04/01/2014

Learning, Memory, Attention

Attention
o Affects learning
No attention = no learning (or very little)
o Spacing out in class
o [Not] seeing a car brake
o You are constantly playing Wheres Waldo
Learning
o How else to get memory?
o Is all learning essentially the same?
o Does what we know already affect what we learn?
Yes!
o Whats the best way to study?
Memory
o Life is memory
o How correct is memory?
o Do we really forget?
o What forms are mental representations in?
Recollection has implications for
o Reliability of eyewitness testimony

o Recovered [?] memories

o Depression
Syllabus:
o Http://quote.ucsd.edu/cogs101b

History, Background
Whats relevant?
o Lot of information in the world
o Some is useful to us, some isnt
o Better to prioritize the relevant information
By evolving so as to do so from the outset
Sensitivity to light, e.g.
By figuring it out as you go along
Look at the stop sign, not the cows (while driving)
Paying attention to whats relevant
o How do you know to look at the stop sign and not a million
other things?
Visual salience (obviousness)
Knowing what stop sign looks like
Learning whats relevant
o An example about learning: Quines gavagai problem (AKA
induction problem)
Rabbit? Hopping? Wiggly nose? Dinner? Pointing finger?

o May have biases in learning

Example 1: Food-borne illness
Rats associate taste w/ illness (Garcia)
Pigeons attribute to visual cues (Shapiro et al.,
1980)
Labels refer to whole objects (rabbit, not pink
nose)
o Big questions:
What are the biases?
Can you derive biases from information present in
the environment?
Are there overarching similarities across different types
of learning?
Remembering Whats Relevant
o Lots of things we could remember but dont
Somehow we sort through overload
o Mistakes tell us about how memory works
Remembering names (notoriously hard; why?)
Science of Cognition
o Intuition is not enough
o Problems with observation
Confirmation bias
Expectations influence (others!) behavior
o Scientific Observation
Public

If it only happens for believers not verifiable

Allows others to replicate your findings
Self-correcting
Testable (you can prove that its wrong)
This is how science works.. ideally
Historical Context to Cognition
o Introspectionism (e.g. Wilhelm Wundt)
Trained observers using intuition
But what about unconscious inferences?
o Behaviorism (John B. Watson)
No mind or consciousness; operationalize
o Skinner: no free will, just reinforcements and rewards
o Essentially studied patterns of learning*
o Observable stimulus: mind isnt observable
o Unable to explain more complex behaviors
o *Because of its association with behaviorism, learning was
considered a dirty word among cognitive researchers
o Behaviorism
Troubling learning effects
Species differences
Stimulus differences
Tinklepaugh (1928, 1932): delayed response-memory

Skinner (1957) wrote Verbal Behavior

Language is response to stimulus
Correct productions get reinforced (e.g. by
parental approval)
Behaviorism vs. Chomsky
o Cognitive revolution
o Chomsky (1959)
Criticism #1: Whats a stimulus?
Criticism #2: Whats a reward?
Criticism #3: How do you talk about things that dont
exist? (Dragons)
Criticism #4: Infinite possible sentences
Historical Context to Cognition
o Cognitive Psychology
The mind is acceptable territory again
Information processing (ca. WWII)
People as senders/receivers of information
Mind-as-computer
Storage, retrieval, processing
Serial processing (something that can only do one
thing at a time)
o Cognitive psychology methods
o Reaction time
o Processing time

o Errors
o Various neuroscientific measures
o Similar questions to introspectionists
Interested in the mind and brain again
o Cognitive Revolution
o Cognitive Science
Mind-as-computer: Turing machines (1936)
o Before real computers
o Tape + tape head
o Mathematical proof: can compute any mathematical function
o Maybe the mind is a Turing Machine
Computations, not hardware
o Turing Test
How can you tell if a computer thinks?
Sustaining Behavior
o ADHD and attention
Ceci & Tishman (1984)
ADHD vs nonADHD kids
NonADHD learned materials better
ADHD learned distracting stuff better

Not intellect: problem is sustaining behavior

Additional consideration: switch cost
Automaticity
o Practice makes task less demanding
Uses less capacity
o What kind of practice reduces capacity?
Consistent vs. varied mapping
Consistent: B is always target, Q distractor
Varied: B, Q each
o Consistent vs. varied: Schneider * Shiffrin (1977)
o With consistent mappings, looked like parallel search
o Shiffrin & Schenider
o Is automacity cost-free? NO
o After consistent training, use old target as a distractor: 22%
drop in detection of real target
o Stroop task: the automatic response isnt the correct one
Dyslexic & ADHD children: apparently worse
Issue of executive control?
o Or Not-omaticity?
o Hirst et al (1980)
Read while dictating sentences
No consistent mapping, so no automaticity

