Psychology 439g 572b
1 / 101

Psychology 439G/572B - PowerPoint PPT Presentation

  • Uploaded on

Psychology 439G/572B. Semantic Memory. What is Semantic Memory?. What is Semantic Memory?. General World Knowledge. What is Semantic Memory?. General World Knowledge Facts. What is Semantic Memory?. General World Knowledge Facts Rules. Thou Shalt Not Cheat. What is Semantic Memory?.

I am the owner, or an agent authorized to act on behalf of the owner, of the copyrighted work described.
Download Presentation

PowerPoint Slideshow about 'Psychology 439G/572B' - salena

An Image/Link below is provided (as is) to download presentation

Download Policy: Content on the Website is provided to you AS IS for your information and personal use and may not be sold / licensed / shared on other websites without getting consent from its author.While downloading, if for some reason you are not able to download a presentation, the publisher may have deleted the file from their server.

- - - - - - - - - - - - - - - - - - - - - - - - - - E N D - - - - - - - - - - - - - - - - - - - - - - - - - -
Presentation Transcript
Psychology 439g 572b l.jpg

Psychology 439G/572B

Semantic Memory

What is semantic memory3 l.jpg
What is Semantic Memory?

  • General World Knowledge

What is semantic memory4 l.jpg
What is Semantic Memory?

  • General World Knowledge

  • Facts

What is semantic memory5 l.jpg
What is Semantic Memory?

  • General World Knowledge

  • Facts

  • Rules

Thou Shalt Not Cheat

What is semantic memory6 l.jpg
What is Semantic Memory?

  • General World Knowledge

  • Facts

  • Rules

  • Mathematics

2x2x2x2 = 16

Thou Shalt Not Cheat

What is semantic memory7 l.jpg
What is Semantic Memory?

  • General World Knowledge

  • Facts

  • Rules

  • Mathematics

  • Categories

2x2x2x2 = 16

Thou Shalt Not Cheat

What is semantic memory8 l.jpg
What is Semantic Memory?

  • General World Knowledge

  • Facts

  • Rules

  • Mathematics

  • Categories

  • Concepts

2x2x2x2 = 16

Thou Shalt Not Cheat

What is semantic memory9 l.jpg
What is Semantic Memory?

  • General World Knowledge

  • Facts

  • Rules

  • Mathematics

  • Categories

  • Concepts

  • Language

2x2x2x2 = 16

Thou Shalt Not Cheat

What is semantic memory10 l.jpg
What is Semantic Memory?

  • General World Knowledge

  • Facts

  • Rules

  • Mathematics

  • Categories

  • Concepts

  • Language

  • Meaning

2x2x2x2 = 16

Thou Shalt Not Cheat

This class l.jpg
This Class

  • No text; readings available from me

  • One paper, due at the end of classes

  • Research Paper (40%)

    • Literature Review on topic in semantic memory

    • Summarize current state of literature

    • 20-25 pages; follow APA guidelines

    • Undergraduates: at least 12 articles, 8 experimental, no more than 4 from class

  • Class Participation (40%)

    • First, you really, really have to show up

    • Second, you really, really have to participate in discussion of articles

    • Third, you have to present an article at least once in the term, and moderate the following discussion

  • Weekly Thought Paper (20%)

Weekly thought paper l.jpg
Weekly Thought Paper

  • Students will be required to write a weekly thought paper on the readings they were assigned for that week.

  • No more than 1-2 pages long

  • Should reflect if/how the various articles agreed or came into conflict,

  • Concerns the student had with methodology or interpretations in empirical articles, or

  • Other issues regarding the assigned articles that the student found interesting.

  • These will be collected at the end of class each week.

What we will be studying l.jpg
What We Will Be Studying…

  • The first week(s) should be review for some, new for others; A general summary of human cognition and especially the memory system

  • Methodologies of studying semantic memory

  • Models of Semantic Memory (including, why bother?)

    • Hierarchical Models, Feature List Models, Spreading Activation, Connectionist Models, High-dimensional Models

  • Perceptual Views of Semantic Memory

  • Effects of Age and Dementia on Semantic Memory

  • Categorization

  • Semantic Memory and the Brain

The articles for the undergraduates l.jpg
The Articles (for the undergraduates)

  • Since articles advance the science, they are the primary focus of the seminar

  • Some articles are review, more are empirical

  • It is important to understand both types of articles

  • Experimental papers

    • What is the theory or hypothesis being tested? What are the competitors?

    • Methods: How was the experiment done? What subjects, equipment, statistics were used? What are the IV and DV

    • Results: What happened? Was the theory supported? What would competing theories predict?

    • Conclusions: How do the results relate to the theory? Do you agree with this interpretation?

A quick intro or review of cognition l.jpg
A Quick Intro (or Review) of Cognition

  • Cognition: The group of mental processes and systems used in learning, thinking, comprehending, remembering, planning, imagining, perceiving, attending, and communicating.

A quick intro or review of cognition16 l.jpg
A Quick Intro (or Review) of Cognition

  • Cognition: The group of mental processes and systems used in learning, thinking, comprehending, remembering, planning, imagining, perceiving, attending, and communicating.

  • Or to be easier, cognition is thought

  • Memory:

  • The cognitive processes of acquiring, maintaining, and retrieving information for later use.

  • Storage “space” in which these processes take place

Memory l.jpg

  • “Memory is the most important function of the brain; without it life would be a blank. Our knowledge is all based on memory. Every thought, every action, our very conception of personal identity, is based on memory…. Without memory, all experience would be useless”

    • Eldridge-Green, 1900, p. 1

  • “Life is all memory except for the one present moment that goes by you so quick you hardly catch it going.”

    • Tennessee Williams

  • “Semantics. The curse of man.”

