COVERT FACE RECOGNITION IN PROSOPAGNOSIA: A DISSOCIABLE FUNCTION? Stefan R. Schweinbergerl. 2 , Thomas Klos2 and Werner Sommer' ('University of Konstanz and 2Bezirkskrankenhaus Erlangen, Germany) ABSTRACT Covert face recognition was investigated in a patient with prosopagnosia without object agnosia. This patient performed well in various face processing tasks like expression analysis and feature processing and had relatively preserved semantic knowledge about persons, but was slightly impaired in the visual matching of unfamiliar faces. In a face-name paired­ associate relearning task, covert face recognition was demonstrated to be above-chance. However, as this task cannot be meaningfully applied to control subjects, results do not necessarily indicate that the degree of covert face recognition is normal. In fact, in contrast to control subjects, the patient showed significantly reduced· associative priming of names by face primes as compared to name primes, suggesting a quantitative reduction of covert face recognition. It is argued that these results support the view that overt and covert face recognition are brought about by the same functional system (Farah, O'Reilly and Vecera, 1993). INTRODUCTION Prosopagnosia is the inability to identify well-known persons by their faces. Prosopagnosic patients can usually distinguish faces from other stimuli, and are often able to recognize common objects. They also remain able to recognize familiar people by voice, gait, or distinguishing features (clothes, glasses), and can correctly describe persons if presented with their names. Thus, it is not the knowledge about familiar people that is lost - it is only the access to this knowledge from the face that is impossible. Moreover, other aspects of face processing, apart from recognizing facial identity, may be relatively intact, for example recognition of emotion displayed by the face (Hecaen, 1981 ), lipreading (Campbell, Landis and Regard, 1986), or matching of simultaneously presented face portraits taken from different perspectives (Bruyer, Laterre, Seron et al., 1983; Malone, Morris, Kay et al., 1982). While prosopagnosic patients will not recognize people by their faces, covert knowledge for overtly unrecognized faces can be demonstrated in some of these patients. Perhaps the earliest evidence was reported by Cole and Perez-Cruet (1964). Among other results, these authors briefly mention a classical conditioning procedure, where presentations of the patient's own face but not of other faces were paired with brief electric shocks. The patient was reported to have developed differential autonomic responses (e.g. changes in heart rate and skin conductance) contingent on his own face. However, differentiation was judged as poor, and the effect could not unequivocally be ascribed to covert Cortex, (1995) 31, 517-529 518 Stefan R. Schweinberger and Others recognition since explicit recognition of his own face could not be excluded. Recently, covert face recognition has been studied more systematically with a number of different experimental techniques (see Bruyer, 1991, for review). Bruyer et al. (1983) showed that their patient could learn to associate correct names with familiar faces better than false names, demonstrating that face identity - although not recognized overtly - continued fo influence behaviour. Rizzo, Hurtig and Damasio (1987) found that their patients' visual scanning behaviour, although not completely normal, was different for familiar and unfamiliar faces. In a series of experiments, DeHaan, Young and Newcombe's (1987) patient P.H. showed covert knowledge of faces at the level of face familiarity, personal semantics, and names. P.H. also showed associative priming of name recognition. That is, he was faster in deciding whether, say, "Nancy Reagan" is a familiar name when preceded by Ronald Reagan's face rather than by Helmut Kohl's face - although he was unable to overtly recognize these faces (Young, Hellawell and De Haan, 1988). Psychophysiological studies have provided further evidence for covert face recognition in prosopagnosic patients. For example, Bauer (1984) showed to his patient L.F. familiar faces along with spoken names which could be correct or incorrect. Although L.F. could not identify the correct names, electrodermal responses (EDRs) were larger for correct than incorrect names. Tranel and Damasio (1985, 1988) observed larger EDRs to familiar than to unfamiliar faces even though their patients were unable to overtly discriminate them. Renault, Signoret, Debruille et al. (1989) measured the P300 component of the event­ related brain potential (ERP) while their patient made facial familiarity decisions. The P300 is usually larger in response to rare than frequent stimulus categories. When familiar