Articles

CME Covert recognition in acquired anddevelopmental prosopagnosia

Jason J.S. Barton, MD, PhD, FRCPC; Mariya Cherkasova, BA; and Margaret O’Connor, PhD

Article abstract—Background: Some patients with prosopagnosia have covert recognition, meaning that they retainsome familiarity or knowledge of facial identity of which they are not aware. Objective: To test the hypothesis thatprosopagnosic patients with right occipitotemporal lesions and impaired face perception lack covert processing, whereaspatients with associative prosopagnosia and bilateral anterior temporal lesions possess it. Methods: Eight patients withprosopagnosia were tested with a battery of four face recognition tests to determine their ability to discriminate betweenfamous and unknown faces. Results: Measures of overt familiarity revealed better residual discrimination in patients withacquired prosopagnosia than in those with the developmental form. With forced-choice methods using famous faces pairedwith unknown faces, no patient demonstrated covert familiarity. However, when the semantic cue of the name of thefamous face was provided, covert processing was present in all five patients with acquired prosopagnosia, including thethree with extensive right-sided lesions and impaired perceptual discrimination of facial configuration. Sorting unrecog-nized faces by occupation was also performed above chance in three of these five patients. In contrast, none of the threepatients with developmental prosopagnosia had covert processing, even though two demonstrated flawless performance onsimilar tests of name (rather than face) recognition. Overt familiarity correlated highly with the degree of covertrecognition. Conclusions: Extensive right occipitotemporal lesions with significant deficits in face perception are notincompatible with covert face processing. Covert processing is absent in developmental prosopagnosia, because thiscondition likely precludes the establishment of a store of accurate facial memories. The presence of covert processingcorrelates with the degree of residual overt familiarity, indicating that these are related phenomena.

NEUROLOGY 2001;57:1161–1168

Prosopagnosic patients deny familiarity with andcannot identify faces of people known to them. How-ever, a range of physiologic and behavioral techniqueshave shown in at least 12 patients that some retain“covert recognition” of these faces.1 Two different co-vert phenomena can be shown: covert familiarity, ordistinguishing known from unknown faces, and co-vert knowledge, or retained information about name,occupation, and other facts associated with a face.

For covert familiarity, differences between familiarand unfamiliar faces have been shown with physiologicmeasurements such as the electrodermal skin conduc-tance test2 and the P300 evoked potential.3 Behavioralmethods can also show covert familiarity. The se-quence of eye fixations of two prosopagnosic patientswas less stereotyped when they viewed familiar ratherthan unfamiliar faces.4 Judging whether two differentpictures portray the same person was more rapid anddependent on internal facial features with familiarfaces in normal subjects and also one prosopagnosicpatient.5 Matching the old and young faces of individu-

als across a 30-year span was easier with famous facesfor another prosopagnosic patient.6 The classification ofa target name as familiar or unfamiliar was affected bywhether a simultaneously shown face was related tothe name, indicating the presence of covert “priming.”7

For covert knowledge, greater electrodermal re-sponses were found in two patients when a name readaloud belonged to a famous face than when it did not.8,9

When forced to choose between a correct and incorrectname, one patient guessed better than chance,6 thoughanother did not for occupation.5 Several prosopagnosicpatients were better at learning to associate a famousface with a correct name than with an incorrectone.5,6,10,11 Covert processing has also been shown witha “facial interference” task in one patient.12 In this, thespeed to classify a name by occupation was faster if asimultaneously shown face was that of either thenamed person or someone with the same occupationthan if the face was of an unrelated person.

Conversely, some studies have shown lack of co-vert recognition in at least some prosopagnosic

See also pages 1151 and 1168

From the Departments of Neurology (Drs. Barton and O’Connor, and M. Cherkasova) and Ophthalmology (Dr. Barton), Beth Israel Deaconess MedicalCenter and Harvard Medical School; and Department of Biomedical Engineering, Boston University (Dr. Barton), MA.Supported by a grant from the National Institute of Neurological Disorders and Stroke (J.J.S.B.).Presented at the 53rd annual meeting of the American Academy of Neurology; Philadelphia, PA; May 9, 2001.Received December 28, 2000. Accepted in final form May 12, 2001.Address correspondence and reprint requests to Dr. J.J.S. Barton, Department of Neurology, KS 452, Beth Israel Deaconess Medical Center, 330 BrooklineAvenue, Boston, MA 02215; e-mail: jbarton@caregroup.harvard.edu

Copyright © 2001 by AAN Enterprises, Inc. 1161

patients.13-18 These include absence of covert famil-iarity, as measured by electrodermal17 or behavioralforced-choice indexes,16 and lack of covert knowledge,as demonstrated by absence of the expected benefitof correct names in learning name–face pairs,11,13-15

lack of priming of name classification by correct fac-es,15,16,19 and chance performance on forced-choicepresentations of names with faces.18 It has been sug-gested that absence of covert processing correlatedwith deficits that “(a) were perceptual in nature or(b) were not limited to defective face processing.”1

Others have also speculated that severe apperceptivedefects explain lack of covert processing.15,17,18

