Neural and genetic foundations of face recognition and prosopagnosia.
ABSTRACT Faces are of essential importance for human social life. They provide valuable information about the identity, expression, gaze, health, and age of a person. Recent face-processing models assume highly interconnected neural structures between different temporal, occipital, and frontal brain areas with several feedback loops. A selective deficit in the visual learning and recognition of faces is known as prosopagnosia, which can be found both in acquired and congenital form. Recently, a hereditary sub-type of congenital prosopagnosia with a very high prevalence rate of 2.5% has been identified. Recent research results show that hereditary prosopagnosia is a clearly circumscribed face-processing deficit with a characteristic set of clinical symptoms. Comparing face processing of people of prosopagnosia with that of controls can help to develop a more conclusive and integrated model of face processing. Here, we provide a summary of the current state of face processing research. We also describe the different types of prosopagnosia and present the set of typical symptoms found in the hereditary type. Finally, we will discuss the implications for future face recognition research.
- SourceAvailable from: Edmund T Rolls[show abstract] [hide abstract]
ABSTRACT: Of 497 single neurones recorded in the cortex in the fundus of the superior temporal sulcus (STS) of three alert rhesus monkeys, a population of at least 48 cells which were selectively responsive to faces had the following response properties: (1) The cells' responses to faces (real or projected, human or rhesus monkey) were two to ten times as large as those to gratings, simple geometrical stimuli or complex 3-D objects. (2) Neuronal responses to faces were excitatory, sustained and were time-locked to the stimulus presentation with a latency of between 80 and 160 ms. (3) The cells were unresponsive to auditory or tactile stimuli and to the sight of arousing or aversive stimuli. (4) The magnitude of the responses of 28 cells tested was relatively constant despite transformations, such as rotation, so that the face was inverted or horizontal, and alterations of colour, size or distance. (5) Rotation to profile substantially reduced the responses of 21 cells (31 tested). (6) Masking out or presenting parts of the face (i.e. eyes, mouth or hair) in isolation revealed that different cells responded to different features or subsets of features. (7) For several cells, responses to the normal organisation of cut-out or line-drawn facial features were significantly larger than to jumbled controls. These findings indicate that explanations in terms of arousal, emotional or motor reactions, simple visual feature sensitivity or receptive fields are insufficient to account for the selective responses to faces and face features observed in this population of STS neurones. It appears that these neurones are part of a system specialised to code for faces or features present in faces, and it is suggested that damage to this system is related to prosopagnosia, or difficulty in face recognition, in man and to the tameness and social disturbances which follow temporal lobe damage and are part of the Klüver-Bucy syndrome in the monkey.Experimental Brain Research 02/1982; 47(3):329-42. · 2.22 Impact Factor
- [show abstract] [hide abstract]
ABSTRACT: To investigate where facial recognition is located anatomically and to establish whether there is a graded transition from unimpaired recognition of faces to complete prosopagnosia after infarctions in the territory of the middle cerebral artery. A computerised morphing program was developed which shows 30 frames gradually changing from portrait photographs of unfamiliar persons to those of well known persons. With a standardised protocol, 31 patients with right and left sided infarctions in the territory of the middle cerebral artery and an age and sex matched control group were compared by non-parametric tests. Facial recognition in patients with right sided lesions was significantly impaired compared with controls and with patients with left sided lesions. A graded impairment in facial recognition in patients with right sided ischaemic infarcts in the territory of the middle cerebral artery seems to exist.Journal of Neurology Neurosurgery & Psychiatry 04/1997; 62(3):261-4. · 4.92 Impact Factor
- [show abstract] [hide abstract]
ABSTRACT: Expertise with unfamiliar objects ('greebles') recruits face-selective areas in the fusiform gyrus (FFA) and occipital lobe (OFA). Here we extend this finding to other homogeneous categories. Bird and car experts were tested with functional magnetic resonance imaging during tasks with faces, familiar objects, cars and birds. Homogeneous categories activated the FFA more than familiar objects. Moreover, the right FFA and OFA showed significant expertise effects. An independent behavioral test of expertise predicted relative activation in the right FFA for birds versus cars within each group. The results suggest that level of categorization and expertise, rather than superficial properties of objects, determine the specialization of the FFA.Nature Neuroscience 03/2000; · 15.25 Impact Factor
Neural and genetic foundations of face recognition
Thomas Gru ¨ter1, Martina Gru ¨ter2
and Claus-Christian Carbon1*
1University of Vienna, Faculty of Psychology, Vienna, Austria
2Nottulner Landweg 33, D-48161 Mu ¨nster, Germany
about the identity, expression, gaze, health, and age of a person. Recent face-processing
models assume highly interconnected neural structures between different temporal,
occipital, and frontal brain areas with several feedback loops. A selective deficit in the
visual learning and recognition of faces is known as prosopagnosia, which can be found
both in acquired and congenital form. Recently, a hereditary sub-type of congenital
prosopagnosia with a very high prevalence rate of 2.5% has been identified. Recent
research results show that hereditary prosopagnosia is a clearly circumscribed face-
processing deficit with a characteristic set of clinical symptoms. Comparing face
processing of people of prosopagnosiawith that of controls can help to develop a more
conclusive and integrated model of face processing. Here, we provide a summary of the
current state of face processing research. We also describe the different types of
prosopagnosia and present the set of typical symptoms found in the hereditary type.
Finally, we will discuss the implications for future face recognition research.
Faces are of utmost importance for human social life.Not only do they tell us the identity
of other people, but they also inform us about their mood, age, health, and their gender.
