We behave effectively with respect to the world. What makes that possible?
May 22, 2015
We behave effectively with respect to the world. What makes that possible?
1
23
4
Problem: the world is at some remove. How do we achieve action at a distance?
We behave effectively with respect to the world. What makes that possible?
What are the mechanisms that allow perceptual agents to achieve action at a
distance?
PhenomenaWhat kinds of properties of the world are perceived?
PhilosophyWhat kinds of properties and theoretical assumptions should
anchor our theory of perception?
PhysicsWhat kinds of properties are present or “recorded” in the
energy media of the world?
PhysiologyWhat kinds of properties can sense organs and nerve cells
“record” and how doe these sense organs influence the perception?
What do we have to understand to understand the mechanisms of perception?
PhenomenaWhat kinds of properties
are perceived?
Problem: the world is at some remove. How do we achieve action at a distance?
Themes from the Overview
We behave effectively with respect to the world. What makes that possible?
PhilosophyWhat kinds of
properties/assumptions should anchor our theory?
PhysicsWhat kinds of properties
are present or “recorded” in the energy media?
PhysiologyWhat kinds of properties
can sense organs and nerve cells “record”?
Ancient notions… (500 B.C. to 1000 A.D.)•eyes receive facsimiles of objects
Indirect RealismWe know the world through
intermediaries
eye is a chamber that captures images
…persist in modern view•copies stand between us and the worldeye as camera•disconnected sensations get connected higher up
•light rays hit eye as pointillistic mosaic
The image is ambiguous, impoverished. (1) It doesn’t match the world.(2) It doesn’t match our experience.
Environment Organism link is bad therefore…
Perception requires processes to elaborate input, constructing a series of representations of the world that increasingly come to resemble it.
Examine the E O link
A B C DA B C D
How good are the images?
Molyneux’s Premise (1692): distance is not perceivable by eye
What gets linked?
What do you need to fix the bad link? Experience Knowledge:Empiricism
An object in the world at some distance from me that goads me or stimulates me to act:
distal stimulus or SD
The pattern at a sense organ caused by an energy pattern in the world:
proximal stimulus or SP
3 Old Guys who set the conceptual agenda
1. Berkeley (1700s)
Eventually, visual image memories of motions idea of distance.
touch is the rationalizer•direct contact with the world•not susceptible to the loss of structure•can help out those senses that are susceptiblemuscular feelings can help fix defects in visual image once I learn which images co-occur with which feelings.
extended arm grasps bottlecontract arm— image expands
45˚90˚
worries about physiology & its psychological counterpart
2. Müller (1826)
Sensory receptors•Are stimulated by energy (light, sound, pressure)•But not mere conductors of those properties
Emphasizes the contribution of anatomy & physiology
ELECTRIC PULSE
PRESSURE CHEMICALS
Visual sensation
OPTIC
NERVE
LIGHT
Eye designed to capture
light
SOUND
AUDITORY
NERVES
Auditory sensation
Ear designed to capture
sound
Müller’s Theory of Specific Nerve Energies impose their own characteristics— “specific nerve energies”—on the mind. This, not the physical properties themselves, is why the qualities for the different senses are different.
SP must be interpreted re: what is normal
•What would normally have produced these features?•What goes with what?
3. Helmholtz (1821-1894): THE MAJOR FIGURE
SD SP
object patchwork of sensations
Helmholtz’s Theory of Unconscious Inference
SP is converted to sensations—a mental response to energy (direct)
Sensations that habitually occur together become linked through memory—associations.
Normalcy is embodied in internal algorithms or rules.
cues + rules = Unconscious Inference
The input is a disjointed, inadequate copy of the world.Perception works by improving the copy via rules.
Perception of the world is indirect.
Attended Stimulus
Environmental Stimulus
Action
Stimulus on the receptors
Transduction
Processing
Perception
Recognition
The Perceptual Process:
Stimulus
Perception
Stimulus Energy
Sensation
Physiological sensation
Perception
Three key relationships
Knowledge cues & rules
Unconscious Inference
Berkeley: coupling of percepts uses meaningfulness of one to explain another
Müller: the sensory apparatus itself contributes its character to the input
Helmholtz: mental computations reflect internalized knowledge of the world and how it affects us.
The nature of E S influences the nature of S P
World Energy Sensations Perception
All share the theory of inadequate input
ignored link
historically important links
currency is converted into currency: How does physical energy map onto psychological
experience? Measurement is the key to making perceptual psychology a science
The nature of E S influences the nature of S P
World Energy Sensations Perception
ignored link
historically important links
currency is converted into currency: How does physical energy map onto psychological
experience? Measurement is the key to making perceptual psychology a science
Energy Sensation Experience
Energy SensationE S
Physical Psychological
Perceive event (cat rubbing leg)
Event in world
Pressure (energy)
Sense properties of pressure () e.g. amount, location.
