Video Module 18: Theories of Categorisation

Question 1

What are the three kinds of theories of how we categorise concepts?

Answer 1

1. classical theory
2. probabilistic theories
3. ‘theory’ theory

Question 2

probabilistic theories of categorisation

Answer 2

propose that category representation is probabilistic—based on similarities between members—rather than deterministic—based on rules of categories.
1. prototype theory
2. exemplar theory

Question 3

prototype theory

Answer 3

suggests that categories are represented by the average of all members or instances of that category, called a prototype
- a prototype does not have to exist in the world: a mental average is an abstract representation
- suggests that we do not rely on specific examples of objects (exemplars)
- we assess for category membership and typicality by comparing the specific instances we encounter to the prototype we store in our minds
—explains the idea of graded category membership: items can be better or worse examples (more or less similar to the prototype)

Question 4

What is the evidence for prototype theory?

Answer 4

A study was conducted by Posner & Keele in which they created two prototype stimuli participants had never seen before, and made distortions of those prototypes. Participants were trained on the distortions until they could identify the two categories, then had to complete a categorisation task that included new, old, and the prototype stimuli.
- Researchers found that immediately after testing, participants were equally accurate in categorising both old stimuli and the prototypes the distortions were created from. They found that two weeks later, participants were most accurate at categorising the prototype stimuli.
- Suggests that participants store an average of their experiences

Question 5

What are criticisms of prototype theory?

Answer 5

1. PT assumes that we do not store info about specific instances (exemplars): simply not the case
2. PT doesn’t account for variance within a category: it defines the “center” (average) of a category, but it does not tell us about the “spread”
   —We sometimes make judgements about which category an object belongs to based on the category variance

Question 6

exemplar theory

Answer 6

suggests that concepts are represented by all of the specific instances of that concept—the exemplars—that we have experienced
- proposes that, in order to categorise an object, we compare it to all of the exemplars in our memory and decide the object’s category based on which exemplars are most similar to it
- we do not necessarily compare new instances to the “most typical” exemplar
- exemplar theory allows for prototypes: we only abstract a prototype if necessary—for example, when we need to categorise a new exemplar
- exemplars exist in the real world
- suggests that category representation is concrete: we do not create “summaries” of categories
- we make judgements of typicality and category membership on frequency: more frequent exemplars are better members of a category; rare exemplars are “worse” examples of a category
- accounts for within-category variance

Question 7

What are criticisms of exemplar theory?

Answer 7

1. ET assumes that individual instances are stored w/o blending into one another: simply not the case; we often confuse two items for the same item, or remember an item as having the features of a different item
2. ET has difficulty accounting for people’s ability to extract general properties of categories to classify new instances: sometimes we think of general features of a category rather than specific examples of a category

Question 8

Is exemplar theory or prototype theory more accurate?

Answer 8

We use both prototype and exemplar theory to categorise objects.
- prototypes carry more general knowledge of objects and more useful descriptions of categories as a whole
- exemplars carry more details of objects
- Every concept we possess is a mix of exemplars and prototypes stored in our mind
- Young children likely remember only exemplars, however as we age and gain more experiences, we may extract common features from exemplars and average instances to create prototypes

Question 9

What are some criticisms of probabilistic theories?

Answer 9

1. Probabilistic theories rely on similarity, which is always relative: there are an infinite # of ways that two things can be similar; we make judgements about similarity on a case-by-case basis
   —similarity can be along multiple dimensions
   —similarity judgements require some assumptions that aren’t always met (e.g. “this painting resembles the original” is a directional statement)
2. Categories can be context dependent: the relevant features of an object can depend on the context that they are presented in
3. Similarity does not explain why two concepts fall into the same category: our beliefs of essential properties of objects strongly influence our judgements of category membership