Knowledge

1
Knowledge

• AIMS OBJECTIVES
• The aim of this lecture is to review current
  approaches to understanding knowledge
• At the end of the lecture you will have learned
• What is meant by the term propositional knowledge
• The difference between a category and a prototype
• Typicality effects
• Taxonomic categories

• CORE READING
• Parkin, A. (2000). Essential Cognitive
  Psychology. Psychology Press, Chapter 8.
• SUPPLEMENTARY READING
• Larochelle S, et al., (2000). What some effects
  might not be The time to verify membership in
  "well-defined'' categories. Quarterly Journal of
  Experimental Psychology-A, 53(4), 929-961.

2
Face recognition
Inferences If x and y then z If 4 legs and
barksdog
3
What is knowledge?

• External representations
• Turtle
• Oldest living creature
• Buries eggs etc..

• Internal representations
• Store of facts (LTM).
• Allows language etc..
• Also, allows us to make inferences about the
  external environment.
• This enables us to solve novel problems (145)
• Adaptive and productive.

4
hierarchy
rule based
automatic
Abstract, verbal statements
Literal, spatial relations
15.00
gillsscalestailfin
5
Exemplars and categories

• We do not encode all objects encountered in
  everyday events.
• We categorise each object as an exemplar of a
  type on the basis that it shares features e.g.
  a turtle
• an amphibious creature
• long living
• slower than a hare.

• Abstract common features form a category.
• A category is a class or description of objects
  or events with common attributes and the members
  of categories are called instances.
• People tend to think in taxonomic categories
• Furniture, vegetable..

6
ambiguous
7
Categories and concepts
Literal perceptual boundaries

• Concepts are categories of mental representations
  stored in LTM memory.
• Concepts are used to perceive, store, act on and
  communicate about objects and events -gt false
  memories.
• How are conceptual categories represented in the
  human mind?

Abstract independent of form
8
Defining feature theory

• Concepts defined by necessary and sufficient
  attributes.
• An instance is a member of a conceptual category
  only if it meets minimal properties e.g., a
  triangle
• a two dimensional geometric figure
• with three straight sides joined at their ends
• with angles adding up to 180o

• Prediction DFT assumes clear boundaries between
  members and non-members of a category.
• All members of category should be equally
  representative of the conceptual category.
• All people should represent categories in the
  same way.
• BUT category boundaries are not clearly defined
  or discrete.
• Typicality effects.

9
Predictions of DFT
reptiles, Crawling animals
fish gillsscalestailfin
amphibious creatures
10
Typicality effects
More typical fish
11
Fuzzy boundaries
A fish?
Fruit or vegetable?
12
Eleanor Rosch

• Typicality effects (1975, JEP, 104, 192-233).
• People tend to decide that some instances are
  better members than others.
• Most typical furniture
• chair
• vase
• drapes

• Instances are not literally defined as members of
  a category on basis of logic.
• Instances have an internal abstract relative
  structure.
• This structure needs an organising principle.

13
Roschs prototype theory

• Conceptual categories are represented by a
  prototype.
• A prototype is a composite or abstraction of
  features.
• This combines all of the characteristics of the
  most typical members of a category (fruit -
  seeds, edible).

• Categorisation tasks
• Is this a fish?
• This is based on overall similarity of an
  exemplar to the abstract prototype rather than on
  the features of the items themselves.
• This theory explains the typicality effect
  because not all instances precisely resemble the
  prototype.

14
Network Theory

• How are representations of knowledge arranged in
  LTM
• Collins and Quillian (1969) taxonomies.
• Knowledge about biological forms is organised in
  a hierarchy.
• General concepts are at the apex and specific
  concepts are at the base.

• The superordinate level is the most inclusive
  category e.g., animal.
• The superordinate category subsumes more specific
  subordinate categories e.g., category animals
• birds, fish etc
• canary, ostrich, shark, salmon

15
Network Theory

• Individual concepts are called nodes and these
  are represented by definitions based on a set of
  properties.
• Subordinate categories can inherit the
  properties of a superordinate category.
• This makes categories very economical and allows
  us to draw inferences.

16
defining features
nodes
This is a type of distributed network
17
Prediction

• Sentence verification YES/NO for semantic
  category decisions should be a function of the
  number of levels that need to be passed to answer
  a question.
• Category Property Levels
• A canary is a canary can sing 0
• A canary is a bird can fly 1
• A canary is an animal has flesh 2
• A canary is a fish has gills false

18
Main effect of number of levels concept
structure
19
Neuropsychological evidence
Temporal lobe
20
Category specific agnosia
animal

• Warrington and Shallice (1984) reported a patient
  JBR who had visual agnosia.
• JBR had a selective deficit when asked to name
  pictures from the semantic category living things
  (e.g., animals) but no impairment with non-living
  things and patient YOT shows the reverse pattern.

21
(No Transcript)
22
Neuropsychological evidence

• Patients with dementia seem to lose subordinate
  information (e.g. canary) before superordinate
  name (e.g. animal) on tests of concept knowledge.
• Data consistent with idea of hierarchical
  knowledge in semantic memory that supports
  Collins model.

• Evidence from brain imaging suggests category
  specific regions in the brain (e.g.
  Thompson-Schill, 1999).

23
Thompson-Schill, et al (1999)
Living things
Non-living things
24
(No Transcript)
25
Summary

• Prototype theory assumes categories are
  represented by exemplars and there is some
  evidence in support.
• Neuropsychological data suggests knowledge about
  the world is represented in taxonomic categories
  in a distributed brain network.