Pressure sensitive nerves
Energy SensationPsychophysics
Anticipated by Weber (mid-late 1800s)If the amount of energy is too small, it’s not
noticeable.
Psychophysics looks at the E S link
Say “now” when you see the gray square.
Absolute threshold
minimum energy that can just be detected
Trial 1 Trial 2
No
No
No
No
No
No
Yes
Trial 5
No
No
No
No
Yes
Trial 3
No
No
No
No
No
Yes
Trial 4
No
YesYesYes
Trial 6
YesYesYesYesYesYes
No
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
No
Intensity (I)
3
4
5
6
7
8
9
10
2
1
11
12
0
654
1
32
Trial
Smallest Intensity Detected
657
6
75
Absolute threshold
Mean 6
6
Determining the Absolute threshold: Method of Limits
Anticipated by Weber (mid-late 1800s)If change in amount of energy is too small, it’s not
noticeable.
Psychophysics looks at the E S link
Weber’s focus was on discriminating two detectable stimuli: How similar could they be and still be sensed as
different?
Not absolute change but relative change
Just Noticeable Difference
I / I = K
A change in intensity
relative to the initial intensity equals a constant.
– =
Just Noticeable Difference (JND)
Different
I2 I1
102 g 100 g
I2 - I1 = I
I: physical intensity
Different
I2 I1
101 g 100 g
Same
I sensed is not absolute!!
Rather, the JND is a constant threshold.
Difference Thresholds: How similar can objects be and still be sensed as different?
Consider what happens when we use different values of intensity (I)
100 + 2 = 102 2/100 = 1/50 (.02)
200 + 4 = 204 4/200 = 1/50 (.02)
400 + 8 = 408 8/400 = 1/50 (.02)
The greater the value of I the greater must be the value of I for a difference to be sensed.
WEBER’S LAW: = K (a constant)II
Intensity I
JND I + I
II
Change in II K
JND’s for all senses: - Vision (e.g. change in brightness) - Hearing (e.g. change in loudness)
I
I
I
I
I = 20
The correspondence between a physical stimulus and our perception of it is systematic
but it is not always 1:1.
I = 500I = 500
.2
.5
smaller K more sensitive lower threshold
K = 1 I = 20
I = 1000I = 500 K
= .5
I = 20I = 10
1Slope
I / I = K 1st truly quantitative law of psychology
Fechner (mid-late 1800s)
Weber’s goal was to study jnds; Fechner’s insight was that such a quantification allows you to probe mental states.
Fechner (mid-late 1800s)
Demonstrated how mental activity could be measured quantitatively!!
… also started to examine whether we can assume a
equivalence of changes in intensity of stimulation?
Steven’s attempted to understand the relationships between and intensity in his examination of Magnitude estimation.
0 1000 2000 3000
Stimulus Intensity
300
200
100
10
Standard = 100
Response = 160
Response = 130
Response = 200
Response = 25
Response = 50
T1
T2
T3
T4
T5
300
200
100
100 1000 2000
3000Stimulus Intensity
Response = 150
Response = 225
Response = 350
Response = 90
Response = 95
T1
T2
T3
T4
T5
Standard = 100
Subjective intensity of magnitide () is some constant multiplied by the intensity (I) to some
power (n).
Stimulus Intensity
Mag
nit
ud
e
Esti
mate
Log Stimulus Intensity
Log
M
ag
nit
ud
e
Esti
mate
= kIn
n = the slope of the line in the log-log plot
Perceptual Sense reflects Power Law Functions
Implicit Metatheory: To say there is a absolute or noticeable threshold is to say
that there are un-noticeable things.
To say there is a just noticeable difference is to say that there are also un-noticeable differences.
To say the perception is power law like is to say that the…
Connection between mind & body is in the quantitative relation between mental sensation & material stimulus
I/I = k and = kIn highlight slippage between & (not 1:1).
How do you measure the change in stimulation?
You need methods…
…that yield quantities that can be put into law form
Fechner formally developed Psychophysics as the methodology,
a methodology that endorses a metatheory
Assumes the some currency has some currency
May not be 1:1, but we must therefore understand:
Understanding these things might help in understanding the processes of
perception.
Lets start with vision…
Psychophysics looks at the E S link
1. The physics of different energy medium,2. How the physiology is designed to transduce this energy
and… 3. How his transduction maps onto sensory and psychological
experience?
currency is converted into currency: How does physical energy map onto psychological
experience?
World Energy Sensations Perception
Understanding these things might help in understanding the processes of
perception.
Lets start with vision…
Assumes the some currency has some currencyMay not be 1:1, but we must therefore understand:
Psychophysics looks at the E S link
1. The physics of different energy medium,2. How the physiology is designed to transduce this energy
and… 3. How this transduction maps onto sensory and psychological
experience?
