Chapter 3: Perceiving Objects & Recognizing Patterns Flashcards
Perception
Take sensory input and interpret it meaningful
Perception of patterns, objects, people, and events
Conscious Experience
Active reconstructive process
What you see, hear, and experience does not directly correspond to the stimuli in the world
Distal Stimulus
Objects and events in the real world
To process info about this stimuli, must receive info from sensory systems
Proximal Stimulus
Reception of info and its registration by a sense organ
E.g. light waves reflect from the trees and cars to eyes (retina)
Retinal image is formed (2D), with size dependent on distance from objects
Retinal images are upside down and reversed
Percept
Meaningful interpretation of proximal stimulus
Size Constancy
Objects appear smaller farther away, but we know that they are time same size regardless of where we look from
Objects that are farther away take up less space on the retina
Pattern Recognition
Recognition of a particular object, event, etc. as belonging to a class of objects, events, etc.
Almost all instances of perception involve pattern recognition
E.g. recognize a bush as part of the shrubs category
Gestalt Approaches To Perception
Gestalt psychologists think that we identify objects as a unit or whole
Form perception
Illusory/subjective contours
Form Perception
Segregation of the whole into objects (figure) and background (ground)
E.g. reversible figure where you either see two sides of faces or a vase
Illusory/Subjective Contours
See a triangle in white space even though there are no lines creating a triangle
Gestalt Principles of Perceptual Organization
Proximity
Similarity
Good continuation
Closure
Common fate
Symmetry/Parallelism
Proximity
we group things that are closer together (e.g. group A in rows)
Similarity
group elements that are similar (e.g. group B in columns)
Good Continuation
group objects whose contours form a continuous straight or curved line (e.g. see two intersection lines in C)
Closure
perceive shapes even when there are gaps in a figure (e.g. see a rectangle in E)
Common Fate
elements that move together will be grouped together (e.g. see F)
Law of Pragnanz
of all possible ways of interpreting a display, we use the organization that yields the simplest and most stable shape or form
Most Gestalt principles are subsumed under this law
Simple and symmetric forms are seen more easily
Can help to explain experience with subjective contours
Some principles are used by infants form 3-6 months
Issues With Gestalt Approach to Perception
Don’t know how these principles are translated into cognitive or physiological processes
Law of Pragnanz can be seen as circular
Bottom-Up Processes
Data Driven/Stimulus Driven
Perceiver starts with small bits of info from environment that are combined in various ways to form a percept
E.g. see edges, rectangular and other shapes, certain lighted regions, put together conclude you see doors and a hallway
Three Bottom Up Processes
Template matching
Feature analysis
Prototype matching
Template Matching
Correspondence between (external) stimulus and stored pattern in memory (internal template)
A pattern is compared to all templates and identified by the template that best matches it
Templates
Previously stored patterns
Need to be an exact match to stimuli
E.g. QR code
Problems With template Matching
Need a lot of templates
Does not explain how we recognize new objects (create new templates?)
Does not deal well with surface variation of stimuli (e.g. writing - everyone’s handwriting looks different)
Featural Analysis
Processing stimuli by breaking them down into components
Use recognition of parts to infer what the whole represents
Fits the strongest with neurophysiological evidence
Objects
Combinations of features
Features
Small (local) templates that can be combined in. many different ways
Edge Detectors
Cells responding strongly to borders between light and dark (Hubel & Weisel)
There are many other cells specified to respond to different elements of stimuli (e.g. certain orientation of lines)
Visual Search Task
Present participants with arrays of letters
Respond if they detected the presence of a particular target
Participants took longer to find a Z than a Q when arrays share features of straight/angular lines
Same has been found for auditory features
Advantages of Feature Analysis
More flexible system
Reduced number of (local) templates
Issues With Feature Analysis
No good definitions of what can be a feature and what cannot
E.g. what are the features of the face? Specific features for right and left sides or more general? Etc.
Prototype Matching
Explain perception in terms of matching an input to a stored representation of info, but the stored representation is a prototype
Matching pattern to stored representation (this time a prototype)
Exact match is not required, only approximate
Prototype
idealized representation of some class of objects or events
Top Down Processes
Theory Driven/Conceptually Driven
Perceiver expectations, theories or concepts guide the selection and combination of the info in the pattern recognition process
Expectations guided where you look, what you looked at, and how info is put together
Problems With Bottom Up Processing
Context & expectation effects
Top down processes have to interact with bottom up processes (otherwise wouldn’t be able to perceive things you weren’t expecting)
Context Effects
Accuracy and length of time required to recognize objects vary with context (e.g. recognize utensils quicker when looking at kitchen scene)
David Marr’s Model
Primal sketch: depicts areas of relative brightness and darkness in a 2D image as well as localized geometric structure (bottom up processes)
2 ½ D sketch: use shading, texture, edges. Etc viewer derives info about what the surfaces are and how they are positioned (relies on bottom up processes)
3D sketch: info from real world knowledge or specific expectations incorporated (top down processes)
Change Blindness
Inability to detect changes to an object or scene especially when given different views of that object or scene
E.g. people are blind to changes in movie scenes where in one scene a child is playing blocks and in other is not, etc.
