Title: Psyc 351: Psychology of Perception Week 8 Overview of Perceptual Organization
1Psyc 351 Psychology of PerceptionWeek 8
Overview of Perceptual Organization
- Adapted from lectures given at
- BISCA, 2001 Bolzano, Italy
James R. Pomerantz Department of Psychology Rice
University
Houston, Texas, USA
September, 2001
Pomerantz, BISCA, Bolzano, Italy
I -1
2Perceptual Organization in the 1930s Karl
Dallenbachs photograph
September, 2001
Pomerantz, BISCA, Bolzano, Italy
I -2
3Perceptual Organization in the 1970s R. C.
Jamess Photograph
September, 2001
Pomerantz, BISCA, Bolzano, Italy
I -3
4Perceptual Organization in the 2000s Bev
Doolittles Painting
September, 2001
Pomerantz, BISCA, Bolzano, Italy
I -4
5Key Issues Raised byDallenbachs Cow
- Where are the objects in the image? Where does
one object stop and the next one start?
- Which edges represent illumination changes and
which reflectance changes?
- What are the objects?
- What is the role of knowledge, top-down
processing in segmentation and identification?
September, 2001
Pomerantz, BISCA, Bolzano, Italy
I -5
6What Is Perceptual Organization?
- the processes by which the bits and pieces of
visual information that are available in the
retinal image are structured into the larger
units of perceived objects and their
interrelations - Stephen E. Palmer, Vision Science, 1999
September, 2001
Pomerantz, BISCA, Bolzano, Italy
I -6
7What Is Perceptual Organization?
- Perceptual Organization is central to the key
question of perception how do we make the leap
from information detected by our sensory
receptors to our perceptions of the world?
This requires not just the detection of
information but the organization of that
information into veridical percepts. - Pomerantz Kubovy, 1986
September, 2001
Pomerantz, BISCA, Bolzano, Italy
I -7
8What Is Perceptual Organization?
- Perceptual organization is the process by which
particular relationships among potentially
separate elements (including parts, features, and
dimensions) are perceived (selected from
alternative relationships) and guide the
interpretation of those elements in sum, how we
process sensory information in context. - Pomerantz Kubovy, 1986
September, 2001
Pomerantz, BISCA, Bolzano, Italy
I -8
9Why Study Perceptual Organization?
- It is arguably among the earliest steps in
perception
- It is an essential tep, solving basic questions
that must be resolved before further image
analysis takes place
September, 2001
Pomerantz, BISCA, Bolzano, Italy
I -9
10Note The Problem of PO Extends Beyond Images of
Animals!
September, 2001
Pomerantz, BISCA, Bolzano, Italy
I -10
11There Are Many, Many Others
September, 2001
Pomerantz, BISCA, Bolzano, Italy
I -11
12More Classic Examples of Perceptual Organization
September, 2001
Pomerantz, BISCA, Bolzano, Italy
I -12
13More Classic Examples of Perceptual Organization
September, 2001
Pomerantz, BISCA, Bolzano, Italy
I -13
14More Classic Examples of Perceptual Organization
September, 2001
Pomerantz, BISCA, Bolzano, Italy
I -14
15More Classic Examples of Perceptual Organization
September, 2001
Pomerantz, BISCA, Bolzano, Italy
I -15
16More Classic Examples of Perceptual Organization
September, 2001
Pomerantz, BISCA, Bolzano, Italy
I -16
17More Classic Examples of Perceptual Organization
September, 2001
Pomerantz, BISCA, Bolzano, Italy
I -17
18Component Issues in Perceptual Organization
- Grouping and segmentation Dalmatian, Laws of
Grouping, Part Whole relations
- Figure ground segregation Rubins Vase
Doolittles ponies
- Emergent Features Subjective Contours,
Configural Superiority Effects
- Perceptual coupling (constancies) Shepards
Boxtops, Ames Room
- Multistability Necker cube Barber Pole
- Globality, Simplicity
September, 2001
Pomerantz, BISCA, Bolzano, Italy