    • Maxwell, 1976

  • Methods of studying memory l.jpg
    Methods of Studying Memory

    • Experimental cognitive psychology:

      • Reaction time experiments, priming, accuracy

      • Well controlled experiments

      • Indirect measures of internal processes

    • Modeling (Cognitive Science)

      • Computational models to understand memory and cognition

      • Semantic networks, connectionist networks

      • Simulations of human performance

    • Cognitive Neuropsychology

      • Brain damaged patients

      • Category specific deficits, dementia, aphasics, lesion data

      • Cannot provide adequate controls

    • Cognitive Neuroscience

      • Imaging technology: PET, fMRI, ERP, MEG

      • Either bad temporal or bad spatial resolution. Localization problematic

    The information processing system l.jpg
    The Information Processing System











    Sensory store l.jpg
    Sensory Store

    • A very complete but brief (<1 second) representation of the stimulus

    • One for each sensory system; different lengths of time for each

    • Stimulus is encoded for further processing

    • Quickly decays; attention must be paid to stimulus or the information will not enter working memory

    • Because it decays so quickly, how can it’s capacity be measured (or properly reported by experimental subjects)?

    Sperling 1960 l.jpg
    Sperling (1960)

    • A clever experiment in which subjects are presented with an array of letters for 50 msec

    • If subjects are asked what they saw, they can only report about 4 letters

    However, in a second condition a tone is played after the array has been removed. Subjects are told that a low tone means report the bottom row, a middle tone means the middle row, and a high tone the high row.

    Subjects can remember 3 out of 4 letters.

    Since the tone is played after the array is no longer present, we can infer that people are actually recalling 9 out of 12 letters, and can only average 4 letters in the first condition because the memory decays before they can complete the report

    The information processing system22 l.jpg
    The Information Processing System











    Working memory l.jpg
    Working Memory

    • Also called Short-term memory

    • Both storage and processor; The more processing that needs to be done, the less storage is available

    • Quite limited capacity, compared to sensory store (very complete but brief) and permanent memory (limitless over a human lifespan)

    • The famous number: 7 ± 2 (Miller, 1956)

    • Once can hold 7 ± 2 items in working memory

    • However, this can be moderated, somewhat, be re-encoding the data. This process is sometimes called “chunking”

    Working memory example l.jpg
    Working Memory Example

    • How many letters can you remember:

    A D F B I G U I B R B C I A B M W T V D C I B M

    Now, try again:

    A D F B I G U I B R B C I A B M W T V D C I B M

    • Re-encoding the letters into larger but more meaningful chunks allows you to remember more of the letters (Miller, 1956). This is why phone numbers, SIN numbers, etc. are separated into sections

      • 519-661-2111

    • This process does require some mental effort (takes longer). It also requires the help of permanent memory.

    Baddeley hitch 1974 l.jpg
    Baddeley & Hitch (1974)

    • A more detailed version of working memory

    The information processing system26 l.jpg
    The Information Processing System











    A competing theory of memory l.jpg
    A Competing Theory of Memory

    • Craik and Lockheart proposed a “levels of processing” system.

    • Any perceived stimulus will receive processing

    • However, those that receive minimal attention are only processed in shallow memory

    • Those that receive more intentional processing are stored more deeply and more meaningfully in memory

    Levels of processing l.jpg
    Levels of Processing

    • Imagine a task in which pairs of words are presented

    queen broom

    In one condition of the experiment, subjects have to mentally count the number of vowels for both words


    In another level, subjects have to mentally make a sentence out of the target words:

    The queen had never held a broom

    Then the subjects are presented with a recall task for the target words. Guess which group did better?

    Permanent memory l.jpg
    Permanent Memory

    • Also called Long-term memory. Informally, it’s what’s gotten past your attentional and working memory bottlenecks and actually been stored permanently in your brain.

    • “Divided” into Episodic Memory and Semantic Memory

    • Episodic memory, simply, is memory tied to specific events in your life

      • For instance, your 13th birthday

      • Your episodic memory and mine are quite different

    • Semantic Memory is all information not tied to a specific episode: language, world knowledge, concepts, etc.

      • Semantic memory can be inferenced: Does a tiger breathe?

      • Your semantic memory and mine are quite similar

    Memory example l.jpg
    Memory Example

    • Bird, cold, antarctic, feathers, beak, eggs, swims, aquatic, warm-blooded, can’t fly, eats fish, waddles, black and white, tuxedo

    Related issues for memory l.jpg
    Related Issues for Memory

    • Automatic vs. Controlled Processing

    • Top-down (conscious) processing vs. bottom-up (data-driven) processing

    • Representation

    • Brain Function

    Automatic vs controlled processing l.jpg
    Automatic vs. Controlled Processing

    • Automatic Processes

      • Performed without conscious awareness

      • Often unintentional (Stroop task) Green, Red

      • Generally performed in parallel

      • Consume few attentional resources

    • Controlled Processes

      • Require conscious control

      • Performed serially

      • Take longer to execute

    • fully controlled<-----------------------------> fullyautomatic (learning affects this)

    • 4 x 4. Now, 444 x 44

    Top down vs bottom up processing l.jpg
    Top-Down vs. Bottom-Up Processing

    • Also called conscious process vs. data-driven processing

    • Examine the following pictures:

    Top-down processing relies on permanent memory to overcome problems with actually present stimuli

    Usually used automatically and without effort

    Representation l.jpg

    • How are things stored in memory? What form do these mental representations take?

    • External representations: words, pictures

    • Internal representations: Symbolic, distributed

    Brain function l.jpg
    Brain Function

    • How the brain functions is much more important to cognitive psychology now than 20 years ago

    • Neural networks use attributes of neuronal function to solve complex problems

    • Neuroimaging studies are exploding; these attempt to locate the neural substrate of cognitive tasks

    • Brain changes in older adults, demented adults, schizophrenia, etc.