Light: The stimulus for vision
Electromagnetic radiation structured in waves* over space
distance
en
erg
y amplitude = Intensity
Wavelength
same amplitudedifferent
wavelengths
multiple wavelength
s(vs. pure)
same wavelengthdifferent amplitude
Complexity or Purity
Different wavelength
shue
multiple of wavelength
saturationbrightness
Differentintensities
Amplitude ≈ Intensity
Wavelength
Wavelength: most relevant for color vision…
How is light structured?
Light travels far•we can know about far
objects
surfaces, substances
source
Some light gets to eye
Light travels fast•we can know them
immediately
reflected scattered absorbed
Light travels straight•good for image-
production
Why light? Vision as a distance sense
The eye captures light reflected from objects and forms an image on the back of
the eye.
How “should” the image be formed?
box with a pinhole as the
eye
Solution #1:
Allow one ray from each part of the object into the eye.
Camera Obscura (Alhazen)
Limitation: Doesn’t let in much light—blurry image
Problem: Spatial ordering of rays reflected from the object have to be recovered from the divergent light.
Allows larger holemore lightsharper image
Limitation: Clear focus depends on the power of the lens and angle of divergence of light rays.
Problem: Spatial ordering of rays reflected from the object have to be recovered from the divergent light.
…while letting in enough light for a clear image.
Solution #2: Use a lens that refracts light so that rays
from the same point on the object converge.
Limitation: Clear focus depends on the power of the lens and angle of divergence of light rays.
Problem: Spatial ordering of rays reflected from the object have to be recovered from the divergent light
…while letting in enough light for a clear image.
Solution #2:Use a lens that refracts light so that rays
from the same point on the object converge.
Different distance of object from eye changes angle of light rays Out of focus for
that lens.
…of objects at varying distances.
Solution #3:
Lens with variable optical power changes shape to accom- modate the distance of the object to the size of the eye.
Problem: Spatial ordering of rays reflected from the object have to be recovered from the divergent light
…while letting in enough light for a clear image
The eye as a chamber for capturing light
Optical parts
Structures for gathering and focusing light
cornea
iris
lens
pupil optic nerve
What properties should the eye have?
Translating parts
Structures for copying light and sending
signals
fovea
Light is focused on light-sensitive photoreceptors of retina that convert or transduce physical energy into neural activity
•Lens shape can be changed to accommodate to the distance of an object for focusing on the back of the eye—the retina—centering on the fovea•Fluid in eye keeps its shape from changing so that focusing properties are reliable
Focusing-relevant features highlight the importance of the retinal image
LVF RVF
Retinal Image is Starting Point for Vision…and we have two
Visual Fields defined relative to fixation x:
Right Visual Cortex
LVF projects to right side of each eye and on to the Right Hemisphere
LVF RVF
Retinal Image is Starting Point for Vision…and we have two
Visual Fields defined relative to fixation x:
RVF projects to left side of each eye and on to the Left Hemisphere
Right Visual Cortex
Left Visual CortexRegions of left eye correspond to regions of right eye
At some point we have to (re)connect visual fields.
LVF projects to right side of each eye and on to the Right Hemisphere
•Philosophy: Begin with objective physical properties•Physics: Intensity, wavelength of reflected light•Physiology: Brightness, color but not 1:1•Phenomena: Product of mental computation
Physics Psycho- Anatomy & Algorithms physics Physiology
Goals: Get copy of world inside headDifficulty: Pointillisitic nature of light (and
sensations)
Solution & Problem: Eyes capture images
world energy sensations pattern perception
Pervasive Themes
Before we had techniques to see cells, we had behavioral data:
Go from bright light into dark room—can’t see at first
Improves for 5 min., levels off… improves again for 15–20 min.
thre
shold
minutes in dark•1st acts fast, adapts less.Kink in function is clue 2 functions•2nd adapts slowly but more.
2 functions 2 types of photoreceptors 2 job descriptions:
Transducing the image
•work in dim and bright light•provide sensitivity and clarity•work in B&W and Color
Location, number, connections differ.
Rods •across periphery
•fovea only has cones
no. of receptors per square mm
Blindspot: no receptors because the optic nerve leaves the eye.
•many:1 connections •1:1 connections
•more plentiful120,000,000
•fewer in number8,000,000
What are consequences?
What are consequences?
So does shape: rods
and cones
Cones
•many:1 with later cells
•1:1 with later cells
Ø!Ø
ØØ
minutes in dark
Threshold(minimal
visible light intensity)
Return to dark adaptation curves for hints
very little light required
more light required
Differ in Sensitivity
2 weak signals that are separate remain below threshold of next cell
less sensitive
2 weak signals that are connected can exceed threshold of next cell
more sensitive
Ø
Ø
ganglia ganglia
How a pattern is experienced depends on where it projects on the retina
Is there a cost to pooling signals?Is there a benefit to keeping signals
separate?
detail is missed: less acuity
detail is noticed: greater acuity
•mosaic of receptors breaks up continuity of world
Array of ≈ 130,000,000 photoreceptors converts the retinal image into a neural image to be transmitted to the brain.