I -18
19Grouping and Segmentation
From Wertheimer, 1923
September, 2001
Pomerantz, BISCA, Bolzano, Italy
I -19
20Lower illustration from Koffka
September, 2001
Pomerantz, BISCA, Bolzano, Italy
I -20
21September, 2001
Pomerantz, BISCA, Bolzano, Italy
I -21
22September, 2001
Pomerantz, BISCA, Bolzano, Italy
I -22
23September, 2001
Pomerantz, BISCA, Bolzano, Italy
I -23
24Grouping of Separated Elements
We can perceive objects well despite interruptions
September, 2001
Pomerantz, BISCA, Bolzano, Italy
I -24
25Segmentation Edges
September, 2001
Pomerantz, BISCA, Bolzano, Italy
I -25
26Figure Ground Segregation
From Rubin
September, 2001
Pomerantz, BISCA, Bolzano, Italy
I -26
27Rubins Figure Brought to Life
September, 2001
Pomerantz, BISCA, Bolzano, Italy
I -27
28September, 2001
Pomerantz, BISCA, Bolzano, Italy
I -28
29From Kanizsa
September, 2001
Pomerantz, BISCA, Bolzano, Italy
I -29
30From Escher
September, 2001
Pomerantz, BISCA, Bolzano, Italy
I -30
31September, 2001
Pomerantz, BISCA, Bolzano, Italy
I -31
32From Bregman
September, 2001
Pomerantz, BISCA, Bolzano, Italy
I -32
33From Rock
September, 2001
Pomerantz, BISCA, Bolzano, Italy
I -33
34Emergent Features
From Wertheimer, 1912 Phi (or beta) apparent
motion, plus the Correspondence problem from
Grouping
September, 2001
Pomerantz, BISCA, Bolzano, Italy
I -34
35From Kanizsa
September, 2001
Pomerantz, BISCA, Bolzano, Italy
I -35
36From Coren,Ward, and Enns, Sensation Perception.
September, 2001
Pomerantz, BISCA, Bolzano, Italy
I -36
37Two Veteran Emergent FeaturesRevealed through
Configural Superiority Effects
September, 2001
Pomerantz, BISCA, Bolzano, Italy
I -37
Slide 37
38Important ControlConfigural Inferiority Effects
Pomerantz, Sager, Stoever, 1977
September, 2001
Pomerantz, BISCA, Bolzano, Italy
I -38
Slide 38
39Virtual lines and edges
September, 2001
Pomerantz, BISCA, Bolzano, Italy
I -39
40Perceptual Coupling
From Shepard, 1981 Two two yellow parallelograms
have identical shapes.
September, 2001
Pomerantz, BISCA, Bolzano, Italy
I -40
41The Ames Room
Note Kubovy and others dont always regard coupl
ing
as a Gestalt problem
September, 2001
Pomerantz, BISCA, Bolzano, Italy
I -41
42Multistability
From Necker, Kopferman
September, 2001
Pomerantz, BISCA, Bolzano, Italy
I -42
43From Ternus Phenomenal Identity,
or, the Matching Unit problem
September, 2001
Pomerantz, BISCA, Bolzano, Italy
I -43
44More Ternus
Short ISI (0 msec)
Long ISI (200 msec?)
September, 2001
Pomerantz, BISCA, Bolzano, Italy
I -44
45From Boring
September, 2001
Pomerantz, BISCA, Bolzano, Italy
I -45
46From Wallach Wallachs overlooked variation po
lka dots
September, 2001
Pomerantz, BISCA, Bolzano, Italy
I -46
47Pointing Triangles In which direction do they
point? cf. control buttons on a cd player, etc
.
September, 2001
Pomerantz, BISCA, Bolzano, Italy
I -47
48From Attneave
September, 2001
Pomerantz, BISCA, Bolzano, Italy
I -48
49From Palmer
September, 2001
Pomerantz, BISCA, Bolzano, Italy
I -49
50Globality
- Configural effects work over large expanses of
the visual field, not just local patches
- Eg color. We achieve color constancies by
comparing wavelength distributions across the
entire visual field.
- E.g. the aperture problem in motion.
September, 2001
Pomerantz, BISCA, Bolzano, Italy
I -50
51Simplicity Prägnanz (the minimum principle)
- At the heart of the Gestalt approach.
- Claim We organize our percepts in the simplest
way that is consistent with the information in
the stimulus.
- Cf. distribution of electromagnetic fields.
September, 2001
Pomerantz, BISCA, Bolzano, Italy
I -51
52Soap Bubble Metaphor
The soap bubble computes an answer to a complex
problem, finding the simplest solution possible.
Question Does the human perceptual system work
in a similar fashion?
