FACE RECOGNITION'

1
FACE RECOGNITION.
2
Why should we study it?
It is different from other processes of object
recognition.
Communication / Socialisation.
Evolution- Babies prefer faces to other visual
stimuli (Johnson Morton, 1991).
Attractiveness.
Some people lack this ability. Tanaka (2003).
Children with Autism.
Practical applications Eyewitness testimony.
3
What do we know about it?
Young, Hay Ellis (1985). 4 types of face
recognition errors Diary Studies Asked 22
subjects to make diary notes of all the errors
they made.

• Failure to recognise someone familiar.

• Misidentification.

• Familiar, but not sure why.

• Familiar, but no recall for name.

• Kaszniak (2004). Feeling of knowing.
• Medial prefrontal cortex shows activity during
  this state.
• But not active when participants either know
  or do not know a face.

4
Seems to be something difficult about names
Young et al (1985). Separate storage for
names. Process involves 3 different systems.
Face Recognition System. stores
representations. Semantic System all other
information about person. Lexical system
vocabulary, including names.
5
Flude (1989). Hardly ever possible to name a face
without knowing something else about the person.
Decisions about familiarity are faster then
decisions about person-knowledge. - which
in turn are faster then producing the name.
(Young 1986).
6
Early Theories.
1.Template Theories. Pattern recognition
approach. Stimulus face is recognised if it
matches a template we have stored.
But - Profiles?
Distance?
Cognitive economy!!
7
Feature Theories
Recognition is achieved if enough features of the
stimulus face match those we have stored about a
person.
Distinctiveness. Light (1979). Prominent
features recognised more. Rhodes et al (1987)
Caricatures rated better likenesses than photos.
8
Configurational v Component Information.
Young, Hellawell Hay, (1986). Component
Information about eyes, nose, mouth, etc.
Configurational Overall arrangement of
features and shape of face.

• Presented participants with pictures of famous
  faces.
• Faces comprised of 2 different halves.
• Close together v apart.

Found Closely aligned faces hard to
recognise. Much easier to recognise when moved
apart. Suggests evidence for a face recognition
unit. Also holistic rather than feature
analysis/processing.
9
Configurational v Component Information.

• Evidence Searcy Bartlett (1996).
• Present subjects with faces 2 conditions.
• Condition 1. Configurational. Move eyes up,
  apart, elongate chin.
• Condition 2 .Component. Blacken teeth, discolour
  eyes, etc.
• Half of the faces in each condition were
  inverted (upside down).

Found - Component distortions quickly detected
in upright AND inverted faces. Configurational
often not detected in inverted faces.

• So -
• Both used in processing of normal (upright)
  faces.
• Inverted rely more on components (no
  configuration to work with.)

Much more complicated than early theories
suggested.
10
Inversion Effect. Perception of a face disappears
when inverted. (Moscovitch 1997).
11
Thompson, (1980) The Thatcher Illusion. Invert
components of the face (eyes mouth). Strong
illusion of the grotesque when upright. Effect
disappears when whole face inverted.
12
But is that all there is to it?
Real Life. Bruce Valentine (1988) 2D versus
3D. Evidence from experimental studies.
Contextual information faces not identified
in a vacuum!
Faces often also identified by expression
emotional states.

• Laurence et al (2003). Turner Syndrome.
• Female patients show impairments in face
  recognition.
• Impaired recognition of emotions particularly
  fear.

13
Prosopagnosia.
Ability to recognise faces is acutely impaired.
Could be genetic, but usually result of head
trauma. No recognition of familiar faces, but
often can identify person on basis of other cues,
e.g. voice, gait, posture, etc. Object
recognition intact.
Capgras Delusion. Ellis Lewis (2001). A
reverse condition? Significant others replaced
by imposters/aliens. Patients may also
recognise themselves as their own
double. Overt recognition intact, but covert
not. Debate structural or psychiatric?
Prosopagnosia overt route interrupted Capgras
Delusion covert route interrupted
14
Face-inversion effect in prosopagnosics.

• Farah (1990, 1994).
• Patient LH given inverted and upright faces.
• Unimpaired present difficulties recognising
  inverted faces.
• Patient showed opposite effect.