They are a major factor of sexual attractiveness. From the eyes, we can judge the
direction of gaze with remarkable accuracy, and the observation of lip movements helps
to understand speech.
Neurons tuned to faces are not limited to human brains, they have been found in the
superior temporal sulcus (STS) of monkeys as well (Perrett, Rolls, & Caan, 1982).
Recently, Tsao, Freiwald, Tootell, and Livingstone (2006) identified a face sensitive area
in the macaque monkey brain, which is topographically homologous to the face
sensitive area on the fusiform gyrus in humans (fusiform face area, FFA).
This indicates that face recognition is evolutionarilyold and alreadyexisted in the last
common ancestor of humans and macaque monkeys.
*Correspondence should be addressed to University of Vienna, Faculty of Psychology, Liebiggasse 5, A-1010 Vienna,
Austria/Europe (e-mail: email@example.com).
Journal of Neuropsychology (2008), 2, 79–97
q 2008 The British Psychological Society
Other face-related abilities are probably inherited as well: the gaze direction
judgment is active from birth (Farroni, Massaccesi, Menon, & Johnson, 2007), and
basic facial expressions as well as their recognition are universal among humans
(Ekman et al., 1987).
The function of the human brain’s face identity recognition system cannot be
substituted in full by other brain regions. Even if the system is damaged in early
childhood, the plasticity of the juvenile brain may not be sufficient to recover the
functionality (Barton, Cherkasova, Press, Intriligator, & O’Connor, 2003). The result is a
life-long prosopagnosia, a selective face recognition deficit.
Prosopagnosia is a scholarly Greek compound word (from prosopon ¼ the face
and agnosia ¼ non-recognition or non-knowledge). It was introduced by Joachim
Bodamer (1947), who published a very thorough description of two soldiers with a
markedly decreased face recognition after suffering severe brain injuries in the Second
World War. In the following years, several case reports were published and
prosopagnosia was regarded as a rare consequence of a damage to the temporal or
There is also a congenital type of prosopagnosia, which is present from birth and
is not accompanied by any gross brain abnormalities, first reported by McConachie
(1976). Recently, this type has been shown to be very common: about 2.5% of the
Caucasian population could be affected, placing the prevalence of congenital
prosopagnosia in the same region as dyslexia or dyscalculia (Kennerknecht et al.,
2006). The condition can be hereditary. De Haan (1999) found a family where the
father and two daughters suffered from a markedly reduced face recognition ability.
Grueter (2004) examined 38 people with congenital prosopagnosia in six families.
Prosopagnosia diagnosis is difficult: there is no test which can establish the diagnosis
of a face recognition dysfunction. Other than in the case of dyscalculia and dyslexia,
face recognition skills are not taught in school and there are no age- and general
intelligence-adjusted standards. The widely used Warrington Recognition Memory for
Faces (RMFs) and the Benton Facial Recognition Test (BFRT) are not sufficiently face
specific (Duchaine & Weidenfeld, 2003). Other tests have been proposed (Duchaine
& Nakayama, 2006; Grueter et al., 2007; Minnebusch, Suchan, Ramon, & Daum,
2007), but none of them can establish but only support a diagnosis. Prosopagnosia is
not always obvious. In Bodamer’s (1947) two cases, the patients never complained of
their face recognition deficit, only Bodamer’s observations, targeted questions and
subsequent tests revealed their impairment. Rosler, Lanquillon, Dippel, and Braune
(1997) tested 31 patients in the second week after a right (16) and left (15) temporal
lobe infarction for face recognition. They found that most of them were impaired,
some of them severely, but none complained of face recognition problems. The
authors conclude that ‘probably these deficits are easy to compensate for, because
recognizing a certain person ‘behind’ a face is made easy by the perception of further
visual and acoustic cues’. If this applies to patients with a recent stroke, how much
more difficult may it be to identify people with a life-long face recognition deficit?
We should always be aware that in the first place face recognition is a tool for person
recognition and people with a defect will have a large reservoir of other tools to
In this article, we will sum up the current research on face recognition and its
dysfunctions with a special focus on the hereditary type of prosopagnosia, because with
a prevalence of 2.5% this type is by far the most frequent one. We will propose a
diagnostic checklist based on the clinical symptoms.
80Thomas Gru ¨ter et al.
Normal face recognition process
Face recognition is a multistage process. First, a face is recognized as such; this stage
is often described as detection phase. In a second step, individual and non-individual
facial information (emotional expression, eye gaze, gender, age, and health) must be
analysed. Individual face recognition includes a structural encoding and a comparison
with stored mental images of faces. The ‘classic’ functional face recognition model of
Bruce and Young (1986) depicts this process (Figure 1). It does not connect the tasks
involved with any cerebral structures, though. A later model by Ellis and Lewis
assumes conscious face recognition and the generation of a feeling of familiarity to be
processed by separate modules. This dissociation can be shown in some cases of
Capgras delusion, a rare psychiatric condition where the patients are convinced that
Figure 1. Functional model of face recognition by Bruce and Young (1986).