September, 2001
Pomerantz, BISCA, Bolzano, Italy
I -52
53PO and the Gestalt Psychologists
- Max Wertheimer
- Kurt Koffka
- Wolfgang Kohler
- They identified the basic problem, in some
respects in its current terms, and uncovered many
of its phenomena.
- They attempted a theory as well, but it has has
less impact.
September, 2001
Pomerantz, BISCA, Bolzano, Italy
I -53
54The Key to Gestalt EffectsNon-Additivity
- Stimulus A activates Representation A
- Stimulus B activates Representation B
- Q Will Stimuli A B activate
- Representations A B
- Something additional beyond this
- Something less than this
- Something simply different from this?
September, 2001
Pomerantz, BISCA, Bolzano, Italy
I -54
55Non-Additivities in Perception
- Widely Heralded Slogan of Gestalt Psychology
- The whole is greater than the sum of its parts?
No.
- No, not the sum
- Summing is a meaningless procedure
- Koffka,
1935
September, 2001
Pomerantz, BISCA, Bolzano, Italy
I -55
56Non-Additivities in Perception
- Rather,
- The whole is different from the sum of its
parts
- Sometime greater than, sometimes less than, often
different from
- Better to say the Gestalt claim was that elements
interact non-linearly in perception.
September, 2001
Pomerantz, BISCA, Bolzano, Italy
I -56
57Examples of Non-Additivities
- All involve the emergence of new features
- Color
- Apparent Motion
- Orientation
- Subjective Contours
September, 2001
Pomerantz, BISCA, Bolzano, Italy
I -57
58Orientation Glass Patterns
September, 2001
Pomerantz, BISCA, Bolzano, Italy
I -58
59Orientation Glass Patterns
September, 2001
Pomerantz, BISCA, Bolzano, Italy
I -59
60Orientation Glass Patterns
September, 2001
Pomerantz, BISCA, Bolzano, Italy
I -60
61Apparent Motion
September, 2001
Pomerantz, BISCA, Bolzano, Italy
I -61
62Apparent Motion
September, 2001
Pomerantz, BISCA, Bolzano, Italy
I -62
63Apparent Motion
Distinguish pure Phi motion from beta motion
September, 2001
Pomerantz, BISCA, Bolzano, Italy
I -63
64More Apparent Motion
September, 2001
Pomerantz, BISCA, Bolzano, Italy
I -64
65Apparent Motion of Holes
September, 2001
Pomerantz, BISCA, Bolzano, Italy
I -65
66Dunckers Rolling Wheel
September, 2001
Pomerantz, BISCA, Bolzano, Italy
I -66
67Dunckers Rolling Wheel
Rolling wheel demo (download requires Flash plu
g-in)
September, 2001
Pomerantz, BISCA, Bolzano, Italy
I -67
68Kanizsas Subjective Contours
September, 2001
Pomerantz, BISCA, Bolzano, Italy
I -68
69Shadow Fonts
September, 2001
Pomerantz, BISCA, Bolzano, Italy
I -69
Slide 69
70Combine Subjective Contours with Apparent Motion
September, 2001
Pomerantz, BISCA, Bolzano, Italy
I -70
71Combine Subjective Contours with Apparent Motion
Reversed direction of apparent
motion in the bottom panel compared with the top
indicates non-additivitity and suggests the domi
nance of subjective contours in deciding what
is moving. Bottom panel from Ramachandran
September, 2001
Pomerantz, BISCA, Bolzano, Italy
I -71
72This time, use dots
September, 2001
Pomerantz, BISCA, Bolzano, Italy
I -72
73This time, use dots
Even here with stimuli formed from discrete eleme
nts, it is possible to see reversed direction of
apparent motion in bottom panel compared with
top.
September, 2001
Pomerantz, BISCA, Bolzano, Italy
I -73
74Blocks forming continuous line
This display also shows reversed motion in the b
ottom panel, indicating that subjective contours
are not critical for producing this
non-additivity.
September, 2001
Pomerantz, BISCA, Bolzano, Italy
I -74
75Gestalt Psychology in 2001?
- Perceptual organization has languished in
relative obscurity since the end of the Gestalt
era In the latter half of the 1900s, vision
scientists have generally ignored organizational
issues, as though such things as grouping,
part-whole relations, reference frames did not
really matter, secure in their belief that linear
systems analysis and single cell recording
studies of cortical area V1 would lay bare the
mysteries of perception The next frontier of
vision science will be to solve the problems of
perceptual organization and its effects on visual
processing.
