PSY 368 Human Memory

1
PSY 368 Human Memory

• Semantic Memory

2
Announcements

• Due date changes
• Data from Experiment 3 due April 9 (Mon, 1 week
  from today)
• Experiment 3 Report due April 16 (2 weeks from
  today)

3
Experiment 3

• Interaction of Episodic and Semantic Memory
  (Exp 3) (Download detailed instructions form
  Blackboard)
• Modification of Anderson, Bjork, Bjork (1994)
• (see Blackboard Media Library Optional Readings
  to download a pdf of this paper if you want to
  read more)
• Question Can the retrieval of some items impact
  the retrieval of others?
• e.g., Suppose that you are studying for a test.
  You decide to study half the material. Does
  studying half the material have an impact on the
  half of the material that you didnt study?

4
Experiment 3

• Interaction of Episodic and Semantic Memory
  (Exp 3) (Download detailed instructions form
  Blackboard)
• Stimuli 4 categories
• Drinks, Weapons, Fish, Fruits
• Six exemplars from each category
• Write out category and exemplar on index cards

Drink vodka

• The full list of 24 items is in the detailed
  instructions
• Subjects find 3 willing participants

Weapon sword
Fish trout
5
Experiment 3

• Interaction of Episodic and Semantic Memory
  (Exp 3) (Download detailed instructions form
  Blackboard)
• Procedure 4 phases
• Study phase subs will study all categories and
  exemplars
• Shuffle all of the cards, read Study phase 1
  instructions, present each card to subject for 3
  seconds in random order

Drink vodka
Weapon sword
Weapon sword
Drink vodka
Weapon sword
Weapon sword
Drink vodka
Weapon sword
Weapon sword
Drink vodka
Weapon sword
Weapon sword
Drink vodka
Weapon sword
Weapon sword
Drink vodka
Weapon sword
Weapon sword
Drink vodka
Weapon sword
Weapon sword
Drink vodka
Weapon sword
Fish trout
6
Experiment 3

• Interaction of Episodic and Semantic Memory
  (Exp 3) (Download detailed instructions form
  Blackboard)
• Procedure 4 phases
• Practice phase subs will attempt to remember
  some of the studied items (half from 2 of the
  categories) by coming up with exemplars with cues
  (category and first letter)
• Give practice phase recall sheet to subject, Read
  practice phase instructions to subject, give subs
  category and first letter (see ordered list in
  detailed instructions) and give them 15 secs to
  practice it before moving to next item
• drinks v, weapons s, drinks r,
  weapons r, drinks g, weapons t

7
Experiment 3

• Interaction of Episodic and Semantic Memory
  (Exp 3) (Download detailed instructions form
  Blackboard)
• Procedure 4 phases
• Distractor phase complete a city generation task
• Read distractor phase instructions, Give
  distractor US Cities Task sheet

8
Experiment 3

• Interaction of Episodic and Semantic Memory
  (Exp 3) (Download detailed instructions form
  Blackboard)
• Procedure 4 phases
• Test phase free recall of all studied items (by
  category)
• Read test phase instructions, give recall test
  response sheets (1 for each of the 4 categories)
• Give 30 seconds for recall for each category

9
Experiment 3

• Interaction of Episodic and Semantic Memory
  (Exp 3) (Download detailed instructions form
  Blackboard)
• Scoring

Subject 1 Data
Practiced recalled
(divide by 6)
Non-practiced recalled
(divide by 6)
Control recalled
(divide by 12)
Sample data
Banana Orange Lemon Tomato Club Sword Bomb Guppy T
rout Ale Rum Vodka Beer
10
Experiment 3

• Interaction of Episodic and Semantic Memory
  (Exp 3) (Download detailed instructions form
  Blackboard)
• Scoring

Subject 1 Data
Practiced recalled 3
(divide by 6) 3/6 50
Non-practiced recalled
(divide by 6)
Control recalled
(divide by 12)
Sample data
Banana Orange Lemon Tomato Club Sword Bomb Guppy T
rout Ale Rum Vodka Beer
drinks v, weapons s, drinks r,
weapons r, drinks g, weapons t
11
Experiment 3

• Interaction of Episodic and Semantic Memory
  (Exp 3) (Download detailed instructions form
  Blackboard)
• Scoring