and unfamiliar faces were equiprobable, P300 amplitude did not discriminate between these face categories. However, when the probability of familiar faces was reduced to 33%, P300 amplitude was larger to familiar than unfamiliar faces, demonstrating a covert categorization. Several hypotheses have been offered in order to explain covert face recognition in prosopagnosia. Bauer ( 1986) suggested that EDR discrimination is based on a dorsal visual-limbic pathway, whereas damage of ventral occipitotemporal projections was held responsible for overt face recognition deficits. This model of independent brain systems for overt and covert face recognition implies that performance in overt and covert tests may, in principle, be completely independent of each other. De Haan, Bauer and Greve (1992) suggested that covert face recognition might be brought about by a single, normally functioning face recognition system which is, however, disconnected from a ''conscious awareness system" necessary for overt recognition. A rather different explanation was recently put forward by Farah, O'Reilly and Vecera (1993), who assumed that prosopagnosia always involves an impairment, although to varying degrees, in visual processing of faces. Most generally,. they hypothesize that covert face recognition tests are more sensitive than overt tests to the residual information encoded in a damaged visual recognition system. These authors showed by computer simulations that a single lesion in the visual recognition module of a neural network may produce dissociations between covert and overt tests similar to those that have been observed in several Covert face recognition in prosopagnosia 519 prosopagnosic patients (for a related approach see Burton, Young, Bruce et al., 1991). One way to decide between these alternative explanations of covert face recognition in prosopagnosia on a empirical basis is to investigate the magnitude of covert face recognition effects. Completely normal effects in terms of some quantitative measure would be consistent with either the hypothesis of independent brain systems for overt and covert face recognition (Bauer, 1986) or the disconnection-from-consciousness hypothesis (De Haan et al., 1992). In contrast, if covert effects in prosopagnosics are invariably reduced in magnitude, the sensitivity-to-residual-knowledge hypothesis (Farah et al., 1993) would be a more parsimonious explanation of the covert recognition phenomenon (see also Schweinberger, 1992). To date, many studies have demonstrated above-chance recognition in prosopagnosic patients but have left open the question of whether these effects are normal in size or merely present to some degree. Only recently, two studies evaluated the magnitude of covert face recognition. Young et al. (1988) showed that, in their patient P.H., the magnitude of associative priming of name recognition by face primes did not statistically differ from priming by name primes. For example, P.R.'s reaction time for deciding. whether "Nancy Reagan" is a familiar name was reduced when it was primed by Ronald Reagan's face (as compared to an unrelated prime face). This priming effect was comparable to that arising from priming by Ronald Reagan's name. Similar data were obtained in a cross-domain identity priming condition by De Haan et al. (1992) and were taken to indicate that the locus of face priming "... is preserved and functions normally.. .'' (p. 88). Of course, one caveat for such a conclusion is that it is based on the acceptance of the null hypothesis, i.e. on the nonsignificant difference between face- and name-mediated priming. While De Haan et al. (1992) did not collect data from controls, Young et al. (1988) have found priming effects of similar magnitude by both faces and names also in normal subjects. However, Farah et al. (1993) have pointed out that these experiments for normals and for the patient differed in several respects, leaving open the possibility that under the precise conditions of the patient study, face-mediated priming might have had an advantage over name-mediated priming. We report an investigation of covert face recognition in a patient with prosopagnosia without object agnosia. This patient's perceptual operations with faces appeared to be remarkably intact as regards expression analysis and feature processing, although he was slightly impaired in matching unfamiliar faces. In contrast, he was severely impaired in recognizing both famous faces and previously memorized unfamiliar faces. Covert face recognition was investigated using both a face-name paired-associate relearning task and an associative priming task analogous to the study of Young et al. (1988). While the face­ name relearning task is a frequently used procedure to reveal covert face recognition in prosopagnosia, the associative priming task should specifically reveal (1) whether or not priming from faces and names was of equal magnitude in our patient, and (2) whether or not his pattern of priming matched that of control subjects tested with exactly the same procedure. 