Nonetheless could write & comprehend

o Logan: instance theory of automaticity
Do effortfully vs. retrieve from memory
39 * 39; driving to friends house
A race model (these are everywhere!)
o Stroop Effect
o You read words automatically
o You dont name colors automatically
o Naming the font color when youre trying to suppress reading
the word itself is hard!
When the font color mismatches the word
Easy when the font color matches the word
The Central Executive
o ( = action selection)
o Picks what youre going to do
o Ability to shift attention
o Cognitive control
o Psychoolgical Refractory Period
Dual test
Task A: L hand respond high or low tone
Stim A
Stim B

Respond to A then B
o Why this bottleneck?
Avoid interference (pat head, rub stomach)
Bottlenecks on the Road
o Things you do while driving
Talk to passengers
Listen to radio/sing along
Eat
Dial phone
o Strayer & Johnston (2001)
Radio, book on tape OK
Shadowing via cell OK
Except word-generation variant
Unconstrained conversation (+ hands) bad
2x increase in missing traffic signals
S&J: cell diverts attention to an engaging cognitive
context other than the one immediately assoaited with
driving
o Levy, Pashler & Boer (2006)
PRP paradigm
Braking (single-response, well-practiced)
A/V identification with A/V response (2AFC)
Modality independence? No

Maybe auditory will interfere less

Strongest delay was at shorest IOI
Regardless of modality, braking was delayed

Context Matters
o Why are we reading about perception?
Perception and Memory
o How much of perception is guided by higher-level (top-down)
influences?
Learned knowledge
o Alternative: maybe its built in
QUIZ
o The Stroop Effects Show
Words are named automatically
o What has been taken as evidence for feature integration
theory?
Conjunction search is slower than a single-feature
search
o Evidence from Sperlings experiment asking participants to
partially report their memory for an array of letter shows
Theories of Object ID
o Feature detection
o Structural descriptions
o Template matching
o Neural networks

Feature Detection
o Objects are composed of separable parts (features)
o Similarity: % of shared features
o Feature:object :: Leture:Word
o Parsimonious: A small # of features can be recombined into
much larger number of objects
o Psychological evidence
Lettvin et al. (1959): frog visual system
Edges
Moving edges
Bug detector (convex edge)
Dimming detector
Hubel & Wiesel cats and edge detectors
People seem able to learn new features (e.g. Goldstone
and colleagues)
o Perceptual Evidence
Visual search asymmetries (Treisman & Gormican,
1988)
Tiltedness is a feature that allows it to pop out
o Problems
Anagram problem
How do the features go together? + T L
AEIOPRSTN
STOP, POST, POTS, OPTS, PTSO

Segmentation problem
To find features of object, must already have
isolated it out
o Structural Description
Features plus conjunctions
One example: Biederman (1988) recognition by
components theory
Some geons (3D shapes)
Biederman: recognition by components
Components: 36 geons
Simple 3D Objects
Structural descriptions are built up by describing
how the geons relate spatially
Nonaccidental properties
Not likely to ccur by chance
Geons: Are objects decomposed into components?
Biederman (1988)
As fast to recognize outline as color pics*
Reed (1974)
People are faster to identify objects that
would be part of the structural description
Geons: Are object relationships important?
Biederman
Remove vertices, hard to ID
Remove other areas, not as hard

Geons: con
Not so great for complex objects
More geons? Eventually loses usefulness
What about things other than shape?
Oliva & Schyns (2000): color helps when
color is diagnostic
Template Matching
o Template = copy of image
o Find closest image
o Uh-oh what about variability?
o Solution 1 (Ullman): align image, then match
o Solution 2: Multiple templates from multiple perspectives
o Pros
No need to extract properties
Use the image itself
Just shrink/stretch
o Cons
Inflexible
Template explosion
Face Recognition
o Complex visual objects
o Special built-in brain area for faces

o Two visual subsystems:

Configural/holistic vs.
Decomposition/analytic
Farah & colleagues

Top-Down vs. Botton-Up

o Comprehension can affect lower level processing
o Effects of knowledge on simple perceptual processing
Work Superiority Effect
o Illustration of top-down effects on perception
o Top-down perception = using context
o Is there a K?
o Faster to detect it in a word FORK or psuedoword GORK
Interactive Activation Model
o McClelland & Rumelhart, 1981
o Accounts for
Word superiority effect
Contextual fill-in
o Word level, letter level feature level
o Connections can be excitatory or inhibitory
o Success of model