Unimpaired cant use holistic/configurational
processing on inverted faces. Prosopagnosics
dont have this ability at all so not
constrained to relying on it.
More support for argument that faces are
processed holistically - (configurational)
rather than analysis-by parts (component)?
But more recently does some processing remain
intact? Covert recognition patient is
consciously unaware. Whats the evidence?
15
Covert Recognition.
Baurer (1988) Implicit knowledge. Patient asked
to match famous faces with appropriate names. No
explicit knowledge, but greater electrodermal
response when names and faces matched.
DeHaan, Young Newcomb (1987) Patient
P.H.. Given random set of faces asked to
classify into politician non-politician. Cat
egorisation times much longer when face was NOT
politician.
Sergent Poncet (1990). Semantic Information can
prompt. But patient has to actively
spontaneously make use of it.
16
Not all prosopagnosics show covert
recognition. Kohler Moscovitch (1997). Occurs
when there is a deficit at perceptual
level. Patient impaired at primary stage of
analysing incoming information.
17
Theories Models of Face Recognition

• Farah (1990, 1994) Two-Process Theory.
• Holistic mainly when recognising faces and
  objects.
• Analysis-by-parts mainly recognition of words,
  letters etc

Completely different types of stimuli. Words/text
needs to be broken down and built up again in
order to identify. But not faces?
How does this all happen?
18
Bruce Youngs Model (1986)
Unfamiliar faces
Familiar faces
View centred descriptions
Expression analysis
Structural encoding.
Facial Speech analysis
Expression independent descriptions
Directed visual processing
Face recognition units

• More comprehensive 8-stage model.
• Explains process, and tries to account for
  recognition errors unfamiliar faces.

Person identity nodes
Cognitive system
Name generation
19
Bruce Youngs Model (1986)
Unfamiliar faces
Familiar faces
View centred descriptions
Expression analysis
Structural encoding.
Facial Speech analysis
Expression independent descriptions
Directed visual processing
Face recognition units

• More comprehensive 8-stage model.
• Explains process, and tries to account for
  recognition errors unfamiliar faces.

Person identity nodes
Cognitive system
Name generation
20
Evaluation of Bruce Youngs (1986) model.
Accounts for differences in familiar and
unfamiliar faces
Explains why names cannot be retrieved without
any other semantic information.
Explains feelings of familiarity we have without
knowing anything else.

• But -
• Little detail about processing of unfamiliar
  faces.
• Cognitive system - poorly specified.
• Doesnt explain covert recognition in
  Prosopagnosia.
• Doesnt explain (fairly exceptional) cases of
  names without any other biographical
  information.

21
3. Burton Bruce (1993) Interactive Activation
Competition (IAC) model.
Based on connectionist networks. Activation does
not necessarily occur in identical sequences.

• Has mainly 3 separate pools of information.
• Face Recognition units (FRUs).
• Person Identity nodes (PINs)
• Semantic information.

Crucial difference is that there is no link from
Semantic system to FRUs. Recognition will occur
if enough activation is achieved at the PIN
level. Accounts for findings that recognition
can occur without activation of FRUs - if
enough activation is received from Semantic
system.
22
Burton Bruce (1993). I.A.C. model (simplified).
Face Recognition Units
PINs can be activated by verbal or visual
information!
Tony Blair
Charles
Diana
Cher
Person Identity Nodes
Semantic Information
Charles
Tony Blair
Royals
Prime Minister
Diana
Cher
Singer
..see also Schweinberger Burton (2003).
23
Support for the model.
1. Bruce Valentine (1986). Priming
Effects. e.g. (Queen Prince William) v
(Queen Mick Jagger).
2. De Haan, Young Newcomb (1991). Names v
Autobiographical Information (PINs). Model
does not store these separately - both in
Semantic system. Explains how amnesic patient
could match 88 of faces with names, without
knowing anything else about them.
3. Covert Recognition in Prosopagnosics. Explain
s how patients can access information without
having to access FRUs.
24
Applications of Face Recognition
Eyewitness Testimony who can identify the
accused?
Levin (2000). Cross-Race (CR) v Same Race (SR).
Feature processing becomes second to basic
visual processing.
Selke Vynda (1985). Age Gender. M/Fs
between age 7 -16 years investigated.

• 7 and 8-year-old children had a significantly
  higher error score than the other age groups.
• Boys reached adult performance levels at 9-10
  years of age.
• Below age 15 - 16 years male subjects show
  significantly better face recognition ability
  than female subjects.
• in contrast to the repeatedly found adult
  female superiority in face recognition (p33).

Useful website of face recognition studies.
http//www.utdallas.edu/otoole/face_try.html