Section 5. Memory
1. What is memory?
2. Memory subsystems
short-term (working) memory
3. Memory processes
levels of processing theory
amnesic syndrome/LTM subsystems
4. Representation of information in memory
analog vs. propositional representation
Terms to know for this section
Reisberg - pp. 123-167 (memory structures), 168-210 (memory
processes), 394-439 (visual memory), 256-303 (theories of memory
organization), 304-345 (concepts), 211-255 (memory problems)
5.1 What is memory?
Functional definition: Memory is the retention of potentially usable
information over some period of time.
Memory subsystems - organization of memory into component parts
Memory processes - the activities that occur in memory
encoding - transduction
retrieval - relationship to encoding
5.2 Memory subsystems
The spatial metaphor for memory - items in memory are like objects located
in a space that must be searched for retrieval to take place.
Problem - If we search memory to retrieve information, why is it sometimes
so easy to say that we don't know something? (Would it not be necessary to
search everywhere in memory to confirm that we did not know something?)
Characteristics of memory subsystems -
capacity - how much information?
duration -how long does the information last?
representation - what kind of information is retained?
information loss - how does forgetting occur?
decay/interference (retrieval failure)
The modal (3-store) model of memory (Atkinson & Shiffrin, 1968)
1. Sensory memory
uncoded, precategorical representation
information loss through interference
Sperling' s (1960) experiments - visual sensory memory - iconic memory
2. Short-term memory (STM)
limited capacity - 7+2 chunks (Miller, 1956)
short duration - limited by rehearsal?
information loss through interference
interaction between sensory information and information from long-term
chunk - a meaningful unit of information
visual STM (deGroot, 1965) - expert vs. novice chess players
phonological STM - Conrad (1964) - visual letter confusions are
3. Long-term memory (LTM)
abstract, highly-organized representation
information loss due to retrieval failure
importance of rehearsal
Evidence for the distinction between STM & LTM: Recall task; stimuli =
interference task / rate of presentation manipulation
serial position curve
Control processes - rehearsal, encoding, etc.
Rundus & Atkinson (1970) - Rehearsal processes in free recall: A
procedure for direct observation
Rehearsal was the primary mechanism considered for the transfer of
information from STM to LTM within the context of the modal model.
& Atkinson utilized a verbal protocol in which subjects spoke aloud while
learning a list of words that were to be recalled later. They showed that
from the beginning of the list received more rehearsals than items from later
in the list. Except for items at the end of the list, the number of rehearsals
was correlated with the number of words recalled.
Criticisms of the modal model -
over-simplified - suggested that STM and LTM were each unitary
role of rehearsal exaggerated
Revisions to the modal model -
Working memory (Baddeley & Hitch, 1974; Baddeley, 1986)
augmented version of STM (based, in part, on neuropsychological
modality-free central executive
articulatory loop (phonological store + articulatory production
visual-spatial sketch pad
5.3 Memory processes
Alternatives to the modal model assume a unitary store that is equivalent to
LTM and in which STM is equivalent to temporarily activated regions of
Thus, this type of model uses an activation metaphor. In addition, less
emphasis is placed on rehearsal as a mechanism for adding information to
LTM. As an alternative, more emphasis is placed on how subjects encode
1. How does the type of encoding affect the retrieval of information?
levels of processing theory (Craik & Lockhart, 1972)
the level or depth of processing of information affects its
deep processing creates more robust memories than shallow
shallow processing -
Is the word in capital letters?
deep processing -
Would the word fit in the sentence? "He met a ___ on the
In general, recognition memory performance was best (80%) for words that
received deep processing whereas performance was worst (20%) for words
that received shallow processing.
Problems with levels of processing theory -
what level of processing does the orienting task utilize? How can
depth be assessed independently of memory performance?
relation between orienting task and memory task (Morris, Bransford,
& Franks, 1977 - semantic/rhyme encoding task; standard/rhyme
2. What happens when we remember something?
1. Retrieval cues
retrieval of information is based on the concept of retrieval cues
What is a retrieval cue? …anything that could serve to activate relevant or
associated information in memory
Sometimes the cues are supplied to you, as in a recognition memory task
(such as multiple choice exam). In that case, in that case you are assumed to
require only some type of decision process to evaluate the information
evoked by the cues.
