Congenital prosopagnosia: face-blind from birth

Congenital prosopagnosia refers to the deficit in face processing that is apparent from early childhood in the absence of any underlying neurological basis and in the presence of intact sensory and intellectual function. Several such cases have been described recently and elucidating the mechanisms giving rise to this impairment should aid our understanding of the psychological and neural mechanisms mediating face processing. Fundamental questions include: What is the nature and extent of the face-processing deficit in congenital prosopagnosia? Is the deficit related to a more general perceptual deficit such as the failure to process configural information? Are any neural alterations detectable using fMRI, ERP or structural analyses of the anatomy of the ventral visual cortex? We discuss these issues in relation to the existing literature and suggest directions for future research. The failure to recognize faces can have devastating consequences for individuals suffering from this deficit.

The failure to recognize faces can have devastating consequences for individuals suffering from this deficit. Prosopagnosia, as this disorder is termed, although rather rare has usually been documented in individuals who have sustained brain damage in adulthood. Acquired prosopagnosia (AP) has been recognized for a long time [1,2] and has provided a unique window into the psychological and neural substrate of face processing. In recent years, attention has been paid to an analogous impairment, congenital prosopagnosia (CP), which refers to the impairment in face processing that is apparent from birth in the absence of any brain damage, and occurs in the presence of intact sensory and intellectual functions. As is true in AP, CP individuals are typically able to acknowledge that a face is present but are unable to identify the face and, hence, rely on voice or other cues such as clothing, gait or accessories for person identification. CP can also be severely debilitating, affecting the recognition of even the most familiar individuals, such as family members or one's self [3]. But the deficit in CP may go beyond recognizing familiar faces -CP individuals also fail to discriminate between unknown faces, suggesting a perceptual, rather than memorial basis, for the deficit. Unlike AP, CP may go undetected as the individual has no means of comparison with normal face processing skills. Also, because CP individuals have had a lifetime to develop compensatory strategies, they are adept at using salient features such as hairline or eye brows for recognition and may even perform normally on standard face recognition tasks if speed is not measured [4,5].
Importantly, the term 'congenital' is used in CP specifically to denote the absence of an acquired lesion or any other neurological concomitant at any stage of development [6,7]. As such, this label excludes individuals with a face processing impairment resulting from visual deprivation, such as in cases of infantile cataracts (see Box 1), or from other developmental problems as in cases of Autism Spectrum Disorder [8]. CP also contrasts with the more general term 'developmental prosopagnosia', which includes individuals with CP but also individuals who have sustained brain damage either before birth or in early childhood [4,[9][10][11]]. An important diagnostic criterion for CP is that face perception was never normal in the lifetime of the individual; in most cases we can only rely on self testimony or parents' testimony, although in some cases there is supporting evidence from formal testing carried out during childhood (see [12,13] for a longitudinal case study and [6,14] for CP in children, although note that with the exception of [6], these cases have abnormal EEGs and so may not be pure CPs). A further crucial criterion is that no positive evidence for any neurological impairment should be present. Finally, there is accumulating evidence that there is a familial factor in many cases of CP (as detailed below) and thus, formal testing establishing face processing impairments in additional family members could further assist in the diagnosis of CP rather then AP.
In recent years, a growing number of CP cases have been reported [15], but whether this is because of increasing prevalence or increased recognition of the disorder is not known. In spite of this flurry of research activity, many questions remain unanswered. In additional, many inconsistencies are apparent, perhaps because of the heterogeneity of the disorder and the varying methods of assessment used. We begin by reviewing the existing behavioral studies. Among the issues addressed here are the nature and severity of the face-processing deficit in CP, the extent to which the disorder is selective for faces and the relationship between the deficit in face perception and other forms of perceptual impairments. There are also a few recent studies investigating a possible neural basis for CP and we review these next. We conclude by examining the potential contribution of studies of CP to understanding normal brain-behavior correspondences and by suggesting some directions for future research.

