Oct 25, 2010
Long term memory: forgetting
o Manner in which it is lost
Forget very rapidly at first
Then less and less
Younger memory traces are more vulnerable to forgetting
o Rate of forgetting does not vary with degree of original learning
Stronger and weaker original memories forgotten at same rate
Learn repeated items to greater extent
• Forgetting rate would be the exact same if it was presented
Memories fade over time without use
Something has to happen over time to cause the decay
• Absence of rehearsal, lose neural connection associated
• Earlier memories interfere with later
• More lists someone has learned the more difficult it is to
recall the last list
• Later memories interfere with earlier memory
Encoding specificity Hypothesis
• Cannot retrieve some memories because do not have
appropriate or specific cue
• Original learning
o Very stagnant
o Good cues for memory
• Non cued
o Gradual decline
o Approx the same or bit better with right cue
• Decay vs interference
Interference has much greater effect
More games = more interference
Same as STM
o Over load retrieval cues
Retrieval depends on cues If too many events are associated with same cue, more difficult to
• Self Relevance Effect
o More relevant the adj were rated the better they were recalled
o Well learned info can be retained over much of our life
• Amnesia following trauma
o Retro disrupts info from before the event
o Antero disrupts info after the event
o Severe epilepsy
o Remove parts of hippocampus
Greatly reduced seizures
o Could not learn new info
Explicit memory effected
• Implicit vs Explicit
o Patients show normal implicit memory
o Pricked patients on hand when shaking upon meeting
o Patient cant remember him the next day but not shake hand
• Amnesia and Implicit Memory
o Korsakoff’s have impaired explicit
o Women with amnesia asked to dial phone number
Ends up calling mother
o Procedural implicit memory still intact
November 1, 2010
• Fred Bartlett
o Memory should be studied in natural circumstances
o No lists of words
o People don’t describe the same thing the same way
• Barletts Methods
o Repeated reproduction
Ask the same person to describe the same event at different times
o Serial reproduction
Many people to transmit the material
o Both show same thing
o Retelling changes over time
• Changes over recalls
Info gets lost
Illogical is often forgotten o Rationalization
Make sense of story
Make inferences to help make sense
o Dominant themes
Some themes are dominant, anchor point
• Reconstructive Memory
o Add personal knowledge of world to help guide recall
o The angry raisins/ grapes of wrath
• Loftus and Palmer / eyewitness testimony
o Estimate of speed of cars based on the harshness of the verb
o Was there any broken glass? (there was none )
o More people in smashed group said there was
• Misinformation effect
o Subjects see slides depicting an auto accident at an intersection
o Test 1: did another car pass the red Datsun while it was stopped at the
For one group makes sense, there was a sign in the image they saw
25% who didn’t see a sign say they did.
• Mcloskey and Zaragoza
o Inconsistent info makes them more inaccurate
• Hyman, husband and Billings
o Contact parents and ask for descriptions of actual events that occurred
when subjects were kids
o Asked subject q;s about events
o Also asked about false events
o Form of interference
o Source confusion
o Warning subject that there is misinformation reduces this effect
• Integration of Prior Knowledge
o ½ the name was Hitler
o Recognition test
5 mins to 1 week later
14 sentences, ½ old
• Very accurate
• Subject who read martin still very accurate
• Subjects who read Hitler confused thematically related new
sentences with old
o When people have prior info they are going to use for reconstruction
o Benefit More easily understand framework of story
Bring prior info
Confuse memory of story with own memories of subject matter
• Anderson and Picert
o Given description of house
o Ask to read in on perspective
o Home buyer perspective
Leaky roof, damp basement
o Burglar perspective
st Rare coin collection, tv, stereo
o 1 recall (of whole story)
Recall detail was greater for details related to perspective
o 2 recall
Same perspective or different
Recall improved when perspective changed
New perspective provided a second way to search memory
• The cognitive interview
o Recount event in more than one sequence, environment and context,
o Provide many cues to help memory, Shown to increase recall
• 7 sins of memory
o Long term memory is transient
Actual forgetting from long term memory
Failure of attention during encoding by relying on automatic or
Temporary loss of access to info
More common in elderly
Tip of the tongue
Attributing events to an incorrect source of context
Incorporate information provided by others into you own
recollection of events
Tendency for knowledge beliefs and feelings to distort memory of
past events and influence current and future judgments o Persistence
Tendency to remember things that we don’t want to or need to
November 3, 2010
Models of Semantic Memory
• Tip of the tongue
o State of knowing a fact or name but not being able to recall or retrieve the
info right away
o Brown and Mcneil