520 Stefan R. Schweinberger and Others Fig. I - CT sca11 of M. T CASE REPORT At the time of investigation, M.T. was a 32-year old male who had worked as a computer assistant before his illness. He reported to be ambidextrous although his writing hand was always the right one. On July 26, 1991, M.T. suffered a right hemisphere stroke following occlusion of the right carotis intema, initially causing complete left hemiplegia, bradycardia and impaired consciousness. One month after the stroke the hemiplegia had reduced to problems with fine movements of the left hand and fingers. A CT scan from August 1991 (cf. Figure 1) revealed a large hypodense area in the right hemisphere. The lesion was mainly temporo-parietal but extended also to frontal and occipital areas, in the absence of evidence for left-hemispheric pathology. M.T. was described by the treating psychologist as introverted and suspicious, with reduced illness insight. Speech functions as assessed with the Aachener Aphasie Test Batterie (Huber, Poeck, Weniger et a!., 1983) were normal. In particular, this held for reading, naming, comprehension, oral speech repetition, and for the Token Test. In the intelligence test IST­ 70 (Amthauer, 1970), M.T. obtained a normal profile with IQ scores for the different subtests ranging between 94 and 108. In the Standard Progressive Matrices test (Raven, 1956) applied in March 1992, he scored at a percentile rank >75. In October 1991, formal tests of visual acuity and visual fields showed no deficits. There was no unilateral neglect as assessed with the Line Bisection test. His auditory-verbal memory span (digits) was 5 and 5 forward and backward, respectively, and his visuospatial memory span (Corsi block test) was 6 and 6. M.T.'s face recognition problems came to our knowledge in May 1992 when he participated in a larger study of memory search for faces and words in brain-damaged patients (see below). When confronted with his face recognition problems, he reported that since his stroke he "sees people differently than before", and that occasionally familiar people complained not to have been recognized by him. Informal investigation with a small set of portraits from the department staff suggested that when unequivocal paraphernalia were Covert face recognition in prosopagnosia 521 available (e.g. dictaphone earphones in case of the department secretary, distinctive hairstyle or beard), he could fairly compensate for his deficit. His performance dropped, however, when paraphernalia were absent or could not be used as effective cues. Even after several months of intensive contact, he frequently mixed up two therapists who look similar in terms of figure and who wear similarly formed glasses, but whose faces are quite dissimilar otherwise. The tests reported in the following were conducted between May and October, 1992. Face Processing Familiar Face Recognition Of 18 portraits of celebrities which were identified almost perfectly by I 0 control subjects (mean= 17.7), M.T. could identify only one. Even in this case, it was only after considerable hesitation that he stated the portrait "might perhaps show Willy Brandt". He claimed that all other depicted persons were completely unknown to him. However, when presented with the names of these persons, he could correctly describe the persons' identities in 13 out of 18 cases. Although this score is also considerable below the control mean (mean= 17.7), performance for names was clearly much better than for faces. Thus, his face recognition problem cannot be simply explained by lacking knowledge of the photographed persons. Unfamiliar Face Processing On the short form of the Benton and van Allen Test (Levin, Hamsher and Benton, 1975), which requires the simultaneous matching of unfamiliar faces, M.T. reached a score of 21/ 27 (low normal). On the long form of the same test, conducted in a different session, he scored 37/54 (moderately impaired). This test is perceptually difficult but has low memory demands since the comparison faces are shown simultaneously. M.T.'s score, although below average, is not worse than the score of many right brain-damaged patients without prosopagnosia. On both occasions he solved the test in a very slow and effortful manner. The performance of M.T. in a memory search experiment (cf. Schweinberger, Klos and Sommer, 1993) is shown in Table I. In this experiment M.T. had to decide by speeded two­ choice responses whether or not a particular stimulus belonged to a set of previously