Documente Academic
Documente Profesional
Documente Cultură
Cognitive
Psychology
Greg Francis
Fall 2016
http://www.psych.purdue.edu/~gfrancis/Classes/PSY200/indexF16.html
Prof. Greg Francis
Lecture 01
Topics Textbook
There is no textbook
For example
Lecture notes are used instead
!What s the deal with left and right brains?
!Why does everyone love Prozac? If you want a book, borrow from a past
class
!Why telephone operators seem rude.
!Why there is a gate at the first floor stairway There are optional readings in the
in the Psychology building. syllabus
!What to do if you are drunk while studying for !Not for every subject
an exam.
!What is the plural of walkman?
Four great mysteries
To me, these are a poor substitute for attending
class web page
Introduction
lecture
Humans face four great mysteries about the
universe
PSY 200 1) Why is there something instead of nothing?
! This is the domain of physics
Greg Francis ! Most of us are not going to understand the ideas
Lecture 01
!Reduced form (6 to a
2) How did life form? 3) Why is there so much
! This question is addressed at the boundary between diversity of life?
chemistry and biology ! This is the domain of
biology
! Evolution and natural
selection answer this
CogLab CogLab
Labs are listed on the syllabus
Homework
They must be completed by 10:00 am at the
You participate in classic experiments date indicated in the syllabus
Total lab grade contributes to 15% of your class ! else you get no credit
grade. ! Better to do it the night before
Grade is based solely on completing the Since I wrote CogLab, you get access to the
experiment, not on the quality of the data experiments for free
! (a $50 value!)
Grading Grading
Last semesters grades (Spring 2016) Last semesters grades (Spring 2016)
Frequency
E1 E2 E3 E4 Final CogLab% A 10
Average 69 69 78 79 73 87 B 25
120.00
Max 92 97 98 100 95 100 C 24
40.00
20.00
0.00
0.0 20.0 40.0 60.0 80.0 100.0 120.0
CogLab score
Purdue University Purdue University
Attitude/Advice
During lectures: turn off cell phones, don t read newspapers, don t play Next time
games
Questions are always welcome. I can adjust my lecturing pace accordingly
Print out the lectures and bring them to class. Take notes during class. Not Cognitive neuroscience
everything is on the slides.
The brain
Everything we talk about in class is important
Work on the study guide every week, so the ideas/answers are fresh in your The modularity hypothesis
mind.
This class is an introductory class, but that does not mean it is easy
! Its like Introduction to Physics or Introduction to Chemistry CogLab on Brain asymmetry due!
! Almost every other subtopic in psychology depends on the ideas in cognitive
psychology
! Everything is at least 10,000 times more complicated than what we discuss
What s the deal with left and right brains?
If you don t find a topic interesting, just wait a week
Cerebellum
Purdue University (muscle control, learning)
Purdue University
Fore-brain Fore-brain
Cortex Cortex
Similar to a thick, Similar to a thick,
crumpled crumpled
newspaper page newspaper page
Grooves (fissures Grooves (fissures
or sulci) separate or sulci) separate
regions regions
Fore-brain Fore-brain
Cortex Cortex
Similar to a thick, Similar to a thick,
crumpled newspaper crumpled
page
newspaper page
Grooves (fissures or
Grooves (fissures
sulci) separate regions
or sulci) separate
regions
http://www.youtube.com/watch?v=ZMLzP1VCANo
Purdue University Purdue University
! subject cannot name it ! Right side: art, music, recognition of faces and shapes, Eastern
mysticism
! subject can pick up nut
with left hand Vast oversimplification
! in a normal brain, both
If nut flashes on right
sides are involved in
side many tasks
! subject can name it Results do support the idea
! subject cannot pick up that different parts of the
nut with left hand until brain are involved in
he says nut out loud different cognitive tasks
(modularity hypothesis)
1981 Nobel Prize for Roger Sperry!
Purdue University Purdue University
CogLab CogLab
In the CogLab + Your task was to +
experiment Brain judge whether the
asymmetry presented word was
You stared at a old (seen on an
central fixation point + wood earlier trial) or + wood
new (not previously
A word was
seen in this
presented to either
experiment)
the left or right side
of fixation + Time + Time
CogLab CogLab
Federmeier & Benjamin (2005) found better memory Other explanations than hemispheric
performance for words presented in the right visual field
specialization
Words in the right visual field go to the left hemisphere
! Which is known to be specialized for language !Reading goes from left to right, from fixation to
right visual field
!Perceptual advantage to right visual field?
!Attentional advantage to right visual field?
CogLab Data
Global (13,231 right-handed participants; 1383 left-
Anatomy
handed participants) The cortex contains large fissures that
No effect: maybe participants are unable to fixate? separate five major areas
!Limbic (already discussed)
!Occipital
!Parietal
!Temporal
!Frontal
Receives Sensations of
information from pain,
the eye temperature,
touch, pressure
!Most investigated
area of the brain
Primary
sensory
area
MRI scans
daily with ever increasing resolution
PET scans
Cognitive neuroscience relies strongly on
Functional MRI
the modularity hypothesis
Putting everything together is very difficult How to study the brain without killing someone.
Non-invasive
Lecture 03 Maps of brain activity
The goal is to relate brain events to cognitive
events
How to study the brain without
killing someone.
Purdue University Purdue University
Resolution Electroencephalogram
For almost every technique we have to worry about its
ability to discriminate differences in
! Space: which EEG
place is active?
! Time: when The brain
does something
happen? produces
Finer resolution electrical activity
is usually better
! But can be Put electrodes
difficult to deal
with so much on the head
data
Brain maps
EEG You can analyze the EEG signals in many different ways
Watch the electrical current change through time while Compare the signal strength for different situations
reading sentences (averaged across many trials) Ayahuasca is a Brazilian psychoactive tea
! Good temporal resolution Semantic
! Kutas & Hillyard (1980) anomaly
! Seconds ! Digestion
! Thinking about exams
! (Silva, 2002)
! ..
Functional MRI
Functional MRI
Connectome Connectome
You can use similar technology (diffusion Gives an
spectral imaging) to focus on particular types of anatomical
cellular material map of how
!E.g., identify axons (discussed later) that connect information
brain cells can travel
Gives an
anatomical
map of how
information
can travel
Purdue University Purdue University
fMRI fMRI
Passive listening vs. active listening Passive
The colors show the
!Vannest et al. (2009) listening
difference maps
Twenty children (ages 11-13) complete three relative to listening to
the tones
tasks
Common activity
!Passive listening: hear a female speaker tell a 30-
(breathing, digestion,
second story
hearing machine
!Active response: hear the same speaker tell a story in
noise,) is
5 second segments of two sentences. Scanning Active
subtracted out
occurred after the sentences (silence). Answer response
questions The colors are not
brain activity!
!Random tones: no task, just listen
Purdue University Purdue University
fMRI fMRI
More signals and Passive Does more signal for the Passive
listening listening
different patterns for active response mean active
active listening response listening is better
compared to passive than passive listening?
listening Tested children on
(Could it be comprehension of stories
otherwise?) ! PL: 75.1% correct, SD=12.7
! AR: 79.1% correct, SD=9.1
Active Active
response No real difference in response
comprehension
Scan
now
Or scan
now
Purdue University Purdue University
Statistics Statistics
There is a significant difference in fMRI activity for some regions of The subject in this study was a mature Atlantic
the brain
salmon (sex unknown)
! Medial brain cavity and upper spinal column
The active regions identified by the fMRI are due to
chance
Even with purely random noise, there will be some
statistically significant findings
! The brain has lots of random noise
Neurons A neuron
Dendrite Axon
The brain cells that are responsible for cognition
! input ! output
are neurons Soma
Myelin sheath
! integrate
! insulate
Myelin A neuron
Diffusion Spectral Imaging detects properties of There are many different types of neurons
the myelin sheath (white matter) We will describe only the most common characteristics
https://www.youtube.com/watch?v=2qTuZlMvFgY
(time 2:55) Purdue University Purdue University
Next time
What is the neural activity that produces
brain scans?
How do neurons transmit information to
other neurons?