Used dictionary definitions to produce this
o 50% of the time people can recall the first letter
o About half are resolved within 1 minute
o Use partial info cue for word or fact
• Episodic vs Semantic
o Structuralist (multiple systems) view
• Different memory systems, distinct from one another
Only one memory system
Different info from one memory system
Specific vs generalized memories
Selecting between instances vs averaging over instances
• Semantic Networks
Represents concepts in network
Structural and semantic nodes
Relations between features and nodes
o Words with more than one meaning (bank)
• Anderson et al
o Verb indicates a certain meaning on the word
Encoding a particular structural node as connected to a particular
Semantic intrudes to influence episodic
o Supports encoding specificity
• Best cue
o Best cue is what matches how it was encoded
• Production vs Verification
o V—indicating the truth of a test item
o P—retrieving an instance from memory when given a cue • Collins and Quillian/ Hierarchical network model
Semantic made of network of basic elements (nodes) has a list of
properties associated with it
• Organization of the info is hierarchical
• Assumption of cognitive economy
o Feature or properties are represented only once at
the highest level of the hierarchy
• It takes time to move from one level of the hierarchy to a
• Answer true or false as quickly as you can
o Time based on # of levels to reach
o Property is slower than category membership
o Facilitation occurs when previous question primes
or activates pathway through networks for next
Category Size effect
• Faster at answering about small categories than larger
• Semantic Features Model
Common to members of the category
• Get rest of NOTES FROM ALYYYYYYYYYYYYYYYYY
o Assumes questionable distinction between characteristic and defining
Can people identify defining features for every category
o Model cannot easily explain priming effects
November 8, 2010
Models of Semantic Memory
SPREADING ACTIVATION MODEL
· Concepts organized in network, but it is not hierarchical.
· Features stored with concepts.
· New: size of the link between two concepts represents the strength of the associations.
· Long distance = weaker association.
· Assumes that activation (when one node is activated by thinking about concept) is spread
to other concepts similar.
Lexical Decision Task · Demonstrates how related items can prime each other.
· If the letters represent a word, respond yes; if a non-word, respond no.
· Measure time it takes to make decision.
· When we have pairs of words in the list related to each other (e.g., butter following bread
– words that have a relation) you are faster at responding.
This is called semantic priming
The Fan Effect
· The number of connections a concept has.
E.G., the concept of robin would have a large fan; aardvarks would have a small fan.
The fan effect determines
· How quickly you can access information
· The likelihood of activating information
· Large fan = slower, but greater probability because of more connection
Advantages to Spreading Activation Model
· Can explain semantic priming.
· Can explain typicality effect.
· Can explain category size and reversals of category size effects.
· Can explain the fan effect.
· Model is very flexible, can explain almost any pattern of data.
Problems of Spreading Activation Model
· Has many assumptions
· Flexibility of model means it does not make many strong predictions and it is therefore
difficult to test this model.
· Model assumes that activation spread from node to node.
· Looked at priming between categories and members of categories.
· First item is always name of the category.
· Second item a member of catendry.
· Expect priming if the 2 item is member.
E.G., if you present “part of the human body” and then “hand” people will be fast in
· Varied time between prime and target.
· Faster to verify target when it matches prime category (e.g., disease à cancer)
· Found greater priming at longer intervals.
· Fast at short SOAs, but even faster at long SOAs.
· Suspected that at short intervals we see automatic priming (SAM)
· Argued 2 aspects to priming (1) automatic priming (2) top-down expectation, which
Did experiment to show this
· Told participants that if you see the prime BUILDING, a part of the body was likely to
follow. E.G., building à hand
· Found that at long SOAs only, building would prime body part. Only come about
· Shares activation model but identifies propositions.
· The smallest unit of knowledge that is either true of false.
· Propositions represented by ellipses, and are formed by the intersection of nodes.
E.G., in semantic memory you have info representing dogs and cat. One piece of info is
that dogs hate cats. In semantic network you have proposition that dogs will chase cats.
According to model dog is the agent, cat is the object, and chase is the relation.
· Combing info of individual nodes to form propositions.
· Network can represent categorical information and specifics.
· Anderson argues that episodic and semantic memory are represented by nodes
· Unit of analysis when we comprehend and the way we are storing information in
· We lose specific details of sentence because of propositions.
