Categorisation

Question 1

what is a category?

Answer 1

• A category defines a set of instances that belong together as a group (e.g., bird, animal).
• Defines the shared attributes of those instances.
• Psychologically we are able to learn about and store a representation of a category in memory.
• Representations of categories are typically referred to as concepts.

Question 2

what are the functions of concepts?

Answer 2

• Classification:
  - Concepts allow us to treat discriminably different things as equivalent.
• Understanding and prediction
  - Anticipate the possible characteristics that will be displayed.
  - Initiate appropriate behaviour to novel instances.
• Communication
  - Knowledge about categories facilitates communication in society.

Question 3

how do we categorise?

Answer 3

• Now we know why, let’s look at how
• How do theorists believe this occurs for human cognition?
• Key questions:
  - Are all instances of a category stored in memory?
  - How are the relationships between instances encoded?
  - Do we store category specific knowledge?

Question 4

what is the classical view?

Answer 4

• Bruner, Goodnow & Austin (1956)
• Mental representations of categories consist of defining features that are necessary for category membership.
• Instances of a category are either in, or they are not.
• There are no boundary instances, or instances between categories.
• Thus, all category members are equivalent members.
• Classical view: not particularly well specified in terms of process…the concept is represented by the defining features.
• Do we evaluate each instance against the definition each time we categorise?

Question 5

what are the failings of the classical view?

Answer 5

• 1. Typicality effects
     
     • Intuitively (and shown experimentally) some instances are more typical category members.
     • E.g., Bach vs. Paris (and Sparrow vs. Penguin).
• 2. Unclear cases
     
     • There is often ambiguity surrounding cases.
       e.g., “sport”
• 3. Failure to specify defining features.
     
     • Triangle quite easy
     • But, Bird? Musical instrument?
• If we struggle to come up with the defining features when we stop and think about it, is this really a plausible theory of natural concept formation?

Question 6

what did McCloskey & Glucksberg 1978?

Answer 6

• Participants asked to determined category membership for items that varied in typicality.
  Participants performed task again after 1 month.
• Typicality correlates with category assignment
• Greatest variation in retest responses for instances of intermediate typicality.
• The results (and intuition) tell us that categories have “fuzzy boundaries”
• Categories are ill-defined
• So what defines typicality?

Question 7

what did Rosch & Mervis 1975 study?

Answer 7

• What makes something typical?
• Frequency of occurrence?
  - Sparrows encountered more often than penguins.
  - But, not always the case
• The more attributes a member has in common with other members, the more it will be considered a good/typical/representative member of the category.
• Think Robin, Sparrow, in contrast to Penguin.
• Evidence in favour of fuzzy categories and against the classical view (set definitions).
• “…members of a category come to be viewed as prototypical of the category as a whole in proportion to the extent to which they bear a family resemblance to (have the attributes which overlap those of) other members of the category. Conversely, items viewed as most prototypical of one category will be those with least family resemblance to or membership in other categories”

Question 8

what is prototype theory?

Answer 8

• This “family resemblance” principle suggests a summary representation is formed.
• Might reflect an “ideal” (average) member of the category.
• This is termed the Prototype.
• As more instances are experienced, the prototype changes.
• The prototype reflects the CENTRAL TENDENCY, the AVERAGE
• Gets around the problem of defining features and allows for fuzzy boundaries.

Question 9

what is the evidence for prototype of abstraction?

Answer 9

• Posner & Keele (1970)
• Participants trained with variations, but NOT the prototype
• Initial data:
  - good performance on exemplars AND prototype
  - Small distortions from prototype learned better than large distortions.
• Data after 1 week:
  impaired performance on exemplars,
  - Relatively good performance on prototype!
• Suggests learning involves representation of prototype abstracted during the training phase?
• Natural categories lead to strange prototypes…

Question 10

how do we define prototypes?

Answer 10

• What is the prototype?
• Can we have a single representation of category “bird”?
  
  • a bird that flies and doesn’t fly?!
• Rather than a “best item”, perhaps more likely is a representation of ideal features.
• Reflecting family resemblance scores

Question 11

what is exemplar theory?

Answer 11

• The alternative to prototype theory is that participants store each and every exemplar.
• Classification is based on assessing how similar the probe item is to each stored item.
• In exemplar theory, decisions are made on basis of computed similarity
• Exemplar theory easily explains typicality and fuzzy boundaries.

Question 12

what is Nosofsky’s generalised context model?

Answer 12

• The GCM (see Nosofsky, 1984,1986) is the most widely cited and tested model of exemplar theory.
• Making a decision
  - Probability that exemplar i is a member of category J:
  - What is the similarity of i to all members of category J, compared to similarity of i to members in all categories.
  refer slides 29-32 for equation
• Prototype is the central tendency of the exemplars in the category
• Similarity of probes are computed with respect to prototype
• The prototype model can be defined in similar terms.
• Here though, we are assessing similarity (calculated in the same way as before), against the prototypical values of the category.

Question 13

what is the evidence for prototype?

Answer 13

• Smith & Minda (2002).
  - Imagine a stimulus can be defined by just two dimensions (e.g., height & width).
• Key predictions:
  - Prototype theory predicts as the probe moves from 8-1, it becomes more similar to the category representation (the prototype).
  
  • Exemplar theory predicts that as the probe moves from 8-3 it becomes more similar to the category representation (the average of the exemplars)
  • However, Exemplar theory also predicts that as the probe moves from 3-1, it will not increase in similarity to the category representation.
• Result:
  - As the probe item became more similar to the prototype (levels 3-1), participants were more likely to endorse as a member of the category (a prototype enhancement effect).
  
  • This uniquely supports the prototype theory of categorisation.