• Stimulus compared to miniature copies (templates) of stored patterns
• Problem: we do not have millions of different templates

• Abstract form representing basic or most crucial elements
• Ex) Bird--> feathers, wings, beak...we compare attributes with our prototype

• Prototypes= circles, stars, and triangles
• Distorted prototype with 1-5 transformations
• Subjects were shows distortions, not prototypes
• Recognition was greater for prototypes than distortions
• Confidence decreased with # of transformations
• We store the idea/concept in our memory

• Prototype face from hair, nose, eyes, mouth
• Distortions had in common 0-3 parts of prototype
• S's were most confident about the prototype faces

• Prototypes are the "average" of all exemplars seen
• Ex) taking two people's faces and blending them together to make a creepy baby

• Prototypes are the most frequently experienced combination of attributes
• Best supported by experiments
• Ex) I have seen crows a lot, so when I picture a bird, it is the size of a crow

• Identify basic visual features, then compare with objects stored in memory
• Uses bottom-up processing

• start with the most basic aspect of something and work your way up

• Feature Theory
• Image demons--> take in visual information from all senses

• Feature Demons: break down info into distinct features
• Cognitive Demons: combine featres into meaningful wholes and compare to memory
• Decision Demons: Decide which items represented by combination of features

• Showed S's pairs of letters
• Quickly answered "same" or "different"
• Faster if fewer features in common (QW= faster response time, EF= slower response time) 
• Problem: context effects 

• 24 distinct geometric "primitives" (shapes) combined to form any object
• Still bottom-up processing, but geons are basic features
• Objects with corners= harder to recognize 
• Objects with midsegments= easier o recognize 

• Recognition based on overall shape, not features
• Top-down processing
• Start with the whole and then break down the image

• Large (global) letters made up of small (local) letters
• No effect on naming global letters (Big "H" made up of small "s")
• Slower reaction time when asked to identify small/local letters
• Process whole before parts

• Learned regular faces and scrambled faces with names
• Identified whole faces or parts of normal or scrambled faces
• Normal faces= better at identifying whole face than parts
• Scrambled faces= better able to identify parts
• Gestalt works if we see image in familiar orientation

• Holistic processing of faces
• Top-down/ parallel processing of parts
• Neurons in temporal cortex respond to intact faces, but not scrambled faces

• Whichever theory gets the job done the fastest
• They are all correct in some way 

Letter Recognition

Visual Coding of Info in STM

• Presented two letters side-by-side
• AA, Aa, AB
• Presented simultaneously, .5, 1, 2 seconds apart
• Press button if letters are the same

• RT greater (more time) for Aa than for AA
• No difference in RT after 2 seconds
• Early coding in STM visual, later coding= auditory
• Name retrieval= one component of pattern recognition

• Easier to identify a letter in a word (WORK) than a non-word (ORWK

• Letter processing is necessary for word processing
• bottom-up processing strategy
• you must process some of the letters before you process the word

Word

^

Letter

^

Feature

• S's judged if 3 letter word or nonword was present in a briefly presented 4 word display
• S's formed illusory conjunctions of letters
• Remembered letters together that had not been together
• Supports letter dependent processing

• Word is processed as a whole
• Direct connection betwenn features and words
• while letter identification may not occur, NOT NECESSARY for word identification

• Letters and words identified in parallel
• If word finished before letters, stop ID and go to the next word
• Subjects asked to circle each letter "t"
• More likely to miss "t" if in correctly spelled word than mispelled
• more likely to miss the "t" in words like "the"
• More likely to miss in common words than rare words
• In familiar words, subjects do not necessarily process all letters