Face recognition and visual agnosias
Bruce and Young’s theory of face
recognition, including case studies and
explanations of prosopagnosia
Face recognition
• Face recognition is a very important aspect of social functioning. Most faces are
broadly similar, in spite of this we are very good at distinguishing between
different faces. Even if the face is unfamiliar we make certain judgements e.g.
mood, general age and gender. Bruce took photos of a large number of men and
women - to avoid obvious cues of gender, men were closely shaved, women
had no make- up and they all wore swimming caps to conceal their hair.
Participants were 96% accurate.
• We are better at recognising faces we are familiar with – Bahrick gave lecturers
set of photos and asked them to pick the students they had taught from those
they had never taught. Recognition was good immediately after the end of term
but dropped significantly after a year and only chance after 8 years. Suggesting
exposure to faces needs to be maintained. One reason we are good at
recognising familiar faces is that we have the opportunity to view them from
different angles and so build up a more complete mental picture.
Is face recognition special?
• Yes
Infant preferences for faces – Fantz –babies as young as 4 days old showed
a preference for a schematic face rather than jumbled up or blocked. If
face preference is innate it would make sense as it would be adaptive – a
newborn who recognises and responds to its own species will better elicit
attachment and caring. But the results from study could be because infants
prefer symmetry and interesting pictures. See also studies of people with
prosopagnosia (after this). Would be adaptive e.g. recognising an enemy,
your own child, decoding facial expressions. MRI scans of brain activity
show that the fusiform gyrus became more active when subjects looking at
faces than when looking at other objects. Suggesting this area of the brain
is specialised for processing faces.
Is face recognition special?
• No
Faces are quite similar so high level cognitive processing is needed to
differentiate them. Gauthier – used MRI to record brain activity of people who
were shown pictures of birds and cars and were asked to identify type of bird/car.
Found fusiform area also active during this task, suggesting area not dedicated to
faces. Experts in bird/car recognition used this area to identify different
categories, we are all experts at recognising faces. So suggests fusiform area is
specialised for the recognition of any object category for which we possess
expertise. Other support – we find it harder to recognise faces of people from
other races – probably due to our lack of experience/expertise. If fusiform area is
specialised for expert processing then people e.g. who are car experts should find
it difficult to carry out a face recognition task and a car recognition task at the
same time. People who are not car experts should be able to do both at same
time as only face recognition task would need fusiform activity. Gaithier found
that this was exactly what happened.
THEORIES OF FACE RECOGNITION
How do you recognise people?
• Do we use feature analysis. Comparing each feature of the face with a
stored list of features. Shepherd, Davies and Ellis briefly showed
participants unfamiliar faces and later asked them to describe the faces
seen. Features most frequently recalled in order were, hair, eyes, nose,
mouth, eyebrows, chin and forehead. Few mentioned shape of face or
expression. But there are problems with a simple feature-based theory. Yin
– inversion effect, could recognise pictures of objects turned upside down
as easily as right way up, but not faces. Suggestion is that upside down
impairs holistic processing but not feature analysis, so face must be
processed in a holistic way. Bruce and Valentine – scrambled faces of
celebrities, scrambled faces harder to identify than normally configured
faces – again suggesting we find it easier to process faces holistically – topdown theory.
Orientation is important
• Young et al. (1987) paired different top and bottom halves of faces.
• They found that recognition of top-halves was easier when faces were
inverted. Where faces were upright performance was better when
the new lower-half was omitted.
• The joined-up upright face led to a ‘new’ configuration which
interfered with the detection of individual halves.
7
The ‘Thatcher Illusion’
(Thomson, 1980)
8
The ‘Thatcher Illusion’
(Thomson, 1980)
9
Why does the ‘Thatcher illusion’ occur?
• Bartlett and Searcy (1993) conducted experiments to
measure face ‘grotesqueness’.
• Their results supported the “configural processing
hypothesis”
• i.e. We have a difficulty in understanding the configuration
of features when faces are inverted.
• We aren’t aware of the odd configuration of elements
within the inverted Thatcher image.
10
Does the inversion effect suggest that
face recognition is special?
• Diamond and Carey (1986) tested recognition for faces and dogs.
• They found that dog judges and breeders were relatively impaired for
inverted faces compared to ‘normal’ individuals.
• This suggests that frequent exposure results in the inversion effect. i.e.
Configuration becomes important through practice?
11
Bruce and Young
• Bruce and Young’s (1986) model is the most influential theoretical
approach to face recognition. The model has eight components:
structural encoding, expression analysis, facial speech analysis,
directed visual processing, face recognition nodes, person identity
nodes, name generation and a cognitive system.
