Categorization
J. L. Borges (1966) Other inquisitions 1937-1952. N. Y.,
Washington Square Press . Quoted in Rosch, E. (1978) Principles of
categorization. In Rosch & Lloyd, Cognition and categorization.
The following is a taxonomy of the animal kingdom…attributed to an ancient
chinese encyclopedia entitled the Celestial Emporium of Benevolent
Knowledge:
On those remote pages it is written that animals are divided
into (a) those that belong to the Emperor, (b) embalmed
ones, (c) those that are trained, (d) suckling pigs, (e)
mermaids, (f) fabulous ones (g) stray dogs, (h) those that
are included in this classification, (i) those that tremble, (j)
innumerable ones, (k) those drawn with a very fine camel’s
hair brush, (l) others, (m) those that have just broken a
flower vase, (n) those that resemble flies from a distance.
If these aren’t good categories, what are?
Ones used frequently in communication
Good categories
Rosch
1. Cognitive Economy
– Reduce number of discriminations in world
Can’t have separate label for each thing
– Informativeness
Knowing that something is an X tells us more
Good categories
1. Cognitive Economy
– Reduce number of discriminations in world
Can’t have separate label for each thing
– Informativeness
Knowing that something is an X tells us more
These goals are at odds
Good categories
1. Cognitive Economy
– Reduce number of discriminations in world
Yields few large categories
– Informativeness
Yields many small categories
Good categories
1. Cognitive Economy
– Reduce number of discriminations in world
Yields few large categories
– Informativeness
Yields many small categories
Where is trade-off optimal?
Good categories
1. Cognitive Economy
– Reduce number of discriminations in world
Can’t have separate label for each thing
– Informativeness
Knowing that something is an X tells us more
2. Reflect perceived structure of world
– Correlated features feathers/beaks
– Utility for us
Vehicles
car
sedan
station
wagon
truck
garbage
boat
dump
row
motor
Furniture
chair
easy
kitchen
lamp
table
rug
floor
Chinese
rag
Animals
snake
garter
rattle
dog
Pekinese
cow
mutt
milk
beef
Two dimensions of categorization
• Vertical: Level of abstraction or
inclusiveness
• Horizontal: Organization within categories
Vertical
• What are you sitting on?
• What is this I am wearing?
Vertical
• What are you sitting on?
Chair, not auditorium chair or furniture
• What is this I am wearing?
Sweater, not wool sweater or clothing
Preferred level of reference, why?
List features common to members of the
following categories:
• Apple
• Tools
• Shirt
• Furniture
• Green grapes
• Watermelon
• Shoes
• Denim pants
Preferred level of abstraction
• Informativeness: Indexed by features
• Highest-->middle: Large increase
• Middle-->lowest: Small increase
Preferred level of abstraction
• Informativeness: Indexed by features
• Highest-->middle: Large increase
Superordinate (vehicle) to basic level (car)
• Middle-->lowest: Small increase
Basic level (car) to subordinate (4-door)
Basic Level distinguished by
convergence of many cognitive tasks:
•
Common attributes
•
Similarity of shapes
•
Identifiability of shapes
•
Imagery
•
Motor programs
•
Communication: labeling
•
Communication: fastest verification
•
Development: first labels in lexicon
•
Development: categorization
•
Language: earliest differentiation within language
•
Language: most frequent, shortest labels
•
Language: basic level terms neutral
Book review of pretentious novel:
“ And so, after putting away my 10 year old royal 470
manual and lining up my Mongol number 3
pencils on my Goldsmith Brothers Formica
imitation-wood desk, I slide into my oversize
squirrel-skin L. L. Bean slippers and shuffle off to
the kitchen. There, holding Decades in my
trembling right hand, I drop it, plunk , into my new
Sears 20-gallon, celadon-green Permanex trash
can.”
Why do these measures converge at
Basic Level?
Basic Level distinguished by
convergence of many cognitive tasks:
•
Common attributes
•
Similarity of shapes
•
Identifiability of shapes
•
Imagery
•
Motor programs
•
Communication: labeling
•
Communication: fastest verification
•
Development: first labels in lexicon
•
Development: categorization
•
Language: earliest differentiation within language
•
Language: most frequent, shortest labels
•
Language: basic level terms neutral
Appearance
Behavior
Communication
Parts & Basic Level
Tversky & H e m e n w a y
• Among attributes, parts proliferate at basic
level
• Parts form a bridge from appearance to
behavior
Vertical Dimension of Categorization
• Goals of categorization
– Cognitive economy
• Informative
• Reduce # discriminations
– Reflect perceived structure of world
• Basic level maximizes informativeness given #
categories that must be kept in mind
• Many cognitive tasks converge on basic level
• Parts underlie convergence
What about categories of other things?