But…
How good is mapping between world and experience?
Which of these problems get fixed higher up? How are they fixed?
•image is upside down
•different parts of retina have different sensitivities
•mosaic of receptors breaks up continuity of world
Array of ≈ 130,000,000 photoreceptors converts the retinal image into a neural image to be transmitted to the brain.
But…
Receptors outnumber cells in the next layer pooling of information, editing, altering before signals are passed along
How good is mapping between world and experience?
Which of these problems get fixed higher up? How are they fixed?
mechanismsrules
•image is upside down
•different parts of retina have different sensitivities
Ganglia respond to receptors, not to light (register differences in light)
Examine the language of ganglia: action
potentials
All or None
Cells beyond receptors condense and reorganize data130,000,000 receptors 1,000,000 ganglia
• many, many rods : 1 ganglion; 1 (or a few) cones : 1 ganglion a lot of editing What kind?
What kind of stimulus does a ganglion prefer?
Use single cell recording with electrodes
oscilloscope
amplifier
•a still eye to keep stimulus where you want it.•stimulate various areas of the retina see effect on a particular ganglion.
microelectrode
time (sec)
volt
ag
e
reveals editing
Without stimulus there is a base level of spontaneous activity.
Task: Find the region on the retina whose stimulation will change the resting level (higher or lower) of Ganglion “A”.
How: Scan retina with stimulus to see where ganglion’s activity is affected (where matters)
Homogeneous gray spontaneous activity.
Spot of light greater than spontaneous activity
Dark spot in area less than spontaneous activity
Within area greater than spontaneous activityOutside area less than spontaneous activity
Spontaneous firing rate is affected up or down
concentric ON/OFF regions Ganglion cell’s receptive
field(a collection of retinal cells)
“ON” response “OFF” response
Distribution of concentric ON/OFF regions arises from connections among preganglion collectors
What happens with light outside the ON/OFF region?
spontaneous rate
excited rate
inhibited rate
spontaneous rate
What is experienced depends on where it hits retinaAssessing various ganglia yields a receptive field map
overlap, producing a mosaic covering the whole retina
(also OFF center/ON surround cells)
How does Center/Surround organization work?
patch of light over entire field
patch of dark over entire
field
Uniform illumination on region modest activity
•ON center “prefers” light; surrounding OFF circle “prefers” dark
Increased activation from ON center countered by decreased activation from OFF surround.
antagonistic responses from center and surround
lateral inhibition
Decreased activation from ON center countered by increased activation from OFF
surround.
Dark edges over OFF surround with light on ON centervigorous response: whole receptive field is getting its preferred stimulusdetects and accents light/dark boundary
Edges are preferred by this kind of cell.
illuminate only the center
dark bar on the surround
What pattern on the retina would be preferred by a Center/Surround cell?
Consequences of antagonistic relationship between center and surround
Intensity of Center
Response
Intensity of Center
Response
Intensity of Center
Response
same response to dif. intensities
dif. responses to the same intensity
Constancy Illusions
Good mapping in a limited range?
From what we know about acuity, how should size vary in different areas of the
retina?large in periphery; small near fovea
Many:1 vs. 1:1—Receptive fields vary in size
Physiological mechanisms are recovering edge and size information building blocks of meaning.
Small receptive fields respond best to small objects; large to large beginning of object size extraction.
From what we know about preferences of receptive fields, how should they respond to objects of different sizes?
Acuity: smallest high contrast detail perceived at a given distance
Receptive fields have consequences for the kinds of patterns that go into Unconscious
Inferences.
What letter is this? F E A H O D P R Identification Acuity
Can you see this? Detection Acuity
1 or 2? Pattern or gray?
Resolution Acuity
K
Note disparities between and
made possible by lateral inhibition — mechanism that highlights edges through sideways connections
among cells.
Illusory consequences illustrate how it works.
dependence on “irrelevant” conditions (e.g., distance)
The “private line” from foveal cones to the brain provides fine detail……but it’s neurologically expensive receptive field organization is important
Mach bands—regions of heightened and reduced brightnesses.
Intensity changes in stepwise fashionLight
Intensity
1 2 3 4 5 6Position
high
low
Light Intensity
100
40
a b c d e f g h
activity w/NO neighbor: 40 40 40 40 100 100 100 100Inhibition from Left: -2 -4 -4 -4 -4 -10 -10 -10Inhibition from Right: -4 -4 -4 -10 -10 -10 -10 -18Total Output: 34 32 32 26 86 80 80 72
Perceived
Lightness
lightness does not.