Purdue University
Greg Francis
Lecture 05
A neuron A neuron
An action potential The action potential then affects the membranes of
generated at the other cells dendrites
soma travels
down the axon to
the terminals
Output Output
Myelin is like insulation for the The output of a neuron is either excitatory or inhibitory on the
cell s axon
other neuron it reaches
! it insures that the signal
generated by the action Excitatory: when our neuron sends an output, the receiving
potential is strong neuron is more likely to produce an action potential
! Jumps electrically rather than Inhibitory: when our neuron sends an output, the receiving
the normal chemical exchanges
neuron is less likely to produce an action potential
In multiple sclerosis the body s
immune system attacks myelin
! physical problems (paralysis)
! cognitive problems (memory,
reasoning, judgement)
! cause unknown (300,000
people)
Networks Epilepsy
Cognitive behavior is related to Disease of central nervous system
groups of neurons working ! causes mostly unknown
together
Seizures
Include excitation and inhibition
! bursts of electrical activity travelling through networks in
! more later
the brain
! brain activity is out of control
! epileptic fits
Epilepsy Epilepsy
One theory (but not yet proven) is that epilepsy
EEG recordings patients inhibitory cells are not working properly
are often used to Excitatory cells activate everything until they
diagnose epilepsy exhaust themselves
Many different
types of epilepsy,
with different EEG
patterns
Neural connections
Molecular structure
http://www.mind.ilstu.edu/curriculum/neurons_intro/neurotransmission_classic_3.0.swf
Purdue University Purdue University
Receptor Receptor
Very large
molecules called When it accepts a neurotransmitter, it
proteins
starts a chain reaction of events
Similar to a filter
! accepts some
!physical, chemical, electrical
neuro- !locally changes the cell membrane
transmitters
depolarization (excitation)
! rejects others
hyperpolarization (inhibition)
Parkinson s Drugs
Lack of dopamine Interact with neurotransmitters in lots of ways, for
example
!Many different causes
! Replace: accepted by receptor and with similar effect
!In extreme cases, patients are frozen
! Production: increase or decrease
Give patients large doses of L-DOPA ! Reuptake: knock out enzymes that remove
neurotransmitter from receptor, neurotransmitter has a
!a precursor of dopamine bigger effect
!sometimes solves the problem ! Blocking: enter receptor but does not trigger reaction,
!lots of side effects partly closes receptor protein so neurotransmitter
cannot enter
Awakenings, by Oliver Sacks
https://www.youtube.com/watch?v=koL0PWCJ4lo
Purdue University Purdue University
Receptors Receptors
A receptor has a simple receptive field
Light sensors (rods and cones) respond to light
Light
Blind spot
Blind spot In CogLab you mapped your blind spot
my
Where nerves leave data
looks
the back of the eye,
like
there are no light this
receptors
!light that hits this spot is
not visible
Light
40 40
Light 35 Light 35
Firing rate
Firing rate
30 30
25 25
20 20
15 15
10 10
5 5
0 0
Time Time
Purdue University Purdue University
-+
(surround)
On-center, off-surround means the cell
is sensitive to
the location of
a small spot of
light
Purdue University Purdue University
Information processing
Information processing
For simple cells, an image like this
For simple cells, an image like this
!is coded something like this
Strong
responses
at edges!
Next time
Networks of neurons
Connections between cells
Feedback resonance
Purdue University
Demonstration Feedback
Feedback in networks can act to clean up
Cell activities do settle down eventually noisy sensory information to make it
Final pattern of activities satisfies consistent with what our systems expect
constraints of the network connections In a very real way, what we see, hear, taste,
smell, touch, and think, is biased by our
Error correction capabilities network s expectation
Can tolerate the loss of some cells A network s expectation is established by its
connection weights
Emergent properties of the network
! excitation -- inhibition
!no single cell has these properties
Seeing things that are not Seeing things that are not
there there
Feedback Conclusions
Similar situation with the illusory Networks of neurons have properties
circle seen here different from single cells
!emergent properties
!stable activities
!multiple constraints
!tolerance to errors and cell loss
Next time
Purdue University
Demonstration
Deep learning
Learning
Google used one version of a neural network to analyze
10 million YouTube stills This may not be the same type of learning
! 1000 computers (16,000 cpus) over 3 days you do when you study for school
The network self-organized to identify common patterns !but it is important just the same
! Cats, faces, tool-like objects oriented at 30 degrees
Consider the length of your arm
!to catch and throw objects your brain must know
exactly your arm s length
!but the length of your arm changes as you age!
And depends on unknown environmental
factors
Cameras Problem
Enhance visual perception
The network coordinating
!MRI overlaid on actual image
of brain for surgeon eye-hand systems, adjusts
highlight tumor itself
!Avoid other brain regions Extended use of the
faster computer cameras makes
the user adapt so his eyes
are where the cameras are!
less
inhibition
more
inhibition
Brightness contrast
Brightness contrast
Thus, the visual system computes brightness as
something like local contrast
! It s a property of the center-surround cells
! Our percept of brightness is determined by the responses
of cells at contrast edges
! As a result, things that have equal physical intensities can
look dramatically different (next slide)
Brightness contrast
Hermann grid
Filling-in
Filling-in
Brightness
information
Carefully fixate
spreads across
the pink center
surfaces
!Craik-O Brien- If you keep your
Cornsweet eyes very still, it
effect will disappear
The yellow fills-in!
Oriented afterimages
Conclusions
Oriented
Visual perception
reset signals
are also !brightness
implicated in !color
an unusual !form
type of
Largely determined by the receptive fields
afterimage
and network structure of visual circuits
Neurophysiology strongly determines
what we see!
Next time
Visual dynamics
! Flicker
! Persistence
! Motion perception
Purdue University
work.
Purdue University Purdue University
CFF Phosphor
Establishes minimum characteristics of The phosphor on a computer screen
electronic devices typically glows less than 10
milliseconds
Lights flicker at 120 Hz
! ten thousandths of a second
! we spend a lot of time in darkness
The gun reactivates the phosphor
Computer (Cathode Ray Tube, CRT)
every 17 milliseconds
monitors and TV s flicker at around 60
Thus, at any given time 1/3 of the
Hz
screen is dark
! better monitors go faster
! the percept persists in your head!
! Liquid Crystal Display (LCD) monitors work
differently
Although some still flicker
Purdue University Purdue University
Persistence Experiment
What is the source of the persisting percept? Bowen, Pola & Matin (1973)
!Receptors in the eye? !subjects adjust duration of a blank
!Receptive fields? stimulus so onset of probe matched
perceived offset of the target
!Network interactions?
!Cognitive (memory)?
Explanation Explanation
Francis, Grossberg & Mingolla (1994) Offset of input from
the eyes produces an
Something has to reset the network
after response
!else it would keep persisting forever
!e.g., due to
Two mechanisms competition from
!(1) new inputs inhibit old responses orthogonally tuned
cells
!(2) afterimages act as new inputs
Offset response
Note: afterimages get stronger as duration
inhibits persisting
and luminance increase!
response Input from eyes
Purdue University Purdue University
As the target s
Explanation Wait a minute
duration or
luminance If visual percepts persist for over 100
increases milliseconds, why doesn t the world seem
! the afterimage blurry?
produced at target
offset increases in !There should be smears of objects as they
strength move or as we move
! so there is stronger
inhibition to break
There must be something else preventing
the feedback such blurring
! so the persistence !masking
of the original
percept decreases
No mask No mask
Write down all the letters you see Write down all the letters you see
F C
U H
N D
Purdue University Purdue University
Write down all the letters you see Write down all the letters you see
Write down all the letters you see Write down all the letters you see
G M X X X X
Q P X X X X
S Y X X X X
Purdue University Purdue University
Significance Motion
In simple animals (like flies and frogs), we know how
The mask appeared after the target turned off motion is detected
The target was presented all by itself for a brief ! Demo on web page
period of time Reichardt detector
However, our visual system is unable to
develop a complete percept of a scene in a
such a period of time
! Thus, the XXX mask interferes with processing of
the letters by shortening their persisting responses
! And prevents perceived blurring of changing scenes
Time1
Time2
Next time
Attention
What is attention?
What does it do?
Purdue University
Attention Attention
Attention
Information processing Part of attention seems to be due to mental effort on
your part
Modern theories see cognition as information processing
! much like a computer ! attending a lecture
Different systems have different capabilities, capacities, and speeds ! ignoring whispering around you
Necessarily, some information is ignored because it is not Part of attention seems a natural side effect of mental
processed effort
! ignoring the uhs and ums from a speaker
! ignoring the feel of clothes on your body
Tse, 2005
Purdue University Purdue University
Next time
Methods of studying attention
What things influence attention
!Timing, features
CogLabs on Attentional blink and Visual
search due!
Should you pay $59.95 for Mega-speed
reading?
Purdue University
Attention Attention
We saw last time that attention can have
PSY 200
!temporal
!featural
Purdue University Purdue University
Attentional blink
Attentional blink
Measure frequency of detection
Turns out that detection of first letter tends to make ! class data (166 observers)
detection of the second letter very difficult
! if it immediately follows the first
! Attentional blink
Implies that
detecting the
M
first letter
P
causes
K
you to miss the
R
J second letter!
S
This often happens when the target has a unique feature This often happens when the target has a unique feature
relative to the distracters relative to the distracters
! shape ! color
blue corners
green arcs
No searching
required!
Interpretation Interpretation
Feature maps: color, shape Feature maps: color, shape
Feature search can identify target within either feature map Conjunctive search cannot identify target within either feature map alone
Requires search
green arcs green arcs by comparison
across feature
maps.
Serial process
No searching that takes time
required!