Conclusions about the implications of covert pro-cessing are hampered by several factors. First, mostreports on covert processing are single case studies.The largest sample of cases was three.18 Second,there are numerous ways to demonstrate covert pro-cessing, and some patients may show covert process-ing on one test but not another.18 This makes itdifficult to generalize by pooling data from severalstudies. Third, the classification of prosopagnosia asapperceptive or associative (in which face perceptionis purportedly intact and the recognition impairmentis attributed to impaired access of perceptual data tostores of facial memories) often depends on circum-stantial evidence whose validity is open to question.To show apperception, some studies rely upon testsof unfamiliar face matching such as the Benton facerecognition test (BFRT). However, many investiga-tors have pointed out that this may not be a validmeasure of the type of processing required to recog-nize familiar faces.18,20,21 Likewise, the ability tojudge expression, age, and sex from a face may provethat some face perception is impaired but not neces-sarily that the perceptual processes required for facerecognition are dysfunctional. This point deservesemphasis as selective defects for face expression thatspare face recognition and vice versa have been de-scribed.22 Even more tenuous in their relation to facerecognition are tests of nonface object perceptionsuch as the Street completion test, noncanonicalviews of objects, and Ghent overlapping figures. As-sertions that these tests are relevant to face percep-tion are based on the assumption that the latterinvolves “holistic” or Gestalt processes.

In this study, we report on tests of covert process-ing and overt familiarity in a series of eight patientswith prosopagnosia. We employed a uniform behav-ioral battery to detect either covert familiarity orsemantic knowledge about faces. Our goal was todetermine if the presence of covert processing corre-lated with either lesion location or status of faceperceptual processing. The latter was tested by psy-chophysical measures of the perception of the config-uration of facial features. There is mountingevidence in normal subjects that facial configuration(or spatial “second-order” relations) is an importantelement of normal face perception and underlies thisprocess’ vulnerability to orientation (the inversioneffect).23-25 Our previous reports have shown that

some prosopagnosic patients have severe problemswith perceiving spatial configuration.26 The guidinghypothesis was that covert processing would be ab-sent in those patients severely impaired in perceiv-ing facial configuration. Furthermore, as it has beenproposed that bilateral temporal lesions cause asso-ciative prosopagnosia and right occipital lesionscause apperceptive prosopagnosia,27 covert process-ing should also correlate with the site of damage.

Methods. Subjects. Eight prosopagnosic patients par-ticipated. Details of their history are provided elsewhere.26

Three (Patients 1 through 3) have developmental prosop-agnosia. Patient 1 is a 22-year-old man whose difficultieshave been attributed to traumatic and anoxic encephalop-athy at age 1 year.28 Patient 2, a 31-year-old woman, hadno defined insult but has had epilepsy since infancy. Herprosopagnosia is associated with a milder general visualagnosia but without alexia. Patient 3 is a 36-year-oldwoman who experienced a respiratory arrest at age 6. She,too, has a more generalized visual agnosia. Patient 1 hasnormal MRI. Patient 2 has a focal medial occipital polymi-crogyria, and Patient 3 has generalized posterior cerebralatrophy. All three are severely impaired on our tests offacial configurational discrimination.23

Five have acquired prosopagnosia. Patient 4 is a 37-year-old woman who had viral encephalitis at age 17. HerMRI showed extensive lesions of the right medial occipito-temporal cortex and lesser lesions of the left parahip-pocampal gyrus. Patient 5 is a 59-year-old man with aright posterior cerebral artery infarction at age 58. Patient6 is a 52-year-old man seen 7 months after a right occipitalhemorrhage and resection of an underlying oligodendrogli-oma. In keeping with their extensive right occipitotempo-ral damage, Patients 4, 5, and 6 were severely impaired indiscrimination of facial configuration.26 Patient 7, a 38-year-old man, sustained bilateral posterior occipitotempo-ral damage from a gunshot wound 18 years previously. He,too, is severely impaired in perceiving facial configuration.In contrast, Patient 8, a 33-year-old woman, had bilateralanterior temporal lobe damage from a combination of asevere closed-head injury and the surgical resection at age23 and has normal perception of facial configuration.

As a lesion control subject, we provide contrasting datafrom a nonprosopagnosic patient, Patient 9, who is a 41-year-old woman tested 8 months after an infarction involv-ing the entire distribution of the right posterior cerebralartery. Also, seven normal subjects of mean age 31 years(range 22 to 54 years) were given the test of overtfamiliarity.

Procedure and apparatus. Among established neuro-psychological tests, we used the BFRT and Warringtonface recognition test (WFRT). The BFRT tests the percep-tion of unfamiliar faces by asking subjects to find thematch to a target face from a row of unknown faces, some-times varying in lighting or angle of view. The WFRT testsshort-term memory for faces by asking subjects to viewfaces and then indicate which of a larger series of faceswere the ones they had just seen.