Receptors
(activity - inhibition)
Mach bands do not exist physicallywouldn’t be picked up by a photometerPhysiology imposes its character on the input•A physical contrast—a border—that does exist is accentuated•Machinery can be inferred from experience
stimuli ≠ perception of those stimuli
Lightness Contrast
Implies interaction in connections between neighboring cells:some signals boosted, some signals reduced
: Central squares reflect same amount of light.: The darker the surround, the lighter they look.: Central squares reflect same amount of light.
B
A
B
AALEFT looks darker than ARIGHT
Initial “strength” of signals (registered by rods)
ALEFT = ARIGHTBLEFT > BRIGHT•excitatory or inhibitory
Signal from BLEFT inhibits signal from ALEFT lateral inhibition
If signal from B exceeds threshold of laterally connecting cells, signal from A will be reduced
Signal from BRIGHT does not affect ARIGHT Consequently, ALEFT < ARIGHT
•sideways
Subsequent connections•end-to-end
•ganglia pool information, change from what is given copies (which are inadequate)
Evidence for assumptions of Indirect Realism
•separate receptors, connected into receptive fields disconnected sensations, mosaic•eye as camera, receptors as film image language
Same mechanism is destructive and constructive
•It distorts relative to Illusions•It enhances the detection of an important feature of the world edges
sends excitatory signal when stimulated
sends inhibitory signal when stimulated
front view
•many:1 with later cellsgreater sensitivity
•1:1 or few:1 with later cellsgreater acuity
side view
Receptive fields care about size & shape…
…but not orientation.
”
reduced rate:stimulus hits both excitatory and inhibitory cells
Orientation influences what objects mean
Pool some more.
To overcome mosaic, connect receptive fields.
Receptive fields overlap
Across a collection of receptive fields, orientation matters
Collection reports to cells in the cortex.
They have receptive fields too
Record from 3 cortical cells
Cortical cells do edge detection but more cleverly
Cortical receptive field shapes are not uniform Hubel & Wiesel (1959, 1962; Nobel Prize 1981)
stimuli must be positioned appropriately
maximal response to stimuli of a particular orientation ±15°.
response
rate
response
rate
Simple cellsreceptive fields look like their preferred orientations
Complex and
Hypercomplex cells
Some cells prefer movement of those features in a particular direction
•don’t care about precise placement•larger receptive fields•no clear ON-OFF regions•prefer moving stimulus•combinations of features
Measure activity in cell that prefers downward moving horizontal 3 cm line
Provide some information about where, what, and what’s it doing.
Are Cortical Cells Feature Detectors?
Rate of firing is the only vocabulary—how is ambiguity resolved?
Reduced response if orientation or size or motion is not exactwhich is it?
More complex response is still ambiguous
directi
o
n
orientation
response rate
Response is ambiguous•Respond maximally, not exclusively, to their feature
responserate ambiguity at level of single cell
• Multiple representations (> 100,000,000 cortical cells) of the retina and visual field
• Each cortical area performs different processing tasks, extracting specific feature
• Integrated with other properties • Hierarchical organization of visual system geared for building up
ever-better representations of world: simple to complex
Activity of cell ensemble reduces ambiguity in coding
•What kinds of images does the eye produce? which qualitative properties are preserved, lost, distorted
Physical Psychologicallight or an object detection thresholdsdifferences in intensity, etc. jnds, acuityIntensity, reflectance constancy, contrastdifferent wavelengths not distinguished in dim lightX-ray, infrared visible spectrum
gives us variables builds from elements
meaningless adds meaning
Strategies of Indirect Realism
•Psychophysics Characterize match to the world quantitatively
Reality question is emphasized in color perception.•Objects have no color. Reflected light is no more colored than are radio waves! •To appear colored, reflected light must be picked up by the right kind of eye and nervous system.
Color is a psychological thing.
Visual stimuli are built hierarchically from simple to complex
Which level is real, the simple end of the hierarchy or the complex end? The variables or the experiences?
The first representation—levels of contrast—is meaningless and must be reconstituted to get more specific and more meaningful.
Is an edge real?
surfacepigmentshadow
wavelengths, not colors
produce; reflect, absorb; transmit
Color is totally subjective
•Light rays, paints, filters, etc. merely use radiant energy selectively.
•Color is a product of the visual system, not the visible spectrum.
how our retinal physiology responds
White light decomposed into spectral components refracted by a prism and split into rays of different wavelengths. amount of refraction determined by wavelength
Newton (with some refinements)
Nonetheless, color sensations are related in consistent and measurable ways to physical
features of light
infrared (not visible)redorangeyellowgreenbluevioletultraviolet (not visible)
Complexity or Purity
Different wavelengt
hshue
multiple of wavelength
saturationbrightnes
s
Differentintensities
Amplitude ≈ Intensity
Wavelength
How do we respond to different kinds of light?