Automaticity Automaticity
The process whereby a task goes from
When a task is unfamiliar it
seems to require a lot of
requiring a lot of attention to requiring little is
attention to perform called automatization
Later it requires less attention Many tasks are automatizable
! riding a bike
!color naming
! driving a car
! typing !word naming
! tying shoelaces Can measure effects by pitting an
http://www.fieggen.com/shoelace/knots.htm
Stroop task
Stroop (1935)
Stroop effect
Demonstration
!measure reaction time
Stroop effect
Stroop effect Word name interferes with ink color
naming
! ink color does not generally interfere with
word naming
! lots of studies on Stroop effect
Explanation Conclusions
Word reading is well practiced Methods of studying attention
!especially among college undergraduates ! attentional blink
Next time
Intersection of attention, perception, and memory
! Iconic memory
! echoic memory
Purdue University
Write down as many letters as you see Write down as many letters as you see
G W X R
T P Q N
B Z Y H
Purdue University Purdue University
Write down letters from the indicated row Write down letters from the indicated row
R W V D
P S C K
H L Z Q
Purdue University Purdue University
! delay ! delay
Y K W X
L B M R
J T C Q
Purdue University Purdue University
4
Number of letters correct
0
0 0.2 0.4 0.6 1
Results Results
We can test on any row and get essentially the same
CogLab data (139 participants) result
! so, the number of letters that actually persist and is available is
found by multiplying by the number of rows
Number of letters available
8 8
6 6
4 4
2
2
0
0
0.02 0.1 0.3 1
0 0.2 0.4 0.6 1
Delay of tone (seconds)
Purdue University Purdue University
Write down letters from the indicated row Write down letters from the indicated row
M H T R
Y V N W
L S C B
Purdue University Purdue University
X X X X
Purdue University Purdue University
Number of letters
(seconds)
4
F
!Smaller 3
H
available
V capacity 2
A D Significant for 1
S E 0
W some memory
X 0 1 2 3 4
tasks
Delay (seconds)
Purdue University Purdue University
0.8
!1) no delay (immediate)
0.6 Primacy Recency
!2) in the correct order (serial)
0.4
!3) no cues (recall, not recognition)
0.2
Plot percentage correctly recalled against
0
position of item in list 1 2 3 4 5 6 7
Position in list
Purdue University Purdue University
Calculate savings 1
0.8
Savings
0.6
Timeoriginal 0
0 5 10 15 20 25 30
Time (days)
Purdue University Purdue University
WRM
Retention
Retention
Vary duration of counting backward
Peterson & Peterson (1959) Numbers are different from letters, you
!Brown (1958) might not expect any interference
Give subjects trigram !but they can have very strong interference
1
!ask them to count backwards by 3 s and Proportion correct 0.8
then recall trigram 0.6
779, 776, 773,... Suggests
0.4
some memories
last only a 0.2
WRM 782
few seconds! 0
0 3 6 9 12 15 18
Retention interval (seconds)
Purdue University Purdue University
Retention Retention
The results of the Brown-Peterson study The results of the Brown-Peterson study also
suggest that some aspects of forgetting are
suggest that some aspects of forgetting
passive
are process driven ! even if you are distracted, you can recall the trigram if
!keeping a memory active requires effort only a short time has passed
! if many seconds have passed, while you are
!if you are distracted by another task, you distracted, you cannot recall the trigram
cannot apply the effort to keep the memory ! memory has decayed , or something like decay, while
!similar to our observations about attention you were doing the distracting task
and processing
Retention Retention
CogLab data CogLab data (global data set)
! 61 subjects ! 9690 subjects
! Not everyone shows interference on this task
Pizza
There used to be a Little Caesar s in W. Modal Model of Memory
Lafayette with a game where you could win a Atkinson & Shiffrin (1968)
pizza
Multiple stages of memory
! must repeat a sequence of flashing lights
(changes every time) STM plays a dominant role in active memory
! The sequence gets longer until you make a Requires transfer between STM (STS) and LTM (LTS)
mistake
! need a sequence length >7 to win much
Counts number of correct button presses
56 (sequence of 11 buttons): win a soft drink
110 (sequence of 15 buttons): win crazy bread
210 (sequence of 20 buttons): win pizza
nearly impossible with STM properties
! http://www.freegames.ws/games/kidsgames/simon/simon.htm
Probability of recall
0.8 RECENCY
0.6
PRIMACY
0.4
0.2
0
1 2 3 4 5 6 7
Position in list
Purdue University Purdue University
Here s the
CogLab data 0.8 PRIMACY:
Use STM
Demo 0.4
0.2
1 2 3 4 5 6 7
Position in list
Purdue University Purdue University
!Sternberg (1969)
NO NO
5329
5329 5329
5329 8 8888
8888
8 8
8
Purdue University Purdue University
3333 3333
3 3 3
Set size
Purdue University Purdue University
8 8 8 88 3 3 3
8 8 3 3
8 3
Purdue University Purdue University
comes out
! You ran a version of the experiment in CogLab
Interpretation Controller
Exhaustive search makes sense if search of STM
Controlling attentional system
is done by some process that is
! supervises
! very efficient (can search very quickly) ! coordinates
! dumb (doesn t bother to stop itself) ! starts and stops relatively independent processes
! initiated by some other system (a controller)
e.g.
! Search short term memory
F
Part 1: spatial mental task
Brooks (1968)
(diagrams)
! two types of tasks (visuo-spatial and phonological)
! visual imagery
! two types of responses (visuo-spatial and
! classify corners (top or bottom
phonological)
corner?)
Identifies two types of systems that are ! yes if top or bottom
relatively separate ! no if not top or bottom
Results
Results Results
! when you have to respond by pointing, it is easier to
Measure time to finish mental task for each
work with sentence information than diagram information
response type
! when you have to respond verbally, it is easier to work
! diagrams -- pointing with diagram information than sentence information
! sentence -- pointing
Mental task
! diagrams -- verbal
Diagrams Sentences
Response task
! sentence -- verbal
Pointing 28.2s 9.8s
Significance Interference
The results suggest that there are two
These system have only limited resources and capabilities
relatively separate systems Asking a system to do two things at once (e.g., pointing and
!one deals with visuo-spatial information and mental diagram) slows down the system
must do the pointing response and mental Splitting responsibilities across the systems (e.g., spoken
diagram task response and mental diagram) can be done quickly
Baddley (1986) put these ideas together into a model of ! rapid forgetting
working memory Processor of information
Central
Visuo- Phono-
spatial logical
sketchpad loop
Purdue University Purdue University
0.8
0.7
0.6
tippy-toe
0.5 Silent
0.4 Suppression
0.3
0.2
PS
0.1
0
Visual Auditory
Purdue University Purdue University
Memory of a list of items is worse ! similar sounding items interfere with each other in the
phonological loop
when the items sound the same ! two possibilities:
1) harder to rehearse (effect in the ACP)
2) fade more quickly (effect in the PS)
B G P T Worse B
F
recall PS PS
H B T
Purdue University Purdue University
ACP ACP
F
PS N
PSB G
H B P T
Purdue University Purdue University
Context Context
The effect of part-set cueing suggests that to measure memory But memory is not exactly the same as visual search
you must consider the conditions at test
Information must be encoded in memory as well as
! Memory is more often about discrimination of memory traces and
not about the strength of memory traces recalled
! Similar to visual search experiments ! Such encoding can alter what features are stored as part of the memory
! Which changes the discrimination of subsequent recall
Feature
search
It turns out, that to maximize recallability
! the effort and conditions at the time of learning must be
consistent with the properties and conditions of the test
! see the same words, but have different tasks ! 1) Normal recognition task
! 2) Shown a word and asked if any of the target words rhymed with this
! This changes the encoding of information in memory
word
9
Semantic judgement Rhyme judgement 8
Number of words
7
Study
recognized
CHEESE 6
CHEESE Semantic
5
4 Rhyme
3
The man 2
threw the ball SNEEZE 1
to the 0
Standard Rhyme
______.
Recognition test
Purdue University Purdue University
Cue No cue 10
Study no cue
8
soar Study cue
6
nurse
4
auto
2
paper
0
...
No cue Cue
Test display
Purdue University Purdue University
EAGLE
HEALTH EAGLE
WHEEL HEALTH
BOOK WHEEL
... BOOK
...
Purdue University Purdue University
Significance Forgetting
Decompression tables for divers
Forgetting is not always a characteristic of a
! want to remember when under water
memory system, or your brain
! generally study while on land
!although it could be in some cases, it is not
Researchers working under water have difficulty
always
recalling their details on land
! E.g., counts of species Forgetting must be defined operationally
How do you know if something is forgotten? !specify the task and context of retrieval
! changing context may allow subject to recall seemingly !You can never be certain that if you are placed
forgotten information
in a different context you will still show forgetting
! forgetting = retrieval problem?
Testing Mood
Mood has a similar effect (Eich et al, 1994)
! Mood induced by music and directed thoughts
So, if you are intoxicated while studying for an
exam
!and you didn t study before
Study mood
You should be intoxicated while taking the
exam
But do not
expect to do GET HELP!
very well
Purdue University Purdue University
Classrooms Classrooms
Is memory better when you are tested in the same room as Subjects recall all words either in the context of Day 1 or Day 2
lectured? (different contexts for different subjects)
! significant for final exams! Recall was best for words that were studied in the test context
Smith et al. (1978)
! Subject studied words in one of two contexts (on separate days)
! Varied classroom and dress of experimenter
Study context
Cues
CogLab on False memory due!
Environment
State How to take a test.
Mood
Classrooms
Purdue University Purdue University
! but this is impossible because it was never shown ! able to be manipulated, to a certain extent, I can make
you have certain memories
CogLab data (68 participants)
! Type of selected items Percentage of recalls Why does the false memory effect
! In original list 80.8 happen?