Nature and extent of deficit in CP
Although an impairment in face processing is de facto required for the diagnosis of CP, face processing is not monolithic and can involve, for example, detecting whether a face is present among non-face stimuli, determining whether two faces are the same or different or recognizing the individual identity of a face [16,17]. Which of these processes is implicated in CP is unclear and this is explored in some recent papers [15]. Table 1 summarizes biographical and performance characteristics from five additional cases from a recent study [18]. Most cases of CP are able to detect faces from among other objects [11]. Also, some CP individuals succeed in relatively easy face matching tasks (perhaps because this can be achieved by matching individual features), but when reaction time or more detailed measures of discrimination such as A prime (A′) is measured and task demands are increased, the deficit is uncovered more clearly [15,[18][19][20]. In these conditions, roughly half the CP cases can discern the gender and age of the face whereas the remaining half cannot [15], and the same holds true for comprehending emotional expressions [6,13,14]. The extent to which CP individuals can recognize famous faces has similarly yielded mixed results, with some apparent successes [21,22] but other clear failures [11,18,19]. One possible explanation concerns the images used in these tests: often these contain cues such as clothing (e.g. Reagan in a cowboy hat) or other salient facial cues (Stallone with a black eye) and reasonable performance may be supported by the presence of these cues. Whether CP individuals can show implicit or covert processing of faces in the absence of explicit knowledge also remains controversial. In one case [6], despite the CP individual's failure to explicitly identify familiar faces, larger amplitude skin conductance response s were measured for familiar compared with unfamiliar faces. This recognition without awareness is not obtained in all cases [13].
A classical distinction made between different forms of prosopagnosia is whether the deficit is 'apperceptive' or 'associative' in nature. The dichotomy, originally offered by Lissauer [23], attributes the former to a deficit in deriving a sufficiently intact percept whereas, in the latter type, the root of the impairment is at the level of recognition or assignment of meaning. Some CP individuals have been categorized as apperceptive [14] whereas others have been diagnosed as being associatively [6] prosopagnosic [24]. Most interesting perhaps is that the disorder may occur in the absence of any apparent deficits in low-level visual abilities [3,11,18,19]. But, as is probably evident, the clearest pattern that emerges from the literature is the heterogeneity of the disorder. Systematic investigations and perhaps standard methods of investigation are urgently needed especially with the recent influx of new cases (one paper reports that they have been contacted by over 150 individuals since they established a website for this purpose [4]) to be able to elucidate the range of behavioral impairments in CP.

Selectivity of the disorder
The extent to which CP (and AP) is specific to faces remains the subject of an ongoing controversy in the field of visual cognitive neuroscience. Neuropsychological case studies have demonstrated a double dissociation between the recognition of faces and objects, suggesting independence or segregation between these processes [ . An alternative view, however, also supported by numerous neuropsychological and imaging investigations, is that there is a single, general-purpose visual process that subserves both objects and faces. The dissociations, then, arise because of the unusual demands placed on the system by faces, the only class of stimuli for which all humans have extraordinary expertise [32]. Because face processing typically involves individual identification (shown a face, one responds with the individual's identity) whereas other objects are usually recognized at a basic level (for example, as a chair or apple or house), faces entail fine-grained discrimination of perceptually similar exemplars within a category. Whether recognition of other visual non-face objects might also be impaired when homogeneous, complex stimuli are used, remains debatable. Several studies have shown that many AP patients have difficulties categorizing exemplars of within-class objects, which share the same complex configuration [13, [33][34][35] although this does not seem to be true for all prosopagnosic subjects [24,[36][37][38]. The controversy between a domain-specific organization of faces versus a more generic system has not been resolved (for further details, see [31,32,[39][40][41][42]).