Pure #1 + Pure #2 = Composite
equivalence is due to nervous system •Different things with identical neural effects: metamers
pure vs. composite light
If Pure #1 + Pure #2 + … = White then 1 & 2 are complementary
Components “cancel” or “blend” psychologically but not physically spectral components would still be detectable
to instruments.
Both are white
: wavelength : hue
You can add pure colors and get one that’s not a spectral color no characteristic wavelength
530
650
600O
Metamers tell us how to organize the optics: Complementary colors are opposites in some
sense
580580
R
G
B
YY
460
???P
490 B-G
GG--YY490490
A color circle, but…
Other experiences suggest organizations more elaborate than a circle
: intensity : brightness
What happens when there is more or less light?
blue . . . heather blue . . . graygreen . . . heather green . . .grayred . . . pink . . . gray
What does the color look like?
The higher the %white the less saturated a color will look.The higher the % other wavelengths, the less saturated a color will
look.: spectral purity :
saturation Maximal at moderate intensities only
bri
ghtn
ess
white
black
Colors on opposite sides gray
Broadest portion appears at medium lightness.
Any cross-section color wheel for a particular lightness
Wavelength + Intensity + Purity Color experience
all colors can be obtained from a few primaries
tells us about the physiology of color perception
The Color Solid
Phenomenological observations: •the color solid, metamers, complementary
colors allow inferences about physiology
A single wavelength is matched by different amounts of 3 primaries
indicate limits of the information senses pick up indicate how the brain uses that information
Metameric Matching:
Present a target color and have observer match it with a mix of others.
Would 1 cone type work? Or 1 cone per color?
3 primaries
ComparisonxR + yG + zB
1 wavelength
Test FieldC =
Young: 3 specialized cones each acts as a channel responsive to specific spectral composition.
Retina as mosaic of independent triads of cones
Light at each retinal point analyzed into 3 components
Return to experience: •Why is there no reddish-green?•Why is color blindness a matter of confusions? most common is red with green•color afterimages show same combinations
S M L100
0
400 500 600
Wavelength
% of Max. Absorptio
n
Young-Helmholtz (and Newton and Maxwell) Tri-chromatic Theory
Helmholtz: 3 types of cones, each with a graded sensitivity function
What is the origin of the pairings?
Color blindness comes in pairs
Trichromatic theory not the whole story
Color afterimages
complementary colors:
R-G & B-Y
Color blindness comes in pairs
Trichromatic theory not the whole story
Color afterimages
Opponent Process Theory: Perhaps outputs of cones are re-coded somewhere into pairs whose members are antagonists (Hurvich & Jameson, 20th Century)
YYB R
G YYB R G
Opponent Organization
•Hue determined by relative activity of R-G and B-Y;
• lightness determined by activity of B-W cells; •saturation by which kinds of cells are most
active:
violet
• Complementary colors cannot coexist because a pair cannot respond actively to both
• Color deficiencies are characterized by pairs
light
optic nerve
fovea
retina
DESIGN OF RETINA
TO OPTIC NERVE
To Brai
n
RODS
CONES
BIPOLARS
GANGLIONS
LIGHTLIGHT
A NEURAL SYSTEM OF OPPONENT PROCESSES
+
––
CONES
GANGLION
IF + > –, THEN “BLUE”IF – > +, THEN “YELLOW”
IF + > –, THEN “RED”IF – > +, THEN “GREEN”
CONES
GANGLION
–
+ +
FOR BOTH OPPONENT PROCESS SYSTEMS:
IF + = –, THEN “GRAY”
(ACHROMATIC)
Wavelength info at retinal level; feed into opponencyFits into the overall theme of the perceptual system missing physical detail, restoring lost structure, making things up as it goes along. But it is also an example of the visual system getting what it needs: There is a biological advantage to seeing color.
Note: We’ve really been limited to sensations. Perception is still to
come.
Color coding is a two-stage process.
Puzzle:
All this happens inside eyes and brain. How do we experience the world as outside?
101
Bananas?3 pointy cone shapes?
Edges are a specialty of the visual system.Specialty implicates a mechanism for perceiving separable objects.
•Neurophysiology: primitives hierarchical combos
Old theories illustrate persistent issues
analytic introspection: specialized technique for observation stimulus error: mistaking the object for experience
Sensations vs. Perceptions
Structuralism was a kind of mental chemistry. Elemental structures combine to build perception
What does the visual system start with? What are the elements?What do you really see, without interpretation?
No, reallyNo, reallyNo, reallyangled planes, ellipse?black lines, yellow patches
Important historically because of •its strategy atomistic & anatomistic•reaction against it
102
How do I know that this pattern is a hand?
•“geons” •and their relationships•applied to SP & to stored representationStructural Description
Examples of types of representations and types of computations
Analysis must provide components abstract enough to match stored representation.
More candidate primitives
Or this? Or this?