Face recognition and prospagnosia81
close relatives are replaced by identically looking imposters. Ellis and Lewis assume
that in the Capgras delusion the conscious face recognition is preserved but no
longer triggers the feeling of familiarity. In acquired prosopagnosia, the opposite
effect might occur: while there is no conscious face recognition, a subconscious
feeling of familiarity and an autonomous reaction, for example, measured by skin
conductance response, can still be shown in some cases (Ellis & Lewis, 2001). The
feeling of familiarity may also be distorted in a different way: Vuilleumier, Mohr,
Valenza, Wetzel, and Landis (2003) reported the case of a young woman who felt an
irritating hyperfamiliarity towards strangers after a left temporal brain lesion. Her
conscious face recognition, though, was undisturbed. A similar case of hyperfami-
liarity in an epileptic patient who suffered a large left occipito-temporal tissue
damage in early childhood was reported by Murai, Kubota, and Sengoku (2000). Their
patient showed hyperfamiliarity in combination with a paramnesia. Obviously, the
feeling of familiarity is generated in a separate module which is triggered by the face
recognition unit. While Ellis and Lewis hold a disconnection responsible for false
feeling of unfamiliarity, we assume insufficient trigger input to cause a lasting
uncertainty as observed in hereditary prosopagnosia. Instead, the false feeling of
unfamiliarity in the Capgras delusion would stem from a damage in the familiarity
feeling itself. Based on functional imaging research, Gobbini and Haxby (2007)
proposed a face recognition model which ties functional units to certain brain
regions. They distinguish between a core and an extended system (Figure 2). The
core system comprises the occipital face area (OFA) in the inferior occipital lobe, the
fusiform face area (FFA) in the middle fusiform gyrus, and the face area in the dorsal
superior temporal sulcus (STS). In several fMRI studies, these areas showed an
enhanced activation for faces as compared with other visual objects. The STS is
supposed to process dynamic facial information (expression, eye gaze, and facial
speech) while the OFA and FFA focus on invariant facial features for identification.
The core system encodes the visual appearance of familiar faces while the extended
system provides the person’s information and the emotional response. The model is
less explicit about the hierarchy of functions and regions than the earlier model by
Haxby, Hoffman, and Gobbini (2000), which it supersedes, and proposes a tightly
integrated network of face-processing areas.
Face recognition network
Several studies have shown that visual face processing activates a whole network of
brain regions (Gobbini & Haxby, 2007; Ishai, Schmidt, & Boesiger, 2005). The
hierarchical model which postulated that the OFA is responsible for the initial
processing of facial features and feeds the results into the FFA and STS is outdated.
Rossion et al. (2003) reported the case of a woman with an acquired prosopagnosia
due to a major tissue loss in the right inferior lateral occipital lobe and the left medial
temporal lobe, mainly in the fusiform gyrus. Though the right OFA was destroyed, the
right FFA reacted to faces. A similar case was reported by Steeves et al. (2006). They
found a face-related activation in the FFA and the STS in a prosopagnosic patient with a
destroyed OFA. Obviously, the FFA reaction to faces is not necessarily missing in
acquired prosopagnosia, though its quality may change. A second study with the first
patient found a missing adaptation reaction to known faces in the FFA activation (Schiltz
et al., 2006). While in the controls, the FFA activation decreased when a face is shown
repeatedly, the prosopagnosic patient showed no such adaptation.
82Thomas Gru ¨ter et al.
A normal FFA activation has also been observedin congenital prosopagnosia (Avidan,
Hasson, Malach, & Behrmann, 2005). Even in late-blind people, an FFA reacted when
they touched a doll’s face; no such activation was seen in congenitally blind people
(Goyal, Hansen, & Blakemore, 2006). This may indicate an activation by the mental
imagery of faces, which has indeed been observed (Ishai, Haxby, & Ungerleider, 2002).
Usually, the right FFA reacts strong to faces than the left, but in some people, the left FFA
dominates and in others, no face-specific activity could be shown in the fusiform gyri
(Kanwisher, McDermott, & Chun, 1997).
Face-selective areas are also connected to voice recognition areas. In a cross-modal
study von Kriegstein, Kleinschmidt, Sterzer, and Giraud (2005) showed that face-
selective areas are activated by voices of familiar persons, if, and only if, the task
emphasized speaker recognition over content recognition. During this task the FFA
was coupled with the STS voice region but not with other cortical regions normally
involved in person recognition.
Figure 2. Model of face processing by Gobbini and Baxby (2007).
Face recognition and prospagnosia83
Face learning and face memory involve a widely distributed cerebral network
(Ishai et al., 2005) including the middle temporal lobe on both sides, hippocampus,
parahippocampal gyri, amygdala, inferior frontal gyrus, and orbitofrontal cortex. Face
recognition and learning are influenced by age, emotional expression, orientation,
and attention (Aylward et al., 2005; Carbon & Leder, 2006; Carbon, Schweinberger,
Kaufmann, & Leder, 2005; Iidaka et al., 2002; Lueschow et al., 2004). Each
processing stage is tightly integrated with and related to other tasks like voice
recognition or expression evaluation (Ishai et al., 2005). It is being debated, whether
a specialized neural network is responsible for face recognition (McKone, Kanwisher,
& Duchaine, 2007) or whether face recognition is just one task of a neural structure
which processes the individual recognition of homogenous object classes (Bukach,
Gauthier, & Tarr, 2006). In the latter case, face recognition would be an object
recognition subtask. In order to prove the latter concept, Gauthier and Tarr (1997)
constructed artificial objects called ‘greebles’. Greebles are club- or column-shaped
artificial objects with pins, beaks, or horizontal rings attached at different locations.
Similar looking greebles form one family. Using this concept, Gauthier, Tarr and their
co-workers defended their idea that face recognition is only an instance of the brain’s
ability to distinguish very similar objects at expert level (Gauthier, Skudlarski, Gore, &
Anderson, 2000). If so, prosopagnosics should have problems to distinguish between
similar objects. Duchaine, Dingle, Butterworth, and Nakayama (2004), though, found
a normal performance in a prosopagnosic for a visual greeble learning and
discrimination task. Also, if face recognition would only be an instant of expert level
visual object discrimination, an object agnosia should always imply a prosopagnosia.