memorized stimuli. Stimuli could be either unfamiliar faces or abstract words, and the number of memorized items could be either one or three. Table I suggests that although his response times were relatively fast, M.T. made an excessive amount of errors for faces (in fact, more than any other of 50 brain-damaged patients who performed the same experiment). When three faces instead of one had to be held in memory, his accuracy dropped to near-chance level. In contrast, his performance in the analogous task for words was remarkably accurate. It thus appears that M.T. had great problems in unfamiliar face recognition, particularly when memory demands were high. M.T. was then given a test for matching emotional facial expressions using a subset of pictures from Ekman and Friesen ( 1976). He was shown two sets of six faces depicting six different basic expressions (happiness, sadness, surprise, fear, anger, disgust), and was required to find pairs with the same emotion, which were always from two different individuals. M.T. performed errorless (6/6) in two parallel forms of this test. He could also name the depicted emotions without problems. When asked to assemble a face puzzle consisting of the separated facial features (cf. Landis, Regard, Bliestle et al., 1988), he also performed errorless. Finally, he performed well in sorting a series of seven portraits of young to old men with respect to age (cf. De Renzi, Bonacini and Faglioni, 1989). Object Recognition M.T. could spontaneously name visually presented objects. In a formal test, he correctly named 21/21 line drawings of common objects and 8/8 animals without hesitation. For some prosopagnosic patients, object recognition has been reported to be defective only if the Stefan R. Schweinberger and Others 522 TABLE I Mean Reaction Times and Error Percentages of M.T. in a Memory Search Experiment for Unfamiliar Faces and Abstract Words. (Note the excessive amount of errors with faces, particularly when three faces had to be held in memory). Words Faces 3 3 Memory Set Size 750 22 RT (ms) Errors (%) 845 44 653 5 547 5 Note: Chance performance is at 50% errors. discrimination of similar objects belonging to the same category was required (e.g. Damasio, Damasio and van Hoesen, 1982). Whe therefore presented M.T. with 10 pictures of fruits and vegetables, some of which were visually very similar (e.g. onion and garlic, champignon and chanterelle). This test was prepared in a coloured and a black-and-white version. M.T. already performed perfectly on the black-and-white version of the test (10110) presented initially. He could also recognize the symbols of German industrial brands (5/6). In the Poppelreuter pictures of the Tiibinger-Luria-Christensen Test (Test G12-G14; Hamster, Langner and Mayer, 1980), he also performed perfectly. In ten items of Street's completion test, he reached a score of 6110, which is low but still within the normal range. A summary of the visual recognition tests in shown in Table II. Other Visual and Mnestic Functions M.T.'s copy of Rey's complex figure as well as drawing the same figure from memory after 90 min was fairly accurate. His ability to name colours also appeared preserved. Topographical memory was tested by requiring him to indicate the location of five large cities on an empty map of Germany, which he did well. EXPERIMENT 1 In the first experiment, we presented M.T. with a paired-associate face-name relearning task. This task, which has become the most widely used measure of covert face recognition in prosopagnosia (cf. Farah et al., 1993), was employed TABLE II M.T. 's Performance in Several Tests of Visual Recognition Controls M.T. Face recognition Famous face recognition Famous name recognition Matching identity (Benton) Matching emotion Face puzzle Object recognition Line drawings Animal photos Fruits and vegetables Poppelreuter Figures Street Test 1/18 13/18 37/54 12/12 Mean S.D. 17.7 17.7 47.7 0.7 0.7 3.1 7.3 1.4 10/10 21/21 8/8 10110 14/14 6110 Covert face recognition in prosopagnosia 523 to reveal whether or not above-chance covert face recognition can be demonstrated in M.T. Here, the patient is asked to learn to associate portraits of famous people with names of famous people, some pairings being correct ("true pairings") and some not ("untrue pairings"). The general idea is that if corrects pairings are learned better than incorrect pairings, the patients must possess implicit knowledge about facial identity. Materials and Method Stimuli and Procedure Ten faces of politicians were selected for this task, based on the criterion that on two prior occasions, none of these stimuli had been recognized or elicited a feeling of familiarity in M.T. Five faces were randomly assigned to the "true" and five to the ''untrue" set (see Appendix). Faces