Word superiority effect (WORK vs K)

Context effects (WOR?)
Quantitative predictions of word competition
A falsifiable modelnice!
Phonemic Restoration
o Spoken-word effect
Coordinate Systems
o Object-centered vs. viewer-centered representations
Object-c: structural description
Viewer-c: templates
Object-centered = viewpoint-independent
o What makes more sense?
o How would you learn object-centered?
o Can tell a chair is a chair despite its orientation
o Evidence for object-centered
Possible to recognize things in orientation weve never
seen
Reed (1974)
People show strong preferences in segmenting
line drawings
o Evidence for viewer-centered
Tarr & Pinker (1989)
Learn small set of objects at specific orientations

 Then test original and new orientations

How fast to recognize each one as A
o Evidence for both
Burgund & Marsolek (2000)
Teach View 1
Test on View 2
Left hemisphere: View 1 = View 2 [objectcentered]
Right hemisphere: View 1 > View 2 [viewercentered]

Learning
Overview
o Biological background
o Types of learning
o Modern versions of learning
What is learning?
o What goes with what
o Profiting from experience*
If something happens to you once and youre able to
learn it, then youre able to avoid it happen again, etc.
Shape your future based on knowledge youve acquired
o Storing information in memory
Without necessarily acting on it immediately

Non-learning Influences on Behaviors

Fixed-action patterns
o Non-learned behaviors
o Organism does reflexively
o Initiated by releasers
o Once initiated
o Babies
Grasping, head-turning, sucking
All aid in nursing
o Adults: yawning?
o Birds
Greylag goose and egg-rolling
Oystercatchers and larger eggs
Supernormal stimulus
Selectively incubate larger eggs
Bigger egg = healthy bird
Critical periods
o Learning happens in a limited time window with extra
plasticity
o Halfway between built-in behaviors and flexible learning
o Found in humans, animals
o Language

Kids are better at learning language to native fluency

but takes it longer than adult
o Another critical period: Imprinting
Baby duck: Follows moving object
Mama duck, person, wooden duck
Very quickly learns to prefer it
o Birdsong learning
Some birds are born with song
Others learn their species song
Ex: white-crowned sparrow
Often compared to (human) language acquisition
White-crowned sparrow (Marler)
Must hear species song between 10 & 50d
Doesnt actually sing til 150-200d
Raised in lab
Can learn from recording of white-crowned
sparrow song
Cant learn from song sparrow song (different
species)
o Language acquisition
Critical period for learning languages?
The island experiment
Illegal..
Test case: late learners of ASL

Late ASL learners (Newport & Suppalla, 1989)

ASL first language
Acquired at different ages
10+ years experienced using ASL
Better performance the earlier it was acquired
Johnson & Newport (1990)
Similar results with L2 learners of Engish
Newport (1990) less is more hypothesis
Language [sign] elements a, b, c
Meaning elements m, n, o
Many ways to combine (7 * 7 = 49)
Kids have worse working memory so they can
only consider small # of combinations
Much more likely to be right
Cochran, McDonald, & Parault (1999)
Concurrent task = better generalization
Other constrained learning
o Some things very difficult to condition to some stimuli (Bolles
1970)
Run away from shock, press bar for food
Sweetness & nausea, click & pain
o Species-specific constraints
Pigeons learn visual cues to illness

Behaviorism & Learning

o Very quantitative
o In the end, couldnt fully account for all real-world learning
o Still useful in some contexts today
Drug treatment
Chemotherapy and food aversions
Learning rules in many computational models
o Habituation
o Classical conditioning
o Operant conditioning
Trial and error learning
Instrumental learning

Classical Conditioning
o Stimulus substitution?
CS takes place of US
Too simple
o Adaptation
Sometimes the CR is opposite of UR
Compensatory response model

Dinitrophenol + O2 + temp
Conditioned response: - O2, -temp
o Compensatory response model
Explanation for drug tolerance?
Need increasing doses to get effect
CR in opposite direction counteracts drugs
Heroin: Gutierrez-Cebollada et. al
Addicts admitted to hospital for OD
50% had injected normal dose in unusual
environment
Addicts admitted for unrelated stuff (controls)
All in familiar environment
No CS (conditioned stimuli) -> lower tolerance
Overview
o Biological background
o Types of learning
o Modern versions of learning
Operant conditioning
o Other names
Instrumental learning
Trial-and-error learning
o Behavior that acts operates on the environment

o A voluntary response is encourages by rewarding the

response
This means the organism has to do the behavior at least
occasionally to begin with
o Reinforcers
Primary reinforcers (food, water, mating)
Secondary reinforcers
Strong association with primary
EX: Primary
o Punishers
Pain
Nausea
Darkness
o In the lab
Skinner box
Behavior: keypeck, bar press
Pigeon, rat will do sometimes randomly
Reinforcement: food
Or punishment: footshock, nausea