This has been disproved: Carlesimo, Fadda, Turriziani, Tomaiuolo, and Caltagirone
(2001) reported a patient with an acquired global visual object amnesia who could
learn and recognize faces at a normal level. These double dissociate findings suggest
independent neural structures for face and object recognition at least to a certain
In the 1970s to 1990s, there was a debate whether a right-sided lesion could
cause a prosopagnosia, or whether lesions in both occipital and temporal lobes were
necessary to disrupt the face recognition system (De Renzi, Perani, Carlesimo, Silveri,
& Fazio, 1994; Meadows, 1974). Since then, in several imaging studies, the right FFA
consistently shows a stronger activation to faces than its counterpart on the left (e.g.
Kanwisher et al., 1997; Maurer et al., 2007). This lateralization may be present from
birth. In a PET study with six 2-month-old infants, Tzourio-Mazoyer et al. (2002)
found that even at this early age, face-processing preferably activated structures in
the inferior temporal gyrus on the right side. A recent study by Gonzalez, Anderson,
Wood, Mitchell, and Harvey (2007) about the lateralization of memory deficits in
children with temporal lobe epilepsy indicated a lateralization of memory deficits
only for faces, but for no other verbal or non-verbal memory tasks. A comprehensive
meta-analysis of Bouvier and Engel (2006) with 100 cases of acquired prosopagnosia
listed 32% unilateral right-sided and only 3% unilateral left tissue defects as a cause of
the disorder; 65% of the patients had bilateral damages. Together, these findings
indicate that the right lateralization of face processing is present after birth and does
not change with age. While both hemispheres can process faces and compensate for
damage to the homologous area to a certain degree, the right hemisphere is
dominant and a functional defect there may lead to a measurable face recognition
84Thomas Gru ¨ter et al.
Face recognition strategies
Though face recognition is obviously present in newborns already (Bushnell, Sai, &
Mullin, 1989), the acquisition of face expertise takes several years (Mondloch, Maurer, &
Ahola, 2006). Humans use different strategies to learn and recognize faces: they
recognize the facial features (feature based; Carbon & Leder, 2005), the configuration of
the features (configural; Leder & Carbon, 2006), and may recognize faces as a complete
unit (holistic; Leder & Carbon, 2005; Tanaka & Farah, 1991). The configural recognition
strategy works best for upright faces. It fails when faces are turned upside down, the
strategy is disrupted (Carbon & Leder, 2006). Peter Thompson (1980) first demonstrated
this dissociate face-processing effect with a picture of the former British Prime Minster
Margaret Thatcher. By selectively turning the eyes and the mouth region upside down, a
grotesque face emerges (Figure 3a and b). When Thatcherized faces are inverted (turned
upside down), the alteration is hardly detectable and the perception of grotesqueness
vanishes (Carbon, Grueter, Weber, & Lueschow, 2007; Carbon et al., 2005). Another
intriguing example of configural face processing was shown by Craig Mooney (1957).
‘Mooney faces’ are high-contrast photographs in which the face is formed of black
shadows and white surfaces concealing any local or featural detail. By inverting them,
they are hardly recognizable as faces at all (Figure 4). Valentine (1991) proposed that
faces might be stored as vectors in multidimensional space representing the difference
from a ‘prototypical’ face. This model would also explain the so-called other-race effect.
Humans recognize faces of people from other races with less confidence than those of
their own race. In face-space models, the faces of other races have similar vectors and
are difficult to tell apart. The prototypical face is generated from the average of faces in
memory. When more other race faces are stored, the standard face is adjusted and the
recognition deficit decreases (Sporer, 2001).
Figure 3. a) Does one of these pictures look bizarre? Due to disruption of configural face processing,
this is difficult to decide when faces are upside down (Carbon et al., 2005). b) When the faces are
shown in the normal upright position (please turn Figure 3 upside-down), the left one looks bizarre.
Face recognition and prospagnosia 85
Dysfunctions of face recognition
Any neural tissue damage in the face recognition network of the brain can cause a
prosopagnosia, a condition which can best be defined as a selective impairment of the
visual learning and recognition of faces. As of yet, there is no generally accepted
classification. A prosopagnosia caused by an accident or stroke in adulthood is mostly
called ‘acquired’ prosopagnosia or simply ‘prosopagnosia’. More confusing are the
classification attempts for the early onset prosopagnosia. Prosopagnosia can be
inherited or acquired by an early brain tissue damage. The term ‘developmental
prosopagnosia’ has been used for both conditions in the past (Barton et al., 2003; Kress
& Daum, 2003a). In a study of three patients who acquired prosopagnosia at an early
age, Barton et al. found that ‘developmental prosopagnosia is similar to the adult-onset
form in encoding deficits for the spatial arrangement of facial elements’. Therefore, it is
somewhat doubtful if a discrimination between prosopagnosia acquired before or after
adulthood makes sense. There is another type of prosopagnosia which has no defined
onset and is not accompanied by any visible brain damage (Behrmann & Avidan, 2005).
This type has also been called ‘developmental’, while we would prefer the attribute
‘congenital’ to show that it was not acquired at some defined point in life (cf. Behrmann
& Avidan, 2005). A detailed discussion of the classification problem can be found
elsewhere (Grueter et al., 2007). Hereditary prosopagnosia is a type of congenital
prosopagnosia which runs in families, as already documented by De Haan (1999). It is
quite common among the congenital cases (Kennerknecht et al., 2006).
