Cognitive - knowledge.ppt © 2001 Laura Snodgrass, Ph.D.1 Knowledge Structure of semantic memory –relationships among concepts –organization of memory –memory search is dependent upon the organization Many specific theory but three general categories –feature models –prototype models –network models
Cognitive - knowledge.ppt © 2001 Laura Snodgrass, Ph.D.2 Knowledge Characteristics of memory to be explained by theories of knowledge structure –meaning is the basis for organization –recall is abstractive and integrative –recall is constructive –specific research findings
Cognitive - knowledge.ppt © 2001 Laura Snodgrass, Ph.D.3 Meaning Is The Basis For Organization Mnemonics set up meaningful relationships Meaning is a cue in memory search Gist over verbatim memory Clustering in free recall Semantic confusions in recall and recognition Free associations based on meaning Lexical decision tasks and semantic priming
Cognitive - knowledge.ppt © 2001 Laura Snodgrass, Ph.D.4 Recall Is Abstractive And Integrative Gist memory integrate across sentences to make inferences –inferences are recalled better than actual sentences –advertising prototypes are averages across instances
Cognitive - knowledge.ppt © 2001 Laura Snodgrass, Ph.D.5 Recall Is Constructive Episodic memory Schemas influence recall
Cognitive - knowledge.ppt © 2001 Laura Snodgrass, Ph.D.6 Research Findings Relationships among concepts vary from weak to strong Sentence verification –category size effect we are faster to say a canary is a bird than an animal –semantic distance effect or typicality effect we are faster to say a horse is an animal than an elk is an animal
Cognitive - knowledge.ppt © 2001 Laura Snodgrass, Ph.D.7 Research Findings Inferences can be used to derive new information –“She slowly stirred her coffee as she sat talking to her sister.” –infer use of a spoon
Cognitive - knowledge.ppt © 2001 Laura Snodgrass, Ph.D.8 Feature Set Theory Knowledge of a concept –know set of features (attributes) Features –component parts e.g. cab, wheels, ladder, hoses –values on a dimension dimensions can be –perceptual (e.g. color = red) –functional (travel = rapid) –abstractions (feel = danger)
Cognitive - knowledge.ppt © 2001 Laura Snodgrass, Ph.D.9 Feature Set Theory Knowledge = set of all features Features can be divided into –defining - must have –characteristic - may have When we think of a concept all of its features are activated For a decision such as “Is a robin a bird?” matching is done in two stages
Cognitive - knowledge.ppt © 2001 Laura Snodgrass, Ph.D.10 Feature Set Theory Stage 1: –all features are compared if many in common - quick yes if few in common quick no if intermediate number in common go to stage 2 Stage 2 –compare only defining features Degree of relationship between concepts based on number of features in common
Cognitive - knowledge.ppt © 2001 Laura Snodgrass, Ph.D.11 Feature Set Theory
Cognitive - knowledge.ppt © 2001 Laura Snodgrass, Ph.D.12 Feature Set Theory Theory can explain –category size effect –semantic distance (typicality) –violations of category size “scotch is a drink” is faster than scotch is an alcoholic beverage –hedges - “a bat is a bird” –typicality ratings predict number of shared features Problem - what is a “defining” feature?
Cognitive - knowledge.ppt © 2001 Laura Snodgrass, Ph.D.13 Prototype Theory Rosch –prototype is an item most typical of a category –can be an idealized example –items in a category differ in their degree of prototypicality e.g. think of a professor –think of a typical vs. non-typical example –can have a graded structure with fuzzy boundaries can be loosely structured (e.g. games)
Cognitive - knowledge.ppt © 2001 Laura Snodgrass, Ph.D.14 Network Theories Quillian’s Teachable Language Comprehender (TLC) –hierarchical network –cognitive economy properties stored at highest node –questions active concepts, nodes and links –predict RT based on distance in network
Cognitive - knowledge.ppt © 2001 Laura Snodgrass, Ph.D.15 Hierarchical Network
Cognitive - knowledge.ppt © 2001 Laura Snodgrass, Ph.D.16 Hierarchical Network Explains –category size effect Problems –assumes knowledge of taxonomy no reason to assume logical order –frequency of co-occurrence not distance in the hierarchy explain difference in canary-sing versus canary skin –properties associated with concepts based on frequency not cognitive economy
Cognitive - knowledge.ppt © 2001 Laura Snodgrass, Ph.D.17 Spreading Activation Network organized by degree of relationship –use length of link to indicate degree of association Organization is highly idiosyncratic –depends on individual experience Activation of a concept spreads through the network –strength decreases with increasing distance Explains –typicality –priming
Cognitive - knowledge.ppt © 2001 Laura Snodgrass, Ph.D.18 Spreading Activation
Cognitive - knowledge.ppt © 2001 Laura Snodgrass, Ph.D.19 Propositional Network Propositions as basic unit –higher level than just concepts –knowledge has truth value –relationships among propositions basis for network –add time and place tags to form episodic memory Explains –inferences
Cognitive - knowledge.ppt © 2001 Laura Snodgrass, Ph.D.20 Propositional Network Experiment - Kintsch & Glass –sentences with same number of words but different number of propositions –e.g. The settler built the cabin by hand The horse stumbled and broke a leg –better recall when fewer propositions
Cognitive - knowledge.ppt © 2001 Laura Snodgrass, Ph.D.21 Propositional Network