The model predicts that:
 familiar and unfamiliar faces are processed differently;
 facial identity and facial expression are processed separately;
 for a familiar face, familiarity information is accessed before
information about identity or name.
Capgras Syndrome
• Arthur was a young man who sustained a terrible head injury in a car
crash.
• Soon after he claimed that his mother and father had been replaced
by duplicates who looked exactly like his parents.
• He also had problems recognising himself in pictures and claimed
they were ‘another Arthur’.
Face Recognition
• Face recognition is a special case of object recognition.
• Unlike objects though faces move and make meaningful noises.
• We can also watch their lips as they are speaking to help us
understand what they are saying.
Bruce and Young’s Face Recognition Model
Structural Encoding
Processing a basic pattern of the
face
View centred description
Expression-independent
descriptions
Face recognition units
(FRU)
Person identity nodes
(PINS)
Name generation
The image of the face is processed,
including the person’s expression
and watching the movements of
their mouth as they speak (facial
speech analysis).
Features of the face that do not
change with people’s expressions
are processed, e.g. size/shape of
nose.
Bruce and Young’s Face Recognition Model
Structural Encoding
View centred description
These store information about the
structure of familiar faces. This is
where we turn the 2D image we
see into a 3D image.
Expression-independent
descriptions
Face recognition units
(FRU)
These store biographical
information about people, their
favourite food, music films, where
you know them from etc.
Person identity nodes
(PINS)
Name generation
Now we can name the person.
Bruce and Young’s Face Recognition Model
Structural Encoding
Expression analysis
View centred description
Facial speech
analysis
Directed visual
processing
Expression-independent
descriptions
Face recognition units
(FRU)
Person identity nodes
(PINS)
Cognitive System
Name generation
Face Recognition
• After structural encoding different types of information from the face
are extracted in parallel.
• Familiar faces are processed separately from unfamiliar faces.
Face Recognition Model
• Expression analysis looking at peoples expression to see what
emotions they are feeling.
• Spot the fake smile
Face Recognition Model
• Facial speech analysis ‘lip-reading’, watching lip movements to help
understand what people are saying.
• McGurk Effect
Face Recognition Model
• Directed visual processing allows visual processing of unfamiliar faces
e.g. looking for features such as if they have a moustache.
Face Recognition Model
holds additional information that might help us
recognise people. It also helps us decide what parts of the face
recognition system we need to attend to.
Bruce and Young’s Face Recognition Model
Structural Encoding
Expression analysis
View centred description
Facial speech
analysis
Directed visual
processing
Expression-independent
descriptions
Face recognition units
(FRU)
Person identity nodes
(PINS)
Cognitive System
Name generation
Face Recognition
• Celebrity blindness article and test (test is about half way down the
page)
• Thatcher test
Research
• Malone et al. (1982) provided evidence for a double dissociation
between recognition of familiar and unfamiliar faces. Much research
supports the assumption that there are separate routes for
processing facial identity and expression.
• Fox et al. (2011) found patients with damage to the face-recognition
network had impaired identity perception but not expression
perception. Patients with impaired recognition of facial expression
may have other emotional impairments. Different brain regions may
be involved in processing facial expressions and identity. Processing
facial identity is associated with the fusiform face area, while
processing expressions activates the superior temporal sulcus.
• MRI
• https://www.youtube.com/watch?v=12GqKmRA3q8
research
Young et al. (1985) found people decided more quickly on whether a
face was familiar than on the identity of the face. Young et al. (1985)
found evidence that the name of a person cannot be accessed without
also having other information about the person available. However,
Brédart et al. (2005) found evidence contrary to the model, which
showed that speed of recall for names was faster than that for personal
information when the faces were those of personal friends.
Evaluation
• Bruce and Young’s model is deservedly influential and many of its
predictions have received empirical support. However, the model also
possesses limitations, mostly because it is oversimplified. Also, the
assumption that facial identity and expression involve separate
processing routes may be too extreme. The assumption that name
processing always occurs after processing of other personal
information may also be too rigid.
Face recognition and visual agnosias
Face recognition is a specialised form of pattern recognition
Face recognition is an important area of research because it has a number of practical applications
We are generally better at dealing with familiar faces rather than unfamiliar faces
There is some debate as to whether face recognition is a highly specialised mechanism or whether it is simply one type of
complex processing
Computer models of face recognition can operate either feature by feature or holistically. It is generally agreed that human
face recognition is largely a holistic process
Research using scrambled and inverted faces suggests that human face recognition depends largely on holistic processing
One of the most influential models of face recognition was proposed by Bruce and Young. This model contains a number of
independent modules working in parallel. So does this make it an holistic or constructivist/hierarchical theory.