Paired Associates
o Human analog to instrumental learning?

o Pair words, flashcard style

Car -> Coche
Stimulus -> Response -> Reward
o Anticipate word 2 til its readily recalled
But bidirectional (word 2 word 1)
o Each time a pair is learned, strengthened associations
o Learn nowadaysnew associations
But depends greatly on existing knowledge
Conditioning wrap-up
o Weaknesses of conditioning approach
Ignore biological influences on behavior
Ignores role of top-down info (knowledge)
o Nonetheless, still useful in many practical contexts including
Conditional flavor (chemo therapy and food)
Infant methods
o Definitely cognitive stuff going on
o But how to investigate in non-verbal organisms?
o Look at behavior
o Condition head turn procedure
Condition to look at a visual direction to get a reward
o Also used with primates

Learning
o Figuring out what things go together
o (And also what things dont go together)
Other cues to word boundaries
o Stress
English: most words are stress-initial
ThePREttyBAbyWANTSaBOTtle
o Jusczyk, Houston & Newsome
7.5 Months
Familiarize infants with stress-initial words
KINGdom
Play passages with KINGdom vs. HAMlet
Infants listen longer to kingdom > hamlet
passages
Recognize words

Memory
Memory
o Is everything (H.M. example)
o Has many different functions
Have you seen this child?
What is James Bonds telephone number?

What car did he drive?

What kind of car does he typically drive?
What would this room look like if I rearranged the
furniture?
I looked up James Bonds phone number and Im trying
to remember it until I get to the phone
Short-term memory
o AKA primary memory
o Attend to info and make it available to
Process further
Phone number
Bad puns
Store long-term
o Processing: [Maitenance] rehearsal
o Losing it:
Interference from other tasks
Peterson & Peterson
See BKF (tri-grams)
Then, count backwards by 3s from 397
(try)
Thus you cant rehearse B-K-F
The longer you count, the more memory
decreases
Forgetting: newer items overwrite
Limited capacity: You can only cram so much into
STM

Dialing an outside number from my university

phone
o Coding: I.E. what form do the representations take?
Computer metaphor
Concerns STM and LTM
Very big deal in LTM
How do you test this?
Interference of various tasks
Conrad (1964)
View sequence of constonants: MSTLJX
Read random digits aloud (no rehearsal)
Ding! (Recall)
Confusions: acoustic (M N T D)
But with articulatory suppression during
presentation, not acoustic
Note: letters interfere with each other
Phonological similarity effect
Harder to remember things with more overlap
Why digit span > letter span
One, seven, five, three, six, two
Bee, gee, dee, eff, em, ess
Chunks
George Miller (1956)

Limit on STM not physical unit but chunk of

information

Working memory
o Baddeley & Hitch
STM isnt just a storage receptacle
Its a workplace
If you like STM youll LOVE WM!
Lets you manipulate, not just hold onto, information
o Components
Phonological store: about 2 seconds
Articulatory control process: refreshes store
o A lot of effects weve talked about
o Specialized for language learning
o Two-second limit
Welsh vs. English digit span (Ellis and Hennelly)
o Visuospatial sketchpad
Maintain and operate on visual images
Farah & colleagues: maybe separate

Language, Modality, & WM

Boutla et al. (2004)
o Q: Do ASL signers have a reduced STM* capacity?

o Typical test: digit span

o Problems:
Digits in ASL very phonologically similar
Length differences between spoken & signed
o Solutions:
Control for length by having people read lists
o Hearing: digit span vs. ASL: letter span
o Hearing > ASL (6.4 vs. 4.9)
o Working memory task (no serial order component)
Hole, fish She fell in the hole; he caught a fish..
Hearing = ASL (= 3)
o Bilingual: English > ASL (7.1 vs 5.2)
o A1: Yes, STM is smaller even controlling for
Word length
Phonological similarity
o A2: However,
Memory: The Big Picture
o Organization
Stuff youre processing at the moment (STM/WM)
May be remembered only briefly (seconds)
Auditory/visuospatial distinction

Stuff you know (Long Term Memory)

Remembered for long period of time (years)
Episodic/Semantic distinction
Declarative/Procedural distinction
Implicit/Explicit distinction
Long-term memory
o How does it get in there?
Some sort of encoding
Factors affecting how good encoding is
Rehearsal time
Depth of processing
Memory for meaning
Organization
Elaboration
o Levels of processing (Craik & Lockhart)
AKA depth of processing
Types of processing
Shallow: font, captitalization (17%)
Medium: rhymes with (37%)
Deep: pleasantness, fits in sentence 65%
Objection: whats deep or shallow?