    • Category specific deficits in lesioned patients and what this can tell us about semantic storage in the brain

    The spatial metaphor for memory l.jpg
    The Spatial Metaphor for Memory

    • The concept of memory as a “space” in which things are “stored” is far older than the study of psychology

    • This space tends to be imagined in three dimensions

    • Things closer in this space tend to be related

    • This concept is even carried on in modern semantic models

      • Spreading activation models

      • High-dimensional models (but more than 3 dimensions)

    • This metaphor has gone hand in hand with the metaphor for memory that was dominant through the 70s and even 80s: the computer

    • None of these simple metaphors for memory is really flexible enough to account for all that memory can do, and do easily

    The information processing system37 l.jpg
    The Information Processing System











    The computer metaphor l.jpg
    The Computer Metaphor

    Hard Disk

    Human Input







    Multiple memory stores l.jpg
    Multiple Memory Stores

    • Increased sophistication over older, unitary stores

    • Most memory researchers still think in these terms, even when their own models are a good deal more complex

    • Multiple stores (sensory store, WM, etc) allow for a relatively easy explanation of why different memory tasks are performed with greater or lesser amounts of difficulty

      • For instance, why do you remember your 5th grade teacher but you can’t remember the phone number you just heard 5 minutes ago?

    • For the most part, these constructs have held up in the intervening 30 years, but each memory store has become considerably more complicated in theory

      • STM  Working Memory

      • LTM  Semantic vs. Episodic Memory; Semantic vs. Associative Memory; Implicit vs. Explicit Memory

    Long term memory l.jpg
    Long Term Memory

    • How do we use long-term memory?

      • Recall: Yes, I remember my ex-girlfriend’s e-mail address

      • Recognition: Hey, isn’t that the prof who failed me?

      • Procedural: It’s like riding a bicycle… you never really forget

      • Musical: How did that song go again?

      • Linguistic: How can you finish the word: dip_ _ _ _

        • It’s DIPLOMA. Get your mind out of the gutter!

      • Episodic: “So, that’s when the police showed up…”

      • Semantic: An ostrich is a bird. A whale is a mammal.

      • Explicit memory: Yes, the word AARDVARK was on the list

      • Implicit memory: How can I complete the word stem WH___

      • Top-down processing: “I’m guessing they meant “Public Worship”

    • Note: there exists some overlap between some of these concepts, like explicit memory and recall/recognition

    Recall vs recognition l.jpg
    Recall vs. Recognition

    • These both involve conscious processes

    • Recognition: Have I seen something before?

    • Recall: What did I see?

      • Cued recall: OK, you saw some animals. What were they?

      • Free recall: Write down all the words you saw on the list

    • Recognition tends to be a lot better than recall. Up to 95% of a 1500-word word list can be identified as “previously seen”, i.e. recognized.

    • Why is recall more problematic than recognition?

      • Two-process model?

        • Recall: 1) search process, 2) recognition process

        • Recognition: 1) Recognition process

      • Accounts for frequency paradox: for recall, high frequency words are better, but for recognition, low frequency words are better.

        • Common words have more associative links; search is easier

        • Uncommon words have less baggage: old/new judgment is easier

      • But, what is recognition? Theory is vague.

    More complex theories of recall recognition l.jpg
    More complex theories of recall/recognition

    • Multiple route approaches: There are 2 or more ways of encoding information in both recall and recognition

    • Recall (Jones, 1982)

      • Direct route: Cue allows direct access of TBR information

        • DOG  CAT

      • Indirect route: Cue allows recall via inference

        • LEASH  DOG  CAT

    • Recognition (Gardiner & Java, 1993)

      • Familiarity: (know response) “I know I’ve seen that guy before”

      • Contextual information (remember response) “Oh yeah, that’s the guy who ran into my car.”

      • Both familiarity and actual recall are part of recognition

    • These processes may be related to the distinction between semantic and episodic memory

    Implicit and explicit memory l.jpg
    Implicit and Explicit Memory

    • Explicit memory: These are the processes tapped by recall and recognition processes: conscious remembering

    • Implicit memory: When previously presented stimuli affect subsequent behavior, though they cannot consciously be recalled.

    • Example:

      • A list of words is presented; the subject studies the list

      • The subject is then given a word-stem completion task:

        • WH______ -- CH______ -- JU______

          and told to complete the words with any valid English word,

          but not to use any words in the list previously studied

  • Results: subjects much more likely to complete the word stems with words in the lists, though A) they were told not to, and B) these completions are not the most frequent for those word stem (e.g. whale vs. whisper)

  • Implicit and explicit memory44 l.jpg
    Implicit and Explicit Memory

    • This process is called “repetition priming” and is similar to the “semantic priming” we will learn much about.

      • An exposure to an instance “primes” that word for later use, though the word is below the conscious threshold

    • The distinction between implicit and explicit memory was first demonstrated in anterograde amnesiacs

      • As expected, the people with amnesia were very poor at conscious (explicit) retrieval attempts.

      • However, they demonstrated normal or near normal performance on implicit tasks (word-stem completion, anagram solution, degraded picture identification)

    • All a good deal of evidence that implicit memory is less affect by aging than explicit (though this is controversial)

    • These results all suggest a distinction between implicit and explicit memory processes.

    Semantic vs episodic l.jpg
    Semantic vs. Episodic

    • Tulving (1972) proposed the distinction between semantic and episodic memory, though concepts of personal and impersonal knowledge had been around for decades.

      • Semantic memory: World knowledge, facts, language

      • Episodic memory: Memory tied to specific events

    • Now Tulving and collaborators (Wheeler, Stuss, and Tulving (1997) argue the essential feature of episodic memory is its autobiographical nature, rather than the type of material encoded or stored

    • Also, the creation of semantic memory necessarily requires episodes, so episodic memory is integral to semantic memory.

