Network Models of Memory
(This page last updated 1 November, 2005.)
Basic Components of Network Models
- Nodes representing individual ideas
- Each node has an activation.
- Associative links or associations connecting nodes to one
another.
- When the activation of a node rises about a threshold:
- the represented idea becomes active in working memory.
- activation spreads across links to other nodes.
- Spreading Activation
- Once one idea is active, spreading activation allows
related ideas to be retrieved and become active.
Aspects of Memory that can be Explained by Network Models
- semantic priming
- Meyer & Schaneveldt (1971)
- Lexical decision faster if related word was heard
recently.
- Spreading activation reaches word node before word is
presented, so less activation is needed to bring it above
threshold and trigger response.
- sentence verification
- Collins & Quillian (1969)
- Faster to verify "Canary is bird" than "Canary is
animal."
- Assume network is organized hierarchically
- Response time depends on time necessary to travel from node
to node.
- fan effect
- Anderson (1974)
- More time needed to verify sentences when its components
are also parts of other sentences.
- When a node has multiple connections, there is less
activation to flow out of each connection.
Increasing Representational Power of Memory Networks
- Label links to represent different types of associations
- Isa ("Is a")
- Hasa ("Has a")
- Other types of links make it possible to represent specific
episodes or events.
- Propositional Networks
- Proposition: smallest unit of knowledge with a truth
value.
- Node represent entire proposition
- Anderson represents this node with an oval.
- Labeled links to other nodes
- Agent: the one performing the action
- Object: the thing being acted upon
- Relation: the relationship between agent and
object
next class: Concepts
Psych 315H: Cognitive
Psychology 
Kyle Cave
Psychology Dept.
U.
Mass.