Transcript The First Cognitive Psychologists

Chapter 1
Introduction to Cognitive Psychology
The Complexity of Cognition
• Cognition involves
–
–
–
–
–
–
Perception
Attention
Memory
Problem solving
Reasoning
Decision making
– All include “hidden” processes of which
we may not be aware
The First Cognitive Psychologists
• Donders (1868)
– Mental chronometry
• Measuring how long a cognitive
process takes
– Reaction-time (RT) Experiment
• Measures interval between stimulus
presentation and person’s response to
stimulus
The First Cognitive Psychologists
• Donders (1868)
– Simple RT task: participant pushes a
button quickly after a light appears
– Choice RT task: participant pushes one
button if light is on right side, another if
light is on left side
Simple RT
Choice RT
The First Cognitive Psychologists
• Donders (1868)
– Choice RT – Simple RT = Time to make a
decision
• Choice RT = 2.1 sec
• Simple RT = 2 sec
• 1/10th sec to make decision
The First Cognitive Psychologists
• Donders (1868)
– Mental responses cannot be measured
directly but can be inferred from the
participant’s behavior.
The First Cognitive Psychologists
• Helmholtz (~1860s)
– Unconscious inference
• Some of our perceptions are the
result of unconscious assumptions we
make about the environment
– We infer much of what we know about
the world
Caption: The display in (a) looks like (b) a gray rectangle in front
of a light rectangle; but it could be (c) a gray rectangle and a sixsided figure that are lined up appropriately or (d) a gray rectangle
and a strange-looking figure that are lined up appropriately.
The First Cognitive Psychologists
• Ebbinghaus (1885)
– Read list of nonsense syllables aloud
many times to determine number of
repetitions necessary to repeat list
without errors
The First Cognitive Psychologists
• Ebbinghaus (1885)
– After some time, he relearned the list
• Short intervals = fewer repetitions to
relearn
– Learned many different lists at many
different retention intervals
The First Cognitive Psychologists
• Ebbinghaus (1885)
– Savings = [(initial repetitions) – (relearning repetitions)]
/(initial repetitions)
– S = (Ri-Rr)/Ri
– Forgetting curve shows savings as a function
of retention interval
Ebbinghaus’s retention curve, determined by the method of
savings. (Based on data from Ebbinghaus, 1885.)
The First Cognitive Psychologists
• Wundt (1897)
– First psychology laboratory
– University of Leipzig, Germany
– RT experiments
The First Cognitive Psychologists
• Wundt (1897)
– Structuralism: experience is determined
by combining elements of experience
called sensations
gustatory
olfactory
visual
auditory
haptic
The First Cognitive Psychologists
• Wundt (1897)
– Analytic introspection: participants
trained to describe experiences and
thought processes in response to stimuli
The First Cognitive Psychologists
• John Watson noted two problems with this:
– Extremely variable results from person
to person
– Results difficult to verify
• Invisible inner mental processes
The Rise of Behaviorism
• John Watson proposed a new approach
called behaviorism
– Eliminate the mind as a topic of study
– Instead, study directly observable
behavior
The Rise of Behaviorism
• Watson (1920) – “Little Albert” experiment
– 9 month old became frightened by a rat
by pairing a loud noise with every
presentation of the rat
The Rise of Behaviorism
• Watson (1920) – “Little Albert” experiment
– Behavior can be analyzed without any
reference to the mind
– Examined how pairing one stimulus with
another affected behavior
In summary: cognitivism was in crisis…
9am
Skinner:
S
R
Operant conditioning: reinforcers [e.g. food]
“Verbal behavior” (1957): language learned via imitation and reward.
Chomsky: kids use untrained sentences; make errors given reward.
COGNITIVE REVOLUTION
MIND = COMPUTER
-information-processing device
-several stages
computer
I
Input
processor
Memory
unit
human
I
Filter
Detector
Arithmetic
unit
To memory
Cherry (1953) experiment
Attend Left
attended sentences remembered
physiological R
mental R
behavioral R
O
Mental rotation: Shepard & Metzler (1971)
RT
Angle difference
Same or different?
Measure brain activity during learning
read 200 words: create an image
“dirty” = “garbage dump”
20 hrs later:
same 200 words
“did you see this word?” Y/N
Result: 54% from the 1st group
remembered
Brain activity
Davachi et al (2003):
remembered
forgotten
RESULT OF MEMORY TEST
10am
sound to electricity
Motor
area
Auditory
area
to arm and hand
+ Knowledge:
1. Alarm will go off again
in 10 min
2. Still time to get to
class
Neurons: building blocks of nervous system
Golgi first to prove how a neuron looks like
receptor
dendrites
S
axon
TRANSDUCTION: energy to energy conversion (just like ATM)
Why study single neurons?
minielectrode
oscilloscope
voltage
axon
spikes = action potentials
1/1000 sec
1/10 sec
Stimulus intensity represented by firing rate, not spike magnitude
axon
SIGNAL PROPAGATION
without decrease in size
HOW NEURONS COMMUNICATE?
