I. Basics

A. perception defined

1. meaningful interpretation of what we sense

2. involves recognition & millisecond processing

B. components of the "classic" approach

1. distal stimulus, proximal stimulus, percept

2. proximal stimuli are not percepts

a. e.g., size constancy

C. top-down (conceptual) and bottom-up (data-driven) processing

D. illusions

A. one can see, but not interpret information

1. involves damage in the right hemisphere

2. may be able to use another sense to "recognize" the object

B. types

1. apperceptive agnosia

a. inability to recognize faces

b. right hemisphere damage

III. Gestalt models of perception

A. interest in the perception of forms

1. e.g., face/vase picture, subjective contours

2. emphasizes the perception of forms as holistic

a. the whole is more than the sum of its parts

B. Law of Prägnanz (goodness of form)

1. forms are interpreted in the simplest and most stable way

2. active participation

3. principles of organization

a. figure-ground, proximity, similarity, good continuation, closure, symmetry, common fate 4. a more recent example: global/local figures

C. concerns

1. translating principles into cognitive &/or physiological processes

2. circularity of explanation

A. defined

1. stimulus-driven

2. little influence from expectations or prior learning

B. Template Matching

1. perception = template matching

a. check reading, supermarket scanning

2. drawbacks

a. number of templates needed

b. how are templates created

c. can’t explain the flexibility of pattern recognition

C. Featural Analysis

1. perception = identification of simpler features (characteristics)

2. physiological evidence

a. Littner et al. (1959)

1. feature detectors in frog retinas

b. Hubel & Weisel

1. excitation of single nerve cells in the visual cortex

3. Geon Theory (Biederman)

a. involves 36 simple geometric primitives

b. recognition by components (RCB)

c. evidence from degraded forms

1. 100 ms presentation

2. 70% with vertices, 50% without

4. behavioral evidence for features

a. visual search

5. Pandemonium: a model of letter perception

          a. levels of detectors (demons)

          b. what it captures about feature analysis

1. loudness of “screams” depends on clarity & quality

2. more important features can carry more weight

3. could involve learning

6. behavioral evidence for features

a. visual search

7. drawbacks

a. difficulty in defining what a feature is

b. are the same sets of features checked for every stimulus?

D. Prototype Matching

1. perception = matching to idealized, abstracted forms

a. fuzzy rather than exact matches

b. the more shared features, the faster and easier the match

2. evidence

a. Posner & Keele (1968)

1. classify dot patterns derived from a prototype

2. "classification" of old distortions & unseen prototypes @ 87%; new patterns @ 67%

b. Solso & McCarthy (1981)

1. presentation of identikit faces

2. rated old + high confidence = pseudomemory

3. drawbacks

a. little physiological evidence

b. what kind of processing is involved with matching

c. in general, a need for top-down processing

E. Top-Down Processes

1. perception = context & expectation dependent

a. but, must interact with bottom-up processes

2. perceptual learning

a. Gibson & Gibson (1955)

b. matching errors were systematic and indicated learning

c. perception involves learning more about stimulus features

d. expertise

3. change blindness

a. research by Simon & Levins (1997) 1. large changes can go unnoticed
  a. lab study: no detection though warned of continuity errors

b. real life study: 50% did not notice change in interviewer
 

2. encode "gist" information rather than "verbatim" info

4. word-superiority effect a. asked to identify which of two letters "D" or "K" were presented (Reicher, 1969)
  1. presented either in context (in a word) or alone

2. better identification in words

b. connectionist model
  1. top-down processing from the word level combines with bottom-up processing from the feature level

c. neuropsychological evidence
  1. real words and pseudowords showed greater activation in the LH & in semantic areas