Subject 1 Data
Practiced recalled 3
(divide by 6) 3/6 50
Non-practiced recalled 3
(divide by 6) 3/6 50
Control recalled
(divide by 12)
Sample data
Banana Orange Lemon Tomato Club Sword Bomb Guppy T
rout Ale Rum Vodka Beer
drinks v, weapons s, drinks r,
weapons r, drinks g, weapons t
12
Experiment 3

• Interaction of Episodic and Semantic Memory
  (Exp 3) (Download detailed instructions form
  Blackboard)
• Scoring

Subject 1 Data
Practiced recalled 3
(divide by 6) 3/6 50
Non-practiced recalled 3
(divide by 6) 3/6 50
Control recalled 6
(divide by 12) 6/12 50
Sample data
Banana Orange Lemon Tomato Club Sword Bomb Guppy T
rout Ale Rum Vodka Beer
drinks v, weapons s, drinks r,
weapons r, drinks g, weapons t
13
Semantic Memory

• What is knowledge?

• Semantic memory is made up of concepts
• How are these nuggets of knowledge accessed,
  stored and manipulated

• Semantic
• Facts Knowledge
• Focus is on how this information is organized
  (rather than on encoding things into semantic
  memory)

14
Semantic Memory

• Systems view
• Subsystem of Declarative system

• Semantic
• Facts Knowledge
• Focus is on how this information is organized
  (rather than on encoding things into semantic
  memory)

15
Semantic vs. Episodic Memory
15

• Semantic Memory

• Episodic Memory

• Declarative memory for general knowledge and
  facts lacking reference to the episodic context
  in which it was learned.
• Examples
• World knowledge
• Vocabulary
• Rules, formulae, and algorithms
• Knowing awareness

• Memory for specific events in context
• Comes with a sense of reliving the event
• Called conscious recollection or Mental
  time-travel
• Self-knowing

16
Semantic vs. Episodic Memory

• Are they really distinct?
• Evidence from Neuropsychological Dissociations

• While all anterograde amnesics have profound
  deficits in episodic memory, most have only minor
  (if any) semantic impairments
• Spiers, Maguire, and Burgesss (2001) reviewed
  147 cases.
• Vargha-Khadems (1997) patients, Jon and Beth
  (impaired as children but developed normal
  semantic memories).

17
Semantic vs. Episodic Memory

• Are they really distinct?
• Evidence from Neuropsychological Dissociations

• Patients with retrograde amnesia often have a
  selective deficit in either episodic or semantic
  memory
• Episodic impairment with spared semantic memory
• Tulvings (2002) patient, KC (intact pre-trauma
  semantic memory)
• Semantic impairment with spared pre-trauma
  episodic memory
• Yasuda, Watanabe, and Onos (1997) patient

18
Semantic vs. Episodic Memory

• Are they really distinct?
• Evidence from Neuroimaging Dissociations

• Different brain areas are activated for semantic
  and episodic memory tasks (Wheeler et al., 1997)
• During memory encoding
• More left prefrontal cortical activity for
  episodic tasks than semantic.
• During memory retrieval
• More right prefrontal cortical activity during
  episodic memory retrieval than semantic.
• This also suggests that episodic and semantic
  memory are different types of memory.

19
Models of Semantic Memory

• How is semantic information stored/organized?
• Network models
• Propositions
• Hierarchical networks
• Spreading activation
• Exemplar and prototype models
• List models
• Smiths Feature overlap model
• Compound Cue Models
• Scripts and Schemas
• How does the organization impact memory behavior?

20
Propositions

• Representing meaning
• Proposition verifiable statement
• Two or more concepts with a relationship between
  them

A mouse bit a cat bit (mouse, cat)
21
Propositions

• Representing meaning
• Proposition verifiable statement (T/F)
• Two or more concepts with a relationship between
  them

A mouse bit a cat bit (mouse, cat)
22
Propositions

• Representing meaning
• Proposition verifiable statement (T/F)
• Two or more concepts with a relationship between
  them
• Networks Propositions can be represented as
  connected nodes
• Concepts nodes arguments
• Times, places, people, objects, etc.
• Linked by relations
• Verbs, adjectives, etc.