Of the nine cases of CP reviewed by Kress and Daum [15], two appear to be impaired in the recognition of nonface objects, one mildly (subject AB) and one severely (subject LG). We note, however, that the methods of assessment vary considerably and that reaction time is generally not used despite the possibility that the CP individuals might be inordinately slow or might even trade speed off against accuracy. Using an old vs. new recognition memory paradigm and measuring both A′ and reaction time, Duchaine and Nakayama [20] showed a clear dissociation between faces and other object categories (tools, cars, horses) in four out of their seven subjects. The five CP subjects studied by us [18] are all impaired at discriminating between common objects and between novel objects ('Greebles') especially when the discrimination is at the individual level (for example, two different chairs or two Greebles of the same family and gender) even when the pair to be discriminated is visible to the subject for an unlimited duration (see Table 1). Importantly, this deficit is less marked than that for faces and its severity varies considerably among the CP individuals [18]. Some CP subjects also read more slowly than their counterparts [14].

The relationship between configural and face processing in CP
The idea that configural processing is required for face processing is not new, and it has long been suggested that a loss of configural processing may underlie prosopagnosia. Recent support for this view comes from the finding that patients with acquired lesions of the fusiform face area are impaired at deriving the spatial relations between the components of the face [43]. The failure to take the spatial relations between elements into account (see Box 1) forces the individual to rely on a more piece-meal or feature-based strategy in constructing face representations [44].
Evidence for the strong relationship between faces and configural processing also comes from studies comparing processing of upright versus inverted faces, relative to processing upright versus inverted objects. In normal individuals, recognition and discrimination of faces is notoriously better for upright than for inverted faces [45] and this holds to a lesser extent for objects. In general, this decrement under inversion or 'face inversion effect' is taken to be a hallmark indicator of the fact that faces in their upright configuration are processed as a whole; when they are inverted, the whole or configuration is no longer available and a more part-based system is utilized, leading to the cost in reaction time and accuracy for inverted compared with upright faces [46][47][48]. Individuals with AP often do not show the decrement under inversion and may even show better performance on inverted than upright faces (inversion superiority effect) presumably because their feature-based strategy can proceed unhampered by attempts at configural processing [49,50] (see Figure 1). Many CP individuals are also not adversely affected by inversion of faces [11,18] and can also show the inversion superiority effect (Figure 1).
The failure to derive configurations is thought to be particularly devastating for face processing relative to processing other objects, given the homogeneity between exemplars and the need to rely on the second-order rather than first-order statistics of the input [51]. A configural impairment may affect other visual stimuli too if, as is true for faces, the spatial relations between the components need to be represented to differentiate perceptually similar exemplars (see Figure 2). But what exactly constitutes configural processing is itself controversial [40] and the subject of an ongoing debate [52] (see Box 1).

Neural investigations of CP
Only a small number of studies have conducted neural investigations of CP. Two studies have used evoked response potentials (ERP) with specific exploration of the N170 potential, recorded from a relatively circumscribed region at the posterior-inferior aspect of the temporal lobe [28,53]. In one study, subject YT, monitored the frequency of occurrence of butterflies in a stream of pictures of faces, cars, furniture and nonsense stimuli (scrambled input). Although a conspicuous N170 was detected [54], it was elicited for objects as well as for faces, in contrast with the known face-selectivity in normal individuals, yielding a reduced amplitude difference between stimulus types. In a second study with two CP individuals [5], the finding was even more dramatic as neither individual showed any ERP difference between houses and faces. This lack of specificity in this early waveform suggests that the initial encoding of input may not be sufficiently precise to yield fine-grained discrimination.
Reduction or elimination of face-selectivity was also reported in a functional magnetic resonance imaging study [55] in which similar activations for faces and houses were found in the fusiform face area (the pre-eminent area for face processing; FFA). In fact, there was no face-selective activation in any area along the ventral visual pathway even when a permissive threshold was used. However, it is important to note that two of the three subjects who participated in this study (GA and RP) and showed the most abnormal pattern of activation, sustained injuries during childhood (excluding them from the diagnosis of CP which we offer) and this might account for the lack of activation or face selectivity.