    • How does semantic memory affect episodic memory?

    Autobiographical and episodic memory l.jpg
    Autobiographical and Episodic Memory

    • Is there a distinction between the two?

    • Nelson, 1997, p. 357: “What I ate for lunch yesterday is today part of my episodic memory, but being unremarkable in any way it will not, I am sure, become part of my autobiographical memory – it has no significance to my life story”

    • There are often errors in autobiographical memories as in episodic memories. For instance, up to 40% cannot remember minor hospitalizations when asked about them only a year later.

    • Structures of Autobiographical memory:

      • Lifetime periods (school, work)

      • General events (holiday in Europe)

      • Event-specific : memories and images for a single event

    Do we forget things from ltm l.jpg
    Do we forget things from LTM?

    • Or are they just “retrieval failures?” Forgetting itself as a concept needs to be defined.

    • First theory: Decay

      • Memories actually fade over time, like an old B&W photo

      • Supported by behaviorist theory: memories are just like habits

      • However, not supported by cognitive theory (or experimental results). This theory would predict that older information is always remembered more poorly than newer information

    • Second Theory: Interference

      • Instead, some researchers argued that intervening information interfered between the encoding and retrieval processes

    An example of an interference study l.jpg
    An example of an interference study:

    • Subjects learn a paired-word list:

      • List A

        tall-bone park-flea

        plan-leaf grew-cook

        nose-fight rabbit-few

    • They learn this list so that when any first word is presented, they can reply the second word

    • However, then they are asked to learn a new list:

      • List B

        tall-safe park-house

        plan-window grew-pencil

        nose-bench rabbit-card

    • When this is done, they are asked to use the List A pairs one more time. As you might guess, they cannot do it well.

    But is this forgetting l.jpg
    But is this forgetting?

    • Let’s define forgetting as actual disappearance from memory. It’s gone, caput, finito. Does this actually happen?

    • Evidence suggests that LTM is actually permanent, except for:

      • Alzheimer's and other dementias

      • Brain injury

    • Instead, the current thought on LTM, now often called “Permanent Memory” (what a giveaway) suggests that retrieval failures are the reason we cannot instantly and perfectly recall information.

    • So, what are these “retrieval failures” and what causes them?

    Retrieval failures l.jpg
    Retrieval failures

    • First, however: Did we really learn it?

      • “We must never underestimate one of the most obvious reasons for forgetting, namely, that the information was never stored in memory in the first place” (Loftus, 1980, p. 74)

      • Attention must be paid to a stimulus. The stimulus must be rehearsed in some manner. Both of these bottlenecks must be overcome in order for information to get into PM

    • OK, so we learned it. What are Retrieval Failures?

    • A RF familiar to you is the Tip-of-the-tongue (TOT) phenomenon.

    • You are in “TOT state” when you are unable to access information you know you have stored in PM

    • However, you usually have access to some information about the word, often phonological

    Slide51 l.jpg

    • Jeez, what’s the word I’m looking for. Um, exercise, estimate. Man, I know this. Has to do with being obscure. Escape. Oh well, I give up. <Hours later> Wait, I know! Esoteric!

    • Often have access to phonological information during such a retrieval failure; some part of the representation has been activated.

    • There is usually a distractor like the ones above which the subject knows is incorrect but cannot stop retrieving.

    • Remembering often only happens after you have stopped thinking about the word

    • TOTs increase with age

    • More common with Proper names

    • Not common in high-frequency words

    Research on retrieval failures l.jpg
    Research on retrieval failures:

    • Tulving and Pearlstone (1966) did a fairly simple memory task:

      • 2 groups of people studied the same list of 48 items

        Animals Crimes Professions

        dog murder teacher

        giraffe treason engineer

        elephant robbery trucker

        penguin extortion programmer

    • The first group was told to recall as many words as possible. The second group was given the category names (animals, crimes, etc.) as a retrieval cue

    • The free-recall group remembered 40% of words, the cued-group: 62%

    • More information was encoded than could be recalled without cues

    Research on retrieval failures53 l.jpg
    Research on Retrieval Failures

    • Tulving and Pearlstone (1966) made a distinction between availability and accessibility

      • Available: actually in memory (learned)

      • Accessible: the degree to which this memory can successfully be retrieved

    • Cues increase the accessibility of information; Availability always depends on how well the information was learned in the first place

    • A metaphor: The misshelved book in the library

      • The book is still there, in the library, but it is temporarily inaccessible . It cannot be located or retrieved while it is on the wrong shelf

    • Lost from memory: No

    • Lost in memory: Possibly

    Encoding specificity l.jpg
    Encoding Specificity

    • Memories are not isolated bits of data

    • Instead, each memory is associated with other relevant memories

    • For instance, the old song that takes you back to your high-school years

    • You don’t just remember the song, but also a great deal of other information: who you were dating, your school, your friends, racing down Highway 178 in a ’71 Mustang. Good times.

    • The other information available at the time of encoding became part of the representation of the song

    • Thompson and Tulving (1970) measured this empirically

    Encoding specificity thompson tulving 1970 l.jpg
    Encoding Specificity (Thompson & Tulving, 1970)

    • T&T presented a list of words. Some of these words were combined with a cue word. The subjects were told that they didn’t have to learn the cue word but that it might help with later recall

      • Some of these cue words were highly associated with the targets:

        • ice – cold

      • Some were less associated

        • wind – cold

    • Results:

      • When high-associate cues were presented, subjects used them as cues to remember the word. They were also helpful as retrieval cues in words in which no cue was originally presented

      • However, when low-associate cues were presented, only these served as effective retrieval cues. High-associate cues added later provided no benefit for later retrieval

      • In other words, if you had studied wind-cold, “ice” was not an effective retrieval cue for “cold”

      • The original encoding context is very important for later retrieval

    Is memory accurate l.jpg
    Is memory accurate?