23
Propositions

• More complex example
• Children who are slow eat bread that is cold
• Slow children
• Children eat bread
• Bread is cold

24
Propositions

• Kintsch (1974)

• Memory better for sentences with fewer
  propositions

• The crowded passengers squirmed uncomfortably
• passengers crowded
• passengers squirmed
• passengers uncomfortable

Three propositions

• The horse stumbled and broke a leg
• horse stumbled
• horse broke leg

Two propositions
25
Propositions

• Bransford Franks (1971)

• Constructed four-fact sentences, and broke them
  down into smaller sentences
• 4 - The ants in the kitchen ate the sweet jelly
  that was on the table.
• 3 - The ants in the kitchen ate the sweet jelly
• 2 - The ants in the kitchen ate the jelly.
• 1 - The jelly was sweet.

26
Propositions

• Bransford Franks (1971)

• Study Heard 1-, 2-, and 3-fact sentences only
• Test Heard 1-, 2-, 3-, 4- fact sentences (most
  of which were never presented)

27
Propositions

• Bransford Franks (1971)

• Results
• the more facts in the sentences, the more likely
  Ss would judge them as old and with higher
  confidence
• Even if they hadnt actually seen the sentence
• Constructive Model we integrate info from
  individual sentences in order to construct larger
  ideas
• emphasizes the active nature of our cognitive
  processes
• So how might we organize this information in
  memory?

28
Hierarchical Network

• Collins and Quillian Hierarchical Network model
  (1969)
• Lexical entries stored in a hierarchy

• Representation permits cognitive economy
• Reduce redundancy of semantic features

Semantic Features
has skin
Animal
Lexical entry
can move around
breathes
IS A
IS A
29
Hierarchical Network

• Testing the model
• Semantic verification task
• An A is a B True/False

An apple has teeth
Use time on verification tasks to map out the
structure of the lexicon.
30
Hierarchical Network
has skin
Animal
can move around
breathes
has feathers
can fly
Bird

• Testing the model
• Sentence Verification time
• Robins eat worms 1310 msecs
• Robins have feathers 1380 msecs
• Robins have skin 1470 msecs
• Participants do an intersection search

has wings
Robin
eats worms
has a red breast
31
Hierarchical Network
has skin
Animal
can move around
breathes
Robins eat worms
has feathers
can fly
Bird

• Testing the model
• Sentence Verification time
• Robins eat worms 1310 msecs
• Robins have feathers 1380 msecs
• Robins have skin 1470 msecs
• Participants do an intersection search

has wings
Robin
eats worms
has a red breast
32
Hierarchical Network
has skin
Animal
can move around
breathes
Robins have feathers
has feathers
can fly
Bird

• Testing the model
• Sentence Verification time
• Robins eat worms 1310 msecs
• Robins have feathers 1380 msecs
• Robins have skin 1470 msecs
• Participants do an intersection search

has wings
Robin
eats worms
has a red breast
33
Hierarchical Network
has skin
Animal
can move around
breathes
Robins have feathers
has feathers
can fly
Bird

• Testing the model
• Sentence Verification time
• Robins eat worms 1310 msecs
• Robins have feathers 1380 msecs
• Robins have skin 1470 msecs
• Participants do an intersection search

has wings
Robin
eats worms
has a red breast
34
Hierarchical Network
has skin
Animal
can move around
breathes
Robins have skin
has feathers
can fly
Bird

• Testing the model
• Sentence Verification time
• Robins eat worms 1310 msecs
• Robins have feathers 1380 msecs
• Robins have skin 1470 msecs
• Participants do an intersection search

has wings
Robin
eats worms
has a red breast
35
Hierarchical Network
has skin
Animal
can move around
breathes
Robins have skin
has feathers
can fly
Bird

• Testing the model
• Sentence Verification time
• Robins eat worms 1310 msecs
• Robins have feathers 1380 msecs
• Robins have skin 1470 msecs
• Participants do an intersection search

has wings
Robin
eats worms
has a red breast
36
Hierarchical Network

• Problems with the model
• Difficulty representing some relationships
• How are truth, justice, and law related?
• No prediction about false sentences
• A whale is a fish vs. A horse is a fish
• Neither whale or horse is a fish (whale is a
  mammal), but people are faster to reject horse
  than fish