! Normal distractor (not in list) 7.2
! Special distractor (not in list) 69.9
Discrimination Interference
Good memory recall usually requires not only Retroactive interference (RI)
recall of an item from memory !new information prevents recall of
previous information
You also must identify the correct item relative
!e.g., Overwriting a computer file.
to the appropriate context or time frame
!The current trial Proactive interference (PI)
!The context of the experiment !prior learning prohibits new learning
0.9
0.7 5.8
recognized
0.6 5.6
0.5
0.4 5.4
0.3
5.2
0.2
0.1 5
0
4.8
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15
1 2 3 4 5 6
Trial
Trial
Purdue University Purdue University
0.8
0.7 Control Experimental
0.6
0.5 XJF Trial 1 XJF
0.4
0.3
WRM Trial 2 WRM
0.2
0.1
0 DBL Trial 3 DBL
1 2 3 4 5 6 7 8 9 10 11
Trial NRX Trial 4 942
Purdue University Purdue University
Trials 1-3 show build up of PI Works for many kinds of memory tasks
80
5, 7, 9, HAND NICE, SUNNY, ENJOY, KILL
60
Control
40 Experiment
20
D, D, D, D
News stories vs
0
D, D, D, F
1 2 3 4
Trial
Purdue University Purdue University
Testing models
Interference at recall
Test location of PI by changing
If PI prevented the last item from being stored your telling a
instructions after the list is presented
subject that the fourth item was an indoor sport, should
Experiment make no difference (other than guessing)
! stimuli are names of indoor and outdoor ! but it makes a big difference, they show release from PI
Trial 1 FOOTBALL
games
100
Percentage correct
! subjects usually do not notice that word Trial 2
on the fourth trial is an indoor game and
SOCCER 80
Purdue University
Trial
Purdue University
Next time
Constructive memory
Flashbulb memories
Memory misattribution
Misleading questions
Purdue University
Greg Francis
Lecture 19
Discrimination No forgetting?
Brain surgeon (Penfield, 1959)
The task is difficult because you have to do ! Epilepsy patients
! Recall information that might be related to the task from memory ! want to know what is being removed
! Determine if the memory is actually for the correct event Conscious patients report vivid memories
! Determine if the memory is actually for the correct moment in time ! unable to recall normally
! Gauge your confidence in the memory s validity "she saw herself as she had been while giving birth to her
baby."
All of this suggests that performance on a memory task ! stimulation of temporal lobes
involves discriminating information
In the image, numbers indicate
To address the discrimination problem, people engage places where stimulation
in a constructive process to report memories evoked different reported
experiences
Eyewitness testimony
I will show you a series of slides and then ask you
some questions
Percentage correct
60
misinformation Immediate
Compare accuracy according to pre-test questions 50
effect gets
! Subjects without a misleading question--90% accurate stronger with a
40
90% say no 0
Consistent None Misleading
Next time
Amnesia
Anterograde amnesia
Retrograde amnesia
Unusual characteristics
Repression
Purdue University
Amnesia Amnesia
Scope and duration
Loss of memory or memory abilities
Retrograde amnesia for one patient
! retrograde: forgetting events prior to the injury
5 months 8 months 16 months
! anterograde: forgetting events after the injury
!memories cannot be wiped clean !they do not generally forget how to walk, talk,
solve problems
!perhaps they are just not directly accessible
Although they may have problems
!forgetting = recall problem?
!Different types of memory systems
controversial!
Anterograde amnesics
Patient HM
Fairly normal STM digit span (~7 items)
Surgery on hippocampus (to control epilepsy) But very difficult to extend digit span
anterograde amnesia ! how many trials to repeat back list correctly?
! unable to learn anything new 30
25
Thought it was 1953
Mean trials to criterion
20
! shocked by age of face in his mirror
Controls
! Could not stand to read newspapers 15
Patients
! reintroduced himself to doctors, nurses, 10
0
0 5 10 15 20 25
Number of digits to remember
70
60 cafeteria
Controls
50
40
Amnesics ! E.g. mirror
30 drawing
20
task
10
0
0 5 10 15
Item number
Purdue University Purdue University
Repression Repression
Psychotherapists (e.g. Freud) suggested that In a laboratory, showing evidence of repression
infantile amnesia occurred because much of requires
childhood is filled with painful events and memory
!being unable to remember something
of the pain is prevented by psychological defense
mechanisms (repression) !being able to recover the memory through therapy
Repression Conclusions
Next time
Encoding specificity
Levels of processing
Judgments of learning
Practice testing
Learning styles
How to improve your memory without spending
$20.
Purdue University
Memory
Improving memory
We seem to be unable to control our
PSY 200 memories
! learn things we don t want to remember
Study style
Level of processing
Time spent studying is also context for memory
Memory can be influenced by depth of processing at the time
retrieval
of study
Generally, more study leads to better memory ! Craik & Tulving (1975)
Style of study matters too Subjects observe words with associated tasks
25
Number of words recalled
20
15
10
0
Capital Rhyme Synonym
Study task
Purdue University Purdue University
Memory
Improving memory 2
We seem to be unable to control our
PSY 200 memories
! learn things we don t want to remember
Associate items in list with a previously Hook to be remembered items to the list
memorized list !visual imagery helps again!
ITEMS Peg word
One is a bun. Six is a stick. recall by
Two is a shoe. Seven is a heaven. milk bun reciting poem
Three is a bee. Eight is a gate.
Four is a door. Nine is a line.
Five is a hive. Ten is a hen. bread shoe
bananas tree
! pato -> Spanish for duck , sounds like pot-o Experimental group learned more words faster and
imagine duck with pot on its head for longer
! zronok-> Russian for bell , sounds like zrahn-oak ! 6 weeks later
imagine an oak tree with bells as acorns experimental (43% correct)
control (28% correct)
Purdue University Purdue University
Photographic memory?
!Few documented cases
!Generally, not happy outcomes
Brain Training
S.: Luria Several companies market activities
to make you smarter
Visual imagery ! Exercise your brain with games that are
adapted from neuroscience
Sleep Sleep
Subjects learn to identify There is a ordered arrangement to the stimuli
order relationships If you know this arrangement, deciding for any pair is easy
between random shapes ! But subjects are never explicitly told about this arrangement
Sleep Sleep
Subjects are split in to three Subjects are split in to three
groups, according to when groups, according to when
they are tested they are tested
! 20 minutes later ! 20 minutes later
! 12 hours later ! 12 hours later
! 24 hours later ! 24 hours later
No differences when tested Big differences when tested
on the originally studied on new pairs that fit the
items ordered structure
! E.g., A>C, C>E, B>D
Sleep Conclusions
Half of the 12 hour group
had sleep and half did not Lots of ways to improve memory
It makes a difference for
pairs of items that are far !Method of loci
apart in the ordered !Imagery
structure
!Mnemonics
1-degree: A>C, B>D,
2-degree: A>D, B>E, !Brain training
Advice: !Sleep
! study early!
! Get some sleep!
Next time
Mental representation
Prototypes
Exemplars
Propositions
CogLab on Prototypes due!
What is a shoe?
Purdue University
Concepts
Representation of knowledge
What is the information in Long Term Memory?
PSY 200 ! May be several different types
Concepts Definitions
We will look at three topics in concepts Plato (and Socrates) spent a lot of effort
!Definitions (don t really work) trying to define terms like virtue and
!Prototypes (closer to how humans think) knowledge
!Exemplars (more likely than prototypes) !they were largely unsuccessful
Definitions Definitions
Consider the concept shoe, you might define it as Consider the concept shoe, you might define it as Webster s
Dictionary does
Webster s Dictionary does
! A covering for the human foot, usually made of leather, having a
! A covering for the human foot, usually made of leather, thick and somewhat stiff sole and a lighter top.
having a thick and somewhat stiff sole and a lighter top. ! Anything resembling a shoe in form, position, or use.
! Anything resembling a shoe in form, position, or use. But now consider some situations and decide if they are really
shoes
Lots of shoes fit this definition
! A shoe that is intended for display only
Definitions Definitions
Consider the concept shoe, you might define it as Webster s
Dictionary does The difficulty is the same one that Plato and Socrates
! A covering for the human foot, usually made of leather, having a had trying to define virtue
thick and somewhat stiff sole and a lighter top. ! for any definition you come up with, I can find examples that do
! Anything resembling a shoe in form, position, or use. not seem to fit the definition
But now consider some situations and decide if they are really
But we all know what a shoe is
shoes
! so our knowledge of this concept must not be based on some
! a shoe filled with cement, which cannot be worn
precise definition
! a covering worn on the hands
of a person without legs who Note, scientists can (sometimes) create precise
walks on his hands definitions (e.g., a dog is defined by a DNA pattern or
! And this? "
by mating abilities)
! but the definition is somewhat arbitrary
Prototypes Prototypes
Perhaps what defines a concept is similarity In prototype theory it is possible for an object to
among its members be more or less a certain concept
!there may be no absolutely necessary characteristics Consider the concept coffee cup
!there may be no absolutely sufficient characteristics
Prototypes Prototypes
In prototype theory it is possible for an object to In prototype theory it is possible for an object to
be more or less a certain concept be more or less a certain concept
Consider the concept coffee cup Consider the concept coffee cup
!and variations (some are cup-ier than others) !and variations (some are cup-ier than others)
Prototypes Prototypes
In prototype theory it is possible for an object to In prototype theory it is possible for an object to
be more or less a certain concept be more or less a certain concept
Consider the concept coffee cup Consider the concept coffee cup
!and variations (some are cup-ier than others) !and variations (some are cup-ier than others)
Prototypes Prototypes
Lots of experiments suggest the role of Prototypes
prototypes
!Posner & Keele (1968): learning category names for
random dot patterns
!Discriminate two sets of random dot patterns
!Each pattern is a variation of one of two prototype
patterns
A B
Prototypes Prototypes
variations are made by moving some of the dots The key test is done after subjects learn to classify the
variants
! reaction time for judgment is recorded for stimuli they have never
seen before
new variants
the prototypes
! reaction time is faster for the prototypes
! which suggests that the mental representation of the categories
(concepts) are built to favor the prototype of the category
variant of A variant of B Look at CogLab data
Prototypes Prototypes
Consider the types of concepts you can have
Results are based on data from 156 participants.