The model has been influential but only offers a limited explanation of face recognition
Prosopagnosia is a relatively rare disorder in which individuals lose the ability to recognise familiar faces
A number of case studies have been reported which provide clues about the face recognition process in normal humans
Prosopagnosia research lends support to the Bruce and Young model
Bruce and Young’s theory of face recognition
• Discuss Bruce and Young’s theory of face recognition.
• (4 marks + 8 marks)
AO1
• Given the marks available it is not necessary for candidates to outline all
components of the Bruce & Young model for full marks. More important is
that they demonstrate an understanding of general features of the model,
and supplement this with reference to subcomponents. For instance, an
outline that includes reference to a stage model, and to some stages being
sequential and some working in parallel, combined with a brief outline of
e.g. expression analysis and face recognition units, would earn 4 marks. A
listing of two or more subcomponents without reference to overall features
of the model would receive a maximum of 2 marks. Candidates need not
refer to the updated version of the model published recently.
• Diagrams can be an effective way of presenting the model, but there must
be some indication of process as well as key components for marks above
Basic, e.g. through use of directional arrows.
• Examiners should be sensitive to the time constraints of this question part.
AO2/AO3
• Commentary/evaluation of the Bruce and Young model should focus on the
wealth of research evidence. There are many studies on both neurotypicals
and on case studies of brain-damaged patients (e.g. prosopagnosics) that
provide both support and some contradictory findings relevant to the
model. The key to marks in the higher bands will be the extent to which
candidates link findings to specific aspects of the model e.g. whether they
provide support for specific modules within the model, such as the
distinction between face recognition units and name generation.
Alternative interpretations of data would also be an effective source of
AO2/3 material. Methodological evaluation of studies is likely to be
popular, but may only earn marks if implications for the theory are explicit
i.e. commentary on ethical issues is unlikely to be creditworthy, while
references to the limitations of case studies may well earn AO2/3 marks.
• Indicative issues, debates and approaches in the context of the Bruce &
Young theory of face recognition include reductionism and holism – the
model breaks face recognition down into several subcomponents and
strategies, as opposed to approaches that emphasise the holistic aspects of
face recognition. Other IDA in this area might include the applications of
findings e.g. to interpreting the effects of brain damage.
• AO2/3 Mark bands – Best fit
• AO2/3 material should first be placed in the appropriate band according to
the descriptors. However, not all the criteria need be satisfied for an
answer to be placed in a particular band. Weak performance in one area
may be compensated for by strong performance in others. In order to
access the top band, issues and debates and/or approaches need to be
addressed effectively.
Visual agnosias
• Visual agnosia is the inability to recognise familiar objects presented visually.
Most neuropsychologists distinguish between “apperceptive agnosia” and
“associative agnosia”.
• “Associative agnosia” is when perceptual ability is intact but there is a failure of
recognition because of difficulty in accessing the relevant knowledge from
memory – a “normal percept stripped of its meaning” (Teuber 1968). An example
of this would be having the ability to copy a line drawing of an object accurately,
but not being able to name the object. See Rubens and Benson (1971). i.e. they
cannot access the information in their long-term memory.
• “Apperceptive agnosia” is a perceptual deficit, i.e. a failure of recognition due to
impaired visual perception. An example of this would be a failure to distinguish
between a square and a circle.
How to recognise which one
• One way to assess the type of agnosia the patient has is to test the
patient’s ability to copy objects that can’t be recognised (Humphreys,
1999). Patients who can copy objects are said to have associative
agnosia, and those who can’t have apperceptive agnosia.
Apperceptive Agnosia
• Patients with apperceptive agnosia do possess some perceptual
abilities. For example, they typically have normal visual acuity and can
reach for moving targets. However, they have deficient visual
processes because they are very poor at recognising or identifying
objects.
Research evidence
• Grossman, Galetta, and D’Esposito (1997) reviewed previous research
on apperceptive agnosia. One patient could only describe a circle as
“lots of dots”, and another couldn’t discriminate between an “X” and
an “O”.
• Grossman et al. (1997) studied two patients, SZ, a 54-year-old male
accountant, and AP, a 65-year-old female teacher. Both of these
patients could recognise regular geometric shapes and colours in
spite of their severely impaired ability to recognise common objects.
However, they were very poor at recognising more complex shapes or
when geometric shapes were presented upside down.
Associative Agnosia
• Patients with associative agnosia have reasonably good basic
perceptual processes but are poor at accessing their stored semantic
knowledge about objects because it is hard for them to access in
long-term memory.
Research
Anaki et al. (2007) described their patient (DBO) who had experienced
brain damage to the left occipital lobe of the brain (at the back). He
had severe visual agnosia as shown by his ability to name only 1 out of
20 common objects. In spite of his visual agnosia, DBO performed well
on various tasks involving aspects of object recognition. The fact that
he still possessed several visual perceptual skills indicated that he had
associative rather than apperceptive agnosia.