37
Hierarchical Network

• Problems with the model
• Conrad (1972) Effect may be due to frequency of
  association
• For most relationships organization and conjoint
  frequency confounded
• Subjects generated properties for concepts
  werent generated according to levels predictions
  (breathes generated for horse, instead of animal)
  
• Also had subjects verify statements - faster
  based on frequency, not level
• A robin breathes is less frequent than A
  robin eats worms

38
Hierarchical Network

• Problems with the model

Animal

• Smith, Shoben Rips (1974) showed that there are
  hierarchies where more distant categories can be
  faster to categorize than closer ones
• A chicken is a bird
• was slower to verify than
• A chicken is an animal

has feathers
can fly
Bird
has wings
Chicken
lays eggs
clucks
39
Hierarchical Network

• Problems with the model
• Assumption that all lexical entries at the same
  level are equal
• The Typicality Effect (e.g., Katz, 1981)
• Which is a more typical bird? Ostrich or Robin.

40
Hierarchical Network
has skin
Animal
can move around
breathes
has fins
has feathers
can swim
Fish
can fly
Bird
has gills
has wings
Verification times a robin is a bird faster
than an ostrich is a bird
41
Spreading Activation Models

• Collins Loftus (1975)

• Spreading activation
• Most popular model
• Recognizes diversity of information in a semantic
  network
• Captures complexity of our semantic
  representation (at least some of it)
• Consistent with C Qs (1969) results
• Consistent with results from priming studies

42
Spreading Activation Models

• Collins Loftus (1975)

• Spreading activation
• Bring back the network model, but make some
  modifications
• The length of the link matters.
• The less related two concepts are, the longer the
  link. This gets typicality effects (put CHICKEN
  farther from BIRD than ROBIN).
• Search is a process called spreading activation.
• Activate the two nodes involved in a question and
  spread that activation along links. The farther
  it goes, the weaker it gets. When you get an
  intersection between the two spreading
  activations, you can decide on the answer to the
  question.
• This model gets around a lot of the problems with
  the earlier network model.

43
Spreading Activation Models

• Collins Loftus (1975)

• Words represented in lexicon as a network of
  relationships
• Organization is a web of interconnected nodes in
  which connections can represent
• categorical relations
• degree of association
• typicality

street
vehicle
car
bus
truck
house
orange
Fire engine
fire
red
blue
apple
pear
roses
tulips
fruit
flowers
44
Spreading Activation Models

• Collins Loftus (1975)

• Retrieval of information
• Spreading activation
• Limited amount of activation to spread
• Verification times depend on closeness of two
  concepts in a network

street
vehicle
car
bus
truck
house
orange
Fire engine
fire
red
blue
apple
pear
roses
tulips
fruit
flowers
45
Spreading Activation Models

• Semantic priming
• A semantically-related word facilitates the
  processing/identification of a target word
• e.g. It is faster to say BUTTER is a real word
  if preceded by BREAD instead of an unrelated
  word like NURSE (Meyer Schvaneveldt, 1976).
• In the model Priming is accounted for by the
  Spreading of Activation between related concepts.

46
Semantic Feature Lists

• Decomposing concepts into smaller semantic
  attributes/primitives

Features father mother daughter son
Human
Older - -
Female - -

• Perhaps there is a set of necessary and
  sufficient features
• Necessary features have to be present for
  inclusion
• Sufficient if these features are present no
  other features are necessary for inclusion

47
Semantic Feature Lists

• John is a bachelor.
• What does bachelor mean?
• What if John
• is married?
• is divorced?
• has lived with the mother of his children for 10
  years but they arent married?
• has lived with his partner Joe for 10 years?
• Suggests that there probably is no set of
  necessary and sufficient features that make up
  word meaning
• (other classic examples game chair)

48
Feature Overlap Model

• Smiths Feature Overlap Model
• Used lists of characteristics instead of a
  network
• Concepts are clusters of semantic features. There
  are two kinds
• Distinctive features Core parts of the concept.
  They must be present to be a member of the
  concept, theyre the defining features. For
  example, WINGS for BIRD.
• Characteristic features Typically associated
  with the concept, but not necessary. For example,
  CAN FLY for BIRD.
• These features are stored in a redundant manner
• The decision of whether one concept is an example
  of an another depends upon the level of overlap