Stephen Palmer, in press
September, 2001
Pomerantz, BISCA, Bolzano, Italy
I -75
76On the Other Hand
Some of the researchers publishing on PO since
1950 Hochberg, Rock, Kanizsa, Treisman, Attneave
, Garner, Spelke, Palmer, Kubovy, Gerbino,
Shepard, Biederman, Metzger, Leeuwenberg,
Metelli, Graham, Julesz, Kahneman, Miller,
Neisser, Shepard, Wolfe. Reads like a Whos Who
in Psychology.
September, 2001
Pomerantz, BISCA, Bolzano, Italy
I -76
77Why is PO Such a Hard Problem?
- Perceptual organization is difficult to
study because it lies on the border between our
experience of the world and unconscious
perceptual processing. Even the term perceptual
organization is ambiguous it means both the
outcome of perceptual processeshow things
lookand the mechanism that produces itthe
psychophysical processes that precede awareness.
Perceptual organization is difficult to study for
a second reason because it Involves both
bottom-up and top-down processes. It islike
respirationa semi-voluntary process Michael
Kubovy, in press.
September, 2001
Pomerantz, BISCA, Bolzano, Italy
I -77
78Grappling with Basic Concepts
- Example What is an object?
object is not an easy term to define. Indeed,
textbooks with chapters on object perception
generally just assume that we all know what is
being talked about.
Jeremy Wolfe
September, 2001
Pomerantz, BISCA, Bolzano, Italy
I -78
79Barriers to Understanding Perception
- The problem of subjective experience
- Complexity of the stimulus, confounds
- Transparency of perception The main barrier may
be our own perceptual system!
September, 2001
Pomerantz, BISCA, Bolzano, Italy
I -79
80The Proximal Stimulus
What object does this figure depict?
How many objects are shown? Where are the object
boundaries?
September, 2001
Pomerantz, BISCA, Bolzano, Italy
I -80
81A Simpler Example
September, 2001
Pomerantz, BISCA, Bolzano, Italy
I -81
82The Answer Revealed
September, 2001
Pomerantz, BISCA, Bolzano, Italy
I -82
83Another example
September, 2001
Pomerantz, BISCA, Bolzano, Italy
I -83
84September, 2001
Pomerantz, BISCA, Bolzano, Italy
I -84
Slide 84
85Organizing a Stimulus Is Work!
- Looking at the last figure, it seems obvious that
there are figures there, and clear how many there
are and in what arrangement
- None of this is given for free, however it must
be computed by the visual system.
- The belief that the organization is in the
stimulus and not computed is a fundamental error,
one of two the experience error and the stimulus
error.
September, 2001
Pomerantz, BISCA, Bolzano, Italy
I -85
86Paradox of PO
- Everybody knows what it is, yet nobody seems to
know what it is
- Its effects are robust and seemingly obvious
- Yet difficult to measure Palmer, Kubovy
- Complicating matters two errors we make.
September, 2001
Pomerantz, BISCA, Bolzano, Italy
I -86
87The Stimulus and Experience Errors
- In psychology we have often been warned
against the stimulus error, i.e., against the
danger of confusing our knowledge of the physical
conditions of sensory experience with this
experience as such. As I see it, another
mistake, which I propose to call the experience
error, is just as unfortunate. This error occurs
when certain characteristics of sensory
experience are inadvertently attributed to the
mosaic of stimuli (Kohler 1929/47, p. 95).
September, 2001
Pomerantz, BISCA, Bolzano, Italy
I -87
88The Stimulus Error
- A presumption of what the stimulus is.
- We know that a physical object is built out of
certain parts, so when we describe our
perception, we use those same parts.
- Or, we enter a room lit with one candle,
illuminate a second candle, and report that the
room is now twice as bright.
- Striking counterexample the Gelb Effect, where
knowledge has little effect on perception.
September, 2001
Pomerantz, BISCA, Bolzano, Italy
I -88
89The Experience Error
- A presumption that our perception is governed by
the stimulus array.
- We experience a percept that is organized, e.g.,
is segmented into regions, objects, and parts.
So we assume that this organization is available
in the proximal (retinal) image, rather than
having to be computed on the image by our visual
system.