! Pattern type Reaction time (ms) ! and how specific they can be
! Prototypes 645 things: bird, dog, chair, shoe,
! Variants 690
actions: walking, running, sleeping,
Unanswered by this (and many other) experiments is what a prototype is:
goal-derived: things to eat on a diet , things to carry out of a house
! a thing that resides in memory and contains information about the category
features? in case of a fire ,...
! the result of processing information? ad hoc: things that could fall on your head , things you might see
A bit of thought suggests it is the result of processing information while in Paris , gifts to give one s former high school friend who has
just had her second baby ,...
When studied, these concepts all seem to have prototype
characteristics
Prototypes Exemplars
We can generate new concepts from old A concept consists of lots of examples of the
concepts concept
!it s inconceivable that every possible prototype exists !e.g., a coffee cup concept might contain lots of
ready to be used examples of coffee cups
!some must just be built as they are needed
!perhaps even the prototypes for simple concepts like
bird or shoe are also just built when they are
needed
Exemplars Exemplars
Comparing an object to see if it is a coffee cup Even if it is a new object, it may match several
involves comparing it to each example in memory exemplars well enough to generate an overall
and seeing if it matches anything well enough response to indicate it is a coffee cup
Exemplars Exemplars
Some coffee cups seem prototypical because Unlike prototype theory, exemplar theory also contains
they match lots of exemplars information about the variability of examples within a
concept
!that s what defines a prototype
Thus, we know that pizzas have an average size of 16
inches but can come in lots of different sizes
And we know that foot-long rulers have an average size
of 12 inches, but essentially no variability in size
Proposition Proposition
Network Representation Network Representation
! The proposition connects the appropriate concept nodes
!The proposition connects the appropriate concept
nodes
Albert Professor
agent agent
threw Albert threw the book gave The professor gave a test
relation relation
object object
book test
Purdue University Purdue University
Proposition Proposition
Network Representation Network Representation
! The proposition connects the appropriate concept nodes
last Spring
time
Dog
pigeons
agent object
in
Jacob
chase Last Spring, Jacob fed pigeons relation
Dogs chase cats
pigeons in
relation agent
Trafalgar Square
object location
relation
Proposition Proposition
Ratcliff & McKoon (1978)
! study phase
One way of combining concepts subjects are asked to memorize a set of 504 sentences
18 - 1 hour sessions!
!there are also other theories of how to do this ! test phase
show words and have subjects decide if they were in the
Used a lot in Artificial Intelligence study sentences or not
measure reaction time for words from the sentences
Do humans represent interactions of
concepts with propositions?
Some experimental evidence The bandit who stole the passport faked the signature
Proposition
Proposition
In the test phase, a word is given and the
Network Representation
subject responds as quickly as possible
passport passport
passport
object object
bandit bandit
agent agent
agent faked faked
Proposition Proposition
In the test phase, a word is given and the The expectation is that activation will flow through the
subject responds as quickly as possible entire proposition that includes this word
passport passport
passport passport
object object
bandit bandit
agent agent
faked faked
Proposition Proposition
So, if the next word is part of the same proposition, a If words are from different propositions, no
subject will respond even faster priming
passport passport
bandit passport
object object
bandit bandit
agent agent
faked faked
Proposition Proposition
In the test phase, a word is given and the Activation will flow through the entire proposition that
subject responds as quickly as possible includes this word
passport passport
passport passport
object object
bandit bandit
agent agent
faked faked
Proposition Proposition
When the next word is shown, its node has not been
Test Phase : Priming Task
primed, so it responds more slowly
! compare RTs for second in a pair of words
! within a common proposition (bandit -- passport)
passport
signature ! between propositions (passport -- signature)
object ! not related in sentence (horizon -- signature)
Conclusions
Next time
Concepts Other types of knowledge
!definitions Mental images
!prototypes ! mental rotation
!exemplars ! mental scaling
! limitations of
Propositions
!Evidence we think in terms of propositions
CogLab on Mental rotation due!
Is a picture in your head like a picture in the world?
Perception Images
When we see this
We have knowledge about, and memories image how do we
of, perceived stimuli represent the
!sights information in the
!smells image?
!touches !analog: copy of
!sounds image in head and
we can retrieve it
Are these converted into propositions, or
!symbology: convert to
concepts
propositions/concepts
!or is there something else?
Purdue University Purdue University
Images Images
If you ask me questions about the previous slide,
I can remember the wears
Runs to
my answers would not necessarily identify the
image on the ball
girl representation
previous slide and it shorts
Propositions Propositions
So this suggests that mental images are not exactly like real
images It is clear that propositional information influences
! and something like propositional information likely influences mental imagery
reports that are ostensibly based on mental images or mental ! but is it all propositions?
San Diego
maps
in ! are there mental images, as we tend to experience them?
agent Nevada
relation Is there any reason to believe that mental images
San Diego is in in
are at all analogous to real images?
California object
relation ! yes
California is west of Nevada
object Reno is in Nevada
agent relation
agent
California
West Reno
Purdue University Purdue University
1200 1200
Imagery Imagery
1000 No imagery 1000 No imagery
800 800
600 600
400 400
200 200
0 0
Small Big part Small Big part
part Purdue University part Purdue University
fMRI fMRI
Hugdahl et al. (2006) had subjects perform a 3-D mental Compared to the 2-D condition, the 3-D condition has
rotation task or a control task (no judgment, just look at 2- more activity in the superior parietal lobes
D oriented bars and press a button) in an fMRI scanner
Males
Females
After Exam 3
information
!instead, language is an instinct But this is not what determines our capability to
have language!
Biology Learning
Like all skills, language needs the proper
Language is a specialized skill of environment to be developed
human animals !blinded birds cannot navigate by the stars
!Darwin (1871) !Atlantic Ocean turtles that navigate by magnetic
fields need to be in the correct ocean
Humans instinctively learn language
!effortless Language development needs exposure to
!unconscious other people for communication
!procedural knowledge !but it needs surprisingly less exposure than you
might suspect
Purdue University Purdue University
Children do not learn language by simply Instead, each child reinvents language
imitating others !difficult to test because we rarely get to see a
!otherwise they would never come up with language created from a non-language
statements like !however, there are cases!
Pidgin
Pidgin The Ten commandments in pidgen
For example, in New Guinea !as translated by the Alexishafen Catholic Mission in
1937
!pidgin is similar to English (rulers of the plantation) ! 1. Mi Master, God bilong yu, yu no ken mekim masalai end ol
tambaran.
woman: meri (Mary, generic word for woman)
! 2. Yu no ken kolim nating nem bilong God.
another man s wife: meri bilong enaderfelo man ! 3. Yu must santuium sande.
! 4. Yu mast mekin gud long papamama bilong yu.
hair: grass bilong hed ! 5. Yu no ken kilim man.
! 6. Yu no ken brukim fashin bilong marit.
helicopter: mixmasta bilong Jesus Christ
! 7. Yu no ken stilim samting.
coffin: die bokus ! 8. Yu no ken lai.
! 9. Yu no ken duim meri bilong enaderfelo man.
piano: bokus bilong teeth yu hitim teeth bokus is cry ! 10. Yu no ken laik stilim samting.
Purdue University Purdue University
Learning Learning
Pidgin is not a true language
In Hawaii at the turn of the century !word order is arbitrary
!workers from China, Japan, Korea, Portugal, !no rules
The Philippines, and Puerto Rico were Me cape buy, me check make.
!no tenses
brought in to harvest sugar
!no prefixes or suffixes
!they developed a pidgin
!can only be understood in context of the
!some were still alive in 1970 and interviewed
conversation
to see how the pidgin worked
He bought my coffee; he made me out a check.
Education Education
But then how do we explain that
uneducated people speak improperly? This person is not speaking with bad grammar, but
he is also not speaking in Standard American
e.g. gang member in Harlem English (SAE)
You know, like some people say if you He s speaking in a dialect called African American
good an shit, your spirit going Vernacular English (AAVE)
t heaven n if you bad, your spirit Both languages have certain rules
goin to hell. Well bullshit! Your spirit
His statements obey the rules of AAVE precisely!
goin to hell anyway, good or bad.
