable sensitive testing for in both visual fields.25 impairment of facial recognition. The studies mentioned were all carried out on patients with relatively longstanding brain defects, whereas in our patients the interval of one week 1 Bodamer J. Die Prosop-agnosie. Arch Psychiatr between the event of infarction and Nervenkrankheiten 1947;179:6-53. the morphing task was presumably too short to 2 Willebrand H. Ein Fall von Seelenblindheit und Hemian- opsie mit Sectionsbefund. Deutsch Z Nervenheilk 1892; allow selective testing for single modalities of 2:361-5. facial recognition. This view is supported by 3 Charcot JM. Un cas de suppression brusque et isolee de la in vision mentale des signes et des objets (formes et the well known finding other neuropsycho- couleurs ). ProgrMed 1883;11:568-74. logical syndromes for example, aphasia or 4 Folstein MF, Folstein SE, Hugh PR. "Mini-mental-state". A a practical method for grading the cognitive state of apraxia-that final distinction of different patients for the clinician. Jf Psychiatr Res 1975;12: 189-98. subforms is not usually possible in the acute 5 Huber W, Poeck K, Weniger D, Willmes K. Der Aachener Aphasie Test. Guttingen: Holzgrefe, 1983. phase of brain damage. 6 Bishop YMM, Fienberg SE, Holland PW. Discrete multi- In general we may state that the complex variate analysis. Cambridge, MA: MIT Press, 1975. 7 Fienberg SE. The analysis of cross-classified categorical data. function of facial recognition requires the Cambridge, MA: MIT Press, 1977. integrity of the right temporoparieto-occipital 8 Kolb B, Wishaw IQ. Fundamentals of human neuropsychol- This anatomical was ogy. New York: W H Freeman, 1990, 236-7. region. region damaged 9 Kandel ER, Schwartz JT, Jessel TM. Principles of neuro- to different extents in our patients. Three of sciences. New York: Elsevier, 1991, 458-9. 10 Meadows JC. The anatomical basis of prosopagnosia. J7 five patients with right temporoparieto-occipital Neurol Neurosurg Psychiatry 1974;37:489-50 1. lesions failed to recognise both target pictures, 11 Damasio AR, Damasio H, Van Hoesen GW. Prosopagnosia: anatomic basis and behavioural mecha- the remaining two of this group received lower nisms. Neurology 1982;32:33-41. scores (28 for Diana and "failed" for Kohl, 28 12 Young AW, Aggleton JP, Hellawell DJ, Johnson M, Brooks P, Hanley JR. Face processing impairments after amyg- for Kohl and "failed" for Diana) than the dalotomy. Brain 1995;118:15-24. group of patients with temporo-occipital 13 Hecaen H, Angelergues R. for faces (prosopag- nosia). Arch Neurol 1972;7:92-100 involvement, in which only one of four 14 Miller B. Visual recognition and recall after right temporal patients failed with Diana and all four recog- lobe excision in man. Neuropsychologia 1968;6:191-209. 15 Evans JJ, Heggs AJ, Antoun N, Hodges JR: Progressive nised Kohl. Despite the small case numbers in prosopagnosia associated with selective right temporal these subgroups, these results suggest a possible lobe atrophy-a new syndrome? Brain 1995;118:1-13 16 Boeri R, Salmaggi A. Prosopagnosia. Current Opinion in key role for the parietal region in facial recog- Neurology 1994;7:61-4. nition. 17 Sergent J, Poncet M. From covert to overt recognition of faces in a prosopagnosic patient. Brain 1990;113: In accordance with the results of Young and 989-1004. Tupler,2026 our patients with lesions involving 18 Bruce V, Young A. Understanding face recognition. Br Jf Psychol 1986;77:305-27. cortical regions were significantly impaired in 19 Ceccaldi M. Visual object agnosia. Handout, teaching our test of facial recognition, whereas the six course Neuropsychology. First Congress of the European Federation of Neurological Societies Marseille: France, patients with subcortical lesions only were not 1995:2-4. http://jnnp.bmj.com/ (tables 2 and 3). 20 Young AW, Newcombe F, de Haan EHF, Small M, Hay DC. Face perception after brain injury. Brain 1993;116: Our patients, like others with right hemi- 941-59. sphere damage, did not complain about diffi- 21 Etcoff NL. Selective attention to facial identity and facial emotion. Neuropsychologia 1984;22:281-95. culties recognising faces. Probably these 22 Suberi M, McKeever WF. Differential right hemisphere deficits are easy to compensate for, because memory storage of emotional and non-emotional faces. Neuropsychologia 1977;15:757-68. recognising a certain person "behind" a face in 23 Dimmond SJ, Beaumont JG. Differences in vigilance per- daily life is made easy by the perception of fur- formance of the right and left hemisphere. Cortex 1973;

9:259-65. on September 24, 2021 by guest. Protected copyright. ther visual and acoustic cues.27 24 Heilman KM, van den Abell T. Right hemisphere domi- Among the tested patients with right sided nance for mediating cerebral activation. Neuropsychologia there were none with 1979;17:315-21. cortical infarctions 25 Posner MI, Walker JA, Friedrich JR, Rafal RD. Effects of prosopagnosia in a "pure" sense, although parietal lobe injury on covered orienting of visual atten- tion. J Neurosci 1984;4:1863-74. there was a clear impairment in recognition of 26 Tupler LA, Coffey CE, Logue PE, Djang WT, Fangan familiar faces; graded disturbances in the sense SM. Neuropsychological importance of subcortical white matter hyperintensity. Arch Neurol 1992;49:1248-52. of a "prosophypognosia" in this common neu- 27 Von Cramon D, Mai N, Ziegler W, eds. Neuropsychologische rological disease do seem to exist. Diagnostik. London: Chapman and Hall, 1995, 29-33. 28 Benton AL, Hamsher K de S, Vamey N, Spreen 0. With the exception of the Benton-Van Allen Contributions to Neuropsychological assessment: a clinical facial recognition test there are only experi- manual. Oxford: Oxford University Press, 1983. 29 Puschel J, Zaidel E. The Benton-van Allen faces: a lateral- mental tests available for the function of facial ized tachistoscopic study. Neuropsychologia 1994;32: recognition.28 357-67.