49
Feature Overlap Model

• Smiths Feature Overlap Model
• Some examples

BIRD MAMMAL
Distinctive Wings Feathers Nurses-young Warm-blooded Live-birth
Characteristic Flies Small Four-legs
ROBIN WHALE
Distinctive Wings Feathers Swims Live-birth Nurses-young
Characteristic Red-breast Large
50
Feature Overlap Model

• Smiths Feature Overlap Model

• Why characteristic features? Various evidence,
  such as hedges

• OK
• A robin is a true bird.
• Technically speaking, a chicken is a bird.
• Feels wrong
• Technically speaking, a robin is a bird.
• A chicken is a true bird.
• The answer depends on the kinds of feature
  overlap.

51
Feature Overlap Model

• Smiths Feature Overlap Model

• Similar concepts stored together

52
Feature Overlap Model

• Smiths Feature Overlap Model

• Answering a semantic verification question is a
  two-step process.
• Compare on all features. If there is a lot of
  overlap its an easy yes. If there is almost no
  overlap, its an easy no. In the middle, go to
  step two.
• Compare distinctive features. This involves an
  extra stage and should take longer.

53
Feature Overlap Model

• Smiths Feature Overlap Model

Easy yes Easy no Hard yes Hard no
A robin is a bird A robin is a fish A whale is a mammal A whale is a fish

• The model can account for
• Typicality effects One step for more typical
  members, two steps for less typical members, that
  explains the time difference.
• Answering no Why are no responses different?
  Depends on the number of steps (feature overlap).
• Hierarchy Since it isnt a hierarchy but
  similarity, we can understand why different types
  of decisions take different amounts of time.

54
Feature Overlap Model

• Criticisms
• No objective way to distinguish defining and
  characteristic feature
• Many items in category do not share a defining
  feature
• Furniture - do all items share a defining
  feature? Games?
• How many of the features of a bird can you lose
  and still have a bird?
• Because features are all thats important in the
  model, forward and backward associations should
  be the same
• Forward vs. backward associations
• So when asked to do word association task, people
  say insect for concept of butterfly, but
  rarely say butterfly as an example of an
  insect

55
Comparing the Models

• The spreading activation model is more flexible
  than the hierarchical network model.
• Pros of flexibility
• The spreading activation model can account for
  more empirical findings.
• Cons of flexibility
• The flexibility also reduces the specificity of
  the models predictions, making the spreading
  activation model more difficult to test.

56
Semantics as Prototypes

• Prototype theory store feature information with
  most prototypical instance (Eleanor Rosch, 1975)

1) chair 1) sofa 2) couch 3) table 12)
desk 13) bed 42) TV 54) refrigerator
Rate on a scale of 1 to 7 if these are good
examples of category Furniture
57
Semantics as Prototypes

• Prototype theory store feature information with
  most prototypical instance (Eleanor Rosch,
  1975)
• Prototypes
• Some members of a category are better instances
  of the category than others
• Fruit apple vs. pomegranate
• What makes a prototype?
• Possibly an abstraction of exemplars
• More central semantic features
• What type of dog is a prototypical dog?
• What are the features of it?
• We are faster at retrieving prototypes of a
  category than other members of the category

58
Semantics as Prototypes

• The main criticism of the theory
• The model fails to provide a rich enough
  representation of conceptual knowledge
• How can we think logically if our concepts are so
  vague?
• Why do we have concepts which incorporate objects
  which are clearly dissimilar, and exclude others
  which are apparently similar (e.g. mammals)?
• How do our concepts manage to be flexible and
  adaptive, if they are fixed to the similarity
  structure of the world?
• features have different importance in different
  contexts
• what determines the feature weights
• If each of us represents the prototype
  differently, how can we identify when we have the
  same concept, as opposed to two different
  concepts with the same label?

