Lecture 10 - Categorisation and Concept Formation Flashcards
Define a concept
Mental representation
Based sharing common properties
Items same class
Outline concept formation
Induction concepts divide items into classes
According their shares properties
Categorisation
Why are concepts not always defined by specific features
Sometimes do not have necessary or sufficient features define them
Polymorphous
E.g. what is a defining feature of a game
What are the 3 types of concept
- Basic level concept
- Superordinate concept
- Abstract concept
Define basic level concept
Similarity or perceptual qualities
E.g. bird, flower
Define superordinate concept
Groups basic level concepts
Not based perceptual similarity
E.g. politician, tools
Define abstract concept
Not refer to individual entity
But to some property, relation or state
E.g. sameness, truth
Outline Bhatt, Wasserman, Reynolds and Knauss 1998 study on basic level concept formation in animals
Pigeons chambers choice 4 response keys
Learned peck different keys for example mars of each 4 categories: flowers, cara, people, chairs
Controlled for colour by using variety coloured
Able respond correctly new exemplars never seen before
Outline RESULTS Bhatt, Wasserman, Reynolds and Knauss 1998 study on basic level concept formation in animals
Birds had formed concept flowers, cars, people, chairs
Performance more accurate with training 80%
Than with novel, test stimuli 50-60%
Outline Exemplar Theory of basic level concept formation
Learn about every instance independently
Classify novel examplars via similarity learned instances
Store everything you see
Always perform best at what you have seen
Outline Prototype Theory of basic level concept formation
Abstract prototype corresponding to central tendency of training exemplars
Don’t actually store images what you have seen = more viable as storing everything you see takes up capacity
Hold central idea what you have seen = averaging
Could be better at something you have never seen before
Summarise Prototype Model in helping us to Categorise
Category judgements made by comparing new exemplar to prototype
Summarise Exemplar Model in helping us to Categorise
Judgements made by comparing new exemplar to all old exemplars of a category or to exemplar that is most appropriate
Relate Exemplar Theory to Bhatt, Wasserman, Reynolds and Knauss 1998 study on basic level concept formation in animals
Animals storing info about training exemplars
Why they were more accurate at training exemplars to novel test stimuli
Explain why Homa et al 1981 argues humans show Prototype effect
Categorise prototype more accurately than training stimuli
Even though never seen before
Outline Aydin and Pearce 1994 prototype effect in pigeons
Artificial positive (ABC) and negative (DEF) prototypes.
Never see test stimuli in training.
Birds trained 3 element displays created by distorting prototypes
Taught 3 positive patterns always paired with food, 3 negative we’re not
Birds pecked more at positive
Outline Aydin and Pearce 1994 RESULTS prototype effect in pigeons
Responded more positive prototype than any of the trained positive stimuli and less to negative prototype than negative trained stimuli
Prototype effect
Outline Whittlesea 1987 study on prototype effects
Lists 1, 2 and 3 all differ from prototype by 2 letters
List 1 more similar to list 2 than list 3
Only trained on list 1. Tested on 1,2,3
Outline PREDICTIONS Whittlesea 1987 study on prototype effects
Prototype predicts: list 1 = list 2 = list 3. All lists should be learnt the same
Exemplar predicts: list 1 best learnt, then list 2, finally list 3
Outline RESULTS Whittlesea 1987 study on prototype effects
Humans show results consistent with Exemplar Theory
List 1 easiest
Conclusions of Prototype Theory and Exemplar Theory
Humans and animals retain info about training items/exemplars
But show prototype effect
Variation Exemplar Theory explain prototype effect
Outline Aydin and Pearce experiment on how the Exemplar Theory can explain the Prototype effect
Explain why birds peck more positive prototype to trained stimuli
Examine learning each component feature
Storing different parts stimulus rather thing as whole
Looking individual components we can identify something we have never seen before quicker
Outline the combined theory of Feature Theory
Rather look stimulus as a whole
Look at individual components
Features associated with category membership
Compare Feature Theory with Exemplar Theory
Exemplar: classifies basis similarity of whole stimulus. Take everything you have seen as a whole
Feature: classified basis sharing features stored exemplars, break down individual components.
Categories form means associative learning. Features category are associated with category label.
Does category learning show blocking?
Blocking key characteristic associative learning.
Pairing only produce association between X and category if category is surprising
Outline Shanks 1990 study of Exemplar theory and Associative Theory
Medical symptoms paired with disease diagnosis
Subjects must predict disease from symptoms
1 disease common - flue
1 rare - NA
1 target symptom, 2 non target symptom
Presented more cards for common disease
Asked which does headache predict more? NA or flu? Same number pairings for each disease
Outline BLOCKING in Shanks 1990 study of Exemplar theory and Associative Theory
If all that matters is number of pairings then headaches should be as good as predicting flu as NA
Headache paired flu, runny nose also present = loads of experience. Headache = blocked.
Headache paired NA rash present = not sure on rash = pay attention and learn
Outline link between Exemplar Theory and Associative Theory view of Shanks 1990 study
Exemplar Theory: just likely predict flu as NA
Associative Theory: more likely choose rare NA.
Outline RESULTS Shanks 1990 study of Exemplar theory and Associative Theory
Proportion common disease flu diagnosis: .37
Proportion rare disease NA diagnosis: .63
Associative learning best explanation performance categorisation task in humans
Outline superordinate categories
Members not necessarily physically similar but share common associate
Outline Wasserman, De Volder and Coppage 1992 study initial training
Pigeons trained slides people, chairs, cars, flowers
Reinforced making pecking response 1 people and chairs, making response 2 cars and flowers = superordinate categories
Outline Wasserman, De Volder and Coppage 1992 study secondary training
Taught make response 3 to people and response 4 to cars
Tested with chairs and flowers to make response 3 and 4
Outline Wasserman, De Volder and Coppage 1992 study RESULTS
Response 3 to chairs and 4 to flowers correct
Animals can form superordinate categories
What is Pearce 1997 argument on superordinate level concept formation in animals
Not true categorisation as seen in humans
But simply associative learning
We are more complex
But we need to specify exactly how what is more complicated so we can test
Outline abstract concept formation in animals using match to sample technique (MTS)
Bird shown sample key e.g. red then given choice red or green
Peck same colour shown
Master this
But poor transferring skill to different colours
Not really learned concept
Do not learn rule
Outline Wasserman, Hungary and Kirkpatrick-Steiger 1995 study on Match to sample technique (MTS)
Pigeons show complex stimulus displays, given red or green key
Trained arrays 1 set of specific icons
Rewarded for pecking red on same trials, green different trials
Perform above Chance for unfamiliar stimuli not in training stage
Evidence some generalisation of abstract concept