To assess overt familiarity, we presented patients witha series of 20 famous and 20 unfamiliar faces in randomorder and asked them to identify which were familiar andto name them, if possible. The famous faces were taken

1162 NEUROLOGY 57 October (1 of 2) 2001

mainly from show business or politics, spanning a largetime period from the 1940s to the present. From their hitrates (famous faces identified as familiar) and false-alarmrates (nonfamous faces identified as familiar), we con-structed measures of their discriminative ability (d�) usingsignal detection theory methods. In addition to confirmingtheir recognition deficit and establishing their baseline fa-miliarity with our stimuli, this test also allowed us toexamine yet another interesting aspect of abnormal facerecognition: the phenomenon of “false recognition of unfa-miliar faces.”29-31

To assess covert processing, we presented three behav-ioral face tests, modeled after a subset of those used bySergent and Signoret.18 These employ a “direct” forced-choice strategy,32 similar to that frequently used by blind-sight studies to probe implicit performance. (This contrastswith indirect strategies, in which face knowledge is re-vealed as an influence on responses in other tasks.)

1. Forced-choice familiarity with paired faces. This testedjudgments of familiarity using 20 pairs of faces. These40 faces were the same faces used to probe false recog-nition of unfamiliar faces. Each famous face was pairedwith an unfamiliar face of similar age and sex and pos-sessing similar local features such as glasses and hair-style. The subject was asked to indicate which face wasthe well-known one.

2. Forced-choice familiarity, cued by name. This testedwhether this familiarity judgment could be augmentedby semantic information about the famous face. Withthe same set of faces, the subject was told the name ofthe famous person in the pair and asked to indicatewhich face belonged to that name. This test was admin-

istered �30 minutes after the first test and after otherintervening experimental tests.

3. Sorting by occupation. The last test used 41 differentfaces, 21 of people known through show business and 20of people connected with political events. The subjectwas asked to sort the faces into two groups by these“occupations.” They were not told how many there wereof each group.

To analyze the data from these three covert tests, wefirst discarded trials or faces in which the subject was ableto name the face or provide other correct biographic data,indicating overt recognition. Of the remaining trials, bino-mial proportions were used to estimate how many correctresponses were needed to exceed the 95% probability thatthe result was obtained by chance.

As controls for nonvisual semantic knowledge, we usedtwo tests that employed written names instead of faces.One was a paired-name test, in which one famous namewas paired with an invented name matched for ethnicorigin. The famous names were again chosen to sampleknowledge across many decades. In the other name test,the names of the 41 people whose faces were used in theoccupation-sorting test were provided, and the subject wasasked to sort the names by occupations. Name controlswere done after the subjects had completed the face teststo avoid inadvertent semantic priming.

Results. All subjects except Patient 2 had a sense offamiliarity with a few of the famous faces: In particular,Patients 4 and 5 rated half of the 20 famous faces asfamiliar (table). However, the patients who had higherrates of claimed familiarity with famous faces also had

Table Performance (fraction correct) on face tests

Tests

Acquired prosopagnosia

Bilateral Right occipitotemporal Developmental prosopagnosia Control

Patient 8 Patient 7 Patient 4 Patient 5 Patient 6 Patient 1 Patient 2 Patient 3 Patient 9

Benton face recognition test, n � 54 25 32 39 35 32 39 39 28

Warrington face test, n � 50 13 32 29 33 30 30 23 29

Overt face familiarity test

Familiar face, n � 20 0.10 0.30 0.50 0.50 0.20 0.05 0.00 0.45 0.80

Unfamiliar face, n � 20 0.00 0.35 0.15 0.25 0.00 0.15 0.25 0.45 0.00

d� 0.68 �0.14 1.04 0.67 1.12 �0.61 �1.29 0.00 2.80

Covert tests using faces

Paired faces, n � 20 0.15* 0.55* 0.50* 0.40* 0.56* 0.24* 0.35* 0.45* 0.90†

Paired faces, name cue, n � 20 0.79† 0.75† 0.75† 0.80‡ 0.81† 0.65* 0.50* 0.65* 0.90†

Occupation sorting, n � 41 0.83† 0.56* 0.55* 0.74† 0.65§ 0.44* 0.41* 0.51* 0.88†

Control tests using names

Paired names, n � 20 1.00 0.90 0.90 1.00 0.95 0.95 0.40* 1.00 0.95

Occupation sorting, n � 41 1.00 0.98 0.93 0.95 1.00 1.00 0.34* 1.00 1.00

Significant values are indicated in bold.* Not significantly different from chance.† Probability of 0.01 that result was obtained by chance.‡ Probability of 0.02 that result was obtained by chance.§ Probability of 0.03 that result was obtained by chance.

October (1 of 2) 2001 NEUROLOGY 57 1163

high rates of mistaken familiarity with unknown faces (fig-ure 1). Hence, it would appear that the “false recognition ofunfamiliar faces” by some of our patients was due mainlyto a criterion shift. Indeed, calculation of d� from the datashowed similar low discriminative performances for allsubjects, with d� ranging from �1.29 to 1.12. Residualdiscrimination was greater in patients with acquired thanin those with congenital prosopagnosia (t-test, p � 0.05)and best in those with predominantly unilateral right-sided lesions. Among the individual patients, only the d�values of 1.04 for Patient 4 (�2, p � 0.05) and 1.12 forPatient 6 (p � 0.02) indicated slightly significant discrimi-native abilities.

Because she had lived in Africa until recently in com-parative social isolation, Patient 2 was also tested with apersonalized battery composed of family members andtelevision personalities from her favorite shows. She did nobetter on this test.

