Exam 3 (Concepts & Semantic Memory) Flashcards
concept
mental representation of a class or category
exemplar
particular instance
prototype
average or typical example
semantic memory
lexical, conceptual, and encyclopedic knowledge of the world
What do concepts allow us to do? (4)
- allow classification
- enable generalization - ex. exotic bird has eyes
- support reasoning and predictions
- support communication - ex. find “chair”
What are the two approaches to concept formation?
- Theory of Subtractive Abstraction
2. Probabilistic View
Theory of Subtractive Abstraction
ignore insignificant differences, find what they have in common
Classical View
research on artificial concepts (e.g. things that are blue and round)
What are problems with the theory of subtractive abstraction?
- properties relative (e.g. tall)
- we have concepts of things we’ve never experienced (e.g. aliens)
- typicality effects - easier to classify something (ex. sparrow -> bat)
- any truly common properties?
Probabilistic view
concepts organized around properties that are typical of members
Theory of Resemblance
all members often have no properties in common - they share a “family resemblance”
e.g. games: card games, board games, video games
What are the models of semantic memory? (2)
- Prototype approach
2. Exemplar Approach
Prototype Approach
based on resemblance to a prototype
e.g. dog = average dog
The prototype approach allows for the ___ ___ & fuzzy ___
internal structure, boundaries
What are the problems with prototype approach?
- category size
- variability
- correlations of attributes - e.g. small dog, high-pitched bark
- memories for individual examples
Exemplar approach
based on resemblance to a collection of examples
e. g. dog = thing that resembles 1 or more dogs
- German shepard more typical than chihuahua
Natural concepts are based on __ not defining features
resemblance
Graded structure
some members are “better” (more representative)
e.g. furniture
Levels of categorization of resemblance and natural concepts?
- basic-level (e.g. pen)
- subordinate (BIC fine-point pen)
- superordinate (writing instrument
fuzzy boundaries
membership ill-defined or uncertain
What evidence supports resemblance models?
- typicality
- semantic priming
- induction
- pseudo-memory
- network models
Typicality effect
- T/F judgments (does a tiger have stripes)
- typical exemplars (fruit = apple)
semantic priming effect
just by saying “bird” or “feathers”, can reduce RT to sentence verification test
Induction
people often predict that all instances will have a property if a typical instance does
- e.g. bird infected, will infect all species
Network Models
concepts formed by a network of interconnections
interconnections
spread of activation effects including semantic priming
Parallel Distributed Processing (PDP)
info stored as network of interconnected nodes
- connections vary in weight (e.g. red more connected with cherries than firetrucks)
3 properties of PDP
- graceful degradation
- retrieval from incomplete information
- default assignment
Graceful degradation
errors or missing info do not lead to complete failures
e.g. tip-of-tongue
retrieval from incomplete info
if a set of features is sufficient to uniquely characterize an “item” it will activate that node
default assignment
educated guessing
