Long Term Memory: Semantic

1
Long Term Memory Semantic

• Kimberley Clow
• kclow2_at_uwo.ca
• http//instruct.uwo.ca/psychology/130/

2
Outline

• Methods
• Explicit Measures
• Implicit Measures
• Theories
• Defining Features
• Prototype Models
• Probabilistic Models
• Network Models

3
Taxonomy
4
Episodic vs. Semantic Memory

• Episodic Memory
• Memory for specific events / episodes
• Where were you when you first heard about the
  attack on the world trade center?
• Semantic Memory
• Memory for general world knowledge
• What is the date of the attack on the world trade
  center?
• In what city was the world trade center located?
• What does the word trade mean?

5
Example of Semantic Information

• Concept of dogs
• Characteristics
• Has fur
• Has 4 legs
• Has a tail
• Barks
• Bites postal workers
• Types of dogs
• Dalmatians
• Poodles
• Terriers

6
Implicit vs. Explicit

• Measures of explicit memory are sensitive to how
  the information is processed / studied.
• Measures of implicit memory usually show
  facilitation regardless of how the information
  was processed / studied

7
Defining Features

• An item belongs to a category/concept if that
  item incorporates the concepts defining features
• Defining Features
• Essential features
• Necessary and jointly sufficient
• Boundaries between concepts are clear cut
• All members of a category are equally
  representative

8
Example
Bachelor
noun
(Human)
(Animal)
(Male)
lowest academic degree
(Male)
young knight serving under the standard
of another knight
who has never married
young seal without a mate during breeding time
9
Criticisms

• Many features are not absolutely necessary
• If a dog is hairless or loses a leg, it is still
  a dog
• Not all apples are sweet or red
• Not all categories have clearly marked boundaries
• What are the defining features of game?
• Research suggests that all members of a category
  are NOT represented equally

10
Typicality Effects

• The typicality of each as fruit (highest to
  lowest)
• Apple 1.3
• Plum 2.3
• Pineapple 2.3
• Strawberry 2.3
• Fig 4.7
• Olive 6.2

11
Explicit Tasks

• Typicality Ratings
• On a scale of 1-6, how typical of fruit is a(n)
• Apple?
• Olive?
• Banana?
• Pineapple?
• Similarity Ratings
• On a scale of 1 - 6, how similar is a(n)
• apple to a plum?
• plum to a lemon?
• apple to a lemon?
• olive to a plum?

12
From these types of ratings
13
Multidimensional Scaling
14
Typicality vs. Similarity

• Typicality ratings seem to reflect similarity
• Bird Robin Chicken
• Flies -
• Sings -
• Lays eggs
• Is small -
• Nests in trees -

15
And What About These Findings

• Some Strange Effects
• Minimality Violation
• Symmetry Violation
• Triangle Inequality

16
Prototype Theory

• A prototype is the best or ideal example of a
  concept
• Categorization is based on similarity between a
  specific instance (exemplar) and prototype

17
Feature List Models

• Membership in a category is based on
  characteristic and defining properties
• Some members have more characteristic properties
  than others
• Defining properties are not necessarily
  singularly necessary and jointly sufficient
• Something belongs to a category if it is similar
  to members of that category
• Category boundaries are fuzzy

18
Smiths Feature Overlap Model
19
How It Works
20
Implicit Tasks

• Sentence Verification Task
• Shown subject-predicate sentences
• A canary is a bird
• Tested different sentence types
• Set inclusion
• A canary is a bird (true)
• A whale is a fruit (false)
• Property-attribute
• A canary has feathers (true)
• A whale has seeds (false)

21
In a sentence verification task
22

• Feature Verification Task
• Shown a concept and attribute (feature)
• LEMON yellow
• Need to indicate whether the feature is ever true
  of the concept
• LEMON sour
• LEMON fruity
• LEMON hard
• Differences in speed indicate how semantic
  information is organized

23

• Priming
• Present two words
• First word called the prime
• Second word called the target
• Repetition Priming
• can be long-lasting (hours)
• Study TRUCK
• Test TRU__
• Semantic Priming
• short-lived (seconds)

24
Example
25
Priming Results
26
Tverskys Contrast Model

• Lemon Orange
• yellow orange
• oval round
• sour sweet
• trees trees
• citrus citrus
• -ade -ade
• navel
• Similarity a(3) - b(3) - c(4)
• Similarity (I,J) a(shared) - b(I but not J) -
  c(J but not I)

27
Criticisms

• What are characteristic vs. defining features is
  not well defined
• Not all concepts have defining characteristics
• Problem for Overlap Model
• Doesnt work too well for property comparisons
• ROBIN has wings (feature verification)
• Problem for Overlap Model
• Cannot account for effects of frequency and
  associative strength
• Problem for Overlap and Contrast Models

28
Collins QuillianAssociative Network Model
29
To Visualize Another Way
SUPERORDINATE
SUBORDINATE
30
Conrad (1972)

• People respond faster to high frequency
  associates
• Distance in hierarchical structure not as
  important as frequency of association

31
Collins Quillian
32
Collins Loftus (1975)

• Modifications
• Concepts are NOT organized as a hierarchy
• Explains lack of hierarchical findings
• Links vary in associative strength /
  accessibility
• Nodes that are closer together are higher in
  associative strength
• Explains typicality effects

33
Connectionist Networks

• Built upon the associative networks
• Distributed processing assumption
• Concept is represented as a pattern of
  distributed features
• Many units rather than one node
• These units are similar to neurons (or groups of
  neurons)
• If a unit is detected, it becomes activated and
  fires to connected units
• Connections between units have weights based on
  associative strength (and vary with experience)
• Positive weights increase activation of linked
  units
• Negative weights decrease activation of linked
  units