Unlike patients with Capgras’ syndrome, these patientsdid not assign any specific identities to the unfamiliar indi-viduals, with the exception of Patient 2. Patient 2 ascribedspecific incorrect identities to some of both known and un-known faces, but these were not accompanied by any delu-sional aspects and were acknowledged to be guesses.

For comparison, Patient 9, with a right-sided posteriorcerebral artery infarction but without prosopagnosia, ob-tained a d� of 2.8, with no false recognition of unfamiliarfaces. Seven normal subjects obtained a mean d� of 2.57(range 1.88 to 2.93). Thus, our prosopagnosic patients dem-onstrate impaired overt discrimination of familiarity offaces, consistent with their subjective claims.

Covert tests. In the forced-choice test of familiaritywith paired faces, none of the eight patients performed bet-ter than chance (range 15 to 56% correct; table). However,with the semantic cue of the name of the individual to beidentified, all five patients with acquired prosopagnosia didsignificantly better than chance (range 75 to 81% correct).Likewise, when asked to divide the other 41 faces by occu-pation, three of the five patients with acquired prosopag-nosia did better than chance. In contrast, the threepatients with developmental prosopagnosia did not demon-strate any covert semantic processing with either test.

The controls using written names rather than facesshowed that all subjects except Patient 2 achieved 90 to100% correct when asked to choose which of 2 names wasfamous and that they achieved 93 to 100% correct whenasked to divide the 41 names (belonging to the faces) byoccupation. As the names and faces used in the sorting byoccupation test belonged to the same individuals, thisproves that all except Patient 2 had sufficient familiaritywith these well-known individuals.

Patient 9, the patient without prosopagnosia, performed

Figure 2. Correlation of overt familiarity with covertsemantic processing. For each subject, the d� score from thetest of overt familiarity is plotted against the percentagecorrect scores from the two tests (filled symbols � pairedfaces with name cues; open symbols � occupational sorting)of covert semantic face perception. For both acquired anddevelopmental prosopagnosia, there is a high correlation ofovert and covert ability. Black squares � name cue(acquired); black circles � name cue (developmental); opensquares � occupation (acquired); open circles � occupation(developmental).

Figure 1. Detection theory analysis of overt familiarity inprosopagnosia. Hit rates (famous faces identified as familiar)are plotted against false-alarm rates (unknown facesidentified as familiar). Percentage correct scores have beennormalized, or z transformed, so that 0 corresponds to 50%correct. Points falling above the diagonal line marking equalhit and false-alarm rates indicate some discriminativeability. Patients with acquired prosopagnosia have betterresidual overt discrimination of familiarity than patientswith developmental prosopagnosia. However, they are stillfar worse than normal subjects and one nonprosopagnosicpatient with a right posterior cerebral artery infarct. Blacksquares � acquired; black circles � developmental; blackx � nonprosopagnosia; open circles � normal.

1164 NEUROLOGY 57 October (1 of 2) 2001

almost flawlessly on all forced-choice tests of name andface recognition. However, this is not an index of covertrecognition in her, as she is able to overtly recognizenearly all the faces in the test.

The accuracy scores on the two tests of covert semanticprocessing, using name and occupation, were comparedwith the d� scores of overt familiarity. There was a highcorrelation for both covert tests (r � 0.90 for paired faceswith name cues and r � 0.72 for occupational sorting)(figure 2).

Discussion. Overt and false familiarity. Ourtests of recognition of familiar and unfamiliar facesrevealed that many patients displayed false alarms,in that they claimed that a number of faces unknownto them were familiar. Within the group, though, thecorrelation of hit rate with false-alarm rate sug-gested that their “false familiarity” with unfamiliarfaces appears to be due to a criterion shift ratherthan to a distinctly different perceptual mechanism.Impaired perceptual input to face-recognition unitsor poor discriminatory function of partially damagedface-recognition units has been held responsible forfalse recognition in other prosopagnosic patients.31

With the exception of Patient 8, who did not haveany false alarms, all our patients have been found tohave poor perception of facial configuration. Falserecognition of unfamiliar faces has also been de-scribed in nonprosopagnosic patients with frontal le-sions, in whom it is hypothesized that the defect liesin inadequate monitoring and decision making inperceptual tasks.29,31

Residual overt discrimination, as measured by d�,was better in patients with acquired prosopagnosiathan in those who had lifelong prosopagnosia andbest in patients with predominantly unilateral right-sided lesions. As expected, discrimination was stillworse than that of the patient with a similar right-sided lesion but no prosopagnosia.

Semantic facilitation of covert recognition. Noneof our subjects had covert familiarity in the paired-face test. All subjects with adult-onset prosopagnosiahad covert familiarity when prompted by the seman-tic cue of the name belonging to the face, and three ofthe five demonstrated covert semantic informationwhen sorting famous faces by occupation. Althoughone might expect that a vague sense of familiaritywould be easier to achieve than the more stringentrecognition required to sort faces by occupation, theresults indicate that semantic cues strongly facilitatecovert processing, even in subjects in whom the per-ception of facial spatial relations is impaired.