    • How well do we remember what we remember?

    • Recognition memory is quite robust; based largely on familiarity

    • Recall is helped greatly by cues

    • Semantic memory can affect episodic memory for events; Memory for specific events can be affected by later information (Loftus & Palmer, 1974)

    • Items not on word lists can easily be manipulated so that they are reported as actually being on the list

    • Eyewitness testimony can easily be affected by improper procedures or leading questions

    Loftus palmer 1974 exp 1 l.jpg
    Loftus & Palmer (1974) Exp 1

    • Is episodic memory affected by later linguistic (semantic) word-choice?

    • L&P E1: Phrasing of the question: How fast were the cars going when they xxx each other (hit, smashed, collided, bumped, contacted)?

    • DV: Estimated speed of the vehicle

    • Results:

      • Significant effect of verb

      • People bad at estimating speed

    • Interpretations of E1:

      • Biasing an uncertain judgment?, or

      • Changing memory itself?

    • E2 attempted to narrow down possible interpretations of E1

    Loftus palmer 1974 exp 2 l.jpg
    Loftus & Palmer (1974) Exp 2

    • As before (E1) but 1/3 of the subjects were given the word hit, 1/3 smashed, and the rest were not asked about speed of the collision

    • A week later, the subjects were tested again. This time, among the filler questions, the critical question was asked: “Did you see any broken glass?”

    • Yes: .32 with smashed

      .14 with hit

      .12 with control (no speed judgment)

    • Conclusion: Because judgment of a specific episode is affected, and because the probability of saying “yes” is not independent within a speed estimate range, L&P concluded that actual memories are being affected, not just biases towards greater speed estimates

    Roediger and mcdermott 1995 e1 l.jpg
    Roediger and McDermott (1995) E1

    • Modeled after a study by Deese (1959), revealing a great deal of false recognition and recall

    • Experiment 1:

      • Used six lists of words grouped by some unifying concept

      • Chair: table, sit, legs, seat, soft, desk, arm, sofa, wood, cushion, rest, and stool (12 associates)

      • Recogn test: 6 critical lures, 12 unrelated words, 12 weakly related words (30 non-studied words) plus 12 studied words

      • Recall test: Done right after each list

      • Results:

        • .40 recalled critical lures

        • .84 recognized critical lure (HR was only .86)

        • Much higher than for non-presented, weakly related words

    • Experiment 2 attempted to replicate this result with more words and longer lists.

    Roediger and mcdermott 1995 e2 l.jpg
    Roediger and McDermott (1995) E2

    • Experiment 2:

      • Used a wider set of materials than E1; 24 15-item sets similar to E1 (included stimuli from E1)

      • Examined recognition rates for lists that had been previous recalled after stimuli presentation and for lists that had not previously been recalled

      • Determined the false alarm rates for critical lures when the relevant list of related words had not been presented

      • Obtained subj judgments about their own experience when reporting lure words as previously seen.

        • Used a remember-know judgment to see how memories are being encoded: Is a false sense of familiarity to blame, or are items actually being misremembered as being on the study lists?

    • Results

      • .55 recall of critical lures.

      • .81 recognition in study+recall cond.; .72 recogn in study+arithm.

    Roediger and mcdermott 1995 conc l.jpg
    Roediger and McDermott (1995) Conc.

    • High levels of false recall and recognition

    • High levels of false “remember” judgments in recogn.

    • False judgments made with high confidence

    • Recall increased both accurate recognition of studied items and false recognition of critical lures

    Bartlett 1932 war of the ghosts l.jpg
    Bartlett (1932) War of the Ghosts

    One night two young men from Egulac went down to the river to hunt seals, and while they were there it became foggy and calm. Then they heard war-cries, and they thought: “Maybe this is a war-party.” They escaped to the shore, and hid behind a log. Now canoes came up, and they heard the noise of paddles, and saw one canoe coming up to them. There were five men in the canoe, and they said:

    “What do you think? We wish to take you along. We are going up the river to make war on the people.”

    One of the young men said: “I have no arrows.”

    “Arrows are in the canoe,” they said.

    “I will not go along. I might get killed. My relatives do not know where I have gone. But you,” he said turning to the other, “may go with them.”

    So one of the young men went, but the other returned home.

    And the warriors went on up the river to a town on the other side of Kalama. The people came down to the water, and they began to fight, and many were killed. But presently the young man heard one of the warriors say: “Quick, let us go home: that Indian has been hit.” Now he thought: “Oh, they are ghosts.” He did not feel sick, but they said he had been shot.

    So the canoes went back to Egulac, and the young man went ashore to his house, and made a fire. And he told everybody and said: “Behold I accompanied the ghosts, and we went to fight. Many of our fellows were killed, and many of those who attacked us were killed. They said I was hit, and I did not feel sick.”

    He told it all, and then he became quiet. When the sun rose he fell down. Something black came out of his mouth. His face became contorted. The people jumped up and cried.

    He was dead.

    Subjects recall l.jpg
    Subjects’ Recall

    There were two men in a boat, sailing towards an island. When they approached the island, some natives came running towards them, and informed them that there was fighting going on on the island, and invited them to join. One said to the other: “You had better go, I cannot very well, because I have relatives expecting me, and they will not know what has become of me. But you have no one to expect you.” So one accompanied the natives, but the other returned.

    …Anyhow, the man was in the midst of the fighting, and was wounded. The natives endeavored to persuade him to return, but he assured them that he had not been wounded.

    I have an idea that his fighting won the admiration of the natives.

    The wounded man ultimately fell unconscious. He was taken from the fighting by the natives.