Consider contractions of words
Rules
Rules
In SAE you can replace some word pairs AAVE allows speakers to drop some words
with contractions ! if you are bad --> if you bad is
! They are --> They re grammatically correct
Language Conclusions
So if everyone is speaking a language, which is
correct? Language is an instinct
!none, they are just different !specialized skill among humans
!they are different dialects of English !children need little tutoring to learn language
!children invent language if one is not readily
Linguist Max Weinreich
available
! A language is a dialect with an army and a navy.
Language follows rules
The dialect you speak may give away your
!even when it doesn t seem to
personal history, but it is not fundamentally
worse than any other dialect.
Purdue University Purdue University
Next time
Grammar
Long term dependencies
Phrases
Language universals
Purdue University
How do we do it?
Dr. Francis says something new!
Two key aspects
!the sound dog has nothing to do with dogs How many combinations?
!compare driving on parkway to parking on !if interrupted anywhere in the middle of a
driveway, blueberries and cranberries, sentence, there are about 10 words one could
hamburger select before finishing it up
Grammar !if sentences average around 20 words, that
means there are 10^20 sentences
!the order of words matters
!Dog bites man. vs. Man bites dog.
Purdue University Purdue University
Grammar Vastness
But in fact, there are infinitely many It is amazing how powerful language is
different sentences
You have probably never heard the
!there is no limit to how long a sentence can
following sentence
be
!moreover, it is probably its first utterance in
For any sentence I give you, you can
human history, but you understand it anyhow
always make it longer by adding
something like The poor student attendance, along with the too big
!Professor Francis said that, . classroom, sometimes makes the 8:30 section of cognitive
psychology feel like the lecture is given in a warehouse.
Grammar Grammar
You not only understand language, you You can also have sentences without
sense when a sentence is ungrammatical meaning that are perceived as grammatical
!Colorless green ideas sleep furiously.
!Is raining. Sometimes you
still understand !If we don t succeed, we run the risk of failure.
!The child seems sleeping.
what was meant! (a not joking Dan Quayle)
!Sally poured the glass with water. ! Twas brillig, and the slithy toves
!It s a flying finches, they are. Did gyre and gimble in the wabe:
All mimsy were the borogoves,
!Rarely is the question asked: Is our children
And the mome raths outgrabe.
learning? (a joking George W. Bush)
Statistics
Nonsense sentences
Think about the sentence If you just learned the statistical
!Colorless green ideas sleep furiously. combinations of words, you might think
What is the probability that in normal life you something like this was a grammatical
would hear the word green follow the word sentence
colorless ? House to ask for is to earn our living by
!it must be close to zero working towards a goal for his team in old
New York was a wonderful place wasn t it
But we recognize it as a grammatically even pleasant to talk about and laugh hard
when he tells lies he should not tell me the
correct sentence!
reason why you are is evident
Purdue University Purdue University
Phrases Phrases
Every sentence is built out of phrases All noun phrases obey certain rules
!rewrite rules
The happy boy eats candy. NP-->(det)A*N the happy boy
Sockets Usefulness
In a phrase tree, a phrase is like a component It is important to appreciate how the phrase tree
that snaps into the right place
approach simplifies the description of language
!any appropriate phrase works! (even nonsense
phrases) Consider how we learn a new word and know
S how to use it
Next time
Words
Mental lexicon
Morphology
Structure
CogLab on Word superiority due!
Words Grammar
The rules of phrases
PSY 200
!rules for combining phrases
Greg Francis !universals for all languages
Morphology Morphology
English can convey this information in as many ways
Other languages have many more as other languages, but we use grammatical phrases
to do so
variations
Simple present tense
!Italian and Spanish have 50 forms of each verb
! General truths: Ducks quack.
!classical Greek has 350 forms of each verb ! Habitual action: I quack like a duck when I wake up.
!Turkish has 2 million forms of each verb Present Perfect Progressive
!some languages build entire sentences around ! To express duration of an action that began in the past, has
one complex verb continued into the present, and may continue into the future:
The duck has been quacking for two hours, and he hasn t
There are rules for these forms finished yet.
Morphology Suffixes
English has lots of these derivational suffixes
On the other hand, English morphology
-able -ify -ance -ism -ous Examples
allows one to easily create new words
of morphemes
from old words -age -ion -ary -ist -y
Compounding Rules
English also allows new words to be created So what are the rules?
out of other words
!and combinations can be combined
One looks to be easy
!to pluralize a noun, add -s
tooth brush unmicrowaveability
N N
bootylicious -s
dog -s wug
Nstem Nstem
dog bite
Vstem Astemaffix Vstem Nstemaffix
brush By the way, ignore what grammar
tooth school might have taught you, these
are not adjective-noun phrases crunch -able crunch -er
Purdue University Purdue University
Root Lexicon
Some morphemes can
N To keep track of what can attach to what,
only be attached to certain
there must be a mental dictionary of
types of words
!a root is a word that cannot Nstem Ninflection morphemes
be split into smaller parts !-able
-s
!some morphemes attach adjective stem affiix; means capable of
Nstem Nstemaffix being X d ;attach me to a verb stem
only to roots
!thus, Darwinianisms is a !-er
-ism
word, but Darwinismians is noun stem affix; means one who X s ;
not Nroot Nrootaffix attach me to a verb stem
-ian must attach to a root !...
Darwin -ian
Purdue University Purdue University
Exceptions
Very special cases
The exceptions generally come from other
languages (with appropriate rules) What is the plural of walkman?
Parsing Linguistics
Why do you say you are sad? I am showing a demonstration for class.
My parents won t give me a new car. Why do you say you are showing a demonstration for class?
I have five brothers and two sisters. Can you elaborate on that? Eliza just picks up
You seem to dwell on your family. Well, I am a professor. on key phrases and
They are important. Tell me about your professor. generates some
stock responses.
Is it because of your plans that you say they are important? I am the professor!
VP No photo
VP
NP
NP PP
PP
V V N
N NP
NP
discuss with discuss with
sex Dick Cavett sex Dick Cavett
Parsing
Mentalese
Parsing is something like building a
phrase tree in reverse
That two different internal thoughts can
give rise to the same language statement Let s parse through a simple sentence
is interesting word by word
!it suggests that we think in some way that is !The dog likes ice cream.
different from language S
!a mentalese, if you will NP VP
det N V NP
N
the dog likes ice cream
Purdue University Purdue University
S PP PP PP S PP
Remarkable is the rapidity of the motion of the wing of the hummingbird. The rapidity that the motion has is remarkable.
Difficult sentences
Word order These sentences are difficult for humans because of
limited memory
This sentence is nearly impossible ! when a phrase tree includes many unfilled branches of the
same type (PP)
! the parser becomes confused as to which phrase is
associated with a new word
! ends up backtracking to sort out the phrases
! sometimes falls apart ( has has has )
The grammar generator and the parser are different
things in your language system
S PP PP PP
! these are grammatically correct sentences
The rapidity that the motion that the wing that the hummingbird has has has is remarkable. ! they are not good sentences Don t
! you make sentences like these make me
show you
your exams!
Purdue University Purdue University
S S
NP VP NP VP
Word pencil is Word pencil is
inconsistent with consistent with
det N det A N
structure created! structure created!
the plastic pencil the plastic pencil
!The plastic pencil marks easily (verb) In the lexical decision experiment, you see a
sequential pair of words/non-words, and we measure
Parsers build phrase trees on the fly, so the reaction time for you to decide if the second
backtracking is often required word is a word
! RT is faster if the second word is semantically related to the
!many times it is so fast that we do not notice
first word
!seems effortless
Sentence ambiguity
Evidence of ambiguity
Interestingly, people often miss ambiguities in
No one notices the ambiguity
sentences
But, give a lexical decision test for words verses non-
! Time flies like an arrow.
words
Humans recognize only one interpretation
! Flashed visually on a screen just after the word was spoken
! Subjects respond faster for words related to either definition of Computer algorithms can find 5 interpretations
bug ! all grammatically correct!
Schemas / scripts
Conclusions
Giving computers the general knowledge of
life needed to create something like schemas Understanding language
is very difficult
Parsing
This is why computers do not carry on
Phrase trees (in reverse)
conversations with you
Ambiguities
Lots of work going on in artificial intelligence to
Computer generated interpretations
address this problem
Missing information / schemas
Next time
Speech
Phonemes
Articulation / coarticulation
CogLab on Categorical Perception
Discrimination due
Speech Language
Lecture 29 !words
Illusions Illusions
When you hear what I say, you think you
The blurriness of speech explains some long-held
hear at least confusions
!separate words ! Oronyms (Mondegreens)
!separate syllables
Phonemes Phonemes
phoneme \'fo-,nem\ n
Speech is made of phonemes
[F phoneme, fr. Gk phonemat-, phonema speech
sound, utterance, fr. phonein to sound](ca. 1916): Different combinations of phonemes
a member of the set of the smallest units of speech correspond to different syllables and
that serve to distinguish one utterance from
words
another in a language or dialect, the \p\ of pat and
the \f\ of fat are two different phonemes in English> We seemingly hear more phonemes than
the ear can actually handle
!how?