• Scenes
• People
• Events
Horizontal Dimension of Categorization
Characterize internal structure of categories
Knowledge representation of categories
“Meaning” of categories
Early view of categorization
• World is full of things varying on many
dimensions
• Different cultures draw category boundaries in
different places, color, corn, parrot, aunt
• If categories are unit of thought, then different
cultures think differently
Early view of categorization
• World is full of things varying on many
dimensions
• Different cultures draw category boundaries in
different places, color, corn, parrot, aunt
• If categories are unit of thought, then different
cultures think differently
• BUT, cultures differ in environments, needs
Sapir- Whorf Hypothesis:
Language shapes Thought
• How do you separate culture from language?
Conditions for testing Whorf Hypothesis
• Languages differ with respect to an attribute
• Physical, culture-free measure of attribute
• Non-linguistic dependent measure
• Prevalent attribute; culture independent
Conditions for testing Whorf Hypothesis
• Languages differ with respect to an attribute
• Physical, culture-free measure of attribute
• Non-linguistic dependent measure
• Prevalent attribute; culture independent
Color
Codability of color predicts memory
Brown & Lenneberg
• Group 1: codability: short, agreed-upon labels
• Group 2:
– See color chip
– Select that color from array of colors
– Memory better for more codable colors
Codability of color predicts memory
Brown & Lenneberg
• Group 1: codability: short, agreed-upon labels
• Group 2:
– See color chip
– Select that color from array of colors
– Memory better for more codable colors
• BUT: didn’t test across languages
Codable colors
Berlin & Kay, De Valois
• Likely to be prototypical colors across languages
• Languages differ on color boundaries, not centers
• Visual system especially sensitive to prototypical
colors
Teaching color & shape names to Dani
Rosch
• No names for colors or shapes; taught names
• Easy to learn prototypical colors, shapes
• Hard to learn peripheral colors, shapes
Whorf Hypothesis: 2 strikes against
• Colors (shapes) that are highly codable are
– Central members of categories
– Remembered/learned better across cultures
• Perhaps these categories are universal
• But, hold on….not out yet
Define:
• Vegetable
• Table
• Vehicle
• Cup
Rate how “good” each exemplar is of
category:
FRUIT: pineapple, grapes, persimmon, apple
VEHICLE: bus, jeep, skateboard, car
FURNITURE: clock, table, couch, ottoman
CLOTHING: belt, shoes, shirt, pants
Write examples of the following
categories:
Odd Number
Color
Musical instrument
Emotion
Tool
Horizontal organization of categories
• Definitions aren’t good
• Boundaries aren’t good
• Agreement on good examples, focal cases:
Prototypes
Horizontal, internal structure of
categories
• Family resemblance
• Set of characteristic features
– No one member has all the features
– Prototypical members have more of the features
Horizontal, internal structure of
categories
• Family resemblance
• Set of characteristic features
– No one member has all the features
– Prototypical members have more of the features
Furniture: legs, seat, back--chair, couch, not rug
Fruit: sweet, seeds, small--apple, not watermelon
Cognitive tasks supporting typicality
• Verification RT: faster to say yes to “car is
vehicle” than “skateboard is vehicle”
• Development: typical learned earlier
• Production: typical produced earlier
• Language: typical more frequently used
Structure of categories
• Vertical: basic level is privileged/neutral
• Horizontal: think of categories in terms of
– Typical examples
– Family resemblance
Rather than necessary and sufficient features
Whorf Hypothesis: Act III
Levinson
• Some widely dispersed languages don’t use left
and right to describe locations; use NSEW
Whorf Hypothesis: Act III
Levinson
• Some widely dispersed languages don’t use left
and right to describe locations; use NSEW
• Task
– Study parade of animals, aardvark….giraffe….zebra
– Order mixed up, participant turned 180 degrees
– Put animals in order
Whorf Hypothesis: Act III
Levinson
• Some widely dispersed languages don’t use left
and right to describe locations; use NSEW
• Task
– Study parade of animals, aardvark….giraffe….zebra
– Order mixed up, participant turned 180 degrees
– Put animals in order
• Results
– Speakers of R/L languages line up animals L/R
– Speakers of no R/L languages line up animals NSEW
Whorf Hypothesis: Act III
Levinson
• Some widely dispersed languages don’t use left
and right to describe locations; use NSEW
• Task
– Study parade of animals, aardvark….giraffe….zebra
– Order mixed up, participant turned 180 degrees
– Put animals in order
• Results
– Speakers of R/L languages line up animals L/R
– Speakers of no R/L languages line up animals NSEW
• BUT: implicit verbalization?