Long-term memory
o Forgetting = insufficient retrieval cues?
Savings in relearning (Nelson et al 1979)
Day X: learn wd/# pairs (e.g. 42-life, 30-great)
Day X + 30: test for recognition
Unrecognized pairs are retaught
Original pairings (42-life, 30-great)
Switched (42-great, 30lifes
If youre retaught original pairings, remember it 57%,
switch 22% accuracy
Distrinctions in memory
Knowledge of how to do motor acts
Episodic and Semantic
o Declarative knowledge
o Knowledge you can state
o Episodic: particular events
H.S. grad dinner
Text message you received
o Semantic: general information
What typically constitutes a dinner

What a dog is like

Semantic
o Categories, event schemas
o Concepts that are related activate each other
Lexical decision task
Prime is related or unrelated word
Fish -> lamp (LD to lamp is normal)
Cup -> bowl (LD to bowl is faster
o Category knowledge thought to be hiearchial
o Categories
Can make inferences
Are horses animals?
Do horses have electrons?
Collins & Quillian: hierarchical model
o Nodes (bird, fish animal)
Concepts, not word-forms
o Links
Labeled
Directed
o Activation tags
o Superordinate/subordinate

o Sentence verification
o Prolem 1: typicality effect
A penguin is a bird
A robin is a bird (Faster)
Ok, stronger link for robin
o Problem 2: Violations of hierarchy
A chicken is a bird
A chicken is an animal (FASTER)
Add another link: chicken animal
o Problem 3: false responses
A bat is a bird (Slower to say false)
A bat is a plant
Relatedness effect
Episodic Memory
o Lots of details
o Temporally specific
Remember things in the order they happen
o Memory for source
o Truth determined by individual
Not by group consensus, like what table refers to

Models of Memory
Adaptive Control of Thought
o ACT Theory (John R. Anderson)
o Extension of hierarchial model
But better!
o Attempts to explain
Learning
Memory
Language
Reasoning
Problem Solving
ACT Model
o Proposition: Kind of like a sentence, but one sentence can
contain multiple propositions
o The professor things the chalkboard is dusty
o Type-token distinction: A type is just a class of
objects/things; a token is a particular instance of that class
o Compare to semantic vs. episodic
o Spreading activation
Nodes activate, intersect if connected to same
proposition
o Have you seen this sentence?
The chalkboard is dusty
The professor is dusty

o The evidence
Getting the gist
Jim told Ed about the fun exam
Jim and Ed talked about the fun test
Jim told Ed about the bad exam X
Nurse primes doctor
Fan effects

Schemas
o Influences on memory
Place shemas
Dorm room; grad student office
Are there books in your TAs office?
Brewer & Treyens (1981) : 30% say yes when no
books were actually present
Markman & Gentner
Activated schemas by juxtaposing similar pictures
o Stereotypes
Scientist
Test tubes & symbols
Caucasian

Male
Botla et al
Of those using pronoun in discussing this on a
problem, all used he/him
Boutla is a woman

Scripts
o Evidence for scripts (Bower et. al 1979)
Study 2
Present 10 lists of actions
Some lists in other, outhers out of order
Asked people to recall actions
In order lists: 50% correct order at recall
Out of order lists
o Problems
What about things that dont have a particularly
stereotyped order of occurrence?
Going to the bathroom at a restaurant
Where does this fit in

Do we need imagery?
o Kosslyn: map-scanning experiments
o Scanning an image = scanning real map

o Procedure:
Learn fictitious map (tree, pond, well)
Then, image map
Focus on tree
Now imagine black speck moving from there to
pond
Sometimes no pond
Longer distance = longer scan time
o Laend & Tedorescu
Eye movements when looking at picture of object
Imagery = Perception?
o One argument : No, imagery is already some sort of
abstracted representation
The imaged object cant be reinterpreted
Chambers & Reisberg : bunny-duck
Yes you can reinterpret mental images
Finke, Pinker, Farah
Imagine D rotated left 90 degrees and put
atop a J