were glossy black-and-white photographs sized 9 by 13 em. The names of the depicted persons were printed in black on white carton sheets sized 4.5 by 16 em. M.T. was asked to learn the associations of faces and names. In .each of ten trials, the 10 stimulus pairs were presented sequentially in pseudorandomized order for about 5 s each, with names placed below the faces. Thereafter, the 10 faces were again randomized, and presented one by one. M.T. was to place the corresponding name card below each face. Results Each trial was scored with a "+" if M.T. was better on true than untrue names, "0" if there was no difference), and "-" if he was better on untrue names (cf. Table III). Clearly, true pairs were learned better than untrue pairs (Sign Test; p<.Ol). Thus, albeit face identity could not be retrieved explicitly, it affected performance at an implicit level. When M.T. was told afterwards that he had learned both true and untrue associations, he was very surprised. Thus, he appeared to have been unaware of any untrue face-name pairs. ExPERIMENT 2 The first experiment had demonstrated above-chance covert face recognition in M.T. However, with the face-name relearning task it is impossible to reveal whether or not the degree of covert face recognition is normal (also cf. Farah et al., 1993). The main reason for this is that the task cannot be reasonably applied to normal subjects, who would easily detect the untrue pairings. In contrast, the associative name priming task as used by Young et al. (1988) TABLE III Number of Correct Responses in Experiment 1 (Paired-associate Learning of True. and Untrue Face-name Pairings) Pairings True Untrue Difference Trial number 5 2 + 2 3 4 5 6 7 8 9 10 2 2 5 2 3 1 3 1 5 3 5 4 5 5 0 5 3 4 3 0 + + + + + + + 524 Stefan R. Schweinberger and Others appears suitable to assess the magnitude of covert face recogmt10n, for two reasons. First, the explicit aspect of the task (familiarity decisions about personal names) can be easily applied to both normal subjects and prosopagnosic patients. Second, the patient can serve as his own control by comparing associative priming of name recognition by face primes with priming by name primes. As mentioned above, this comparison is justified if it is demonstrated that priming by faces and names does not normally differ. Although this has already been shown by Young et al. (1988), Farah et al. (1993) have criticized this study for having recorded the patient and control data in two experiments differing in several variables like trial number and exposure duration. We therefore also tested a group ·of· control subjects with the same procedure. Materials and Method Subjects In addition to M.T., 11 control subjects without neurological history (mean age 40.0 years; 6 mell/5 women) were tested. Stimuli and Apparatus Faces were 30 black-and-white portrait photographs of famous persons and 15 portraits of unfamiliar persons which resembled the famous faces with respect to physical appearance and general portrait style. Names were those of 45 famous persons and 45 unfamiliar persons which resembled the famous ones with respect to orthographic criteria. The face photographs were video-digitized and displayed on a computer monitor, with a mean luminance of 38 cd/m2 at a size of 4.3 X 4.8 em. The names were shown at about equal luminance in the centre of the monitor, with christian name placed above family name. Letter height was 0.6 em and width of the names ranged between 1.0 and 4.8 em, depending on the number of characters. All stimuli were shown on a black background. Procedure M.T. was instructed that he would see personal names on the monitor, and that he should press as fast and accurate as possible one of two keys if a name represented a famous person, and the other key if not. He was told that the names could be preceded by faces or by other names which might, or might not be related to the target names but that he should only respond to the target names. Prime names were shown in light-blue to distinguish them from the target names, displayed in white. The experimental trials were preceded by twenty practice trials, ten for face primes and ten for name primes. Stimuli from these practice trials were not used subsequently. Thereafter, 90 trials with face primes were shown, followed by 90 trials with name primes. These 90 trials were always presented in a fixed randomized order. Short breaks were allowed after every 45 trials. Prime stimuli were presented for 745 ms, followed by a blank screen which was replaced after 455 ms by the target name. Target names were presented until a key was pressed. Errors were indicated by a brief tone (500Hz, 71 dB, 110 ms duration). The prime stimulus of the next trial was presented 1600 ms after the response. M.T. used the