Packing Packing
If the ear can only distinguish up to 20 sounds per If phonemes are being smashed together
second
there must be some blurriness
! and we can interpret speech that seems to contain 50
phonemes per second !and this can lead to misinterpretations
! then the speaker must be combining many phonemes
together to overcome the limits of the ear This is also why computer speech sounds
The listener hears the 20 (or so) sounds in a second, funny
but interprets them as more than 20 different !The programs do not combine phonemes in the
phonemes right way
Speech Physiology
So what are phonemes?
Lungs push
All speech is made of sounds
air out to
! sound is a pattern of pressure on the ear
! a tuning fork vibrates back and forth to make the sound of a make a
pure tone sound
! Frequency of vibration corresponds to pitch of the sound
!other
Speech consists of lots of patterns of this sort
organs
! With many different overlapping frequencies
shape
sound
Example Example
Note what your lips do as you say
Note where your tongue is as you say ! boot book
!bet butt The lips add additional frequencies to make different
!beet bat sounds
Thus, you can hear someone smile across a
The position of the tongue shapes the vocal
telephone!
tract and makes different sounds!
Vowels are all distinguished by the shape of the vocal
!this is true for all vowels tract
Consonants Consonants
Consonants are more complicated (3) Manner of articulation
!different type of control of air flow !/d/, /t/ (stop)
! /p/, /t/, /f/ (not voiced, or unvoiced) Each consonant is uniquely identified by its
(2) Place of articulation: voice (or not) and its place and manner of
!/d/, /t/ (upper gum)
articulation
!/m/, /b/, /p/ (lips)
!/f/, /v/ (lip and teeth)
Purdue University Purdue University
Consonants
Some languages have other characteristics as well Fun
(e.g., tone, timing)
Why do we say razzle-dazzle instead of
For example, in English, the difference between /ba/
and /pa/ is the timing of the release of air for the dazzle-razzle?
consonant and the voicing of the vowel !for phrases like this, people always first say the
Voice Onset Time (VOT) is short for /ba/ and longer word with a leading consonant that impedes air flow
the least
for /pa/
super-duper willy-nilly walkie-talkie
CogLab data: sounds It s a
differ in VOT, judge if helter-skelter roly-poly namby-pamby rule!
same or different sounds
harum-scarum holy moly wing-ding
Rules
Phonemes
To say a word, we must combine phonemes
English uses 22-26 (it depends on how you count)
combinations of voicing, place, and manner of In every language there are rules (trees) that
articulation (and 20 vowels) describe what phonemes can follow other
! Rotokas (Papua New Guinea) uses 6 (and 5 vowels) phonemes
! Khoisian (Bushman) uses 141
Uses clicks as consonants Thus, we can identify possible words from
No language uses some possible sounds impossible words
! raspberries, scraping teeth, squawking,
!plast ptak
! Note, these sounds are used for communication, but not as
part of language! !vlas rtut
Japanese does not distinguish /r/ from /l/ !thole hlad
Purdue University
!nypip dnom Purdue University
Compression Coarticulation
Moving the tongue (and other We generally do not notice these
articulators) around is difficult and takes adjustments
time !we are tuned to recognize the new sounds
!to say sounds faster, people use as coarticulation
coarticulation
This is the main reason computers have
!shape tongue in advanced preparation for
a hard time recognizing human speech!
the next phoneme
!this influences the sound of phonemes
Coarticulation Coarticulation
Notice that your tongue body is in different There are rules for how to coarticulate
positions for the two /k/ sounds in
When a stop-consonant appears between
!Cape Cod
two vowels, you do not actually stop
Note too, that the /s/ becomes /sh/ in !flapping
!horseshoe
slapped --> slapt
And /n/ becomes /m/ in
patting --> padding
!NPR
writing --> wriding
You can enunciate these correctly , but in
casual speech you do not!
Purdue University Purdue University
Spelling Spelling
It is true that English spelling does not seem to agree
We have often observed that written language with pronunciation
is different from spoken language ! a problem for learning how to read!
Nor should it
George Bernard Shaw (among others)
! if words were spelled the way they were pronounced, we
complained about spelling in English would lose the visual connection between words
!he noted you could spell fish as g-h-o-t-i ! slap --> slapped would become slapt
! write --> writing would become wridding
! National Public Radio --> NPR would become MPR
gh -- tough o -- women ti -- nation
Other approaches
There are other written forms of language that avoid
Conclusions
some of these problems
The most sensible written Speech
language is probably the
Blurring
Korean hangul
! Drawn characters indicate Phonemes
how consonants are
pronounced
Articulation
Coarticulation
Spelling
Next time
Learning language
Babies
Children
Learning a second language
CogLab on Age of Acquisition.
Purdue University
Lecture 30 Learning
!you do have to learn some specifics for your
native tongue
When should you learn a foreign !rules
language? !words
How do you learn a second language? Repetition of the same sound leads to
boredom and fewer sucks
!ba, ba, ba, ba, ba, ba, ba, ba,
What do babies do?
Purdue University Purdue University
CogLab data:
Identification task
Discrimination task
BA PA
Babbling Babbling
Babbling sounds are the same in all languages
Babbling teaches child how sequences of muscle
! patterns are common across languages
combinations lead to different sounds
By the end of the first year babies combine syllables ! necessary to produce speech
to sound like words By about 10 months babies learn the sounds of their
! neh-nee
native tongue
! da-dee
! they can no longer distinguish phonemes that are not part of
! meh-neh the language
Babbling is important ! Part of learning is forgetting!
! children who do not babble often show slower speech
development
! deaf children babble with hands, if parents use sign
language
Purdue University Purdue University
!3) One word utterances (~1 year) The average high school graduate knows
about 60,000 different words (not counting
!4) Two-word utterances and telegraphic speech
compound words and such)
(1-3 years)
!means that in 17 years of life (not counting the first
!5) Basic adult sequences with grammar one), they learned an average of 10 new words
(~4 years) each day (one word every 90 waking minutes)
Errors
Expected errors Children do make errors, but the errors are consistent
Consider a child hearing adults talk and how with rules of language
they might incorrectly apply what they learn Children often over generalize a rule
! -s to pluralize a noun
Out of 66,000 sentences, children never made Mouses, leafs
! -ed to make the past tense of a verb
these errors
My teacher holded the baby rabbits and we patted them.
He seems happy. --> He is smiling. --> Hey, Horton heared a Who.
Does he seem happy? Does he be smiling? I finded Renee.
Once upon a time a alligator was eating a dinosaur and the
dinosaur was eating the alligator and the dinosaur was eaten
by the alligator and the alligator goed kerplunk.
Overgeneralization Overgeneralizations
We know this is the most difficult part of language because adults
These past tense forms sound wrong because English make the same kind of mistakes
has around 180 irregular verbs ! tread - trod strive -strove
! inherited from other languages ! dwell - dwelt slay-slew
! These past-tense forms are not derived from rules ! rend - rent smite - smote
Irregular forms have to be memorized, word by word Sound weird because we do not often hear them
! many adults regularize
If a child cannot remember (in its lexicon) the words
! s/he defaults to the rule ! treaded, strived, dwelled,
slayed, rended, smited
These errors are for the most difficult parts of a
! thus language changes!
language to learn
! Because they don t follow the normal rules
! This new golf ball could obsolete many golf courses. What accounts for the difference?
! If she subscribes us up, she ll get a bonus. ! most likely it is age
! Boiler up! ! there seems to be a critical period during which
language can be learned
Children s errors tend to track the more difficult
! beyond age six (or so) it becomes more difficult to learn
aspects of a language, relative to other languages
a language (first or second)
! Adults make the same kinds of mistakes for still more difficult
to remember cases
Next time
Language & brain
Broca s aphasia
Wernicke s aphasia
Anomia
Language ability of chimps
Purdue University
Wernicke s area ==> Wernicke s aphasia ! omitted function words (or, be, the)
! skipped function words when reading (or, be, the) but read
similar sounding words (oar, bee)
! named objects and recognized names
! high (nonverbal) IQ
Boy, I m sweating, I m awful nervous, you know, once H.W.:But, oh, I know. She's waiting for this!
in a while I get caught up, I can t mention the tarripoi, a Examiner:No, I meant right here with her hand, right where you can't
month ago, quite a little, I ve done a lot well. I impose a figure out what she's doing with that hand.
lot, while on the other hand, you know what I mean, I H.W.:Oh, I think she's saying I want two or three, I want one, I think, I
think so, and so, so she's gonna get this one for sure it's gonna fall down
have to run around, look it over, trebbin and all there or whatever, she's gonna get that one and, and there, he's gonna get
one himself or more, it all depends with this when they fall down...and
that sort of stuff... when it falls down there's no problem, all they got to do is fix it and go
right back up and get some more.
Purdue University Purdue University
C.B. I can't tell you what that is, but I know what it is, but I
don't know where it is. But I don't know what's under. I know
it's you couldn't say it's ... I couldn't say what it is. I couldn't
say what that is. This shu-- that should be right in here. That's
very bad in there. Anyway, this one here, and that, and that's it.
This is the getting in here and that's the getting around here,
and that, and that's it. This is getting in here and that's the
getting around here, this one and one with this one. And this
one, and that's it, isn't it? I don't know what else you'd want.