For practical purposes, we would also like to define another type, the ‘symptomatic’
prosopagnosia which is part of other diseases like a pervasive developmental disorders
(PDD), psychiatric diseases, dementia or congenital impairments of vision (Elgar &
Campbell, 2001; Geldart, Mondloch, Maurer, de Schonen, & Brent, 2002; Serra et al.,
2003). This classification is not disjunctive and complete in a mathematical sense.
Figure 4. Mooney faces: While the left picture appears, more or less, as meaningless black and white
areas, the left one is a face. This also depicts the disruption of configural face processing at unusual
86 Thomas Gru ¨ter et al.
For example, a hypoxia at birth can cause a congenital and acquired prosopagnosia at
the same time. Still, the classification covers the vast majority of cases.
An acquired prosopagnosia should not necessarily be considered a disease, but as one of
several visually selective residual impairments after a brain tissue damage. It can be
defined as a dissociation of face and object recognition impairment to the disadvantage
of face processing. The individual range of impairments depends on the extent and the
location of the damage, causing a unique pattern of symptoms in each patient. An
acquired pure prosopagnosia is rare, only very few cases have been published
(Kleinschmidt & Cohen, 2006; Rossion et al., 2003). The widely differing patterns of
impairment make it difficult to identify the face-processing brain regions. In a
comprehensive meta-study of 100 cases of acquired prosopagnosia, Bouvier and Engel
(2006) found a high lesion overlap near the OFA while the FFA and the STS were less
often affected. Instead, they identified an area medial to the FFAwhichwas damaged in a
number of cases. They cautioned, though, that the image slices used to identify the
extent of the damage did not always cover the ventral cerebral surface with the FFA.
Therefore, a damage to the FFA may be more frequent than they could demonstrate.
Acquired prosopagnosia is quite rare, most neurologists have told us that they never
saw a single case during their professional career. De Renzi et al. stressed that ‘the
inability to recognize familiar faces is a rare disorder, not manifested by the majority of
patients with right temporo-occipital injury’. Hier, Mondlock, and Caplan (1983)
examined 41 patients within 7 days after a right-sided stroke and diagnosed
prosopagnosia in 19 of them, mostly accompanied by more dramatic symptoms like
visual field defect, hemineglect, anosognosia (denial of illness) or weakness of arm and
leg. Therefore, a temporary prosopagnosia after a right-sided stroke may well be hidden
by more apparent symptoms and not be recorded at all.
In some cases of acquired prosopagnosia, a residual autonomic (covert) response to
familiar faces has been reported (e.g. De Haan, Bauer, & Greve, 1992; Tranel & Damasio,
1985). Some authors therefore postulated a second pathway for the covert, unconscious
recognition of faces (De Haan et al., 1992). Farah, O’Reilly, and Vecera (1993) offered a
different explanation which they based on a neural network simulation of face
recognition. Stone and Valentine (2003) argued that emotional preference rather than
individual recognition may trigger an autonomic response and confound the study
results (see also Schweinberger & Burton, 2003).
Congenital and hereditary prosopagnosia
McConachie was the first to report a case of prosopagnosia without any history of brain
tissue damage (retested by De Haan & Campbell, 1991). Twenty years later, Ariel and
Sadeh (1996) published a second case. By 2002, less than a dozen cases of congenital
prosopagnosia were published worldwide, three of these had been found in one family
(De Haan, 1999). Other researchers also mentioned affected family members
(Behrmann, Avidan, Marotta, & Kimchi, 2005; Bentin, Deouell, & Soroker, 1999; De
Haan, Young, & Newcombe, 1991; Duchaine, 2000; Galaburda & Duchaine, 2003;
McConachie, 1976; Temple, 1992), but did not assess the formal genetics. From about
2002, prosopagnosia received more public interest. An internet mailing list for
prosopagnosics was started by affected persons and some web sites were dedicated to
Face recognition and prospagnosia87
the condition (e.g. www.faceblind.org, www.prosopagnosie.de). Duchaine and
Nakayama (2005) wrote that within 2 year after they set-up their web site, more than
175 self-diagnosed prosopagnosics contacted them. Our research group systematically
investigated the relatives of people with a congenital prosopagnosia and found seven
pedigrees with 38 cases of congenital prosopagnosia (Grueter, 2004; Grueter et al.,
2007). All pedigrees were in accordance with a simple autosomal dominant mode of
inheritance. A survey by Kennerknecht et al. (2006) found a prevalence of 2.5% in the
German population (17 out of 689) for congenital prosopagnosia. In 14 cases, first-grade
relatives were affected as well, three prosopagnosics did not want the researchers to
interview family members. This indicates that the hereditary type is quite frequent if not
prevailing among the cases of congenital prosopagnosia. It should be noted that
hereditary prosopagnosia is the first identified hereditary disorder to affect the central
visual cognition system. The familial segregation of hereditary prosopagnosia in a simple
autosomal dominant mode suggests that it may be caused by a single mutation in one or
more genes (point mutation). This would imply that at least within families, but
probably also across families, the underlying defect is the same. This is supported by the
quite uniform scope of clinical symptoms in hereditary prosopagnosia. To our
knowledge, the responsible gene has not been identified yet.
Though the affected people have a lifetime to compensate for their deficit, a careful
interview with a set of standard questions showed a set of typical symptoms (Table 1).
The table summarizes the results of semi-structured interviews with 54 people with
hereditary prosopagnosia (45 females, 9 males, mean age: 45.0, range 19–92) and 56
controls (35 females, 21 males, mean age: 47.4, range 16–78). The interviews lasted
between 1 and 2 hours. The interviewer asked open questions from a questionnaire
with at least three or four questions regarding each diagnostic item. Interviewers were
held to embed the questions into the conversation and to make sure that questions
about the same diagnostic item were not asked in a row. All interviews were done by
experienced physicians. This proved to be helpful because questions about face
recognition deficits can be embarrassing and most interviewed people played the
symptoms down. The diagnostic procedure is also described in Kennerknecht et al.