right index and middle fingers for operating two response keys assigned to famous and unfamiliar target names, respectively. For famous target names preceded by both face and name primes, there were three conditions with 15 trials each: (1) REL-target name preceded by a related prime; (2) NED­ name preceded by a neutral (unfamiliar) prime; (3) UNREL-name preceded by an unrelated prime. To exclude a predictive value of the prime for the response, unfamiliar target names occurred with the same frequency as famous target names and could be preceded by the same types of primes. Covert face recognition in prosopagnosia 525 Results M.T. The performance of M.T. can be seen in Table IV. As associative priming may in principle be due to both facilitation from related primes and inhibition from unrelated primes, differences between the REL and UNREL conditions normally produce the largest and most reliable priming effects. It appears that for face primes, M.T. shows no substantial priming. In contrast, there seems to be a consistent effect of priming by names. To verify these observations, the reaction times of M.T. to correctly classified famous names were submitted to a two (prime domain: face versus name) by three (prime type: related, neutral, unrelated) analysis of variance (ANOV A). In order to reduce the effect of outliers, RTs which exceeded the mean RT in each individual condition by more than 2 standard deviations were omitted. There was a main effect of prime domain (F = 5.18; d.f. =I ,35; p<.05), reflecting faster RTs for trials with name than face primes. Most importantly, there was a significant interaction between prime type and prime domain (F = 4.07; d.f. = 2,35; p< .05). When tested for each prime domain separately, the effect of prime type was significant for name primes (F=3.85; d.f. =2,37; p<.05) but not for face primes (F= 1.67; d.f. =2,38; p<.20). That is, the priming effect for name primes proved to be reliable, whereas the differences in RTs for face primes to be seen in Table IV did not. Control Subjects An ANOV A of the median response times of the controls also yielded a significant effect of prime domain (F= 82.01; d.f = 1, 10; p<.OI). Moreover, there was also a highly significant main effect of prime type (F = 28.25; d.f.= 2, 20; p<.O 1) but no interaction between prime domain and prime type (F = 2.59; d.f. = 2, 20; p<.ll). Although the interaction did not reach significance, it can be seen (cf. Table IV) that a possible difference between prime domains would reflect larger, rather than smaller priming effects for faces as compared to names. TABLE IV Median Reaction Times and Percentages of Errors in Experiment 2 (Associative Priming of Name Recognition) Face primes Name primes REL NEU UNREL REL NEU UNREL M.T. RT Errors (%) 829 0 759 0 876 13 747 0 745 0 929 20 Controls RT Errors (%) 709 3 752 3 831 12 681 673 720 8 1 I 526 Stefan R. Schweinberger and Others DISCUSSION The results of Experiment 2 demonstrate that M.T. shows a preserved pattern of associative priming by name primes. In contrast, priming by face primes was significantly smaller, suggesting that covert face recognition in the associative priming task is reduced in M.T. Both in M.T. and controls, RTs were generally faster in the name prime condition. This result can probably be ascribed to practice effects as a result of the blocked, rather than counterbalanced presentation of face and name prime conditions (also cf. De Haan et al., 1992, Experiment 3). For the same reason, the numerical - although not significant - difference between face and name priming effects in controls need not counter the notion that associative priming effects by faces and names do not normally differ (cf. Young et al., 1988). It does, however, corroborate the reduction of face priming in M.T. in demonstrating that under the particular conditions of the present experiment, face priming is normally as large as, or even larger than, priming from names. The present associative priming effects include a substantial inhibitory component, that is, longer RTs in the unrelated as compared to the neutral priming condition. This is in line with Bruce and Valentine (1986, Experiment 2) but in contrast to the study of Young et al. (1988). It has been argued that whereas facilitation reflects automatic activation within the semantic system, inhibition is related to strategic expectancies regarding the target (Posner and Snyder, 1975). If this is true, we might assume a considerable contribution of strategic expectancies to the present priming effects. On the other hand, it appears unlikely that this can entirely explain the reduction of priming effects for faces in M.T. As Table IV suggests, it is mainly the absence of facilitation which distinguished M.T.'s performance in the face