In contrast to these findings, subject YT, who showed reduced face selectivity in the ERP study reported above exhibited face-related activation in ventral visual cortex that mirrored that of normal individuals in terms of site of activation, activation profile and hemispheric laterality [56]. The only possible difference between YT and the controls was a slightly reduced degree of selectivity for faces over objects in the lateral occipital cortex. This finding has been replicated and extended in a group of four CP individuals [57] all of whom evinced a normal pattern of fMRI activation in the fusiform gyrus (FFA) and in other ventral occipito-temporal (VOT) areas, in response to faces, buildings and other objects, shown both as line drawings and as more natural images (see Figure 3). These CP individuals also showed normal adaptation levels and, like control subjects, exhibited evidence of global representation of faces in the FFA. The absence of a BOLD-behavioral correlation (impaired behavior, normal BOLD pattern) suggests that facerelated activation in ventral cortical areas might be necessary but may not be sufficient for normal face identification. Interestingly, these CP subjects exhibited robust bilateral face-related activation in prefrontal regions (pre-central sulcus, inferior frontal sulcus, anterior lateral sulcus) in the context of a one-back, working memory task. This was in sharp contrast to their control group who only exhibited some right lateralized activation in the pre-central sulcus. The poor performance exhibited by CP subjects (Figure 3b) on the one-back task specifically for face stimuli suggests that this prefrontal activity is related to working memory [58-60]. Alternatively, this activity could reflect the recruitment of compensatory mechanisms under taxing perceptual conditions by the CP subjects. Further research is required to determine the exact role of this activation in face processing in CP individuals. As is apparent, the neural mechanism underlying CP remains unspecified and structural scanning typically reveals no observable impairment [6]. We should note that detailed structural MRI volumetric measurement of the ventral cortex suggests a smaller right temporal lobe. Relative to controls, YT had a smaller right temporal lobe [19]. We have now replicated this pattern in other CP individuals (Behrmann et al., unpublished data); when the volumes are normalized for difference in head size by taking intracranial region into account, relative to matched controls, the CP subjects have a smaller right anterior fusiform region and a larger right mid/posterior-fusiform gyrus. No differences are observed in the hippocampus or parahippocampal region, suggesting that it is not the entire temporal cortex that is reduced but rather the fusiform gyrus, the pre-eminent structure for face processing. Whether this structural difference is a result of CP, however, revealing atrophy from under-use or whether this is the underlying predisposition that triggers CP in the first place remains unknown and distinguishing between cause and effect will be crucial.

Conclusion
Aside from the intrinsic interest in the study of neurodevelopmental disorders, CP has the potential to inform our understanding of the psychological and neural mechanisms underlying face processing and their relationship to non-face processing. For example, the fact that CP exists at all and that, over the course of their lifetime the deficit is not resolved in these individuals, suggests that there is a limit on the plasticity of the human ventral visual system; this is in marked contrast with the rather widespread plasticity reported in other sensory domains as well as in higher-cognitive skills such as language [61,62]. Studying CP individuals also enables us to examine the neural mechanisms mediating face recognition using fMRI or other imaging techniques, unaffected by damage that might disrupt normal blood flow or neurovascularization in cases of AP [63]. Hence the fact that the CP individuals show normal BOLD activation of the fusiform face area and other ventral cortical structures is highly provocative and challenges us to generate a more refined and precise notion of the computational properties of these occipito-temporal areas. We should also note that CP might not be as rare as previously thought and, like other developmental disorders (see Box 2), questions about intervention and training will become more pressing.