    Then, I think it is, the natives describe what happened, and the seem to have imagined seeing a ghost coming out of his mouth. Really it was kind of materialization of his breath. I know this phrase was not in the story, but that is the idea I have. Ultimately, the man died at dawn the next day.

    Bartlett 1932 l.jpg
    Bartlett (1932)

    • Bartlett wanted to determine how people’s memories were shaped by their expectations.

    • He used a Indian folk tale with structure and ideas that ran contrary to typical Western storytelling (and therefore, expectations)

    • He found that subjects tended to remember the story in a format much closer to Western expectations. They “normalized” the study, removing more mystical or non-linear elements. Plus, info was added.

    • Precisely, Bartlett said that subjects reconstructed the story rather than reproducing it verbatim, and that they used scripts (schemas) to fill in details that they could not recall.

    • Semantic memory (expectations, or scripts) is affecting episodic memory (ability to accurately recall the story)

    Sulin and dooling 1974 l.jpg
    Sulin and Dooling (1974)

    • Authors presented subjects with a short history:

      • “Gerald Martin strove to undermine the existing government to satisfy his political ambitions… He became a ruthless, uncontrollable dictator. The ultimate effect of his rule was the downfall of his country”

    • Other participants were given the same paragraph, except Gerald Martin was replaced by Adolf Hitler.

    • The subjects who read the paragraph with Hitler were far more likely to (incorrectly) remember reading the sentence: “He hated the Jews particularly and so persecuted them.”

    • In this case, the subject’s semantic knowledge of Hitler cause distortions in their episodic memory of the passage that was read. When Hitler was absent, these distortions were not present.

    Other experiments l.jpg
    Other Experiments

    • Kintsch (1977) argued for another type of memory, constructive memory. Consider the following sentences:

      • The careless driver tossed his cigarette out the window.

      • The forest fire raged for days.

    • It is an inference that the forest fire was started by the cigarette, but after a delay of more than a day, subjects reported explicitly being told that the cigarette caused the fire, rather than remembering that they inferred it.

    • So, Kintsch’s theory is for three types of memory

      • Reproductive: Accurate recall of details

      • Constructive: In which inferences made during language comprehension become part of memory

      • Reconstructive: In which additional information is incorporated into the memory after the fact; semantic memory, schemas, elaboration, leading questions

    Semantic memory as aid to episodic l.jpg
    Semantic Memory as Aid to Episodic?

    • Read the following passage

      • With hocked gems financing him, our hero bravely defied all scornful laughter that tried to prevent his scheme. “Your eyes deceive,” he had said, “an egg not a table correctly typifies this unexplored planet.” Now three sturdy sisters sought proof, forging along sometimes through calm vastness, yet more often over turbulent peaks and valleys. Days became weeks as many doubters spread fearful rumors about the edge. At last from nowhere welcome winged creatures appeared signifying momentous success. (Dooling & Lachman, 1971, p. 217)

    • Memory for this passage was rather poor

    • However, when the subjects were told that the name of the passage was “Christopher Columbus Discovering America” before reading, they remembered the passage much better

    • Knowledge aids episodic recall of this passage.

    Current thought in semantic episodic memory l.jpg
    Current Thought in Semantic/Episodic Memory

    • Loftus (1992) has argued a moderated version of her earlier position:

      • Misinformation acceptance: subjects can accept misleading information after an event. Likelihood of this happening increases with amount of time since event.

    • Source misattribution: Information from one source gets confused with information from another source. The more similar the two events, the more likely this is to occur (Allen & Lindsey, 1998)

      • Two unrelated slides, but with similar features (e.g. a Coke can in one and a Pepsi can in another) can lead subjects to substitute information from the post-event into the original event

    • Trivial details? It is harder for experimenters to distort highly memorable events (robberies, etc) (Fruzetti et al., 1992) but it is not impossible (Loftus, 1992).

    Methods of studying memory again l.jpg
    Methods of Studying Memory (again)

    • Experimental cognitive psychology:

      • Reaction time experiments, priming, accuracy

      • Well controlled experiments

      • Indirect measures of internal processes

    • Modeling (Cognitive Science)

      • Computational models to understand memory and cognition

      • Semantic networks, connectionist networks

      • Simulations of human performance

    • Cognitive Neuropsychology

      • Visual-field experiments, cerebral assymetries

      • Brain damaged patients

      • Category specific deficits, dementia, aphasics, lesion data

    • Cognitive Neuroscience

      • Imaging technology: PET, fMRI, ERP, MEG

      • Either bad temporal or bad spatial resolution. Localization problematic

    Inferences based on reaction time l.jpg
    Inferences based on Reaction Time

    • Why do we use reaction time to infer mental processing?

    • Example:

      • 2 x 8 takes less time to solve than 7 x 9 for both young children and college students

      • But why does it take less time? What differs between these two problems that causes RTs to rise?

        • Any difference in numbers of digits in problem or answer? No.

        • Harder to perceive/encode higher digits? (2 vs. 9) No.

        • Does it take longer to start saying “16” vs. “63”? No.

      • We can then conclude (unless we think of some other reason to be tested) that the increase in RT is due to mental processing differences. But what causes these differences?

        • frequency? Maybe.

        • Retrieval vs. computation? Maybe.

        • Initial learning? Maybe.

    • Point: RT differences lead to generation of hypotheses, and then can be used again to test these hypotheses.

    Semantic priming and rt experiments l.jpg
    Semantic Priming and RT experiments

    • Semantic priming occurs when a earlier word activates another, semantically related word.

    Semantic priming and rt experiments75 l.jpg
    Semantic Priming and RT experiments

    • Semantic priming occurs when a earlier word activates another, semantically related word.

    • This process is demonstrated by faster reaction time to the second word on some task.