59
Semantics as Exemplars

• Instance theory each concept is represented as
  examples of previous experience (e.g., Medin
  Schaffer, 1978)
• Make comparisons to stored instances
• Typically have a probabilistic component
• Which instance gets retrieved for comparison

dog
60
Compound Cue Models

• Info stored with context
• To retrieve info, cues are used to match with
  stored contexts
• Can also account for episodic memory
• SAM, MINERVA 2, TODAM
• Math models that predict sets of results based on
  strength of cue associations
• Also popular models among researchers

61
Compound Cue Models

• Compound-cue model must be combined with theory
  of memory
• Make predictions about performance in memory
  retrieval tasks
• In SAM (search of associative memory), a matrix
  of association among cues and memory traces,
  which are called images
• Cues are assembled in a short-term store, or
  probe set, which is the match against all item in
  memory
• In TODAM (theory of distributed associative
  memory), to-be-remembered items are represented
  as vectors of features
• Sum of vectors, convolution
• The resulting scalar can be mapped into
  familiarity and, in turn, into response time and
  accuracy
• Examine mechanisms of priming and extent to
  explain of priming effects

62
Schema Theory

• Scripts and schemas (Bartlett, Schank)
• Knowledge is packaged in integrated conceptual
  structures.
• Scripts Typical action sequences (e.g., going to
  the restaurant, going to the doctor)
• Schemas Organized knowledge structures (e.g.,
  your knowledge of cognitive psychology).
• It would be possible to describe these with nodes
  and links.
• For example, a schema could be a sub-network
  related to a particular area.

63
Schema Theory

• Restaurant Schema
• Enter - seated by maître d
• Read menu - order from waiter
• Waiter brings food
• Waiter brings check
• Pay check - leave

64
Schema Theory

• Restaurant Schema

Bower, Black, and Turner (1979)

• 73 of respondents reported these common events
  when going to a restaurant
• Sit down
• Look at menu
• Order
• Eat
• Pay bill
• Leave

• 48 also included
• Enter restaurant
• Give reservation name
• Order drinks
• Discuss menu
• Talk
• Eat appetizer
• Order dessert
• Eat dessert
• Leave a tip

65
Schema Theory

• Bartlett (1932)
• Read unfamiliar story
• Remembered differently depending on expectation

66
Schema Theory

• Scripts and schemas (Bartlett, Schank)
• Evidence When people see stories like this
• Chief Resident Jones adjusted his face mask while
  anxiously surveying a pale figure secured to the
  long gleaming table before him. One swift stroke
  of his small, sharp instrument and a thin red
  line appeared. Then an eager young assistant
  carefully extended the opening as another aide
  pushed aside glistening surface fat so that vital
  parts were laid bare. Everyone present stared in
  horror at the ugly growth too large for removal.
  He now knew it was pointless to continue.

67
Schema Theory

• Scripts and schemas (Bartlett, Schank)
• And you ask them to recognize words that might
  have been part of the story, they tend to
  recognize material that is script or schema
  typical even if it wasnt presented. Lets try
• Scalpel?
• Assistant?
• Nurse?
• Doctor?
• Operation?
• Hospital?

68
Schema Theory

• Scripts and schemas (Bartlett, Schank)
• People also tend to fill in missing details from
  scripts and schemas if they are not provided (as
  long as those parts are typical).
• When people are told the script or schema that is
  appropriate before hearing some material they
  tend to understand it better than if they are not
  told it at all or are told it after the material.

69

• Rocky slowly got up from the mat, planning his
  escape. He hesitated a moment and thought. Things
  were not going well. What bothered him most was
  being held, especially since the charge against
  him had been weak. He considered his present
  situation. The lock that held him was strong but
  he thought he could break it. He knew, however,
  that his timing would have to be perfect. Rocky
  was aware that it was because of his early
  roughness that he had been penalized so severely
  - much too severely from his point of view. The
  situation was becoming frustrating the pressure
  had been grinding on him for too long. He was
  being ridden unmercifully. Rocky was getting
  angry now. He felt he was ready to make his move.
  He knew that his success or failure would depend
  on what he did in the next few seconds.

Prison
Wrestling
Other ?
70

• Every Saturday night, four good friends get
  together. When Jerry, Mike, and Pat arrived,
  Karen had just finished writing some notes. She
  quickly arranged the cards and stood up to greet
  her friends at the door. They followed her into
  the living room and sat down facing each other.
  They began to play. Karen's recorder filled the
  room with soft and pleasant music. Her hand
  flashed in front of everyone's eyes and they all
  noticed her diamonds. They continued for many
  hours until everyone was exhausted and quite
  silly. Jerry made his friends laugh as he
  theatrically took a bow, entertaining them all
  with the wildness of his playing. Finally,
  Karen's friends went home.

