Semantic Memory: Collins

1
Semantic Memory Collins Quillian

• Kandise G. Viar
• Senior Seminar
• February 26, 2008

2
Semantic Memory

• Semantic Memory is knowledge of facts and general
  knowledge of the sort learned in school.
• Component of long-term memory.

3
Semantic Networks

• A semantic network is a network model in which
  each node has a specific meaning.
• Semantic networks therefore employ the local
  representation of concepts.
• Adopted by cognitive psychologists as a way to
  explain the organizational and retrieval of
  information in long-term memory.

4
Characteristics of Semantic Networks

• A nodes activity can spread outward along links
  to activate other nodes. These nodes activate
  others and so on
• This is called spreading activation.
• Spreading activation may lose strength as it
  travels outward from its point of origin.
• Energy decreases with increasing distance because
  in theory, energy encounters resistance as it
  passes through succeeding links and nodes.
• Distance between two nodes is determined by their
  relatedness.
• Ex Concepts such as shoe and high-heels are
  semantically related and therefore close in
  proximity to each other. Whereas, shoe and
  flower are not closely related and therefore
  further away from each other in the semantic
  network.

5
Characteristics Continued

• Priming
• The process of a stimulus is facilitated by the
  networks prior exposure to a related stimulus.
• Ex If we are asked to think about the word
  doctor then we automatically think about things
  that associated with the word doctor like
  hospital or drugs or nurse.

6
A Hierarchical Semantic Network Collins
Quillian (1969)

• Suggested that semantic networks have a
  hierarchical organization, with different levels
  representing concepts ranging from the most
  abstract (broad) to down to the most concrete
  (specific).

7
Collins and Quillians Method

• Used a sentence verification task.
• Participants were asked to respond to individual
  sentences that appeared on a computer screen.
• For example, A canary is a bird and A canary
  is an animal.
• If the participant judged that the sentence was
  true then he or she was instructed to hit one
  button.
• If he or she judged the sentence false then
  another button was to be pressed.
• Participant response times were recorded.

8
Collins Quillians Theory

• Theorized that a correct response to one of these
  sentences required the overlap of spreading
  activation.
• If the nodes have a close semantic relation, they
  should be in close proximity within the network.
• To know whether a canary is a bird would require
  the activation of both a canary node and a
  bird node. Recognition of both concepts would
  activate these nodes. Activation would then
  radiate outward through the network from each
  node until each individual units activation
  would mutually affect one another. When the
  activation of these two nodes overlaps then the
  participant can confirm that these two things are
  related. They can confirm the sentence.
• Responses will be fast because spreading
  activation will have less distance to cover.
• If the nodes are less related, the distance
  between them will be greater and response times
  will be longer.

9
Results

• Based on the reaction times obtained from the
  study Collins and Quillian (1969) came up with a
  hypothetical memory structure for the knowledge
  of animals.

10
Hierarchical Model
11
Hierarchical Model Explained Superordinate,
Ordinate, and Subordinate.

• The idea of an animal is abstract (broad) and
  therefore encompasses all kinds of animals.
  Animal is a superordinate category.
• At the next level there are classes of animals
  birds, cats, dogs, fish. These are placed in the
  ordinate category.
• The subordinate category is the most concrete
  (specific) and corresponds to the exact species
  of an animal. Properties of the species are also
  at this level.
• Ex A node corresponding to canary would also
  have links to the superordinate level connecting
  it to is yellow and can sing.

12
Evaluation of the Hierarchical Model

• The model is intuitively pleasing but it fails in
  practice because some concepts are more common
  than others.
• Ex A robin is a more common bird in comparison
  to a flamingo. People reacted faster to a
  sentence like, A robin is a bird vs. A
  flamingo is a bird.
• We have prototypes or idealized versions about
  what a bird is.

13
Evaluation Continued

• Participants are faster to make a correct
  response to, A dog is an animal than to A dog
  is a mammal.
• This happens even though animal is farther away
  from dog in this hierarchical scheme.

14
Evaluation Continued

• Collins Quillian attempted to preserve the
  principle of cognitive economy. This principle
  states that nodes should not have to be coded for
  more times than necessary. They only placed
  property nodes at their appropriate levels in the
  hierarchy.
• Ex Can fly is positioned as a property of
  birds only, even though it could be linked to the
  canary node and to the node representing every
  other kind of bird as well.
• The relation is implied.
• Research has found no difference in response time
  for the sentences, A bird can fly and A canary
  can fly.

15
References

• Friedenburg, J. Silverman G. (2006) Cognitive
  science An introduction to the study of the
  mind. Thousand Oaks, California Sage
  Publications. (P. 131, 224-230)
• http//psych.rice.edu/mmtbn/language/wordComp/memo
  ry.gif (image)