      • Prime word is presented

      • No response to prime word

      • Target is presented

      • Subject makes a response to target word

      • Reaction time (RT) is measured

      • Times to respond to target are compared across conditions of semantic relatedness or unrelatedness

    • Dominant method of measuring the processes involved in lexical access and retrieval. First done by Meyer and Schaneveldt (1971).

    Reaction time exp terminology l.jpg
    Reaction Time Exp. Terminology

    • SOA: Stimulus Onset Asynchrony; the amount of time between the beginning of the first stimulus (prime) and the beginning of the second stimulus (target)

    • ISI: Interstimulus Interval: amount of time between end of prime stimulus and beginning of target stimulus. This value can be zero

    • So, SOA = Prime Duration + ISI

      • Or, if your prime is present for 250ms, and your ISI is 150, then your SOA is 400ms (a medium-sized SOA; responding to a light takes about 150-200ms)

    • Lexical Decision: The task: Is it a real word?

      • aardvark vs. gertunic

    • Naming: Say the word.

    • Semantic Task: Example: Is this word animate or inanimate?

    Priming terms l.jpg
    Priming Terms







    • Priming: RT advantage for target with related prime vs. same target with unrelated or neutral prime

    • Related: cat – dog 650 ms

      Unrelated: table – dog 700 ms

    • Prime advantage: 50ms

    • Four word pair relationship groups:

      • Semantic: bear-cow

      • Associated: cheese-mouse

      • Unrelated dog-table

      • Mediated lion-stripes

    So why is this done l.jpg
    So, why is this done?

    • Tasks chosen are simple; if not speeded, accuracy rate should be 100%. Even with speeded component, accuracy is very high

    • Reaction time more important than accuracy

    • No response to the prime allows for measurements of extremely brief processes

    • Indirect measure of word activation processes, but at a temporal resolution still unmatched by modern imaging processes

    • Very robust effect; easy to set up task computers

    • Can be used in conjunction with visual field differences to study cerebral asymmetries

    • Can show priming effect at SOAs so brief the subject is unaware a prime occurred (Marcel, 1980)

    • Can show which meaning of an ambiguous word is being accessed at what time (bank vs. bank)

    Time course of priming l.jpg
    Time Course of Priming

    • Neely (1977) studied the process of semantic priming over time.

    • Subjects were told to expect a body part (ARM) when a bird (ROBIN) was presented as a prime. This is an expectancy effect

    • SOA was an IV in the experiment; SOA ranged from 250 ms to 2000 ms.

    • At the shortest SOA, only semantically related concepts are primed: This is called facilitation

    • At longer SOAs, the expectancy condition kicked in, and priming was found for body parts, but not for other birds. The depression of priming for birds is called inhibition.

      • Inhibition normally occurs to reduce competitors or unrelated concepts, but can be overridden by controlled processes to inhibit sem. related concepts

    • At short SOAs, only automatic processes can occur. At longer SOAs, controlled processes can override automatic processes

    • So, priming can be both automatic and controlled, depending on the task

    Swinney 1979 l.jpg
    Swinney (1979)

    • Used reaction time and priming to detect which meaning of an ambiguous word is being presented.

    • For instance, bug (insect) vs. bug (listening device)

    • Used context (sentences talking about spies) and lexical decision to determine what meanings were active

    The spy had been careful; no trace of his visit to the Russian embassy was visible. Cleverly hidden under the desk, however, a bug recorded everything said in the conference chamber

    • At various points after “bug” a word would appear over the text and the subject had been instructed to make a lexical decision to it:


    Swinney 197981 l.jpg
    Swinney (1979)

    • Analysis of RT data demonstrated that the dominant meanings (based on frequency) were primed at short SOAs

      • Bug-termite

    • Subordinate, but contextually appropriate meanings were activated at longer SOAs

      • Bug-agent

    • And that both meanings were activated at medium SOAs (450ms)

    • This finding suggests that when ambiguous words are encountered, all possible meanings are activated. At short SOAs, the dominant meaning is activated first, and this usually wins out, but controlled processing (i.e. context of the paragraph) can suppress this meaning and activate the subordinate meaning.

    What about the brain l.jpg
    What about the brain?

    • What can studies about the brain tell us about the memory systems?

    • Studies on source amnesia, in which the source of a fact or piece of information is lost, demonstrated that frontal lobe damage was typically present.

    • Damage to the prefrontal cortex causes episodic memory problems

    • In PET studies, the prefrontal cortex is more active during episodic memory retrieval than semantic memory retrieval

    • Different areas of the brain also appear to be active during implicit and explicit memory tasks

    • Category specific deficits: suggests semantic memory at least partially localized

    How can we look at brain function l.jpg
    How Can We Look at Brain Function

    • Cognitive Neuropsychology

    • Category Specific Deficits

      • Warrington and Shallice (1984)

        • Patients with herpes simplex encephalitis

        • Damaged the medial temporal lobes

        • Patients very good at identifying inanimate objects

          • by verbal description or picture

        • Very bad at identifying living things or foods

        • Other studies have shown this effect with other types of objects, and also evidence of double dissociation.

        • These effects have also been simulated with connectionist models

      • What does this tell us about the organization of concepts in memory?

        • One interpretation has been a sensory/functional dichotomy, where living things are more defined by their sensory characteristics and man-made objects better defined by their function.