Sergent and Signoret18 noted similar effects ofname facilitation in Subjects P.C. and P.M. Theirstudy also found no covert effects for familiarityjudgments, including a paired-face forced-choice testsimilar to ours. Covert effects were not found with anoccupation test but were present with a test in whichone face was shown and the subject asked which oftwo names belonged to the face. When two faceswere shown and one name given, though, they were

at chance. However, in distinction from our test withname cues, which paired a famous and nonfamousface, their “face pairs with names” used two famousfaces from the same occupational category. The closesemantic relation of the two faces may have madetheir task more difficult than our task. Indeed, inter-ference effects between faces of close semantic rela-tion have been documented in another prosopagnosicstudy.33

Names probably have special value as semanticcues to covert processing in prosopagnosia. In stud-ies of their subject P.H., both names and occupationscovertly facilitated learning of paired associations.5However, more specific semantic data such as theparty affiliation or nationality of politicians were noteffective covert cues.34 The authors’ explanation thatonly general semantic data are accessed by covertprocesses is difficult to reconcile with P.H.’s learningfacilitation by true names and the findings of ourstudy and others. One can hardly have a more spe-cific semantic cue than a first and last name. Wesuspect that the strength as well as the specificity ofthe semantic connection are important in covert ac-cess. Given the social importance of linking names tofaces, the facilitatory effects of names on face recog-nition processes are likely among the most robust ofsemantic effects.

Covert processing: functional and anatomic ba-sis. It has been hypothesized that covert processingmay be related to the type of prosopagnosia. Patientswith perceptual dysfunction may not form a suffi-ciently distinct face percept to trigger covert recogni-tion, whereas those with associative prosopagnosiaand intact perceptual mechanisms can.6,15,17 This,then, should relate to another hypothesis: that obli-gate bilateral temporal lesions cause associative de-fects and unilateral right-sided occipitotemporallesions cause apperceptive prosopagnosia.27 How-ever, we found no distinction between Patient 8, withbilateral anterior temporal lesions and normal per-formance on our tests of spatial relation perception,and Patients 4, 5, 6, and 7, all of whom had moreextensive occipitotemporal lesions and perceptualdysfunction.

Sergent and Signoret18 did note a correlation ofcovert processing with better perceptual processing,in that their subjects P.V. and P.C. had covert se-mantic priming of face recognition by names,whereas R.M., who had more severe perceptual dys-function, did not. This also correlated to some degreewith anatomy, as P.C. had a more anterior right-sided lesion of the parahippocampal gyrus, whereasR.M. had a right occipitotemporal lesion includingthe lingual and fusiform gyri.35 However, this corre-lation was not perfect. P.V. had a lesion of the lin-gual and fusiform gyri, similar to R.M., and did havesome perceptual dysfunction, yet still displayed co-vert processing, unlike R.M. Hence, the authors con-cluded that “a perceptual impairment, provided it isnot too severe, does not necessarily prevent implicitaccess to facial knowledge.” Our data on covert pro-

October (1 of 2) 2001 NEUROLOGY 57 1165

cessing in Patients 4, 5, 6, and 7, all of whom hadclear defects on our tests of face perception, supportthis conclusion.

Others have also questioned the proposed correla-tion between perceptual status and covert processingin prosopagnosia.11 That study described three pa-tients, two of whom were impaired on face-matchingtests and one of whom was not. On their test oflearning paired names with faces, the two patientswith impaired face matching showed covert process-ing, whereas the one with better face matching didnot. The authors argued that covert ability was morelikely with a lesion that disconnected face-recognitionunits than with one that destroyed such units.

It is likely that at least some face perception mustsurvive for covert abilities. Those working with elec-trodermal indexes of familiarity have suggested thatthis remnant processing may be located in an alter-native pathway, perhaps a dorsal one via the supe-rior temporal sulcus.17 However, this “dorsal”hypothesis does not explain the lack of covert seman-tic processing in patients without parieto-occipitaldamage.18 Also, there is mounting evidence fromfunctional imaging that the fusiform gyri but not thesuperior temporal sulcal regions are specifically in-volved in tasks requiring recognition of facial identi-ty.36 Last, our data show that residual overtfamiliarity and the degree of covert semantic facialprocessing are quantitatively correlated, arguingagainst the hypothesis that covert processing is me-diated by an anatomically separate mechanism.

Alternatively, the residual processing for covertperception may emerge from the surviving normalstructures in a partly damaged ventral recognitionnetwork,37 a proposal better supported by the corre-lation of overt and covert abilities we found. Func-tional imaging has revealed candidate regions forthese structures, particularly the right fusiform areaand an inferior occipital locus, both specifically activeduring face rather than object perception tasks.38-40

However, the complete destruction of these areas inPatients 4, 5, 6, and 7 does not eradicate covert pro-cessing. Hence, partial survival of these areas is notmandatory for covert semantic knowledge.

Though the right fusiform gyrus has alwaysshown greater and perhaps more specific activationby faces, all imaging studies have shown bilateralactivation.39-41 Familiar faces especially activate awide network of left- and right-sided regions in theanterior temporal, medial temporal, and frontal re-gions.42,43 This is consistent with neuropsychologicstudies of split-brain patients showing face percep-tion in both hemispheres,44 even though the rightmay be more proficient and important, as evidencedby the proliferation of reports of prosopagnosia withunilateral right-sided lesions.45,46 Based on this infor-mation and our data, we hypothesize that covert se-mantic processing may require residual function ofat least one fusiform/inferior occipital region. In theface of extensive destruction of the right occipitotem-poral cortex up to and including the fusiform gyrus

in Patients 4, 5, 6, and 7, their covert processing maydepend on residual face processing by left occipito-temporal regions.