Alternatively, you have to supply the cues yourself, as in a recall task (such
an essay exam). Here you must supply both retirival cues and evaluate the
relevance of the information evoked.
2. Encoding specificity principle (Tulving, 1982)
the similarity between the information available at the time of encoding and
the time of retrieval determines memory performance
+similarity(encoding; retrieval) —> +memory
state-dependent learning (mood / drugs)
physical context (Godden & Baddeley, 1975) - wet vs. dry learning,
wet vs. dry testing
3. Limits on the use of recall and recognition task
doctrine of concordance (Tulving, 1989)
Tacit assumption of many memory researchers that generally there is
similarity "…between what people know, how they behave and what they
experience. Thus conscious awareness is required for, and therefore
accompanies, the acquisition of knowledge, or its retrieval from the
store; retrieved knowledge guides behaviour, and when this happens,
people are aware of the relation between the knowledge and the behaviour."
3. Amnesia and LTM subsystems
Evidence from cognitive neuropsychology may be useful when considering
memory structures and processes.
syndrome - a group of symptoms that collectively characterize a
specific condition or disease
amnesic syndrome (source - Korsakoff' s syndrome, head injuries,
- anterograde amnesia generally present (a lack of memory for events after
- retrograde amnesia sometimes present (a lack of memory for events
before the trauma)
- amnesics generally have normal intelligence
- generally intact STM
- patients can usually learn certain things after the onset of amnesia
What abilities do amnesics retain? (residual learning capacity)
most have good STM performance
most can learn to do tasks involving motor skills (but may be slower
learning than normals).
Patient H.M. & broken drawings
amnesics are susceptible to perceptual priming (repetition priming) -
conscious awareness vs. performance
Theories of amnesia and memory must account for the abilities that
amnesics have lost and those that they retain.
Theories of amnesia and memory must account for these effects.
Theories of memory and amnesia
1. episodic vs. semantic memory (Tulving, 1972)
episodic memory - contextual memory (autobiographical knowledge)
semantic memory - factual knowledge that is independent of the
setting in which it was originally learned
Originally, amnesics were believed to have an impaired episodic memory
system but an intact semantic memory system.
Are episodic & semantic memory really two separate memory processes?
Probably not. Amnesics usually exhibit intact episodic and semantic
for events learned prior to their amnesia.
2. declarative vs. procedural memory (Cohen & Squire, 1980)
declarative memory - "knowing what"; includes episodic and
procedural memory - "knowing how"; performance of skilled actions
that do not require conscious awareness of recollection
Cohen & Squire (1980) argue that amnesics have an impaired declarative
memory but an intact procedural memory.
However, 1) some amnesics are able to acquire declarative knowledge, 2)
most tasks incorporate a mix of declarative and procedural memory.
3. explicit vs. implicit memory (Schacter, 1987)
explicit memory - information available to conscious recollection
implicit memory - information unavailable to conscious recollection;
not usually tested in traditional measures of memory
Support from experiments - compare recall/recognition w/ word fragment
normals (Jacoby & Dallas, 1981)
amnesics - performance is often close to normal in implicit memory
tasks whereas performance is generally worse than normal in explicit
descriptive rather than explanatory
Conclusion? No one theory of memory accounts for all aspects of memory
performance in amnesics and non amnesics. Nor is it likely that memory
be neatly dichotomized as in these theories. Instead, memory is more like a
series of integrated systems that contains some of the characteristics of the
theories described here.
4. Representation of information in memory
What is a representation?
In what form is information represented in memory?
How is information organized so it can be efficiently used?
A representation re-presents or stands for something. It is a transformation
of information in one form to another form that preserves characteristics of
the original information.
Understanding the representation and processing of information are the two
basic goals of cognitive psychology.
Fundamental characteristic of all representational systems - there is less
information contained in the representation than in the thing being
Internal representations (Mental representations)
1. Discrete symbols 1. No discrete symbols
2. Explicit, needs symbol for relation 2. Implicit, no separate symbol for relation
3. Clear rules of combination for types of 3. No clear rules of combination or symbol
4. Abstract 4. Modality-specific
Propositions vs. imagery
Imagery - analog representation based on visual information; manipulation
of visual information
mental rotation (Shepard et al.)
image scanning (Kosslyn et al.)