· General knowledge structures that represents organized clusters of knowledge.
· Represents general procedures, objects, events, or sequence of events.
One specific kind of schema is called:
· Knowledge organized around routine activities (e.g., going to restaurant, visiting a
doctor or dentist).
E.G., when going to a restaurant there is a protocol when someone is serving you.
· Wanted to make sure that people share most aspects of a script.
· Write down all actions and events for different kinds of scripts.
· People had similar script.
E.G., going shopping
· Not every action or event included in stories that participants read (they read 18)
· Read 3 different versions of the same script.
· Subjects then given recognition test for memory for stories.
· Incidental memory experiment – did not know memory would be tested.
· Asked to rate on a scale of 1 – 7 how much they thought the item had been presented in
· Script items that had been stated in 1, 2, or 3 stories got high marks (people confident it
was in the story) · Items not stated in the script increased as the number of stories went up (confidence
increased over the number of scripts)
· False memories
· Tells us that when a script is activated in a story, all the related items in the script are
activated in memory.
· Then later, you are going to get confused between items that were actually mentioned or
items that are part of your script.
Smith & Graeser (1981)
· Atypical actions remembered better because they stand out
· When you encode a story based on a scripted routine, you use script as a foundation –
atypical actions stand out because they are not part of our general script.
Connolly, Hockley, & Pratt
· Presented 6 different stories based on scripted routines.
· Typical and atypical actions presented 0-4 times across different stories (varying times
that typical and atypical times mentioned in the stories).
· Given surprise memory test – asked people to tell her how many times a certain action
· Accurate at remembering whether atypical items were presented across the different
· Far less accurate at remembering typical items
When a script is activated:
· Store script in episodic memory.
· Script represents both stated and inferred typical events.
· “Tag on” atypical actions of story to script.
November 10, 2010
• Benefits/ Function of Categorization
o Reduces the complexity of the environment
o The means by which objects of the world are indentified
o Reduces the need for constant learning
o Allows us to decide what constitutes appropriate action
o Enables us to organize knowledge
o Categorization: The skill by which we group objects and events into
classes so that unique things can be responded to similarly
o Concept Identification: determining what a set of stimuli have in common
• Explicit Concept
o One that you intentionally set out to acquire and have probably developed
specific rules for
• Implicit Concept o One that was incidentally acquired and is not necessarily conscious
• Concept Identification
o Artificial concepts
Used to study how people learn concepts or classify
Make up fake concept with rules that define relevant features
Wont influence memory
o The experimenter display stimulus and they must guess, may get feedback
o See all stimuli at once, choose one to ask the experimenter about, feedback
• Indentifying concepts
o Hypothsis testing: generating rules and testing them
• Consider all values of the first positive trial then eliminate
dimensions as you go
• Choose 1+, test until fail then change
o Natural categorize instead artificial
Tend to organize things in hierarchy
• Subordiante (vehicle)
• Basic (car)
• Subordinate (Honda)
Categories are fuzzy
• Ill defined
Features correlated with each other
• Not independent
Some exemplars share more attributes with category than atypical
Some are more basic than others
• Theories of Categorization
o Prototype Models
Store common features of exemplars during learning to form
o Categorization based on instance
Store separate instances of categories during learning
Retrieve instances and compare novel item to them
• Prototypes in Natural Language Categories
o Natural language categories ill –defined, fuzzy no set of absolute rules to
Centre of category rather than boundaries
Membership based on family resemblance Graded membership, some closer to prototype
• Role of Exemplars
o Preserve info about correlated feature
• Theories of Categorization
o Use both prototype and instance-based methods of categorization
o Different methods used for different types of stimuli and at different
stages of category learning
November 15, 2010
• Dual Coding Theory
o Paivio, Smythe, and Yuille
Which pair of words are easier or harder to remember
Differ in imagery ability
Two coding systems (memory is better when in both systems)
• Verbal/abstract system
o Word meaning
• Imagery system/non verbal spatial system
o Generate and represent visual images
o Theories of Representation
• Imagery is epiphenomenon
o Its all in your mind
o All info is in propositional form
Simple relational statement
o No mental pictures
• Links between perception and mental imagery
• Mental rotation task
• Decision time is a function of how many degrees you have
to move the object
• Happens in real time
• True for 2D or 3D
• Imagery leaves a trail of movement