Direct contact? (touch?) NO!
SYNAPSE (space between axon and next neuron)
Early 1900s: action potentials DO NOT travel across synapses
-they TRIGGER a chemical process
-synaptic VESICLES open and release chemicals
(NEUROTRANSMITTERS)
E
increased firing
electrodes
I
decreased firing
1
2
3
4
5
6
7
EXCITATORY
NEURAL
CIRCUIT
A
C
Properties:
CONVERGENCE
INTERACTION OF E & I
Firing rate (B)
B
4
3-5 2-6 1-7
Receptors
1
2
3
4
5
6
7
INHIBITORY
NEURAL
CIRCUIT
A
C
Firing rate (B)
B
4
3-5 2-6 1-7
Receptors
time
Optic nerve
time
time
time
brain areas: neurons ever more specialized
Hubel & Wiesel (1965): feature detectors
simple cells
complex cells
End-stopped cells
convergence
+
excitation
+
inhibition
…but how to recognize a specific face?
NEURAL CODES
SPECIFICITY CODING: representation of a specific stimulus
firing rate
GRANDMOTHER CELL: responds to only one stimulus
firing rate
neurons
neurons
DISTRIBUTED CODING: the pattern matters, not cells
11am
BRAIN LOBES: outer covering = cerebral cortex
Motor function
Language
Thought
FRONTAL
Memory
Vision
Attention
Touch
PARIETAL
TEMPORAL
Language
Memory
Hearing
Form perception
OCCIPITAL
Vision
SUBCORTICAL AREAS (INSIDE THE BRAIN)
THALAMUS
Vision
Hearing
Touch
HIPPOCAMPUS
Memory
AMYGDALA
Emotions
Emotional memory
CEREBELLUM
Sensory integration
Motor control
NEUROPSYCHOLOGY: behavior after brain damage
Behavioral breakdowns specific to brain damage.
SINGLE DISSOCIATION: STM intact, LTM lost
DOUBLE DISSOCIATION:
Person 1: STM intact, LTM lost
Person 2: STM lost, LTM intact
Proof that
1) STM & LTM have different mechanisms
2) STM & LTM independent of one another
COGNITION: how to measure it in the brain?
BRAIN IMAGING
PET (positron emission tomography):
-blood flow indicates cognitive process
-radioactive stuff injected into blood
-machine measures radioactivity levels
SUBTRACTION TECHNIQUE:
stimulation
_
control
=
difference
fMRI: functional magnetic resonance imaging
-no radioactive material involved
-hemoglobin carries oxygen
-contains iron molecules
-have magnetic properties
Active area
less oxygen, more iron
BRAIN IS ADAPTIVE, FLEXIBLE
“greebles”
firing rate
EXPERIENCE-DEPENDENT PLASTICITY
teaching
neurons new tricks
before
training
after
training
1pm
“perception is simple and easy”
1960s: “will build robot within 10 years that can see, feel and act
like human”
Stimulus energy & [knowledge, context, experience]
briefly flash stimulus
Palmer (1975):
Context influences perception
Related
83% correct
Misleading
40% correct
Unrelated
50% correct
Follow the lead of early cognitive psychologists…
TASK: perceive letters
THEORY: TEMPLATE MATCHING (perception based on features)
K
K
How would a machine do it?
K
need template for every orientation
INTERACTIVE ACTIVATION MODEL (McClelland & Rumelhart, 1981)
Word
FORK
Letter
F
ROOF
K
O
Feature
Stimulus
strongest activation wins
K
R
WORD RECOGNITION
Word
FORK
Letter-position
FFFF
ROOF
KKKK
OOOO
RRRR
Feature
F,R,K
K
Stimulus
strongest activation wins
R, K
FORK
O
O, R
F
WORD SUPERIORITY EFFECT (Reicher, 1969)
a)
b)
c)
stimulus
flashed
mask
present
Which appeared?
FORK
XXXX
K
XXXX
M
QUICK & ACCURATE
XXXX
K
XXXX
M
SLOW & INACCURATE
XXXX
K
XXXX
M
SLOW & INACCURATE
K
RFOK
-LETTERS IN WORDS AFFECTED BY CONTEXT
-LETTERS IN WORDS NOT PROCESSED LETTER BY LETTER
FEEDBACK ACTIVATION explains this result:
FORK
F1
ROOF
K4
O2
R3
-no feedback when standalone letter presented
-word level sends FB to letter level as reinforcement
WORDS
TOP-DOWN
FB
LETTERS
FEATURES
BOTTOM-UP
PROCESSING
2pm
FEATURE INTEGRATION THEORY (TREISMAN, 1986)
object
preattentive
stage
focused
attention
stage
-analyze into features
-not conscious
combine
features
perception
Do we really break objects into features?