    Positron emission tomography l.jpg
    Positron Emission Tomography

    • Require administration of a positron-emitting radiotracer

    • Used to measure cerebral blood flow (CBF) or cerebral metabolic rate for glucose (CMRgl)

    • Water containing 15O (bombarded with protons so it is now a positron emitting radioisotope) is commonly used to measure CBF

    • Glucose containing 18F is used to measure metabolic rate

    • When emitted, a positron travels a short distance before encountering an electron

    • When the two meet, they are annihilated, and two high energy photons are emitted in opposite directions

    • These photons are picked up by detectors; by detecting many of these photons, CBF or CMRgl can be localized

    Pet scans l.jpg
    PET Scans

    • Spatial resolution still limited compared to fMRI, but better than ERP or MEG

    • Temporal resolution poor; need to study stimuli in blocks to overcome 20-60 second image time

    • Radiotracer limits number of possible measurements

    Typical pet experimental method l.jpg
    Typical PET Experimental Method

    • 6-12 Subjects

    • Block design: experimental stimuli are presented together, then control stimuli are presented together

    • Compare brain states for control task vs. brain state for experimental task

    • “Subtract” similarities so that only experimental brain activations are shown

    • Smooth images; generate 3D model; compare across subjects; sometimes combine with MRI data to improve resolution

    • Compare statistically; correct for multiple measurements

    • Compare findings to localization information from lesion and other imaging studies

    Functional magnetic resonance imaging l.jpg
    Functional Magnetic Resonance Imaging

    • No radiation; strong magnetic fields

    • Uses BOLD imaging (Blood Oxygen Level Dependent), measures hemodynamic response

      • Measure of how much oxygen a particular brain location needs during some task

    • Excellent spatial resolution (~ 1mm)

    • Adequate temporal resolution (~ 1-2 seconds in best)

    • More available than PET

    • More expensive, more data to interpret

    • Temporal resolution still not as good as other methods, but very good for an imaging technique

    • Since no metal can be inside the chamber, experimental design sometimes difficult

    • Subject must be very still

    Fmri process l.jpg
    fMRI Process

    • Powerful magnet aligns hydrogen atoms to all spin in the same direction

    • When released, the atoms return to their original spin

    • Oxygenated blood and deoxygenated blood have different spin characteristics. Deoxygenated hemoglobin is paramagnetic and suppresses signal from surrounding tissue

    • Therefore, fMRI can detect levels of oxygenated blood to deoxygenated blood

    • Information is used to generate voxels. Voxels combined to make 3D image of brain

    • Activation (as measured by blood flow) can be constructed from this information

    Typical fmri experimental method l.jpg
    Typical fMRI Experimental Method

    • Items presented to subject through mirror system or non-electronic earphones

    • Head immobilized by bite bar fixed with dental impression

    • Subject cannot speak, so response (if necessary) is to squeeze a ball (gets picked up in motor areas of the brain)

    • In other experiments, subject is simply asked to mentally generate stimuli (like a verb generation task)

      • What do you do with scissors? Pencil? Basket? Keyboard?

    • Originally, activations between experimental and control stimuli were subtracted from each other. Now, more sophisticated sampling techniques are employed

    • Still, much data transformation required to obtain image

    Erp event related potentials l.jpg
    ERP: Event Related Potentials

    • Based on electrical signals from neurons, rather than blood flow, glucose use or neurotransmitter release

    • Direct measure of neural activity

    • Because electrodes placed on skin, however, spatial resolution is very poor

    • Temporal resolution, on the other hand is quite good

      • Can study traditional cognitive tasks with little modification

    • Brain electrical patterns at various periods in response to various stimuli

    • For instance, an incongruous sentence ending leads to a large N400, a negative voltage spike at 400ms

      • The tired man kicked off his shoes and flopped on the alligator.

    • Less expensive than other imaging techniques

    Erp process l.jpg
    ERP Process

    • A number of electrodes are placed on the scalp

    • People perform the cognitive task being studied

    • ERP waveforms (brain activity) measured

    Magnetoencephalography meg l.jpg
    Magnetoencephalography (MEG)

    • Like fMRI, based on magnetism

    • However, unlike fMRI, which measures blood flow and oxygen usage, MEG measures the ionic current generated by neural activity itself

    • Like ERP, has extremely fine temporal resolution

    • Unlike ERP, can provide localization information

    • Information from MEG scan typically placed over a brain map; can see localization and size of magnetic currents over time

    • Like ERP, can be used to examine normal cognitive tasks without blocking and with normal verbal or LDT response

    Not stored in a single place connectionism l.jpg
    Not stored in a single place: Connectionism

    • Also called neural networks, parallel distributed models

    • In connectionist models, information is distributed across a network of nodes and connections.

    • This highly parallel process is much more similar to brain function than to the old computer metaphor

    • Connections have weights

    • Nodes have activation

    • Input is summed weights x activations of input nodes

    Not stored in a single place connectionism98 l.jpg
    Not stored in a single place: Connectionism

    • Networks can have layers: Input and output layers that can receive and provide data, and hidden layers that process information

    • A representation for a given concept is the pattern of all node activations and all connection weights throughout the entire network

    • The same network can store the patterns for many different input/output patterns at once

    Connectionist advantages l.jpg
    Connectionist advantages:

    • Connectionist networks behave in ways that are similar to human memory performance:

      • Content-addressable

        • Show with the vampire, blue monster, and the bird?

      • Can create prototypes and categorize.

        • What do bed, doze, pillow, dream,pajamas, and nap have in common?

      • Can learn from experience and generalize

      • Degrade gracefully

      • Complex behaviors come from simple constraints

    • Some problems:

      • Complex models often learn by knowing what the correct output should be and comparing the models’ performance.

      • This is not similar to neuron function, or frequently, learning performance in humans

    Readings for next week l.jpg
    Readings for Next Week:

    Background for the Unfamiliar:

    Chapter 1 of Ashcraft (1998) and

    Chapter 6 of Eysenck and Keane (2000)

    Methodologies of Studying Semantic Memory:

    Meyer and Schaneveldt (1971)

    Neely (1991). Semantic priming effects in visual word recognition

    The goal of this course l.jpg
    The Goal of This Course

    • Have fun reading some papers about memory, particularly semantic memory

    • Shine some light on some fairly obscure processes

    • Get you thinking about the central problem of cognitive psychology: semantics, or meaning