Propositions - a language-like representation of an event. e.g., predicate
Predicate (arguments); e.g., ON(BOOK, TABLE)
Does one form of representation have priority over other forms? Can one
form of representation be reduced to another form?
Pylyshyn' s criticisms of imagery
Images can' t just be 'pictures in the mind' s eye.' e.g., If an image has
a missing part, it is usually a meaningful part that is missing. This is
necessarily true of a picture.
If the 'mind' s eye' perceives images, what kind of images does it use?
…an infinite regression.
Images can be reduced to propositions. (images ® propositions)
Results from imagery experiments are due to demand characteristics.
Kosslyn' s response
The idea that images are the same as pictures is a 'straw man' .
Propositional accounts of image data are cumbersome.
Images have distinct characteristics that make them different from
Kosslyn' s theory/model of imagery
1. images are represented in a special spatial medium that has the following
limited extent of medium
area of highest resolution at centre
medium has 'grain' that obscures fine detail in small images
once an image is generated it begins to fade
2. Imaginal information is often linked to corresponding propositional
3. Generating images involves both imaginal and propositional information.
Evidence for Kosslyn' s theory/model
experimental - e.g., image tracing task (granularity)
neuropsychological - e.g., dissociation between imaginal and
Why do we have concepts?
generalization (recognition of novel things )
recognition of the relationship among instances
Theories of conceptual organization
1. Defining-attribute view (Classical view)
Basically, a concept has defining attributes that determine the members of a
concept. Defining attributes are also known as properties or features.
the meaning of a concept is the conjunction of a set of attributes
each attribute is singly necessary and jointly sufficient for an item to
be considered a member of a concept
explanation for singly necessary and jointly sufficient:
singly necessary = each characteristic must be present
for the object to belong to the category
jointly sufficient = only these characteristics (and no
others) are required for the object to be considered a
member of the category
definition of a bachelor - male, unmarried, adult
Is Mr. Smith a bachelor
age (beyond adulthood) unimportant
hair colour unimportant
clear-cut boundaries exist between items that are members of the
concept and those that are not members
all members of a concept are equally representative
concepts are nested within superordinate concepts
A model of the defining-attribute view
Collins & Quillian (1969) semantic network
characteristics as defined above
hierarchy is searched in response to inquiries about concept
membership (search times are related to hierarchical distance)
hierarchy is searched in response to inquiries about concept
attributes (search times are related to hierarchical distance)
sentence verification task
"Is a canary a bird?"faster than "Is a canary an animal?"
"Does a canary fly?"faster than "Does a canary have skin?"
Problems with the defining-attribute view
some attributes are more important for determining concept
membership than others
Feature listings for 2 concepts
all members of a concept are not treated equally (typicality)
Typicality ratings for birds (Rips, Shoben, & Smith,
difficulty in determining the defining attributes of concepts (family
Family resemblance analysis (Rosch & Mervis, 1975)
concept membership is not always clear-cut (fuzzy boundaries)
predictions based on the nesting of concepts are not always confirmed (e.g.,
"Is a chicken a bird?"vs. "Is a chicken an animal?")
2. Prototype theory
Designed to account for the problems associated with the defining attribute
model, such as typicality effects and fuzzy concepts (compare with work on
prototype theory in perception, section 4).
What is a prototype? A representation that captures the central tendency of
a set of similar items. It can be a) a summary description that is an
abstraction of many instances, or b) a collection of the 'best instances' of a
there is no set of necessary and sufficient attributes
concept boundaries are fuzzy
instances are graded according to typicality
concept membership is defined by similarity to the concept' s
apparent universality in the categorization of colours
typicality ratings are good predictors of classification times
typical concept members are listed first
children generally learn about typical concept members first
typical members tend to serve as cognitive reference points
family resemblance scores can be calculated and typical members
tend to have the highest scores
Are some levels of abstraction in prototype theory more important than
Rosch et al. (1976)
specific objects - slotted screwdriver, Robertson screwdriver, Phillips
basic level - screwdriver
superordinate level - tool
Individuals tend to use basic level category terms.