(2006). It was validated by face recognition tests in two samples with a total of 21
prosopagnosics (Carbon et al., 2007; Grueter & Grueter, 2007). The test results
confirmed the diagnosis of hereditary prosopagnosia in each case without exception.
The leading symptom of hereditary prosopagnosia is a lack of confidence about the
familiarity of faces. The affected people do not complain that all faces look unfamiliar,
but that they cannot determine the familiarity on a valid basis. Therefore, they not only
overlook familiar people, but also confuse strangers with familiar persons (T. Grueter &
Grueter, 2007). This symptom is found throughout and therefore should be regarded as
a diagnostic hallmark. People with hereditary prosopagnosia also take longer to
recognize familiar faces and to learn new faces. They cannot remember an onset of the
problem. Nearly, all of them find gaze contacts unnecessary for social communication –
as did 8 out of 56 controls. Eye-tracking experiments showed a peculiar scan pattern for
faces in hereditary prosopagnosia. The gaze was more dispersed and more directed to
external facial features (Schwarzer et al., 2007). The development of an adaptive
behavioural pattern (avoiding critical situations, a ready set of excuses, etc.) indicates a
long-standing perceptual deficit. Nearly, all hereditary prosopagnosics admitted
problems with the visual recognition of objects and scenes. This is consistent with
the observation by Behrmann et al. (2005) in five congenital prosopagnosics with
affected family members. They all had problems with the judgment of global
88Thomas Gru ¨ter et al.
configurations from simple visual stimuli. As configural processing is of foremost
importance in face processing (Behrmann et al., 2005; Carbon & Leder, 2005, 2006), a
dysfunction will especially impair face recognition. Not all studies, though, confirmed a
(general) configural processing problem in congenital prosopagnosia (Duchaine, 2000;
Le Grand et al., 2006). In the latter publication, the authors did not find a uniform
pattern of visual impairments to accompany the face recognition problem and stated
that congenital prosopagnosia is a heterogeneous condition. Of the eight prosopagno-
sics tested reported in these publications, only one is reported to have affected family
members. Therefore, some of them may suffer from congenital prosopagnosia other
than the hereditary type.
In hereditary prosopagnosia, there is no subjective or objective perceptual deficit in
the recognition of facial emotions (Duchaine, Parker, & Nakayama, 2003; Humphreys,
Avidan, & Behrmann, 2007). While in acquired prosopagnosia, a central achromatopsia
and a quadrantanopsia (loss of vision in one quarter of the visual field) is quite frequent
(Bouvier & Engel, 2006), it has not been observed in hereditary prosopagnosia. The
affected persons also reported that their judgments of attractiveness and gender are not
different from other people. They had no problems to recall semantic information about
persons and recognized other people easily from non-facial clues (Table 2).
The functional and anatomical peculiarities in congenital prosopagnosia remain
unclear to date. A single case study by Bentin et al. (1999) showed an unusual N170
wave in the EEG. Normally, the N170 is particularly face sensitive, but in the studied
person, it was also strongly elicited by objects. Kress and Daum (2003b) reported a
similar lack of N170 face specificity in two more prosopagnosic patients. While
Bentin et al.’s patient had an unusually small right temporal lobe, both patients
described by Kress and Daum (2003b) had no structural brain abnormalities. In an
fMRI study with four congenital prosopagnosics, Avidan et al. (2005) could not
demonstrate conclusive differences in the anatomy or in the FFA activation to faces.
A recent structural imaging study by Behrmann, Avidan, Gao, and Black (2007) with
six prosopagnosics found them to have a significantly smaller anterior fusiform gyrus,
Table 1. Most discriminative symptoms for the diagnosis of prosopagnosia
Lasting and irritating subjective uncertainty of
Face recognition deficit especially in crowded
places or out-of context encounters
False-negative and false-positive face recognition events
Face recognition time longer than socially accepted
Prolonged face learning time longer than socially accepted
Onset in childhood
Development of adaptive behaviour
No gaze contact necessary
Use of explicit learning strategies for
visual person recognition
Impaired visual recognition of objects and scenes
Face recognition and prospagnosia 89
Comparison of hereditary and acquired prosopagnosia types
The hereditary and acquired types have a completely different aetiology. The acquired
type is the consequence of a localized tissue damage, while the hereditary type is more
probably caused by a defective neural development. As most cases of acquired
prosopagnosia are caused by occipito-temporal tissue losses, adjacent anatomical
structures are quite often damaged as well, though they are not functionally linked to
face recognition. This explains the frequent coincidence of acquired central
achromatopsia and prosopagnosia. Also, parts of the lateral geniculate nucleus may
be damaged, causing a quadrant- or hemianopsia (Bouvier & Engel, 2006). While
emotion recognition is always spared in hereditary prosopagnosia, it is frequently
defective in acquired prosopagnosia (Sergent & Signoret, 1992; Takahashi, Kawamura,
Hirayama, Shiota, & Isono, 1995). Hereditary prosopagnosics show a frequent
impairment in the visual recognition of objects and scenes, indicating a general
impairment in visual recognition most prominent in face recognition. While acquired
prosopagnosia causes a loss of familiarity feeling (Spillmann, Laskowski, Lange, Kasper,
& Schmidt, 2000; Verstichel, 2005), people with hereditary prosopagnosia suffer from a
lasting lack of confidence (Kennerknecht et al., 2006). They cannot judge familiarity, all
faces seem vaguely familiar or vaguely unfamiliar. In her report on the prosopagnosic,
woman Dr S. Temple (1992, p.200) stated about their second meeting: ‘She failed to
recognize me... until I said her name. She would also report that as she knew I had
blonde hair, she had moved expectantly towards several other people before I arrived,
thinking them to be me’. Obviously, seeing faces did not trigger the familiarity decision
at all in Dr S. Temple. She cannot even reliably classify unknown faces as unfamiliar. This
may lead to false recognition of unknown people as familiar, and indeed did all
hereditary prosopagnosics report those events in our interviews. Their familiarity
feeling per se, though, is not impaired, it can reliably be triggered by familiar voices, for
example. False recognition is not common in acquired prosopagnosia. Rapcsak and his
colleagues (1996) studied a group of 21 patients with right hemispheric damage and
found no significant association between false recognition and prosopagnosia severity.