Do features exist independently of objects?
1
8
Treisman & Schmidt (1982)
1
8
Task: ID the numbers, then the rest
Interesting errors… RED CIRCLE, GREEN TRIANGLE, etc.
ILLUSORY CONJUNCTIONS (18% of responses)
“redness”
“curvature”
“tilted line”
…
not yet associated with a specific object
Attention part of conscious perception: no errors if asked to focus on figures
Skeptic: “Hey, but I still don’t buy it! I SEE objects, not features!”
Answer: BALINT’S SYNDROME (case of R.M.)
parietal lobe damage
can’t focus attention on individual objects
TASK: identify colored letters
T
O
23% of RESPONSES = illusory conjunctions (“red T”, “blue O”)
CONCLUSION: you need attention, otherwise ONLY features perceived
TOP-DOWN influence HELPS to reduce errors:
Experimental condition: Illusory conjunctions LESS LIKELY
CARROT
TIRE
LAKE
Control condition (objects not labeled):
illusory conjunctions occur
Features = lines, curves, colors
What about 3D object perception?
We have a theory for that, too!
RECOGNITION-BY-COMPONENTS THEORY (Biedermann, 1987)
?
GEONS ARE (MOSTLY) VIEW INVARIANT:
But…
3 parallel edges seen from many angles
GEONS ARE (MOSTLY) DISCRIMINABLE:
each geon can be distinguished from the others
GEONS ARE (MOSTLY) RESISTANT TO VISUAL NOISE: low light, fog, occlusion
Strength of Biederman’s theory:
9 geons
3 geons
3pm
Perceptual organization
Gestalt psychology
Structuralism: image consists of dots (sensations)
Sensations combined to result in perception of the glasses
Overall PATTERN matters. But how do you combine sensations?
BOTTOM-UP
TOP-DOWN
ONE
STIMULUS
ONE
STIMULUS
THREE
PERCEPTIONS
ONE
PERCEPTION
X
X
GESTALT LAWS OF PERCEPTUAL ORGANIZATION
LAW OF PREGNANZ (LAW OF SIMPLICITY):
simplest configuration perceived
GESTALT LAWS OF PERCEPTUAL ORGANIZATION
LAW OF SIMILARITY:
similar things grouped together
GESTALT LAWS OF PERCEPTUAL ORGANIZATION
LAW OF GOOD CONTINUATION:
smoothest path determines sameness
GESTALT LAWS OF PERCEPTUAL ORGANIZATION
LAW OF PROXIMITY (NEARNESS):
closely spaced things grouped together
GESTALT LAWS OF PERCEPTUAL ORGANIZATION
LAW OF COMMON FATE:
same direction of movement groups things together
GESTALT LAWS OF PERCEPTUAL ORGANIZATION
LAW OF FAMILIARITY: if the collection of parts is
meaningful, it forms a group
GESTALT LAWS describe our everyday world
…really, they’re HEURISTICS (rules of thumb):
work most of the time, not always
contrast with ALGORITHMS: always correct
Slow, analytic process
Fast, perceptual process
Perception is INTELLIGENT, although naively considered easy and trivial
[we see patterns, where there were none]
[irreversible]
4pm
1960s: perception is simple; will build robot that sees within 10 years
1997: computer beats human in chess
2005: computer-driven cars navigate 130+ miles of desert road
Navigation
Cognition & computation
Object recognition
???
RETINAL IMAGE IS AMBIGUOUS
-inverse projection problem
DISTINGUISH OBJECT-BACKGROUND
SEGMENTATION IN SPEECH PERCEPTION
-voice identification devices
BLURRY IMAGE: what is the threshold for perception? How do we do it?
Change in LIGHTNESS due to object properties or illumination:
Computer can’t tell whether (a) and (b) are part of the same object
“I scream, you scream, we all scream for ice cream.”
Same sound, different context for meaning
TRANSITIONAL PROBABILITIES:
whether two neighboring syllables are part of the same or different word
Saffran et al (1996): infants sensitive to things that occur together
regularly in the environment
Stimulus: …bidakupadotigolabutupiro…
…golabutupirobidakupadoti…
…
Within-word transition probability =100% (da-ku)
Between-word transition probability = 33% (ku-pa)
When head turns to light  sound starts
When head turns away  sound stops
2 minutes
listening

TEST
STAGE
“tibida” (part)
“padoti” (word)
Saffran et al (1996) results:
-never heard these words before
-no pauses between words
-only 2 minutes to learn
Listening time (sec)
LEARNING
STAGE
Whole
word
Part
word
Inborn capacity or learned? This is the knowledge that computers need?