103
corner of front surface Keep track of vertex-connected surfaces eliminate those that are inconsistent
not always
3-D corner of one objectnot always
For complicated—natural—scenes, occlusion is a problem
Even if these are identified as cylinders, how do we know they are part of the same object?
We need rules about what’s likely.
…with what you know about objects
Given experience, assign to SP the SD that is most likely to have caused it.
Overcomes problem caused by occlusion
104
=
Different neural activity = different forms… HOWEVER… Form (object) recognition still presupposes a solution…an internal representation.
Still doesn’t answer the how question. How (where) does perception occur?
105
A cross … But why? Perception is not simple addition of elements •“Crossness” not derived from adding circles•Defects (e.g., gaps) in RI would be preserved in brain
Structuralism ignored interactions in nervous system play organizing role
The Gestaltists
Why do things appear as they do?
What do you see?
Gestalt is German for whole form“The whole is different from the sum of its parts.”
Principles of Organization encourage grouping, segregation form emerges
Yes! But what are they?
i. They are what they are?
iii. The brain processes are what they are?
ii. SP are what they are?
is 9 circles
1 SP 2 experiences
106
Grouping or organization is not in the stimulus.
Seen as 3 pairs of lines•SP does not dictate that over
Seen as intersecting shapes•SP does not favor
The Law of Proximity does
or
Seen as alternating columns•SP does not dictate that over
The Law of Similarity does
or
The Law of Good Continuation does
over
Seen as triangle in front of 3 circles•SP does not dictate that over
The Law of Closure does
107
What counts as a form or grouping distinct from a background?
3-D or 2-D?
In general, Simplicity
108
What is the figure can be complicated
Ambiguous Figures: For same SD and same SP, two percepts are possible
How can you have 2 simultaneous, incompatible representations of the same thing?
•Figure-ground is reversible•Can be shifted by attention
Figure vs. Ground: What does the edge belong to?
109
Of the alternatives allowed by the proximal stimulus… infer the more likely.
Figure vs. Ground: What does the edge belong to?
The Helmholtzian Solution:
Use knowledge of which configurations are likely: Principle of maximum likelihood.Make Unconscious Inferences about the world.
Contrast detection is not enough identify which form the edge belongs to(Pattern recognition presupposes a solution)
110
Structuralists emphasized identifying primitives as adding or associating sensations…theirs were too subjective, too qualitative
Gestaltists emphasized emergent properties or organizing through grouping lawstheirs were descriptive, not predictable.
But, how it looks ≠ what it is.Form requires further processing.
Teacher through your glasses?
Teacher in your locket?
far
near
For example, distance matters
111
Pattern recognition = ƒ(distance, size, shape)
Is there reasonably reliable structure that might be used to solve Molyneux’s problem? A B C
D
How do we know both relative distance and absolute distance?
Back to Helmholtz and Unconscious Inference
How does SP come to indicate a particular SD given that SP is2-dimensional and, therefore, ambiguous?
Reasonably reliable in a Helmholtzian what-is-normal sense is provided by cues
RI for large, far objects = RI for small, near objects
112
•texture gradient
•interposition
•SP was probably caused by a particular SD
•Become associated with properties of SD through experience.
II. Attempts to simulate depth: pictorial cues
•relative size
Aspect of SP
I. Berkeley “on-line”: oculomotor cues
X
X
convergence
accommodation
X
X
Distance Perception
•linear perspective
113
III. 2 eyes that receive slightly different views: binocular disparity
Left thumb behind,Right thumb in front;Both far away
Thumbs close together
R
Amount of disparity indicates relative distance, separation
RL
Thumbs far apart
LL
Both up close
RL R
Motion Based Cue for Depth:
Ever look out the window while riding in a car?
Direction of Travel
Objects in the foreground move by faster than objects in the background - Very distant objects appear to remain stationary
Motion Parallax
115
Size Perception
But there’s ambiguity…
RIpumpkin = RIapple Which is bigger?
RIpumpkin > RIapple pumpkin is larger than apple
I. Retinal Image Size: Larger objects project larger RI
II. Familiar Size pumpkins are larger than apples…
•except for small pumpkins compare to similar objects in the scene: III. Relative Size
Unconscious Inference uses knowledge about what’s normal
116
Knowledge is needed to make sense of ambiguous cues to attribute the proper cause to your SP.
Helmholtz’s Unconscious Inference
In natural scenes cues co-occur strengthen the impression of depth & size.
In experiments, cues are manipulated alter the impression of depth & size.