Purdue University Purdue University
Some patients have difficulty with only certain However, the right hemisphere can also work with
types of nouns language
! left handed people
!concrete vs abstract (chair vs trust)
! hemispherectomies (age matters!)
!nonliving vs living (table vs dog)
!animals and vegetables vs food and body parts
!colors
!proper names
Purdue University Purdue University
Chimpanzee language
Problems
In the 1960s several research groups reported
teaching chimpanzees American Sign Language Just like with Eliza (the computer therapist) it is
(ASL) easy to attribute language ability where it does
! after failure to teach spoken language not really exist (9 month old children)
! other groups taught chimps to press symbols on a computer
keyboard or string magnetized plastic shapes on a board You can teach an animal a lot using simple
Claimed to teach chimps hundreds of words conditioning tricks
! and chimps created new compound words Researchers were quick to excuse mistakes
swan -> water bird
stale Danish -> cookie rock
as play , jokes , puns , metaphors ,...
See video:
http://www.npr.org/templates/story/story.php?
storyId=90516132
Purdue University Purdue University
Grammar Grammar
Likewise, the chimps never produced complex sentences
Chimps failed to learn the rules of ASL They tended to say things like the following
grammar ! Nim eat Nim eat.
they communicate
! Drink eat me Nim.
!unable to understand complex signs but not with real
! Tickle me Nim play.
language
Seemingly able to understand complex ! Me eat me eat.
sentences ! Me banana you banana me you give.
! Banana me me me eat.
!Would you please carry the cooler to Penny?
! Give orange me give eat orange me eat orange give me eat
But really, the chimp need only understand two orange give me you.
Evolution Evolution
Note, it would have been interesting if chimps Chimps are the closest evolutionary relatives of
could learn language humans
!and not inconsistent with the idea that we have a !so if any non-human animal could learn language it
language instinct
would probably be chimps
But the failure of chimps to learn language But in evolutionary history, chimps and humans
does not go against the idea that language split from a common ancestor millions of years
evolved in humans ago
!as some people have proposed
Humans evolved a language skill and chimps
did not
Purdue University Purdue University
History Materialism
Nearly all scientists are materialists, but
Descartes dualism (Cartesian dualism) old ideas die hard
! pineal gland link between body and spirit
A lot of work (PET scans, fMRI,) looks
! how they could connect was a real problem
for the site of consciousness
!a special physical transformation
!thalamus
!reticular formation
Mind-body problem
!quantum mechanics
Materialism (the brain is the mind)
! or the mind derives from the brain !distributed awareness
Purdue University Purdue University
An analogy An analogy
When did the British empire learn of the end of the War of For complicated systems like the British empire (and
1812? human brains)
! treaty signed in London months before the Battle of New ! different parts know different things at different times
Orleans ! there is no official moment of knowledge
! word was not received by British troops in America until two ! no official moment of consciousness!
weeks after the Battle of New Orleans (January 8, 1815)
Demonstration
! when does the class know/understand?
same logic to a
something else to put it all together ? computer
! if a conversation with a
!can consciousness arise from non-conscious computer is
processors? (artificial intelligence?) indistinguishable from a
conversation with a
human
! Then conclude the
Purdue University computer is intelligent Purdue University
2. Qualia 2. Qualia
Some researchers object to the very idea that
Consider two people who see the world in
computers could become conscious color opposites
! They argue that some things in consciousness are not
just computation Qualia for person 1 A red apple with a
! e.g., consider the color red green leaf
! There seems to be a particularly subjective experience
of seeing something red
2. Qualia 2. Qualia
Qualia proponents argue, for example,
Clearly, there s a big difference in the
! you can learn all there is to know about light waves,
perceptual experience of these people, photoreceptors, neural transduction and coding of
but their behavior is essentially the same color,
! But suppose you never see any red objects
!And there seems no way to distinguish one
! Your knowledge will not tell you what you will
experience from the other
experience when you first see the red of an apple
!It s the unmeasureable experience that is a ! Indeed, you could be tricked into believing a green
qualia apple was red (if you had never seen green either)
2. Qualia Conclusions
But this is a defeatist argument, or a pointless one Consciousness
! if I knew everything about light, photoreceptors, and
neural representation of colors, then I would be able to
distributed processing in the brain
know what I will experience when I see red !no site of consciousness
! it is difficult (maybe impossible for any single human) to
!no time of consciousness
know (or even imagine knowing) all that information in an
academic sense
Chinese room
! but that doesn t mean that such information does not
exist Qualia
It s partly an empirical question
Artificial Intelligence
! But no one can do the experiment
Daniel Dennet Consciousness Explained
Purdue University
(1991) Purdue University
Next time
Review for exam 4
After exam 4
! Decision making
! Framing effects
! Risks
! Alternatives
Purdue University
Lecture 33 !housing
!job
!cancer treatment
What every consumer should know
What affects our choices?
before buying.
How do we make choices?
Purdue University Purdue University
So Parent B is the choice to award custody Your decisions are influenced by the way a
and to deny custody set of choices is presented
!but one necessarily precludes the other! The Asian disease problem
Subjects are biased by the task at hand !two versions, essentially the same
!focus on different characteristics depending on !lead to different choices
whether they are considering awarding or
denying
B) 50% chance to lose nothing and 50% Selecting B) in either situation means you end
chance to lose $200. 64% up with either $500 or $300
!$300 + $200 or $300 + $0
!$500 - $0 or $500 - $200
Subjects tend to prefer the risky option
!risk seeking with perceived losses People do not just look at the bottom line
!which is why businesses emphasize that approach
Purdue University Purdue University
Consider choosing a job Job A: Limited contact with others, commuting time 20
minutes.
Job B: Moderately sociable, commuting time 60 67%
Purdue University
minutes. Purdue University
Insight
What does that aha! feeling mean?
Topics Experts
We will look at a number of factors
Some people learn how to solve particular
that influence our ability to solve types of problems
problems
What makes an expert different from a
!expertise
novice?
!analogy
Experts know how to describe problems
!set effects
priming !other than that, there seems to be no
fundamental difference (even for geniuses!)
incubation
functional fixedness
!insight
Purdue University Purdue University
Correct pieces
Correct pieces
real game 20
20 randomly
Experts have schemas that allow Master
15
Master
them to organize the piece
15
Beginner Experts try to use the Beginner
10
positions 10 schemas, but they end up
5
! They only need to remember the 5 misremembering the actual
schema 0 piece positions 0
1 2 3 4 5 6 1 2 3 4 5 6
Trial Trial
Purdue University Purdue University
Analogy Analogy
The solution is the same Analogies are actually very difficult to apply
!split up the army and have them converge to !need to identify what is common between two
the fortress from different parts of the kingdom problems
Glick & Holyoak (1980) Analogies are often applied after two
!subjects read stories like these and were problems are solved and well understood
asked to solve the problems
!it is then easier to see what is common
!even when shown one solution and told that it
could be applied by analogy to another One of the problems handed out can be
!subjects used analogies only 20% of the time solved by analogy to these two problems
Easy solution!
Insight Conclusions
Warmth stays mostly steady, right up to Effects on problem solving
proposing a solution Expertise
!the aha feeling Analogy
Set effects
Insight
Unfortunately,
the feeling does Solve remaining problems
not necessarily ! Tumor problem
indicate a
correct solution
Next time
Wrapping up the course
Other courses to take/avoid
Paths to pursue
Graduate school
Purdue University
Further study
Useful background
Brain characteristics
Mathematics ! PSY 222: Introduction to behavioral neuroscience
! Many psychologists have little mathematical ! SLHS 304: Anatomy and physiology of the speech and hearing
background mechanism
! But it is especially useful for cognitive psychology ! PSY 320: Behavioral neuroscience of sensation & arousal
! Take as much mathematics as you can, especially ! PSY 322: Neuroscience of motivated behavior
Calculus (MA 161, 165 or 223) ! PSY 324: Introduction to cognitive neuroscience
Finite (discrete) mathematics (not easy to get at Purdue
West Lafayette) ! PSY 512: Neural systems
Linear (matrix) algebra (MA 262, 265)
Differential equations (MA 266)
! PSY 520: Attention & performance ! SLHS 401: Language & the brain
! PSY 426: Language development
Memory:
! PSY 526: Psycholinguistics
! PSY 311: Human learning & memory
! PSY 314: Introduction to learning Problem solving & decision making
! PSY 410: Animal memory & cognition ! PSY 285: Consumer behavior
! PSY 518: Memory & cognition ! PSY 318: Problem solving & decision making
! PSY 514: Introduction to mathematical psychology
Purdue University Purdue University
Hot topic 1
Further study
Relating cognition to the brain (and vice-
Problem solving & decision making versa)
!PSY 285: Consumer behavior
Several big initiatives
!PSY 318: Problem solving & decision making
!
!PSY 514: Introduction to mathematical
https://www.humanbrainproject.eu
psychology
1.2 billion over 10 years
!PSY 390 with Pizlo, Proctor, Schweickert Develop technologies to bring together
disparate neurophysiological, anatomical,
molecular, and behavioral data
Database (big data)
Modeling (supercomputers, specialized
hardware)
Purdue University Purdue University