Evaluation
• Empirical evidence supports variations. The empirical research provides convincing
evidence that the precise nature of visual agnosia varies considerably from one patient
to another.
• Individual differences. The distinction between apperceptive and associative agnosia
accounts for some individual differences. However, it is oversimplified to try to categorise
into just apperceptive or associative agnosia. For example, it is not easy to categorise
integrative agnosics as having either apperceptive or associative agnosia
• Lack a clear understanding of the cause. It is generally assumed that the main problem
in associative agnosia lies in accessing the stored information about objects that they
possess. However, it is hard to obtain clear evidence that that is in fact the main
problem.
• The distinction lacks relevance to specific deficits. The distinction between apperceptive
and associative agnosia is more relevant to general deficits in object recognition than
specific deficits. For example, prosopagnosics are generally reasonably good at object
recognition for most objects other than faces.
Prosopagnosia
PROSOPAGNOSIA
This is the inability to recognise faces, even ones that are very familiar
such as those of friends and family. Thus, this condition is also know as
“face blindness” and is usually caused by brain damage. Most
prosopagnosics can recognise objects reasonably well but do have
some problems with object recognition.
• Prosopagnosia is an example of an associative agnosia. There are
different types and levels of prosopagnosia, and this suggests that
face recognition occurs in stages - supporting Bruce and Young's
theory. It also suggests that different stages of face recognition
happens in different parts of the brain, and it is damage to these
specific parts that causes the different types and levels of
prosopagnosia.
Case studies of prosopagnosia are useful because they help to
understand how face recognition occurs.
There are 2 different explanations of prosopagnosia -
There are 2 different explanations of
prosopagnosia 1. It's a unique face specific problem.
- Farah et al - brain scans of people doing number of cognitive tasks - FFA (fusiform
gyrus) activates during face recognition but less so during object recognition .
- Barton et al - people with prosopagnosia had damaged FFA - suggests specific
processing just for face recognition - supports B+Y.
- Farah (again!) - faces are special - case study - LH - recognition test - discriminate
between objects but not faces
2. Not just faces!
- Gauthier et al - some people with prosopagnosia had problems that extended
beyond face recognition - inferior to recognising complex objects rather than faces
- expertise - car lovers FFA activated when seen cars, bird watchers birds, suggest
FFA activated not just for faces but for things people are interested in.
• Moscovitch, Winocur, and Behrmann (1997) studied CK, a man with
impaired object recognition. He performed as well as controls on
face-recognition tasks regardless of whether the face was a
photograph, a caricature, or a cartoon provided it was upright and the
internal features were in the correct locations.
• Evidence suggests that the fusiform face area (especially the one in the
right hemisphere) is specialised for processing faces. Prosopagnosics
typically have damage to this brain region (Farah, Tanaka, & Drain, 1995).
• Further support is provided by Downing et al. (2006) who presented
participants with faces, scenes, and 18 object categories (e.g. tools, fruits,
vegetables). The fusiform face area responded significantly more strongly
to faces than to any of the other 19 stimulus categories.
• However, Grill-Spector, Sayres, and Ress (2006) suggest the fusiform
specialisation may not be so pronounced. In this study observers saw faces
and three categories of objects (animals, cars, and abstract sculptures).
More elements in the fusiform face area responded to faces than to any of
the other types of objects, but the differences were not very large.
• Gauthier and Tarr (2002) suggest there are two reasons why faces
appear special even though they aren’t.
• First, we typically recognise faces at the individual level, whereas we often
recognise objects at a more general level. Recognising specific examples of a
category (i.e. faces) is harder than recognising the general category to which
an object belongs.
• Second, nearly all of us have considerably more experience (and thus
expertise) in recognising faces than in recognising individual members of most
other categories.
• Thus, Gauthier and Tarr concluded that the fusiform face area is a brain area
we use for any objects for which we possess expertise not just for processing
faces.
EVALUATION OF EVIDENCE INTO
PROSOPAGNOSIA
Most people do possess much more expertise about faces than almost
any other object category as predicted by the expertise theory. We
have more experience of identifying individual faces than individual
members of most other categories. It thus seems possible that
differences between face and object recognition might be attributable
to differences in expertise.
The greatest limitation is that none of the specific hypotheses of the
expertise theory has been supported. The fusiform area is not
consistently more active when recognising specific objects in the way
that it is when recognising faces. It seems then that the processes
involved in face recognition differ substantially from those involved in
object recognition.
Gestalt!
Gestalt, a German word for form or shape, may refer to: Holism, the
idea that natural systems and their properties should be viewed as
wholes, not as collections of parts.