6. Object and Face perception Flashcards
(22 cards)
What challenge does object recognition face in vision?
Vast variability in retinal images due to changes in viewpoint, lighting, size, and occlusion
Requires invariant representation of object identity
What is the role of the ventral visual stream in object perception?
Processes “what” information: object identity and form
Projects from V1 → V2 → V4 → lateral occipital complex (LOC) → inferior temporal cortex (IT)
How do neurons in V4 contribute to shape representation?
Selectively tuned to combinations of oriented edges and curvature
Provide intermediate complexity features for object models
What deficits arise from damage to LOC/IT regions?
Apperceptive agnosia: impaired perceptual grouping and shape discrimination
Associative agnosia: intact perception but inability to assign meaning or name objects
How do template and feature-matching models propose object recognition works?
Template matching: compare retinal image to stored whole-object templates
Feature matching: detect diagnostic features (e.g., corners, edges) and assemble identity
What are the key principles of Biederman’s RBC model?
Objects decomposed into volumetric primitives (“geons”)
Recognition via identifying geon types and their spatial relations
Tags: Structural Description, Biederman 1987
Study Reference:
Authors & Year: Biederman (1987)
Method: Behavioral tests with geon-based stimuli under viewpoint changes
Key Finding: Recognition robust when geon structure preserved
What behavioral findings support structural-description models?
Deletion of concave regions (geon joints) severely impairs recognition
Deletion of contour fragments (non-geon parts) has less effect
How do view-dependent models explain recognition across viewpoints?
Store multiple 2D snapshots of objects
Recognise novel views via mental interpolation or rotation of stored views
What did Gauthier & Tarr show with Greeble training?
Training on individuating novel “Greebles” induces inversion effects and FFA activation akin to faces
Tags: Expertise, Gauthier & Tarr 1997
Study Reference:
Authors & Year: Gauthier & Tarr (1997)
Method: Greeble individuation training; measured inversion effect and fMRI
Key Finding: Expertise can recruit face-like processing mechanisms
What evidence indicates a specialized face-recognition system?
Strong inversion effect for faces but weaker for other objects
Selective fusiform face area (FFA) activation to faces vs. objects
How do IT neurons respond to faces?
Clusters of face-selective cells in monkey IT (“face patches”)
Exhibit view-invariant tuning to individual identities
What characterizes the human FFA?
Right-lateralized fusiform gyrus region with higher BOLD responses to faces
Lesions cause prosopagnosia; stimulation produces face distortions
What distinguishes acquired and developmental prosopagnosia?
Acquired: Following brain injury (often right fusiform) → lifelong inability to recognize familiar faces
Developmental: No obvious lesion; selective face-recognition deficit from early life
How can we dissociate innate face mechanisms from expertise?
Inversion and composite effects in experts for non-face categories
Mixed evidence: some experts (e.g., car aficionados) show FFA activation, others do not
What is the Other-Race Effect and how can it be modified?
Easier recognition of own-race vs. other-race faces
Training and individuation tasks can reduce ORE magnitude
What criteria define holistic face processing?
Sensitivity to relationships among features over individual features
Measured via composite, inversion, and part-whole effect
Do inversion effects extend to expert object categories?
Yes for well-trained experts (e.g., bird or car experts), but the magnitude is typically smaller than for faces
How does the composite illusion reveal mandatory holistic processing?
Aligned top/bottom face halves create interference, reducing part-based judgments
Misalignment or inversion abolishes the effect
What does the Thatcher illusion demonstrate about upright vs. inverted faces?
Feature inversions are grotesque in upright faces but almost undetectable when inverted
What is the norm-based “face space” model of identity?
Faces encoded as vectors relative to an average face; dimensions capture configural variations
How do adaptation aftereffects support face-space coding?
Prolonged exposure to one identity shifts perception of morphs away from the adapto
How does experience shape the dimensions of face space?
Exposure to own-race and familiar faces tunes channel sensitivities, affecting ORE and adaptation patterns