Finally, many of the individuals with CP tested so far, both in our studies (see Table 1) and in the literature [20], have a family member who is also impaired at face processing. This familial component is of great interest and studies of the genetic predisposition and mechanism will allow us to start building bridges between behavior and cortical development. The only known genetic investigations of which we are aware to date (Grüter, M., MD thesis: Genetik der kongenitalen Prosopagnosie, University of Münster, Germany, 2004) reports that the cumulation segregation ratios are compatible with a simple autosomal dominant mode of inheritance but the molecular DNA and linkage studies have not been reported. Despite the recent progress in the characterization of CP, many outstanding questions exist (see Box 3) and a full explanation of the psychological and neural mechanisms giving rise to CP remains elusive. Note that despite the fact that this was a perceptual discrimination task that did not require face recognition, both patient groups were dramatically impaired compared with the control group. Interestingly, CP subjects were even slower than the AP subjects. Both AP and CP groups exhibited the inversion-superiority effect (although not all individual subjects exhibited this effect), which supports the notion that their configural face processing is impaired. . As is evident, CP individuals show normal speed in identifying local letters but are slow at deriving the global whole from the local elements, suggestive of a failure in representing the spatial relations between components of a display and this form of configural processing may have direct impact on face processing. Note that when the global letter is inconsistent with the local identity but must be identified, interference from the local onto the global (additional RT cost) is observed in CP. In a second, related experiment (see [18]), CP individuals, in contrast with normal controls, are primed by the elements in a display rather than by the overall shape. Thus, when shown an ambiguous display such as a square made of small circles, performance was enhanced for a subsequent display containing circles rather than for a display containing a square. Controls were primed for the square and showed a bias for global shape perception [64]. Note that the CP group was both less accurate and slower than the control group on performance of this task. However, there was some variability within the group with some subjects trading off speed for accuracy or vice versa (compare subjects TM and MT). (c) Averaged face (red) and building (green) activation maps of the CP subjects and control subjects projected on an inflated brain representation shown from a ventral view. The activation is projected on the same brain and is shown in the same statistical threshold (multi-subject GLM statistics, faces: p<0.005 buildings: p<0.05 random effects) to enable direct comparison between the two groups. Note that both controls and CP subjects exhibited bilateral face-related activation in the fusiform gyrus and lateral occipital region and building-related activation in the collateral sulcus. Abbreviations: FG, fusiform gyrus; CoS, collateral sulcus; Ant., anterior; Post., posterior. (d) Face and building activation maps are shown for each of the CP subjects and for four representative control subjects. Note the substantial similarity in the activation pattern exhibited by all CP subjects and controls.

Box 1. The contribution of early visual experience to the development of intact configural face processing.
Faces are complex visual stimuli and their processing therefore involves several different psychological processes [40,48,65]. Le Grand et al.
[66] recently used a sensitive behavioral task to study some of these processes in a group of individuals born with congenital bilateral cataracts.
Crucially, the cataracts precluded retinal stimulation until surgical removal (62-187 days following birth in these individuals) and so studying this patient group allows the assessment of early visual deprivation on the development of normal face processing (see also [67]. In the task they used, faces could be differentiated from each other in two ways (see Figure I): either the face features were changed (featural set) or the features remained constant but the spacing between them varied (configural set). Interestingly, bilateral cataract patients were substantially impaired compared with control subjects on the configural but not on the featural task, suggesting that very early visual experience (first few months of life) is necessary for the normal development of expert face processing which largely relies on second-order, configural processing.
That these individuals do not acquire these abilities even after many years of exposure to faces (like case with CP) implies that plasticity is limited and that the normal development of face processing is to a large extent experience-dependent.
In a second study, Le Grand and colleagues [68] explored hemispheric differences in face processing by testing congenital left eye deprivation (right hemisphere) or right eye deprivation (left hemisphere) cataract patients. Crucially, left but not right eye cataract patients exhibited a specific right hemisphere dependence on second-order configural but not featural processing. This suggests that the right hemisphere is more crucial for the development of intact configural processing. Whether the impairment in configural processing exhibited by these patients is unique to the domain of faces or whether it will also be evident for other stimulus categories when configural processing is required remains to be explored. These findings, however, advance our understanding of the development of different components of normal face processing [40] and raise several interesting questions in relation to CP. First, if and to what extent might CP evolve from lack of exposure or 'training' of the visual system and, if so, what underlying neuronal impairment induces such 'deprivation'. Furthermore, why does face processing and its neural substrate lack plasticity to a large extent [10], and given this, what training regimes might ameliorate these processes in different patient populations (cataract, CP, AP). Future studies combining behavioral and imaging techniques will be required to address these fascinating issues [69].

