Lecture 4 - Face Recognition (not MCQ) Flashcards
Why study face perception?
Nothing communicates as efficiently as human face
Age, gender, mood, speech, gaze/attention, identity
What are the demands of facial recognition?
Generally same as object recognition, recognition in context, object invariance, specificity
Faces require more specificity (exemplar vs category level) + in most situations we need to recognise specific individual face rather than category
What did Ellis/Shepherd/Davies study?
Effectiveness of Photofit system
Found we have not very good way of forming representations of face, people have difficulty reproducing likenesses of even familiar faces
Photofit has inherent belief we process local features individually (do we?)
How do we represent faces?
Configurally, can mean:
Spatial relationship between features important as features themselves, face features interact w/ one another, faces processed holistically
What did Tanaka/Farah study?
Features vs configurations
if parts of faces represented separately memory for parts presented in isolation should be as good as when presented w/ face, participants learned names to go w/ face, asked to pick out Larry from 2 alts
When learn from normal face –> better in learning context
When learn from scrambled face –> better from feature alone
Representation of whole faces based at least partly in holistic
What did Young study?
Upright/Inverted faces, asked to name person on bottom/top half w/ either aligned/misaligned halves, RT measured
People slower to name aligned halves –> perception of novel facial configuration interfered w/ identification of constituent parts
No difference found when face was upside down, interference only for upright
What does the Thatcher illusion tell us?
Lack of configural processing for inverted faces
Importance of surface properties
Why are surface properties important?
Play a role in disambiguating similar looking objects – orange & grapefruit…
Face recognition relies on more specific processing mechanisms & involves different brain regions
Face representations seem to preserve surface characteristics – pigmentation
What are the 2 effects of negation?
2 sources of info adversely affected:
1. Pigmentation: skin/hair colour
2. Pattern of shading/shadow which specify 3D structure of face (line drawings diff to recog unless info about pigmentation/shading preserved in some way)
How does distinctiveness/caricature affect recognition?
Not all faces equally easy to recognise, distinctive recognised more accurately/quickly than those rated as typical
What did Valentine/Bruce find?
Showed pictures of famous/unfamiliar faces + asked to rate how well it would stand out in a crowd
Easier to classify face as a face when typical
Explained using ‘face space’ model by Valentine (distinctive faces represented in sparsely populated regions of brain)
How do we recognise faces based on Bruce/Young?
Structural encoding (expression analysis/facial speech analysis/direct visual processing) –> face recognition units (directed visual processing/cognitive system) –> person identity nodes (cognitive system) –> name generation (cognitive system)
Form structural description –> matched to stored representations (Face Recognition Units FRUs) –> access semantic info (Person Identity Nodes PINs) –> recall person’s name
What is evidence for Bruce/Young’s model?
Efficacy of diff cues in helping resolve errors/difficulties of face recognition
Naturally occurring/induced errors (failed to recognise, misidentified, etc.)
Latencies of diff decisions made to faces (familiarity decision faster than semantic/naming decisions)
Neuropsychological patients w/ diff patterns of face recognition difficulty
Associative prosopagnosia (FRUs)
Semantic impairment (PINs)
Names (anomia)
What does Bruce/Young’s model suggest about independent processing?
Expressions and visual speech processed independently and in parallel to identity
What are problems for Bruce/Young’s theory?
Independence: found RTs for identity judgments independent of variation in expression/facial speech but RTs for expression/facial speech influenced by variation in identity
Covert recognition: some prosopagnosic patients w/ no overt recognition showed covert responses
