organization in semantic memory Flashcards
How is knowledge represented and organized in semantic memory?
3 Models…
Hierarchical Network Model (Collins & Quillian, 1969)
Spreading Activation Model (Collins & Loftus, 1975)
Feature comparison model (Smith et al., 1974)
Knowledge is…
information about the world that is stored in memory, ranging from the everyday to the formal.
Knowledge enables …………….., the ability to establish that a perceived entity belongs to a particular group of things that share key characteristics.
categorization
An early approach to knowledge representation centred on the idea of….
defining attributes (features).
objects can be grouped together in terms of certain attributes (features) that are common to all of them in order to form a concept.
what is this idea centered on?
defining attributes (features).
who put forward a hierarchical network model of how concepts are represented by grouping defining features.
Collins and Quillian (1969)
Semantic memory is organized into series of hierarchical networks:
concepts that are connected with each other in a web-like form.
The Hierarchical network model (Collins & Quillian, 1969) is an example of what network representation, and what are the three heirarchies>
semantic network representation.
Superordinate
Subordinate
Sub-subordinate
Hierarchical network model
Nodes:
correspond to concepts (mental representations of objects, events and ideas)
Hierarchical network model
links:
correspond to relations among concepts and properties of concepts
Hierarchical network model
hierarchy:
general concepts at top and specific concepts at bottom
Hierarchical network model
The relationship between the hierarchical levels:
superordinate:
subordinate:
coordinate:
superordinate: at a higher level in the hierarchy
subordinate: at a lower level in the hierarchy
coordinate: at the same level in the hierarchy
Hierarchical network model
propositional links:
property links:
propositional links: specify superordinate-subordinate relations among concepts (“is a”).
property links: specify properties of concepts, including “has” and “can”.
Hierarchical network model
cognitive economy:
cognitive economy: properties are stored only once at the highest possible node; nodes inherit the properties of superordinate nodes
Hierarchical network model
Nodes (concepts):
living thing, plant, animal, tree, bird, etc.
Hierarchical network model
Links (propositional):
‘is a’ – Links (property): has skin, can swim, etc.
Hierarchical network model
How to test the model?
Sentence verification task:
Participants are presented with statements about concepts or properties
- A canary is yellow (True)
- A canary can fly (True)
- A whale is a fish (False)
Participants report as fast as possible whether the statements are true or false.
Problems with the hierarchical network model:
Familiarity: Controlling for familiarity greatly reduces the hierarchical distance effect.
E.g., longer to verify “ A chicken is a bird” than “ A chicken is an animal”.
Typicality effect: Verification is faster for more representative member categories, independent of hierarchical distance (Rips, Shoben, & Smith, 1973).A robin is a bird” is faster to verify than “An ostrich is a bird”.
Computed production frequency
Conrad (1972)
i.e., how often participants reported a property when probed with an instance name (e.g., how often ‘curly tail’ is produced when probed with ‘pig’).
Hierarchical network model can provide….
straightforward predictions towards response times in sentence verification tasks.
Hierarchical network model
The model failed to predict how people
performed in certain sentence verification tasks, especially the familiarity and typicality effects.
Spreading Activation Model (Collins & Loftus, 1975)
Assumes semantic memory is organized by semantic relatedness:
web of interconnected nodes (concepts) rather than a network of hierarchy.
Links between units of information can vary in length: more strongly associated (similar) concepts are connected via shorter links.
Shorter response times to shorter links.
Spreading activation model
The length of the link between two concepts can….
define their relatedness.
Spreading activation model
Spreading activation:
A node is activated when one thinks of a concept.
Activation spreads to related concepts.
Spreading activation decreases as the distance from the original node is increased.
Spreading activation can account for the priming effect.
Spreading activation model
Semantic priming: A semantically-related word facilitates the processing/identification of a target word.
Spreading activation model
Lexical decision task (Meyer & Schvaneveldt, 1971):
Participants are presented with a pair of words
The spreading activation model is more flexible than the hierarchical network model.
tue or false
true
The spreading activation model can account for more….
empirical findings (typicality effect, priming effect).
spreading activation model
The flexibility also reduces the specificity of
of the model’s predictions, making the spreading activation model more difficult to test.
Feature comparison model (Smith et al., 1974)
Concepts are represented as lists of features:
defining features and characteristic features
A defining feature is
one possessed by all members of a category (e.g. all birds have feathers).
Characteristic features are
are attributes possessed by most members of a category (e.g., Birds can fly).
Feature comparison model (Smith et al., 1974)
Problems:
The model explains how we verify statements but does not account for how we extract the meaning of what we see or hear.
“a canary is a bird” and “a bird is a canary” would both be verified as they yield the same feature overlap. But “a bird is a canary” is not true!
The model cannot explain how we verify statements involving concepts that have no featural overlap, e.g. “the man has a brick”.
Hierarchical network model (Collins & Quillian, 1969)
summarise
Accounts for most of sentence verification task results
Fails to predict typicality and familiarity effects
Spreading Activation Model (Collins & Loftus, 1975)
summarise
Explains typicality and priming effects
Difficult to test due to great level of flexibility of the model
Feature comparison model (Smith et al., 1974)
summarise
Limited to verifying certain statements and does not offer a general theory of how meaning is represented.
