Perception Flashcards
On-off Receptive Field
Ganglion: Light at center on small region, increase rate of firing; around region, decrease rate.
Off-on Receptive Field
Ganglion: Light at center on region decreases rate of firing; light in surrounding increases rate
Edge/bar detection
Multiple On/Off ganglion cells in LGN make pattern that stimulates detector in visual cortex
Goal of Perception? Why difficult?
To infer structure of external world from information available to senses.
Problem: information is variable and ambiguous
Information: ambiguous? variable?
1) Different things can appear the same (obstruction or uncertainty of position)
2) Same thing can look different (viewed from different angles, farther, nearer)
Monocular depth cues
interposition, linear perspective, texture gradient, shading and contours, atmospheric blur, motionparallax, relative height, relative size
Binocular depth cues
retinal disparity (gives rise to stereoillusions, apparent size illusion)
Solution to perception problem
Perceptual system makes reasonable guesses based on reliable cues. Ignores unreliable cues and give rise to perceptual constancies
How does stereogram generate perception of depth?
Images with slight disparity fused. Visual system assumes disparity due to retina and stereopsis, perceive as 3D
Does apparent size track retinal size or actual size?
Actual size. Demonstrated by apparent size illusion, where farther person is moved to same plane as closer person (appears smaller as opposed to farther away)
Perceptual Constancies
Objects are perceived as actual size rather than retinal size (for size constancy, person is not actually smaller, just farther away). Visual system ignores unreliable cues to produce perceptual constancies.
Low-level vision
Edge detection
Color
Spatial localization
Mid-level vision
Object features
Fig/ground segmentation
Perceptual organization
High-level vision
Object recognition
Face recognition
Scenerecognition
What are invariants? Importance?
Invariants are edges that are distinguishable regardless of orientation of depth. Helps classification despite drastic variations in the object’s silhouette, specific contours, and occlusion of large regions of the object.
Evidence that feature relations used in recognition
Biedelman(1987). Experiment testing RT for original images, missing lines, and missing vertices. RT increases from former to latter.
Template Matching Theory
Objects identified by matching percepts with stored representations (“templates”). Given object can have multiple templates and alignment must occur before matching. No need to define features!
Problem: Templates can be mismatched, wrong size, wrong part of retina, wrong orientation etc.
Feature Analysis Theory
Stimuli are combination of elemental features
Pros: features are simple, less difficulty matching; can specify important features that “pop out”; reduces complexity and overload of templates
Con: Relationships between features not represented;
Structural description/recognition-by-components Theory
Recognize object as configuration of simpler components/subobjects (“geons”), classify subobjects, recognize pattern.
Object-centered Representation
Perception more dependent on object and features. Structural description theories
Viewer-centered Representation
Perception more dependent on observer and limiations. Template matching theories
Face recognition evidence
Fusiform gyrus respond to faces, face cells in monkeys, damage causes selective difficulty in recognizing faces (prosopagnosia)
Inversion effects
Better at recognizing upright faces, worse when upside down (Margaret Tatcher Illusion). Not true for other objects
Greebles, Tarr’s Experiment
Can be trained to recognize other complex objects and activate same fusiform, not just face processing. Also shows inversion effect
Top-Down Processing
Occurs when context guides perception; higher level knowledge contributes to interpretation of lower-level perception
Bottom-Up Processing
Stimulus provides information by itself, regardless of context
Word Superiority Effect
When presented with ambiguous stimuli, more accurate when identifying word than letters alone. Word context supplements lone feature information
Gestalt Principles of Grouping
Objects tend to be organized into units according to set of principles. Proximity, good continuation, common region, similarity, common fate, closure