View
215
Download
0
Embed Size (px)
Citation preview
Knowledge
• information that is gained and retained
• what someone has acquired and learned
• organized in some way into our memory
Semantic Organization
• put items that are related in some way into a cluster or a group.
• Cognitive Models - assume that detailed congitive structures represent the way semantic info is organized in memory
Semantic Memory: Cognitive Models
• Set-theoretical model
• semantic feature-comparison model
• network models
• propositional networks
How to study semantic memory
• Association Tasks:– Free association: Used by Freud to study
personality, but may tell us more about the structure of knowledge.
– Category association: People are asked to give associates to a category name.
• fruit: _________
• fruit: a ________
How to study semantic memory• Tip of the tongue (TOT):
– A sensation we have when we are confident we know a word we are searching for, but we are unable to recall it
– Brown & McNeill (1966) research 1.read definitions of infrequent words
2.subjects asked to raise hands when they had a TOT
3.subjects then asked:
What is a similar word? What does the word sound like?
How many syllables? What is the word’s first letter? 4.Results: subjects often could supply partial information
How to study semantic memory• Sentence verification task:
Present sentence: "Is a robin a bird?"
Measure RT to correctly respond
• Category verification task:
bird-robin ("yes")
bird-tree ("no")
Measure RT to correctly respond
How to study semantic memory• Lexical Decision (word/non word) Task:
Present a word (brain) or a non-word (shup).
Ask subjects to decide, as quickly as possible, if the item is a word.
RT tells us how long it takes subjects to search their mental dictionary.
Set-theoretical model
• Concepts in memory are collections (sets) of info.
• Sets include:– instances of a category
• category car has instances of Volkswagon, Saab, Mercedes,…
– attributes or properties of a category • category car has properties of tires, engine, trunk,
metal, windshield…
Set-theoretical model
• Retrieval is a function of verification– must search through 2 or more “sets” to find
overlapping information– more overlap = quicker decisions
Feature Comparison Model
• Basic Assumptions– Concepts are represented as a set of features,
similar to Set-Theoretical model– unlike previous model, differentiates between:
1. Defining features (essential components)
2. Characteristic features (accidental, not always present)
– verification is based more on defining features
Feature Comparison Model
Feature Comparison Model
• Features are ordered according to "definingness"
characteristic features defining features
birds fly birds have wings
birds sing birds have feathers
• Relations between concepts computed based on shared features
Two stage decision model of sentence verification:
Feature Comparison Model
Predictions:
1. Category size effect: A robin is a bird. vs. A robin is an animal.A dog is mammal. vs. A dog is an animal.
2. Typicality effects A robin is a bird. vs. A penguin is a bird.
3. Quick rejection of false sentences: A bat is a bird vs. A pencil is a bird
Feature Comparison Model
• Problems:
1.Defining Features?
2.Semantic Priming?
3.Quick rejection of false sentences?
people are trees
a bat is a bird
a dog is a cat
Network Models
• Hierarchical Network Model -Collins and Quillian - early work
• Spreading Activation Theory - Collins and Loftus
Hierarchical-Network Model
• Representational Assumptions– hierarchically organization of concepts– cognitive economy: properties are stored at the
most general, or highest level possible.
• Processing Assumptions:– intersection search: enter the network at two
concepts, and search for a connection. – type of connection determines yes/no response
Hierarchical-Network Model
Hierarchical-Network Model
• Tests of the model:– Category-Size Effect:
compare: A robin is a bird.
to: A robin is an animal.
– Cognitive Economy: compare: A bird has feathers
to: A bird has skin.
Hierarchical-Network Model
Hierarchical-Network Model
• Challenges to the Hierarchical Assumption: 1) reversals of the category size effect
A dog is a mammal vs. A dog is an animal.
2) typicality effects: A robin is a bird. vs. An ostrich is a bird.
• Challenges to Cognitive Economy
• Negative sentence RT’s not predicted by the model
Spreading Activation
• New assumptions: 1.Not hierarchical: length of links represent
degree of relatedness. Search time depends on link length
2.Spreading Activation: retrieval (activation) of one of the links lead to partial activation of connected nodes. Degree of activation decreases with the distance.
3.Activation decreases with time.
Spreading Activation
Spreading Activation
• New predictions:– Typicality effects:
• A robin is a bird. vs. A chicken is a bird.
– Semantic Priming:
type of trial prime target RT
related prime bread butter 600
unrelated prime nurse butter 670
Propositional Network Models
• HAM and the representation of Knowledge (Human Associatve Memory)
• ACT (Adaptive Control of Thought