Lecture 8: Knowledge

1
Lecture 8 Knowledge
Psyc 317 Cognitive Psychology
2
Todays agenda

• Approaches to Categorization
• Definitions
• Prototypes
• Exemplars
• Is there a special level of category?
• Semantic Networks
• Connectionism
• Categories in the brain

3
Categories are important

• Categories are pointers to knowledge
• If you encounter something new but can categorize
  it
• Know some properties of the object
• Can make inferences about the objects
  relation to other objects
• If we didnt have categories, wed be very
  confused

4
So you encounter something new
5
Definition approach to categories

• Something is a category member if it
• meets the definition
• Seat
• Back
• Four Legs
• Its a chair!

6
Definitions work, dont they?
All chairs! Hmmm
7
So definitions dont work

• Imagine the category of games
• Sports, video games, board games, etc.
• But there still are some similarities
• Wittgensteins (1953) concept of family
  resemblance
• Things in a particular category resemble each
  other in some ways
• Family resemblances allow for variation

8
Prototype approach to categories

• Prototype of a category Formed by
• averaging category members
• An object belongs in a category if it is
• similar to the prototype

?
9
Prototype Example
10
Differences in Prototypicality

• High prototypicality Object closely resembles
  category prototype
• Low prototypicality Object does not resemble
  prototype
• Rosch (1975) - 50 members of different
• categories
• Rate how good an example the object is of
  that category

11
Rosch Typicality Ratings
12
Prototypical Objects Have High Family Resemblance

• Prototypical objects share lots of common
  features
• Example
• What are the common features between chair and
  sofa?
• What about mirror and drawers?

13
Statements About Prototypical Objects Are
Verified Rapidly

• Example Sentence Verification Technique
• Are the following sentences true?
• An apple is a fruit.
• A pomegranate is a fruit.
• A tomato is a fruit.
• Apple is prototypical, RT to Yes is fast
• Pomegranate and tomato are less prototypical,
  RT to Yes is slow

14
Smith et al. (1974) Results
15
Prototypical Objects are Named First

• Mervis et al. (1976)
• Demo Name all the canines (NOT breeds) you
  can think of
• People name dog before coyote

16
Prototypical Objects More Affected by Priming

• Rosch (1975)
• Hear category name as priming stimulus
• Then see two objects from the category
• High prototype objects
• Low prototype objects
• Are the objects same or different?

17
Rosch Results
610 ms
Hear Green
780 ms
18
Why does this occur?

• Green activates prototype for green
• Good greens lead to faster RTs
• Bad greens are not activated, lead to longer
  RTs

19
Prototype SummaryHigh prototype objects

• Family resemblance Have more in common
• Typicality Faster in sentence verification
  tasks
• Naming First to be named
• Priming Are faster judged as similar
• when category is given as a prime

20
Exemplar approach to categories

• Prototypes Average case of a category
• Exemplars Actual common examples

21
Exemplars Prototypes

• Good exemplars and prototypical
• objects are often really similar
• Good exemplars sparrow, robin
• Prototypes look like a combination
  sparrow-robin (sparrobin?)
• High family resemblance

22
Teasing apart a difference

• Question How do subjects place new objects in a
  category?
• Prototype Average together given cases
• Exemplar Use information about each case to
  categorize

23
Medin et al. (1992) Burlosis!

• Create a new category a fake disease called
  burlosis
• Give subjects example case studies with
  symptoms
• How do people create the burlosis category?
• Prototype Average cases together
• Exemplar Use information about each case

24
Training Category creation

• Correlated condition Nosebleed discolored gums
  always occurred together
• Uncorrelated Those two symptoms did not occur
  together

Correlated Condition
25
Burlosis Predictions

• If prototype, average case should be created
  (all symptoms together)
• All symptoms were similar in both conditions
• Correlation should not matter
• If exemplar, common example should be used
  (correlation of two symptoms)
• Correlated symptoms were good exemplars of
  burlosis
• Correlation should matter

26
Test New case categorization

• Subjects given test cases to diagnose (classify
  new objects in a category)
• All test cases had the same average number of
  symptoms
• Prototype Correlation should not matter
• Exemplar Correlation should matter

27
Burlosis Results
Correlation matters!
Participants based diagnosis on
patterns/exemplars
28
What does this show?

• People were using exemplars because they were
  using the correlated information in diagnosis
• If people were using prototypes, correlated
  information should not have mattered in training
  or in test

29
Fine line between prototypes and exemplars

• How do the two coincide?
• Common thinking
• As we first learn category exemplars, we
  average them to form a prototype
• Add exemplars as exceptions to the
  prototype

30
Outline

• Approaches to Categorization
• Definitions
• Prototypes
• Exemplars
• Is there a special level of category?
• Semantic Networks
• Connectionism
• Categories in the brain

31
Levels of categories
Rosch says basic level is special
32
Why is the basic level special?

• Demonstration!
• List as many features that are common
• to all or most items in each category
• Furniture
• Table
• Kitchen table

33
Rosch et al. (1976)
Average number of features listed
34
Name each object
35
Object naming demo

• Everyone used the basic level guitar, fish,
  pants
• Not superordinate instrument, animal, clothing
• Not subordinate rock guitar, trout, Levis

36
Basic level is fast(Rosch et al., 1976)

• Show participants category label (car, vehicle,
  etc.)
• Show picture of object from category
• Time to verify picture is a member
• Basic level RT lt Superordinate level RT

37
Knowledge affects categorizationTanaka Taylor
(1991)

• 2 groups bird experts and nonexperts
• Ask participants to name pictures of all kinds
  of objects (tools, flowers, birds)
• How do bird experts and nonexperts categorize
  birds?

38
Tanaka Taylor Results
39
Culture affects categorizationColey et al. (1997)

• Maya Itza - Guatemalan tribe
• Live close to their natural environment
• Identified local plants and animals at
  subordinate category (oak tree)
• Identified non-native plants and animals as the
  basic level category (flower)

40
Outline

• Approaches to Categorization
• Definitions
• Prototypes
• Exemplars
• Is there a special level of category?
• Semantic Networks
• Connectionism
• Categories in the brain