Several testable predictions follow from this hy-pothesis. First, bilateral lesions of inferior occipitalcortex, which likely cause a more severe general ap-perceptive visual agnosia as well as prosopagnosia,should eliminate covert semantic abilities. Indeed,this was the case for M.S.15 and G.Y.17 Of note, in ourstudy, the two subjects with some damage near theleft fusiform region (Patients 4 and 7) were the oneswho could not sort faces by occupation, though theystill displayed a facilitatory effect of names onforced-choice guessing of famous faces. More anteriorbilateral lesions of the fusiform gyri may have lessprofound general visual defects but still eliminatecovert processing. This may have been the case forSubject 3 of McNeil and Warrington,11 who had aposterior dementia. Verification of these predictionsfrom the literature on patients with covert process-ing is more difficult. Among these, several have uni-lateral right-sided lesions,3,7,18 as we would predict,but more are stated to have bilateral lesions. How-ever, determining the status of the fusiform and infe-rior occipital regions specifically is not easy, as mostreports do not show imaging. The descriptions of atleast some bilateral cases suggest that the fusiformregion may have been spared on one or bothsides,6,10,11 but in others this cannot be determined.2,4,47

Also, a number of these bilateral cases had covert fa-miliarity, demonstrated mainly with the electrodermalskin conductance test, which may not be functionallyequivalent to covert semantic effects.

Whereas this hypothesis also suggests that pa-tients with lesions sparing one fusiform regionshould always have covert semantic processing, thismay not necessarily follow. At least one patient witha unilateral right lesion had absent covert process-ing.18 The degree of lateralization of face processingprobably varies among individuals, just as handed-ness and language do. Functional imaging studieshave provided support for variable lateralization offace processing.39 Subjects with highly right-dominant face processing may not have covert pro-cessing even if the left occipitotemporal regions arespared. Conversely, patients with less lateralizedface processing may not become prosopagnosic aftera single right-sided lesion, consistent with reports ofprosopagnosic patients with obligatory sequential bi-lateral lesions.48

It must be stated that our findings are limited todirect forms of semantic covert processing, usingforced-choice techniques. As others have noted, thepresence of one type of covert processing does notnecessarily mean that all types of covert processingare present.18 Currently, it is not known whetherdirect behavioral, indirect behavioral, and physio-logic forms of covert processing are mediated by sim-ilar or different mechanisms or merely differ in theirefficacy in revealing residual ability, as is apparentin the blindsight literature.49

1166 NEUROLOGY 57 October (1 of 2) 2001

Covert processing and developmental prosopagno-sia. It is also likely that adequate development offace-recognition units before the prosopagnosic lesionis an important prerequisite for covert processing.Whereas the performance of Patient 2 on our name-recognition tests indicates that her absent covertface processing may be due simply to her ignoranceof the people whose faces we used, lack of familiaritycannot explain the equally poor performance of Pa-tients 1 and 3. Prior experience is required to de-velop perceptual expertise50 and has been shown topromote the development of similar perceptual pref-erences in clusters of neurons in temporal cortex.51

As Patients 1 and 3 have had difficulty recognizingfaces since childhood, they may have never devel-oped a region of face expertise that could supportcovert processing. Indeed, three of the eight reportedprosopagnosic cases with absent covert face process-ing had a developmental disorder,13,16,19 and anotherhad hemispherectomy at age 13 for childhood epi-lepsy with preexisting diffuse cortical damage.14 Thefrequency of rare childhood-onset cases among proso-pagnosic patients without covert processing and thelack of such cases among those with this ability de-serve comment. Finally, it is probably unwise to usedevelopmental cases in formulating hypothesesabout the functional determinants of covert abilities,as any perceptual, amnestic, or disconnection deficitwould undoubtedly prevent the acquisition of a storeof facial memories in these subjects.

AcknowledgmentThe authors thank Drs. A. Pascual-Leone for referring Patient 8,A. Budson for Patient 1, F. Drislane for Patient 2, D. Press forPatient 5, D. Pearson for Patient 3, and B. Dworetsky and A.DiBernardo for Patient 7.

References1. Bruyer R. Covert facial recognition in prosopagnosia: a review.

Brain Cogn 1991;15:223–235.2. Tranel D, Damasio A. Knowledge without awareness: an auto-

nomic index of facial recognition by prosopagnosics. Science1985;228:1453–1454.

3. Renault B, Signoret J-L, DeBruille B, Breton F, Bolgert F.Brain potentials reveal covert facial recognition in prosopag-nosia. Neuropsychologia 1989;27:905–912.

4. Rizzo M, Hurtig R, Damasio A. The role of scanpaths in facialrecognition and learning. Ann Neurol 1987;22:41–45.

5. de Haan E, Young A, Newcombe F. Face recognition withoutawareness. Cogn Neuropsychol 1987;4:385–415.

6. Sergent J, Poncet M. From covert to overt recognition of facesin a prosopagnosic patient. Brain 1990;113:989–1004.

7. Schweinberger S, Klos T, Sommer W. Covert face recognitionin prosopagnosia: a dissociable function? Cortex 1995;31:517–529.