Word Superiority Effect Experiment
Reicher
Participants asked to identify which of two letters was briefly presented on screen
Later, presented with two alternatives for what letter might have been
Sometimes presented with a single letter, sometimes with a word, sometimes with three other letters (did not spell word)
Participants better identified letters presented in context of words than alone or nonwords
Word Superiority Effect (Word Advantage)
Letters are easier to perceive in a familiar context (words)
Missing Letter Effect
When readers read connected text quickly divide words into content words and function words
Focus attention more on moderately familiar content words, more likely to miss letters in highly familiar function words
Connectionist Model of Word Perception
McClelland & Rumelhart
Input is processed at several different levels (e.g. features, letters, phonemes, words)
Many connections (excitatory or inhibitory) between each ‘node’
Once a node is activated, activation spreads along node’s excitatory connections to other nodes
E.g. T node will become more active with the Trap node (for word prediction game)
Explains word superiority
Common Assumptions For Perception Models
Perceiver must acquire info about distal stimulus (presumably by interpreting proximal stimuli)
Perceiver does something to the proximal stimuli (probably because it doesn’t contain all the info needed to identify it)
All models describe the act of perception as the construction of mental representations of objects
Constructivist Approach to Perception
People add to and distort the info in the proximal stimulus to obtain a percept, a meaningful interpretation of incoming info
James Gibson
Rejected the idea that perceivers construct mental representations from memories of past encounters with similar stimuli
Perceiver does little work, because the world offers so much info, leaving little need to construct representations and draw inferences
Central question of perception is not how we look at and interpret a stimulus, but instead how we see and navigate among real things in the world
Direct perception
Affordances
Direct Perception
Light hitting the retina contains highly organized info that requires little or no interpretation
Affordances
Acts or behaviours permitted by objects, places, and events
E.g. chairs afford sitting, handle affords grasping
Affordances of objects are directly perceived (e.g. we see a chair is for sitting as easily as we see a chair is made of wood)
Critiques of Gibsonian theory
Without sharp definitions of affordance, theory is not helpful to explain perception
Gibson fails to specify what kinds of things are invariant and what kinds are not
Visual Agnosias
Impairments in ability to interpret visual info
Can’t recognize objects by sight
Can sometimes recognize them by sound, touch, or smell
Problem lies in creating a percept from the proximal stimulus
Apperceptive Agnosia
Can see contours or outlines of a drawing/object but have difficult time matching one object to another or categorizing them
Unable to form stable [pre semantic] representations of objects
Failure to interpolate missing contours or recognize objects from unusual angles
Some can’t name any objects
One case that couldn’t distinguish X from O
Associated with damage to the right hemisphere occipital lobe
Associative Agnosia
Can form percept of object but cannot identify [cannot achieve a correct semantic description]…category specific
Can match objects/drawings and copy drawings, tend to be very slow at it
May become distracted by small details
Cannot readily name the objects they see or draw
Associated with bilateral damage to temporal region (both cerebral hemispheres)
Failed to identify objects presented visually (either confusing them with other objects or has not idea what it is)
Semantics intact given testing in another modality
Category Specific Agnosia
Difficulty identifying items of a particular category, but ability to identify other types of objects is unaffected
Affected categories tend to include clusters of objects that are highly visually and semantically similar
Prosopagnosia
Visual agnosia for faces
Can see details of faces, but cannot recognize a face as a coherent unit
Fail to recognize well known faces of friends/family or yourself
Damage to region in right hemisphere (potentially some left hemisphere involvement as well)
Normal people look at eyes to recognize faces
Proposagnosia people look at incorrect info when judging faces (e.g. mouth or nose)
May have intact object recognition abilities but can’t recognize family or well know people
Right fusiform gyrus (Fusiform face area [FFA])
Unilateral Neglect (Hemineglect)
Ignore stimuli on the opposite side
Damage to parietal cortex
E.g. damage to right parietal cortex means they ignore stimuli on the left
Covert (Unconscious) Face Recognition
There can still be evidence of covert (unconscious) face recognition
GSRs discriminate between familiar and unfamiliar faces
Can recognize people using voices, hair, posture, gait, etc.
Capgras Syndrome
These patients are NOT prosopagnosic - face recognition abilities are intact
Fail to show a GSR response that discriminates between familiar and unfamiliar faces
They claim that their spouses are imposters E.g. insist that identical looking persons had taken the place of her family
Capgras Syndrome Explanations
Results from an attempt to reconcile;
The fact that there is no (unconscious) feeling of familiarity (e.g. no GSR difference between familiar and unfamiliar faces)
The fact that they do agree the person LOOKS identical to someone they know (e.g. spouse)