Tanaka & Taylor (1991) - Object categories and expertise: Is the basic level
in the eye of the beholder?
Problems with prototype theory
not all concepts have prototype characteristics - deals best with
concrete concepts, not abstract concepts
does not incorporate information about the relationship among
concepts (pets & fish; pet fish)
does not account for the coherence of concepts - Why do members of a
concept group together? It is not just similarity - we can form arbitrary
concepts (women, fire, & dangerous things; things in a yard sale, etc.).
How is information about individual concepts combined?
How does the collective organization of information about individual
concepts allow us to deal with everyday situations?
schema - structured cluster of concepts (plural schemas or schemata);
coined by Bartlett (1932)
Early work on schema - Bartlett
Subjects read story with non-Western construction (War of the Ghosts);
recall story at later date. Subjects impose a Western organization on the
information contained in the story.
Subjects' recall was affected by their expectancies; memory was
reconstructed according to past experiences.
Characteristics of schema
schema consist of relations, variables/slots, and values
relations can be of the form 'is-a,' 'eat,' 'cause,' etc.
variables contain other schema or concepts
values are the instances that fit the variables
schema contain general knowledge that can used in many specific
situations (e.g., 'restaurant schema' )
slots may have default values
Theories of schema organization
Schank & Abelson' s script theory
A script is a type of schema that describes the actions that are part of
e.g., restaurant script
walk into restaurant
look for table
decide seat location
go to table
look at menu
order given to server
order given to cook
cook prepares food
script organization allows the efficient use of cognitive resources
criticisms - unprincipled (ad hoc); inflexible
Implications of schematic/reconstructive processes for memory
1.gist vs. verbatim recall
2.John Dean' s memory
3.eye-witness testimony - Loftus
4.enhancing eye-witness testimony
1. gist vs. verbatim recall
gist = meaning
verbatim = surface structure; exact wording
in general, memory for gist is maintained while verbatim
memory is lost
He sent a letter about it to Galileo, the great Italian scientist.
Galileo, the great Italian scientist, sent him a letter about it.
(comparison sentence; different meaning)
He sent Galileo, the great Italian scientist, a letter about it.
(comparison sentence; same meaning, different word order)
2. John Dean's memory
who was John Dean? Richard Nixon' s counsel; Watergate
Neisser (1981) compared what he remembered during the Watergate
trial & Nixon' s tapes
some of gist maintained - if information rehearsed in advance
reflected Dean' s self-image (his recollection was that he
played a more central role than he did in reality)
gist preserved across episodes - underlying tone that permeated multiple
encounters with the same people, discussing the same topics
3. eye-witness testimony
memory is subject to distortions that can occur at the time of encoding and
distortions that can occur at the time of retrieval
Loftus & Zanni (1975) - typical experiment
show people film of vehicles involved in accidents
"Did you see a broken headlight?" vs. "Did you see the broken
(no broken headlight was shown in the film)
results: those asked "Did you see the broken headlight?"were twice as likely
to report that the headlight was broken.
4. enhancing eye-witness testimony
standard police interview
open-ended questions - what happened? - short-answer questions about
Cognitive interview (Fisher & Geiselman, 1987) - increase overlap between
encoding & retrieval; increase number of retrieval paths
1.mentally reinstating the environmental and personal context
2.reporting everything, regardless of the perceived importance
3.recount events in a variety of temporal orders
4.report events from a variety of perspectives (e.g., the witness and
another significant figure in the event)
General mnemonics: if you can' t remember a name, how many
syllables, frequency of occurrence, ethnicity, etc.
cognitive interview produces 25-35% more correct information than
standard interview (also superior to other procedures, such as hypnosis).
PEMA - personal external memory aid
Principles of mnemonics
2. Key word method
foreign languages - link sound / meaning
e.g., Russian: zronok (bell) - zero-oaks
120 Russian words / 3 sessions
control group retains 30% 6 weeks later
key word method retains 45% 6 weeks later
3. method of loci
4. Peg word system
one is a bun
two is a shoe
three is a tree
ten is a hen
You have reached the end of section 5.