Nearly, all hereditary prosopagnosics told us that they do not need no gaze contact in
social interaction. Many added that they use gaze contact nevertheless because they
noticed that other people expect it. Still, the unusual gaze behaviour is among the most
striking symptoms for the interviewer. Many prosopagnosics remember having been
criticized by schoolteachers for not listening, because they did not look at the teachers.
We could not find any data about unusual gaze behaviour in acquired prosopagnosia,
Table 2. Cognitive symptoms usually not impaired in prosopagnosics (two prosopagnosics and one
control were not asked about facial attractiveness, one control was not asked about gender
Normal recognition of facial emotions or emotions in general
Unimpaired recognition of gender from faces
Normal colour recognition
Normal field of vision
Normal judgment of facial attractiveness
Normal semantic memory for persons
Recognition of persons from non-facial clues
90Thomas Gru ¨ter et al.
though. No paper mentioned a sudden change of gaze behaviour in acquired
prosopagnosia. As the symptom is quite obvious, we would assume that it would have
been mentioned somewhere.
Hereditary prosopagnosia is accompanied by awell-defined pattern of symptoms not
always related with ‘pure’ prosopagnosia. False recognition, a familiarity decision
deficit, no need for gaze contact in social interaction, and a general visual recognition
impairment define this condition, while other capabilities like emotion recognition are
not affected. Functionally independent but anatomically adjacent structures are never
affected, ruling out a blood vessel malformation or a birth trauma as a cause of the
condition (Table 3). Point mutations have been shown to cause pervasive functional
disorders of the brain (Lai, Fisher, Hurst, Vargha-Khadem, & Monaco, 2001). The unique
symptom pattern and the simple autosomal wayof inheritance in all reported hereditary
prosopagnosia cases suggest a functional brain development disorder caused by a point
mutation. This offers a twofold opportunity: firstly, finding the gene (or the genes)
would increase our knowledge about the genetics of brain development. Secondly, as
the underlying defect is the same at least within families and probably across families, a
comprehensive study with a large number of hereditary prosopagnosics promises new
insight into cerebral face processing.
Prosopagnosia in autism spectrum disorders (ASDs) and social
developmental disorders (SDDs)
Prosopagnosia has been suggested to be linked to ASD in two ways. ASD may cause
prosopagnosia because the lack of social interest may prevent the acquisition of facial
experience. In this case, the face recognition structures themselves would work at a
normal level, but not be employed sufficiently. On the other hand, a failure to extract
sufficient information from faces may interfere with normal social development. The
same could happen, if the brain cannot build up the normal facial attention preference
levels, because the corresponding structures are abnormal (Sasson, 2006). Schultz
(2005) suggested that a defective function of the fusiform face area and the amygdala
may represent a ‘core mechanism’ for the pathobiology of autism.
This view is highly controversial. Teunisse and de Gelder (2003) studied face
recognition abilities in a group of 17 high-ability autistic adults and found that they
formed a normal configuration-based face representation. People with a congenital
Table 3. Difference between acquired and hereditary prosopagnosia
SymptomsHereditary PA Acquired PA
Feeling of familiarityNot triggered
Lost, no feeling of familiarity
False-negative and false-positive
Emotional expression recognition
Gaze contact necessary
Impaired visual recognition
of objects and scenes
Face recognition and prospagnosia91
prosopagnosia are known to have a deficit there (Carbon et al., 2007). In a
comprehensive study, Barton et al. studied the face recognition performance of 24
people with SDD in comparison with neurotypical and prosopagnosic people. Their
results showed, as they wrote, that ‘SDD subgroup membership by face recognition did
not correlate with a particular SDD diagnosis or subjective ratings of social impairment’.
They concluded that SDD does not necessarily cause a face recognition impairment
(Barton et al., 2004). Hadjikhani and her colleagues (2004) compared the FFA activation
in normal and autistic adults. They assumed that the earlier described FFA activation
deficit in autistic persons (Schultz et al., 2000) might be caused by an attentional deficit
and used a fixation point to make sure that the faces received attention. With this
experimental setting, they found no significant differences in FFA activation to faces
between autistic and neurotypical study participants. It should also be noted that the
prevalence of congenital prosopagnosia (2.5%) is much higher than that of ASD (,0.5%;
Yeargin-Allsopp et al., 2003).
As of today, a causal dependency of SDD or ASD and face recognition abnormalities
cannot be considered as established. The aetiology and pathogenesis of ASD remain
elusive (for a review see Santangelo & Tsatsanis, 2005).