prime condition from that of controls. · It is noteworthy that above-chance covert face recognition could be demonstrated in M.T. in the face-name relearning task, whereas no statistically significant priming effects by face primes were observed in the priming task. This might be taken to indicate that these tasks tap into different components of covert recognition. It might as well, however, reflect a smaller sensitivity of the priming task. This is at least plausible if one considers the numerical trend for priming by faces both in RTs and errors to be seen in Table IV. In our opinion, the statistical insignificance of priming by face primes does therefore not suffice to infer a dissociation between the two covert recognition tests. The results from the face-name relearning task indicates, however, that M.T.'s impaired covert face recognition in the priming experiment does not simply reflect the fact that his perceptual operations with faces were so massively impaired as to make any covert recognition impossible at all. This is also suggested by M.T.'s remarkably preserved performance in several tests of face and object processing, which is certainly much better than the performance of those prosopagnosic patients whose failure to show covert face recognition was ascribed to perceptual disturbances (e.g. Newcombe, Young and DeHaan, 1989). The explanation most near at hand for the observed pattern of reduced covert recognition is that covert and overt face recognition depend on the same Covert face recognition in prosopagnosia 527 functional system. In this line, M.T.'s deficit might be explained both by weakened connections from facial representations to modality-independent person identity nodes (Burton et al., 1991), and by an impairment in encoding facial representations themselves (Farah et al., 1993). However M.T.'s impairments in visual matching and memory search tasks for unfamiliar faces appear to be more :::onsistent with the latter hypothesis, suggesting a deficit already at the encoding :>f facial representations. This conception also counters previous suggestions of intact perceptual :>perations as a prerequisite for covert face recognition (e.g. Bruyer, 1991). [nterestingly, Schweinberger (1992) showed that in 20 prosopagnosic patients reported in the literature, perceptual operations as assessed with a face matching test were judged to be relatively normal in only 7 patients. Covert recognition, in contrast, could be demonstrated not only in each of these 7 patients, but also in 6 out of 13 patients with clear perceptual impairments. Thus, rather than mggesting that completely intact perceptual operations with faces are necessary for covert recognition, this pattern may be taken to argue for the remarkable iensitivity of covert recognition tests to even degraded visual input. It should be kept in mind that observing reduced covert face recognition in :tssociation with prosopagnosia does not exclude the possibility that these functions may be fully dissociated in other patients. On the other hand, we lack iefinite evidence that covert face recognition may be fully preserved in Jrosopagnosia (cf. Farah et al., 1993). Also, no cases with intact overt face ~ecognition but abolished covert recognition seem to have been reported to date. [n combination with the already reported cases of relatively preserved covert ~ecognition without overt recognition, such cases would form a double lissociation, which would be strong evidence for independent brain systems mbserving overt and covert recognition. In the absence of such cases, however, Ne think that the covert face recognition phenomenon is most parsimoniously tccounted for by the greater sensitivity of these tests to residual information Nithin the same functional system. Acknowledgements. We are grateful to the Siidkurier Konstanz for providing the face ;timulus material, to Dr. H. LOBlein for his advice in evaluating the CT, and to the colleagues 'rom the Bezirkskrankenhaus Erlangen for varied forms of support. We would also like to hank two anonymous reviewers for their helpful remarks. Finally, we thank M.T. for his >atience and cooperation throughout this study. REFERENCES ~AUER, R. lntelligenz-Struktur-Test (IST-70). GOttingen: Hogrefe, 1970. lAUER, R.M. Autonomic recognition of names and faces in prosopagnosia: A neuropsychological application of the Guilty Knowledge Test. Neuropsychologia, 22: 457-469, 1984. lAVER, R.M. The cognitive psychophysiology of prosopagnosia. In H.D. Ellis, M.A. Jeeves, F. Newcombe and A. Young (Eds.), Aspects of Face Processing. Dordrecht, Boston: Martinus Nijhoff Publishers, pp. 253-267, 1986. lRUCE, V., and VALENTINE, T. Semantic priming of familiar faces. The Quarterly Journal of Experimental Psychology, 38A: 125-150, 1986. lRUYER, R., LATERRE, C., SERON, X., FEYEREISEN, P., STRYPSTEIN, E., PIERRARD, E., and RECTEM, D. A case of prosopagnosia with some preserved covert remembrance of familiar faces. Brain and Cognition, 2: 257-284, 1983. 