8. Bauer R. Autonomic recognition of names and faces in prosop-agnosia: a neuropsychological application of the guilty knowl-edge test. Neuropsychologia 1984;22:457–469.

9. Bauer R, Verfaellie M. Electrodermal discrimination of famil-iar but not unfamiliar faces in prosopagnosia. Brain Cogn1988;8:240–252.

10. Bruyer R, Laterre C, Seron X, et al. A case of prosopagnosiawith some preserved covert remembrance of familiar faces.Brain Cogn 1983;2:257–284.

11. McNeil J, Warrington E. Prosopagnosia: a re-classification. QJ Exp Psychol 1991;43:267–287.

12. de Haan E, Young A, Newcombe F. Faces interfere with nameclassification in a prosopagnosic patient. Cortex 1987;23:309–316.

13. Young A, Ellis H. Childhood prosopagnosia. Brain Cogn 1989;9:16–47.

14. Sergent J, Villemure J-G. Prosopagnosia in a right hemi-spherectomized patient. Brain 1989;112:975–995.

15. Newcombe F, Young A, de Haan E. Prosopagnosia and objectagnosia without covert recognition. Neuropsychologia 1989;27:179–191.

16. de Haan E, Campbell R. A fifteen year follow-up of a case ofdevelopmental prosopagnosia. Cortex 1991;27:489–509.

17. Bauer R. The cognitive psychophysiology of prosopagnosia. In:Ellis H, Jeeves M, Newcombe F, Young A, eds. Aspects of faceprocessing. Dordecht: Martinus-Nijhoff, 1986:253–267.

18. Sergent J, Signoret J-L. Implicit access to knowledge derivedfrom unrecognized faces. Cereb Cortex 1992;2:389–400.

19. Bentin S, Deouell L, Soroker N. Selective visual streaming inface recognition: evidence from developmental prosopagnosia.Neuroreport 1999;10:823–827.

20. Farah M. Visual agnosia: disorders of visual recognition andwhat they tell us about normal vision. Cambridge: MIT Press,1990.

21. Davidoff J, Landis T. Recognition of unfamiliar faces in pros-opagnosia. Neuropsychologia 1990;28:1143–1161.

22. Young A, Newcombe F, de Haan E, Small M, Hay D. Faceperception after brain injury. Brain 1993;116:941–959.

23. Barton J, Keenan J, Bass T. Discrimination of spatial rela-tions and features in faces: effects of inversion and viewingduration. Br J Psychol (in press).

24. Leder H, Bruce V. Local and relational aspects of face distinc-tiveness. Q J Exp Psychol 1998;51:449–473.

25. Bruce V, Doyle T, Dench N, Burton M. Remembering facialconfigurations. Cognition 1991;38:109–144.

26. Barton J, Press D, Keenan J, O’Connor M. Lesions of thefusiform face area impair perception of facial configuration inprosopagnosia. 2001 (in press).

27. Damasio A, Tranel D, Damasio H. Face agnosia and the neu-ral substrates of memory. Annu Rev Neurosci 1990;13:89–109.

28. Kosslyn S, Hamilton S, Bernstein J. The perception of curva-ture can be selectively disrupted in prosopagnosia. Brain Cogn1995;27:36–58.

29. Young A, Flude B, Hay D, Ellis A. Impaired discrimination offamiliar from unfamiliar faces. Cortex 1993;29:65–75.

30. Rapcsak S, Polster M, Comer J, Rubens A. False recognitionand misidentification of faces following right hemispheredamage. Cortex 1994;30:565–583.

31. Rapcsak S, Polster M, Glisky M, Comer J. False recognition ofunfamiliar faces following right hemisphere damage: neuro-psychological and anatomical observations. Cortex 1996;32:593–611.

32. Young A. Covert recognition. In: Farah M, Ratcliff G, eds. Theneuropsychology of high-level vision. Hillsdale: Lawrence Erl-baum, 1994:331–358.

33. Dixon M, Bub D, Arguin M. Semantic and visual determi-nants of face recognition in a prosopagnosic patient. J CognNeurosci 1998;10:362–376.

34. Young A, de Haan E. Boundaries of covert recognition in pros-opagnosia. Cogn Neuropsychol 1988;5:317–336.

35. Sergent J, Signoret J-L. Varieties of functional deficits inprosopagnosia. Cereb Cortex 1992;2:375–388.

36. Haxby J, Hoffman E, Gobbini M. The distributed human neu-ral system for face perception. Trends Cogn Sci 2000;4:223–233.

37. Farah M, O’Reilly R, Vecera S. Dissociated overt and covertrecognition as an emergent property of a lesioned neural net-work. Psychol Rev 1993;100:571–588.

38. Puce A, Allison T, Gore J, McCarthy G. Face-sensitive regionsin human extrastriate cortex studied by functional MRI.J Neurophysiol 1995;74:1192–1199.

39. Kanwisher N, McDermott J, Chun M. The fusiform face area:a module in human extrastriate cortex specialized for faceperception. J Neurosci 1997;17:4302–4311.

