Textbook Notes (280,000)
CA (170,000)
UTSC (20,000)
Psychology (10,000)
PSYB57H3 (300)
Chapter 3

PSYB57H3 Chapter Notes - Chapter 3: Eleanor J. Gibson, Template Matching, Edge Detection

Course Code
Gabriela Ilie

This preview shows pages 1-3. to view the full 10 pages of the document.
Chapter 3
Perception: take sensory input and interpret
Brain responsible for visual processing occupy half of total cortex
Types of perception: visual, auditory, olfactory, haptic and gustatory
Classic approach to perception
Object to be perceived is a distal stimulus
Reception of information and registration make up the proximal stimulus
Retinal image: image at the back of the retina that is formed
o2D image, and size depends on distance from window and object
oLeft right and up down are reversed.
Meaningful interpretation = percept (seeing and recognizing)
Size constancy: size does not change even if the size in retinal image changes
oPerception is something beyond retina images
Pattern recognition: recognition of a particular object, even to belong to a
certain class – most percepts involve classification
Ames room: trapezoid design, brain knows room as parallel
Experience and expectation contribute to how we see our environment
Bottom up Processes
Data/stimulus driven
Stimulus determines percept – direction is stimulus to onput
Building percepts from small units
Template matching, feature analysis, prototype matching
Perceiver starts with small bits of information that combine to form a percept.
One direction: input to interpretation, system cannot go back, and incidents right
now are unaffected by later processing.
oInformation about a stimulus, and relatively uninfluenced by expectations
or previous learning.
Posner/Raichele argue bottom up are automatic and reflexive even when person is
passively regarding the information.
Perception involves segmentation or parsing of visual stimuli into objects and
backgrounds – focus on how we recognize objects as forms
Form perception: segregation of whole display into objects and backgrounds
Salvador Dali: Slave Market with Disappearing Bust of Voltaire
Figure: definite shape, ground: shapeless, less formed, far in space

Only pages 1-3 are available for preview. Some parts have been intentionally blurred.

Illusionary contours/subjective contours: Illusory outline created by visual cues
that lead to erroneous form perception.
oPerception is a construction process, and requires participation
Perceive units as a whole, not by its parts
Gestalt principles of perceptual organization
oProximity – close things, nearness
oSimilarity – similar objects group
oGood continuation: form straight or curve lines
oClosure: illusionary closures
oCommon fate: elements that move together will group together.
Law of pragnanz: This law holds that objects in the environment are seen in a
way that makes them appear as simple as possible.
Minimal model theory: formalizing law of pragnanz
Drawbacks: unsure how principles are translated into cognitive or physiological
Law can be circular without further specification
Template Matching
Correspondence between external and stored patterns in memory (templates)
oPattern is identified and compared to templates to identify the best match
Example: UPC bar codes for products
Process of perception involves comparing incoming information that are stored
oFurther processing is needed to find the most appropriate fit
oMany models exist in knowledge
oNeed impossibly large number of templates
oDoes not explain how new objects are recognized, and how they are kept
track and created
oDoes not explain how patterns that are similiar, even when patterns differ
Surface variation of stimuli
Unlikely used in every day perception
oHow does a perceiver know which orientation and template to use?
oPossible if stimuli is relatively clean and known ahead of time to be
oDoes not explain noisy patterns and objects, such as blurred, blocked, etc.
Feature Analysis
Analysis of a whole into parts
oObjects = combination of features
oFeatures are small templates that can be combined in many ways
Recognize features to recognize the combination
Break down an object into many components

Only pages 1-3 are available for preview. Some parts have been intentionally blurred.

Lettvin: implant electrodes into frogs and found that certain stimuli can make
them fire more frequently
oSome cells responded to borders – edge detectors
oBug detector respond to black dot in a field of vision
Huble/Wiesel: visual cortex of cats and monkeys respond to particular orientation:
Lettin and H/W research show that detectors look for certain features and retinal
cortical cells have these in existence – confirm feature analysis model applicable
Eleanor Gibson tabulated features of Roman Alphabet (C and G similar, not F)
Neisser: people use features to identify letters
oPerform visual search task easier to find letters in a group that did not
share features (Q/Z in “round letters” and” straight letters” non targets
make search harder
Similar findings in audition syllables which share articulatory features are more
likely to be confused (da/ta confused more than da/sa)
Articulatory features include:
oVoicing (vibration of vocal chords)
oNasality (passing of air through nasal)
oDuration (how long sounds are)
oPlace of articulation (where in mouth)
Selfridge – Pandemonium Model
oDifferent levels of demons that function as feature detectors
Lower level process input
Higher level scan output from lower level
Demons scream response of what they find
1st stage: Image demon convert proximal stimulus into
representation/depictions of information for higher demons to
2nd stage: Feature demon each demon looks for a particular
feature, then screams if found
scream based on confidence (loud/soft)
3rd stage: Letter demon/Cognitive demon pay attention to the
feature demon that is associated with their own features. If feature
demon convinces them the letter is represented, scream based on
confidence (loud/soft)
4th Stage: Decision Demon – Figures out screaming and decides
oScreaming of demons matches how real life stimuli have subjective clarity
and cquality degraded or incomplete but can still recognize
oFeature demons can link to cognitive demons to give more weight, so
some features are weighted more heavily – ( - in A weighted more)
oWeights of features can change to allow learning
You're Reading a Preview

Unlock to view full version