Recognising Faces

Face Identification

Face identification is crucial for legal contexts, using methods like police photo-fits and identity parades. Understanding face recognition can help improve these techniques.

The Challenge of Face Recognition

Faces are hard to distinguish because they share similar parts in similar arrangements. Differences between instances of the same person's face can be greater than differences between different people's faces.

Human Expertise

Humans can recognize thousands of faces, averaging around 5000, indicating significant individual variation in facial recognition abilities.

Pareidolia

People are predisposed to seeing faces everywhere.

Face Inversion Effect

Yin (1969) found that faces are disproportionately harder to remember when upside down compared to other mono-oriented objects, indicating a special factor related to faces.

Thatcher Illusion

Shows that faces are susceptible to inversion, because when a face is upside down, the configural coding is compromised, and the features are emphasized.

Composite Face Effect

Faces are processed as a whole, not just by their individual parts.

Part-Whole Effect

Tanaka & Farah (1993) showed that recognizing parts of a face is easier when seen within the whole face.

Configural and Holistic Processing

Some theories emphasize:

  • Second-order spatial relations: Constant in all faces.

  • Holistic representations: Represent the whole face, including parts and spatial relations.

Neuropsychology Data

Faces behave differently from objects, which suggests a face-specific processing system. The study of patients with impaired face recognition but normal object recognition can help understand this system.

Prosopagnosia

Acquired prosopagnosia, often from brain damage, impairs face recognition while other object recognition remains normal. Some patients might have subtle impairments to object recognition, but some have claimed that there exist "pure prosopagnosic" patients. Patient F.B. (Riddoch et al., 2008) couldn't name famous faces or match faces across different views but was normal at naming objects.

Brain Imaging Studies

fMRI BOLD (Blood Oxygen Level Dependent) response shows brain activity in response to stimuli. Kanwisher et al. (1997) identified the Fusiform Face Area (FFA) in the ventral temporal lobe, primarily in the right hemisphere, which shows selective response to faces.

Bruce and Young Model (1986)

Functional model of face recognition involves:

  • Structural encoding: Computing visual properties useful for recognition.

  • View-centered descriptions: Specific views of a face.

  • Expression-independent descriptions: Capture critical aspects to differentiate people across views.

Identity and Expression

Separate pathways exist for processing identity (static cues) and expression (dynamic cues). Etcoff (1984) and Calder et al. (2000) showed independent composite effects for identity and emotion.

Haxby et al. (2000) Model

Two pathways:

  • Visual cortex via FFA for identity.

  • Visual cortex via superior temporal lobe for expression.

Stages of Face Identification

  1. Face recognition: Recognizing a face as familiar, and uses Face Recognition Units (FRU).

  2. Person identification: Associating semantic information with the person, and uses Person Identity Nodes (PIN).

  3. Name generation: Retrieving the person's name.

Temporal sequence:

  • Familiarity is fastest.

  • Occupation decision is next fastest.

  • Naming is slowest.

Priming and Face Processing

Priming studies show that previous exposure to stimuli affects performance. Repetition priming (Bruce & Valentine, 1985) shows that Face Recognition Units code faces across views but not names, which contributes to familiarity decisions.

Expertise Hypothesis

Gauthier & Tarr (1997) suggest faces are special due to our expertise in individuating them. Training people to be experts at individuating novel stimuli reveals