Box 2. Relationship to other neurodevelopmental disorders
Interestingly, just as a deficit in visual perception may arise congenitally and without an obvious neural basis (i.e. CP), parallel disorders exist in other visual and non-visual domains. The most obvious parallel is with developmental dyslexia (DD), also known as specific reading disability.
DD is also characterized as a disorder in which sensory and intellectual functions are intact and in which motivation and opportunities for acquiring reading are normal [70]. As with CP, understanding the biological origin of DD is of major importance and detailed analysis of the phonological disorder, its specificity, sensitivity to remediation and neural correlate is the topic of many recent investigations. The increased prevalence of DD (roughly 5-17% of the population) relative to CP may be partially attributed to the enormous obstacle faced by an individual in a society that demands literacy whereas the consequences of CP are not as pressing. Many functional imaging studies have been conducted in individuals with developmental dyslexia but much controversy exists concerning the cortical structures that are implicated in this disorder [70].
A further disorder, which seems remarkably similar to CP but which occurs in the auditory domain is congenital amusia. This disorder manifests in dramatic difficulties in appreciating, perceiving and memorizing music, but no difficulties recognizing and processing non-music material such as voices, spoken lyrics or environmental sounds. Like CP, congenital amusia cannot be explained by sensory or brain anomalies, low intelligence or music deprivation. Similar questions to those raised in relation to CP, concerning the specificity of the disorder to music and the possibility of a more fundamental perceptual impairment, giving rise to the amusia have been studied [71]. It has been suggested, for example, that a more fundamental deficit in pitch discrimination may underlie congenital amusia but as in the case of CP the disorder may not be homogeneous, as a problem in the temporal domain of music perception has also been reported in some individuals. An anatomical MRI scan of one such individual's brain failed to reveal any macrostructural abnormalities [71]. Finally, congenital amusia also seems to run in families, suggesting the possibility that again, similar to CP, this defect may be genetically determined.
There is one final congenital disorder that is relevant and it affects language and speech production as well as oromotor control and articulation [72]. Of particular interest for the case of CP is that genetic testing in the KE family, half of whom suffer from this disorder, has identified a mutation in the FOXP2 gene [73,74]. Functional MRI studies have shown differences between affected and unaffected family members in Broca's area with the affected members revealing a more posterior and more extensively bilateral pattern of activation in all speech tasks. Detailed volumetric analysis has shown that the volume of the caudate nucleus, especially in the superior portion, was reduced bilaterally in the affected family members compared with both the unaffected members and the group of age-matched controls. These findings suggest that the FOXP2 gene is involved in the development of the neural systems that mediate speech and language. Whether similar mutations (and accompanying cortical volumetric changes) will be found in CP remains to be determined but the possibility of doing so lies tantalizingly before us. To what extent do CP individuals exhibit evidence for covert recognition of faces? Can we find behavioral (such as increased priming) and neuronal correlates of covert recognition in these individuals? • Given that CP individuals never develop normal face recognition despite seemingly normal exposure to faces throughout their life, it is important to understand whether and to what extent face processing in such individuals can be improved following specifically designed training regimes. Related questions are what regimes will be maximally effective to achieve this goal and what underlying neuronal changes will accompany such behavioral improvement. • There is growing evidence suggesting that there is a familial factor in CP. It will be crucial to determine whether there is an underlying genetic basis for CP. Establishing a link between a specific genetic makeup and high-level cognitive functions such as face processing will be a dramatic breakthrough in the way we understand human cognition. • In addition to the visual cortex, face processing involves other brain regions such as prefrontal cortex. It will be important to understand the role of these regions in face processing and representation in both normal and CP individuals. This issue is of great interest as there is now evidence suggesting that prefrontal cortex might be engaged differently in face processing in CP individuals compared with normal individuals. Non-face discrim.