528 Stefan R. Schweinberger and Others BRUYER, R. Covert face recognition in prosopagnosia: A review. Brain and Cognition, 15: 223-235, 1991. BURTON, A.M., YOUNG, A.W., BRUCE, V., JOHNSTON, R.A., and ELLIS, A.W. Understanding covert recognition. Cognition, 39: 129-166, 1991. CAMPBELL, R., LANDIS, T., and REGARD, M. Face recognition and lipreading. A neurological dissociation. Brain, 109: 509-521, 1986. CoLE, M., and PEREZ-CRUET, J. Prosopagnosia. Neuropsychologia, 2: 237-246, 1964. DAMASIO, A.R., DAMASIO, H., and VAN HOESEN, G.W. Prosopagnosia: Anatomic basis and behavioral mechanisms. Neurology, 32: 331-341, 1982. DE HAAN, E.H.F., BAUER, R.M., and GREVE, K. Behavioral and physiological evidence for covert recognition in a prosopagnosic patient. Cortex, 28: 77-95, 1992. DE HAAN, E.H.F., YouNG, A., and NEWCOMBE, F. Face recognition without awareness. Cognitive Neuropsychology, 4: 385-415, 1987. DE RENZI, E., BONACINI, M.G., and FAGLIONI, P. Right posterior brain-damaged patients are poor at assessing the age of a face. Neuropsychologia, 27: 839-848, 1989. EKMAN, P., and FRIESEN, W.Y. Pictures of Facial Affect. Palo Alto: Consulting Psychologists Press Inc., 1976. FARAH, M.J., O'REILLY, R.C., and VECERA, S.P. Dissociated overt and covert recognition as an emergent property of a lesioned neural network. Psychological Review, 100: 571-588, 1993. HAMSTER, W., LANGNER, W., and MAYER, K. Tiibinger-Luria-Christensen Neuropsychologische Untersuchungsreihe (TULUC). Weinheim: Beltz, 1980. HECAEN, H. The neuropsychology of face recognition. In G. Davies, H.D. Ellis and J. Shepherd (Eds.), Perceiving and Remembering Faces. London: Academic Press, 1981, pp. 39-54. HUBER, W., POECK, K., WENIGER, D., and WILLMES, K. Aachener Aphasie Test (AAT). Giittingen: Hogrefe, 1983. LANDIS, T., REGARD, M., BLIESTLE, A., and KLEIHUES, P. Prosopagnosia and agnosia for noncanonical views. An autopsied case. Brain, Ill: 1287-1297, 1988. LEVIN, H.S., HAMSHER, K., and BENTON, A.L. A short form of the test of facial recognition for clinical use. Journal of Psychology, 9I: 223-228, 1975. MALONE, D.R., MORRIS, H.H., KAY, M.C., and LEVIN, H.S. Prosopagnosia: a double dissociation between the recognition of familiar and unfamiliar faces. Journal of Neurology, Neurosurgery and Psychiatry, 45: 820-822, 1982. MEADOWS, J.C. The anatomical basis of prosopagnosia. Journal of Neurology, Neurosurgery and Psychiatry, 37: 489-501, 1974. NEWCOMBE, F., YouNG, A.W., and DEHAAN, E.H.F. Prosopagnosia and object agnosia without covert recognition. Neuropsychologia, 27: 179-191, 1989. PosNER, M.I., and SNYDER, C.R.R. Attention and cognitive control. In R.L. Solso (Ed.), Information Processing and Cognition: The Loyola Symposium. Hillsdale, New York: Lawrence Erlbaum Associates, 1975, pp. 55-85. RAVEN, J.C. The Standard Progressive Matrices. A Non-verbal Test of Person's Present Capacity for Intellectual Activity. London: Lewis, 1956. RENAULT, B., SIGNORET, J.-L., DEBRUILLE, B., BRETON, F., and BOLGERT, F. Brain potentials reveal covert facial recognition in prosopagnosia. Neuropsychologia, 27: 905:912, 1989. Rizzo, M., HURTIG, R., and DAMASIO, A.R. The role of scanpaths in facial recognition and learning. Annals of Neurology, 22: 41-45, 1987. SCHWEINBERGER, S.R. Funktionelle und neuroanatomische Aspekte der Prosopagnosie. Zeitschrift fiir Neuropsychologie, 3: 106-119, 1992. SCHWEINBERGER, S.R., KLOS, T., and SOMMER, W. Performance deficits and ERPs in unilaterally brain damaged patients during face and word recognition. Psychophysiology, 30: S59, 1993 (Abstract). TRANEL, D., and DAMASIO, A.R. Knowledge without awareness: An autonomic index of facial recognition by prosopagnosics. Science, 228: 1453-1454, 1985. TRANEL, D., and DAMASIO, A.R. Non-conscious face recognition in patients with face agnosia. Behavioural Brain Research, 30: 235-249, 1988. YoUNG, A.W., HELLAWELL, D., and DEHAAN, E.H.F. Cross-domain semantic priming in normal subjects and a prosopagnosic patient. The Quarterly Journal of Experimental Psychology, 40A: 561-580, 1988. Stefan R. Schweinberger, Fachgruppe Psychologie, Universitiit Konstanz, Postfach 5560, 78434 Konsanz, Germany. Covert face recognition in prosopagnosia APPENDIX True and Untrue Face-name Pairs Used in Experiment 1 Name 'rue pairs Bjorn Engholm John F. Kennedy Helmut Kohl Konrad Adenauer George Bush Bjorn Engholm John F. Kennedy Helmut Kohl Konrad Adenauer George Bush Jntrue pairs Ronald Reagan Oskar Lafontaine Leonid Breschnew Hans-Jochen Vogel Helmut Schmidt Helmut Schmidt Ronald Reagan Oskar Lafontaine Leonid Breschnew Hans-Jochen Vogel 529