Strategy: isolate cues or put them in conflict to assess
•cues’ relative potency•illusions gain insight into normal processes, algorithms
We perceive constant object properties despite variability in the SP that is our contact with them. Perceptual Constancies
117
SizePerceived = ƒ(RI Size, DistancePerceived)
Tested by Holway & Boring (1941)
2 hallways:•target disk on L
Indicate size by method of adjustment
Targets chosen to project same RI @ every distance
Percept-Percept Coupling
•response on R
118
Manipulate available cues to distance and assess effect on perceived size:
•fewer cues to distancemore reliance on RI
Breakdown of Size Constancy
F
M
S
P
Actual Size
Perceived Size
constancy
slope = 1
retinal size
slope = 0
Full Cue
Monocular
Static
Peephole
119
Helmholtz
•SP is bad ( shows it)•Perception good (experience shows it)•fixed by rules (experiments show it) Perception is indirect
SP1 ≠ SP2
Percept @ d1 = Percept @ d2
Size Constancy
d1
d2
Ex. I. Perceived Size is unaffected by distance
Perceptual Constancies
120
Rule: RI size decreases with D. So…•If A & B project same size SP•And if A is farther than BA must be bigger than B
percept-percept coupling
A looks bigger than B
A
B A = B
Cue: SP:A = SP:BCue: linear perspective
says “A is farther than B”
Perc’d size is derived from perc’d distance
121
SP1 ≠ SP2
Percept @ p1 = Percept @ p2
Shape Constancy
p1
p2
Ex. II. Perceived Shape is unaffected by perspective
122
Appearance is affected by interpretation!
Parallelograms look similar in size and shape(one is rotation of other)
ShapePerc’d is derived from distancePerc’d
Adding distance cues changes inferred shape.percept-percept coupling
Manipulate depth cues & assess consequences for ShapePerc’d.
123
Fixing SP is logically prior to pattern recognition. What form is it?
Fundamental problem is normalization of image for memory.
Compare to stored representations•How abstract can the representations be? •How tolerant are matching processes?
How would a representational account handle movement?
Objects in the world often move…they are not not static!!
124
125
126
Motion Perception
t2
t3
t1
Physical Event
Perception
Apparent (Stroboscopic) Movement
Do we see things as they are because of the proximal stimulus? No
Do we see things as they are because of brain states? Yes
Or…
A temporal property (change over time) is derived from a succession of static retinal images.
127
Interpret with respect to likelihoods.
Apparent motion of the disk induced by assumption that enclosures don’t move.
The same assumption would underlie our experience of non-illusory motion, too.
Again, illustrated by an illusion: “induced motion”
128
1
23
4 1 4
23
BUT…
Helmholtzian account: compare RI with commands to muscles Did I tell myself to move?•ambiguity if I’m being moved
passively•ambiguity if we’re both moving
Stimulation of successive retinal locations motion
1
23
4
Same pattern for car moving to left and me moving to right.
1
23
4
Puzzle: How is leftward movement of something distinguished from rightward movement of me (or my eyes)?
129
t1
Form Perception Motion Perception
Sequence of static RI infer motion•xyz at t1 vs xyz at t2 vs xyz at t3•deduce rotation about Y
How do you know which points correspond if you don’t already know the type of motion?
t2t3
Helmholtz asked “what is likely?”
What if both are likely?
130
Shortest-Path Constraint… simplicity again!!!
131
Themes from the Overview
•Phenomena: What kinds of properties are perceived?•Physics: What kinds of properties are “recorded” in the energy media?•Physiology: What kinds of properties can sense organs and nerve cells “record”?•Philosophy: What kinds of properties should anchor our theory?
We behave effectively with respect to the world. What makes that possible?
WHAT WHO HOW
The Phenomena of ATTENTION: Orienting
What is the girl feeling…emotion?What is the information for making these judgments?
• Attention drawn to most informative aspects of picture.
Eye Movements
The Phenomena of ATTENTION:
8 8
8 8
8 8
s c
z k
e t
8 8
8 8
8 8
s k
z c
ft
e
s c
z k
e t
s k
z c
ft
e
8 8
8 8
8 8
Stimulus Driven Attention
Search for the letter ‘c’
The Phenomena of ATTENTION:
Goal Directed – Intentional/Change Blindness
Does the number of white T-shirt players change?
Any Gorillas???????
135
Movie
The Phenomena of ATTENTION:
Selective ListeningDICHOTIC LISTENING
WHAT IS KNOWN ABOUT UNATTENDED?Only physical characteristics (speech like sounds). Not meaning.
I cannot tell a lieNever kill a snake
SHADOWING
I cannot tell a lie
“unattended” “attended”
COCKTAIL PARTY EFFECT
Except when important or relevant (e.g. name)
The ever present Unconscious at work yet again….
137
Proximal Stimulus
MeaninglessSensations
Association/CuesDistal Stimulus
Incomplete“Percepts”
Incomplete“Percepts” Perception
Unconscious Inferences, information processing, Laws of organization
Indirect Perspective of Perception Assumes:
Both relinquish the responsibility of perception to an “internal”, “mental”, knower… a homunculus… who organizes and isolates cues and compares percepts and representations.