Representation/organization in LTM

Typical empirical testing paradigm: propositional verification task – rt to rose is flower, vs. rose is plant.Set Theoretical Models: categorical – hierarchical storage. Along with items, feature list are stored. Decisions are made based on comparisons of feature lists – feature overlap. Support found in some vs. all comparisons and most sub vs. immediate super. Problem with atypical category members (whale is mammal vs. whale is animal).

Categorical/Hierarchical Model of Memory

• Some birds are robins vs. all birds are robins• Crow-bird; crow-animal whale-mammal; whale-

animal (atypical problem)

Representation/organization in LTM

• Semantic feature comparison model: two stage process – 1. General feature overlap; 2. (if necessary) comparison of defining features

• Characteristic vs. defining features• Generally solves atypical category member

problem• Problem raised: cognitive economy

Encode comparison

Compare all featuresLow overall similarity (rock is bird)

High overall similarity (robin is bird)

Intermediate similarity (ostrich is bird)

Compare just defining featuresMismatch Match

“No”“Yes”

Representation in LTM

• Network model (Collins & Qullian): retains categorical – hierarchical organization, but only unique features stored with each item. Items assumed to retain features of superordinate category members. Spreading activation key concept. Return of atypical category member problem.

• Modified semantic network model (Collins & Loftus). – removes hierarchical assumption. Distance represents semantic relatedness

Collins & Quillian Spreading activation model

Modified Spreading activation model

Representation in LTM

• Propositional network model: Memory composed of interconnect network of fundamental propositions

• Proposition: smallest unit of information that can be judged t/f; typically represents relationship between concepts.

ACT* Propositional Network Model: J.R. Anderson

• Susan gave Maria, the president of the club, a white cat.