40. Sergent J, Ohta S, MacDonald B. Functional neuroanatomy offace and object processing. Brain 1992;115:15–36.

41. Puce A, Allison T, Asgari M, Gore J, McCarthy G. Differentialsensitivity of human visual cortex to faces, letterstrings, and

October (1 of 2) 2001 NEUROLOGY 57 1167

textures: a functional magnetic resonance study. J Neurosci1996;16:5205–5215.

42. Leveroni C, Seidenberg M, Mayer A, Mead L, Binder J, Rao S.Neural systems underlying the recognition of familiar andnewly learned faces. J Neurosci 2000;20:878–886.

43. Gorno-Tempini M, Price C, Josephs O, et al. The neural sys-tems sustaining face and proper name processing. Brain 1998;121:2103–2118.

44. Gazzaniga M, Smylie C. Facial recognition and brain asymme-tries: clues to underlying mechanisms. Ann Neurol 1983;13:536–540.

45. Landis T, Cummings J, Christen L, Bogen J, Imbof H-G. Areunilateral right posterior cerebral lesions sufficient to causeprosopagnosia? Clinical and radiological findings in six addi-tional patients. Cortex 1986;22:243–252.

46. de Renzi E. Prosopagnosia in two patients with CT scan evi-dence of damage confined to the right hemisphere. Neuropsy-chologia 1986;24:385–389.

47. de Haan E, Young A, Newcombe F. Covert and overt recogni-tion in prosopagnosia. Brain 1991;114:2575–2591.

48. Ettlin T, Beckson M, Benson D, Langfitt J, Amos E, Pineda G.Prosopagnosia: a bihemispheric disorder. Cortex 1992;28:129–134.

49. Marcel A. Blindsight and shape perception: deficit of visualconsciousness or of visual function ? Brain 1998;121:1565–1588.

50. Diamond R, Carey S. Why faces are and are not special: aneffect of expertise. J Exp Psychol 1986;115:107–117.

51. Erickson C, Jagadeesh B, Desimone R. Clustering of perirhi-nal neurons with similar properties following visual experi-ence in monkeys. Nat Neurosci 2000;3:1143–1148.

CME Face memory impairments inpatients with frontal lobe damageS.Z. Rapcsak, MD; L. Nielsen, MA; L.D. Littrell; E.L. Glisky, PhD; A.W. Kaszniak, PhD;

and J.F. Laguna, MD

Article abstract—Objective: To determine whether damage to prefrontal cortex is associated with face memory impair-ment. Background: Neurophysiologic and functional imaging studies suggest that prefrontal cortex is a key component ofa distributed neural network that mediates face recognition memory. However, there have been few attempts to examinethe impact of frontal lobe damage on face memory performance. Methods: Patients with focal frontal lobe lesions andnormal control subjects were administered two-alternative forced-choice and single-probe “yes/no” tests of recognitionmemory for novel faces. Retrograde memory was assessed by using famous faces as stimuli. Results: Compared withcontrol subjects, patients with frontal lobe lesions showed evidence of marked anterograde and relatively mild retrogradeface memory impairment. In addition, patients with right frontal lesions demonstrated increased susceptibility to falserecognition, consistent with the breakdown of strategic memory retrieval, monitoring, and decision functions. Conclusions:Prefrontal cortex plays an important role in the executive control of face memory encoding and retrieval. Left and rightprefrontal regions seem to make different contributions to recognition memory performance.

NEUROLOGY 2001;57:1168–1175

Converging evidence from neurophysiologic record-ings in patients with epilepsy1-3 and from functionalimaging studies in normal individuals4-7 suggeststhat face memory is mediated by a predominantlyright-lateralized neural network that includes infero-temporal cortex, medial temporal lobe, and prefron-tal cortex. Consistent with these findings, a numberof neuropsychological studies have documented se-vere face memory impairment following right ventro-medial temporal lobe damage.8-10 By contrast, therehave been few attempts to investigate face recogni-tion memory in patients with focal frontal lobe le-sions. Despite the paucity of empirical data, there

are some indications that anterograde memory fornovel faces may be impaired following frontal lobedamage, though not to the extent seen in patientswith right temporal lobe lesions.8,11 Another smallgroup study found that patients with dorsolateralfrontal lesions (n � 6) performed worse than controlsubjects in naming famous faces from photographs.12

However, performance on this test of retrogradememory was essentially within normal range whenpatients were asked to match the faces with the cor-rect names. These results, plus the fact that the ma-jority of patients had left frontal lobe damage,suggest that the difficulty in naming famous faces

See also pages 1151 and 1161

From the Neurology Section (Dr. Rapcsak), Southern Arizona VA Health Care System, Tucson; and Departments of Neurology (Drs. Rapcsak, Kaszniak, andLaguna) and Psychology (Drs. Rapcsak, Glisky, and Kaszniak, and L. Nielsen and L. Littrell), University of Arizona, Tucson.Supported by the Cummings Endowment Fund to the Department of Neurology at the University of Arizona.Received April 11, 2001. Accepted in final form July 17, 2001.Address correspondence to Dr. Steven Z. Rapcsak, Neurology Section (1-127), Southern Arizona VA Health Care System, Tucson, AZ 85723; e-mail:szr@u.arizona.edu

1168 Copyright © 2001 by AAN Enterprises, Inc.