Playing music
Playing cards
Other ?
71
Summary of Semantic Memory

• Semantic memory knowledge
• Some evidence for a separate system
• Early models suggested hierarchical network -
  cognitive economy
• Results suggest no strict hierarchy or cognitive
  economy
• But current network models suggest loosened
  hierarchy (spreading activation)
• Other ideas schemas, compound cues

72
If memory for speech is episodic, what are
linguistic symbols?

• Reply Maybe linguistic symbols (words,
  phonemes, etc) are like prototypes.
• Many categories have a prototype, an ideal mean,
  centroid token that best represents the category
  (Rosch, 1978). Prototype members of a category
  come to mind faster, are recognized more quickly,
  etc.
• Categories that are more abstract have fewer
  features than concretes.
• Granny Smith apple gt apple gt fruit
• Fluffy gt tabby cat gt housecat gt cat gt pet
• Bob saying tomato gt English word tomato
• HOWEVER,
• mathematical models of memory exhibit the
  behaviors that support prototypes and
  abstractions. But do it by storing rich detail
  and computing abstractions and prototypes
  whenever needed.

73
Minerva 2 Storing Episodes

• Lets look closer at a specific model.
• Minerva 2 Model (Doug Hintzman, 1986) Every
  episode or exemplar is stored as a trace a
  long vector of features, added to memory. For
  words, the features represent many kinds of
  information. The features can only be 1, -1 or 0
  in Minerva 2.
• Exemplar Memory a matrix of feature vectors for
  each exemplar in the experiment.

1
0
1
0
-1
-1
1
1
1
0
-1
1
0
-1
1
1
-1
1
1
0
0
-1
1
pronunciation ftrs orthographic ftrs
semantic ftrs contextual ftrs
74
Minerva 2

• Probing Memory. Each new episode is a probe into
  the memory matrix.
• The similarity of the probe is computed to all
  traces.
• Traces of the most similar episodes become highly
  active.
• The memory response (or echo) can show greater or
  lesser activity overall (intensity) and a certain
  prominent pattern of activity (content).
• Echo Intensity. Stimulate memory with a probe.
  The more activation across features and traces,
  the greater the intensity of response. So if
  there are many similar copies, the higher
  familiarity of the probe.
• Recognition Task For a new/old recognition task,
  you set a threshold. If total Intensity is
  above threshold, say old, if below, say new.
• Prototypes If the probe is an abstract category
  (eg, fruit), the most intense traces are its
  prototypes.

75
Minerva 2

• Echo Content. The probe activates a subset of
  traces. The common features across this set are
  computed. These features specify an abstract
  pattern similar to the probe but generic a kind
  of abstract category for the probe.
• The features not shared cancel out leaving an
  abstract vector with fewer features a prototype
  or schema or abstract object.
• Thus hearing the word tomato activates the
  prototype pronunciation and the abstracted
  meaning of tomato.
• Our intuitions about abstract symbols words,
  phonemes, etc may reflect integration of
  content across traces.

76
Other models

• TODAM
• Associative Theories
• ACT-R, TODAM
• Search Models
• SAM, REM
• Trace Theories
• Perturbation Model
• Connectionist Models
• PDP, EPIC
• Biological-Based Theories
• HERA, CARA

77
Priming Propositions

• Ratcliff and McKoon (1978)

The mausoleum that enshrined the tsar overlooked
the square.

• Involves two propositions
• P1 OVERLOOK, MAUSOLEUM, SQUARE
• P2 ENSHRINE, MAUSOLEUM, TSAR.

78
Priming Propositions

• Ratcliff and McKoon (1978)

• Results in a cued memory task (how long does it
  take to verify square was in the sentence)

Condition Examples RT to Target Priming Effects
Across sentences Between two propositions in the same sentence Within a single proposition square-clutch square-Tsar square-mausoleum 671 msec 571 msec 551 msec None baseline 100 msec facilitation 120 msed facilitation