Semantic Memory

Question 1

Hierarchical network model

Answer 1

Collins & Quillan (1969): allows inferals and rapid decision making about the truth of related sentences.
- They predicted that a longer response time would be observed in a semantic variation task if the 2 concepts (or concept and property) are differentiated from each other by a level (and even more if they were seperated by 2). An example of levels are animal - bird - canary, properties can only be accessed by first thinking of a concept category - the properties are only represented once at the highest appropriate level - cognitive economy.
It explained priming as well as most results in semantic verification task

Question 2

Semantic verification task

1. ) Canary can sing
2. ) Canary has skin (also true for categories)
3. ) Pig to mammal/animal
4. ) Bird is a table
5. ) Typical/representative member of a category

Answer 2

1. ) 1300msecs
2. ) 1470msecs
3. ) Inconsistency - participants quicker naming a pig as an animal than mammal (Rips et al (1973)
4. ) Part. were very quick at naming false sentences even if hypothetically distant
5. ) Quick reaction time and therefore led to feature models

Question 3

Feature models

Answer 3

Rips et al (1974): speed of reaction in semantic verification task now depends on the amount of matching required.

2 stages hypothesised: Compare all features and if only few overlaps then look for distinctive features (core/defining features of a concept)

Concept is a group of semantic features.

Question 4

Problems with the Feature Model (Rips et al (1974))

Answer 4

No cognitive economy - no explanation of how the distinctive features are extracted - and whether an object can be missing some distinctive features and still be a member of a concept.

Question 5

Spreading activation model and problems

Answer 5

Collins and Loftus (1975): whenever a person thinks about a concept, the approprie node is activated - spreads to related concepts (first strongly related and then more weakly connected concepts)

The lengths of the links in the model represented the relatedness of concept, however there was no rules to determine them. They were measured by asking people how close they perceived them to be related or simply to ask them to name members of categories and the amount of times a concept was placed within a category determined the closeness of the relationship. This leads to a problem of individual differences.

This theory is not falsifiable, however does not carry much explanatory power.
It is still widely used.

Question 6

PDP model

Answer 6

McClelland & Rumelhart (1985): This feedforward model assumes that semantic information is not stored but reconstructed in response to probes - pattern completion.

The model assumes that processing takes place through interaction of large numbers of simple processing elements whis send ex/in signals.

Question 7

Scripts

Answer 7

Bower et al (1979): script and tag - don’t have to remember all that happened - only the deviations from the norm. They showed that expectations had been formed through asking participants about visiting a restaurant.

Bertson & Rubin (2004) - Life Script

Question 8

Schema - reconstruction and help with understanding

Answer 8

Well integrated framework about the world, events, people and actions.

Bartlett (1932) - folktale.

Anderson and Pritchett (1978 - certain schema activated - affects details recalled.

Brewer and Treyens (1981) - graduates office.

Bransford & Johnson (1972) - learning context before passage allows semantic memory to form associations which aids future recall.