Face processing plays a decisive role in effective social interactions. As it has not been
designed or engineered, it is highly efficient, but not necessarily structured in a simple
or logical way. Current research results show that a complex interplay between different
highly face-responsive brain areas is involved in encoding, integration, and
interpretation of the huge amount of different information we draw from a face
(Gobbini & Haxby, 2007). Some structures in the occipital (OFA) and temporal lobes
(STS, FFA) havebeen found to be specialized on facerecognition. Tissuedamage in these
regions lead to a lasting face-processing deficit, called (acquired) prosopagnosia. In
contrast to many other visual abilities, the brain’s functional plasticity cannot
compensate for this deficit. While the acquired type is rarely observed, the congenital
type without any gross brain abnormalities is quite frequent. In this condition, deficits in
face processing are apparent from early childhood on while other sensory and
intellectual functions are preservedand macro-spatial brain abnormalities are discreet or
absent (Behrmann & Avidan, 2005). Only recently, a hereditary type of congenital
prosopagnosia (De Haan, 1999) was identified. The segregation is compatible with a
simple autosomal dominant way of inheritance suggesting a point mutation as a cause of
defect (M. Grueter, 2004). First studies indicate a very high prevalence (2.5%) of
hereditary prosopagnosia (Kennerknecht et al., 2006). The condition is characterized by
a great homogeneity of clinical symptoms, including deficits in learning and recognition
of faces, while other facial information is processed normally (e.g. expression and gaze).
Obligatory symptoms are, for instance, lasting and irritating subjective uncertainty of
face recognition, time-inefficient recognition strategies, and false-positive and –negative
face recognition events (see Table 1). In the light of such impairments, apologetic,
compensating, and preventing strategies are typically and intuitively developed by
people with prosopagnosia. The high prevalence of hereditary prosopagnosia
demonstrates its societal importance. Face research will have to analyse the
neurocognitive bases and develop better diagnostic criteria. This will create the pre-
condition for educational and training programs in a sensitized society. Face recognition
is a complex and highly integrated task employing large parts of the brain. A unitary and
92Thomas Gru ¨ter et al.
common impairment of face perception like the hereditary type of prosopagnosia opens
a big window of opportunity to improve our knowledge about face processing and its
This research was supported by a research grant of the City of Vienna (MA7) to CCC. We
appreciate the participants’ efforts in taking part in the study and the reviewers’ comments for
improving the manuscript.
Ariel, R., & Sadeh, M. (1996). Congenital visual agnosia and prosopagnosia ina child:A case report.
Cortex, 32(2), 221–240.
Avidan, G., Hasson, U., Malach, R., & Behrmann, M. (2005). Detailed exploration of face-related
processing in congenital prosopagnosia: 2. Functional neuroimaging findings. Journal of
Cognitive Neuroscience, 17(7), 1150–1167.
Aylward, E. H., Park, J. E., Field, K. M., Parsons, A. C., Richards, T. L., Cramer, S. C., et al. (2005).
Brain Activation during Face Perception: Evidence of a Developmental Change. Journal of
Cognitive Neuroscience, 17, 308–319.
Barton, J. J. S., Cherkasova, M. V., Hefter, R., Cox, T. A., O’Connor, M., & Manoach, D. S. (2004). Are
patients with social developmental disorders prosopagnosic? Perceptual heterogeneity in the
Asperger and socio-emotional processing disorders. Brain, 127, 1706–1716.
Barton, J. J. S., Cherkasova, M. V., Press, D. Z., Intriligator, J. M., & O’Connor, M. (2003).
Developmental prosopagnosia: A study of three patients. Brain and Cognition, 51(1), 12–30.
Behrmann, M., & Avidan, G. (2005). Congenital prosopagnosia: Face-blind from birth. Trends in
Cognitive Sciences, 9(4), 180–187.
Behrmann, M., Avidan, G., Gao, F., & Black, S. (2007). Structural imaging reveals anatomical
alterations in inferotemporal cortex in congenital prosopagnosia. Cerebral Cortex, 17,
Behrmann, M., Avidan, G., Marotta, J. J., & Kimchi, R. (2005). Detailed exploration of face-related
processing in congenital prosopagnosia: 1. Behavioral findings. Journal of Cognitive
Neuroscience, 17(7), 1–19.
Bentin, S., Deouell, L. Y., & Soroker, N. (1999). Selective visual streaming in face recognition:
Evidence from developmental prosopagnosia. Neuroreport, 10(4), 823–827.
Bodamer, J. (1947). Die Prosop-Agnosie. Archiv fu ¨r Psychiatrie und Nervenkrankheiten, 179,
Bouvier, S. E., & Engel, S. A. (2006). Behavioral deficits and cortical damage loci in cerebral
achromatopsia. Cerebral Cortex, 16(2), 183–191.
Bruce, V., & Young, A. (1986). Understanding face recognition. British Journal of Psychology,
Bukach, C. M., Gauthier, I., & Tarr, M. J. (2006). Beyond faces and modularity: The power of an
expertise framework. Trends in Cognitive Sciences, 10(4), 159–166.
Bushnell, I. W. R., Sai, F., & Mullin, J. T. (1989). Neonatal recognition of the mother’s face. British
Journal of Developmental Psychology, 7, 3–15.
Carbon, C. C., Grueter, T., Weber, J. E., & Lueschow, A. (2007). Faces as objects of non-expertise:
Processing of Thatcherised faces in congenital prosopagnosia. Perception, 36, 1635–1645.
Carbon, C. C., & Leder, H. (2005). When feature information comes first! Early processing of
inverted faces. Perception, 34(9), 1117–1134.
Carbon, C. C., & Leder, H. (2006). When faces are heads: View-dependent recognition of faces
altered relationally or componentially. Swiss Journal of Psychology, 65(4), 245–252.
Face recognition and prospagnosia93