September, 2001
Pomerantz, BISCA, Bolzano, Italy
I -89
90An Example of These Errors?
- This stimulus is described as a box with a line
drawn across it, both containing a gap. Do we
see it that way? If so, is it because we
constructed it that way? And are these the real
physical parts? - (From Duncan, 1984)
September, 2001
Pomerantz, BISCA, Bolzano, Italy
I -90
91An Example of These Errors?
- Recall that this is a better description of the
proximal stimulus? Are the parts directly
represented here? Does this array support the
notion of a box and a line with gaps?
September, 2001
Pomerantz, BISCA, Bolzano, Italy
I -91
92Another Example?
- Steve Palmer, Element Connectedness
September, 2001
Pomerantz, BISCA, Bolzano, Italy
I -92
93The Link Between Perceptual Organization
Representation?
- The key issues are
- What in the stimulus is represented?
- How to represent elements in and out of context?
- PO defines the fundamental units and establishes
the hierarchy in which they are later organized.
The representation of a stimulus defines its
organization.
September, 2001
Pomerantz, BISCA, Bolzano, Italy
I -93
94Example Multistability
- Any one stimulus can have multiple
representations
- Each of these may respond to a different
organization
- Thus, representation and organization issues are
fundamentally intertwined.
September, 2001
Pomerantz, BISCA, Bolzano, Italy
I -94
95Roadmap to the Phenomena, Methods, and Findings
of PO
- Phenomena Grouping, figure-ground segregation,
multistability, constancies
- Methods demonstration vs. behavioral measures
- Findings Many examples, but based organizational
phenomena appear to govern how the rest of
perception functions.
September, 2001
Pomerantz, BISCA, Bolzano, Italy
I -95
96Theories of Perceptual Organization
- Lists of Laws
- General summarizing principles
- Full Blown Theories (Marrs categories)
- Computational
- Cognitive Neuropsychological
September, 2001
Pomerantz, BISCA, Bolzano, Italy
I -96
97Laws of Grouping
There are many, many
September, 2001
Pomerantz, BISCA, Bolzano, Italy
I -97
98Helson, 1933
- Listed 114 Gestalt Laws!
- Others estimate that over 700 have been proposed
over the decades.
- Boring, 1942 narrowed them down to 14.
September, 2001
Pomerantz, BISCA, Bolzano, Italy
I -98
99Borings 14
- Naturalness of form
- Figure and ground
- Articulation
- Good and poor forms
- Strong and weak forms
- Open and closed forms
- Dynamic basis of form
September, 2001
Pomerantz, BISCA, Bolzano, Italy
I -99
100Borings 14, continued
- Persistence of form
- Constancy of form
- Symmetry of form
- Integration of similars and adjacents
- Meaningfulness of forms
- Fusion of forms
- Transposition of forms.
September, 2001
Pomerantz, BISCA, Bolzano, Italy
I -100
101Borings 14, continued
- Note that Borings list of 14 omits some
important principles, including
- Good continuation
- Area
- Convexity
- Common fate
- One-sided function of contour
September, 2001
Pomerantz, BISCA, Bolzano, Italy
I -101
102Vagueness of These Laws
Number 5 Strong Weak Forms A strong form cohe
res and resists disintegration by analysis into
parts of by fusion with another form.
Many of these laws are a bit vague
September, 2001
Pomerantz, BISCA, Bolzano, Italy
I -102
103General Principle of Grouping
- Similarity
- Proximity similar location
- Common fate similar motion
- Similarity is too broad a term to be useful?
September, 2001
Pomerantz, BISCA, Bolzano, Italy
I -103
104General Principle of Grouping
Prägnanz (global minimum principle) return of
the soap bubbles Dynamic self-distribution if t
he kind of function which Gestalt Psychology
believes to be essential in neurological and
psychological theory.
Köhler, 1929
September, 2001
Pomerantz, BISCA, Bolzano, Italy
I -104
105Definition of Prägnanz
Psychological organization will always be as
good as the prevailing conditions allow. In
this definition the term good is undefined. It
embraces such properties as regularity, symmetry,
simplicity and others Koffka, 1935, building o
n Wertheimer
September, 2001
Pomerantz, BISCA, Bolzano, Italy
I -105
106Issues Entailed in Prägnanz
- Nature of simplification
- How much distortion to allow (illusions, etc.) at
the expense of veridicality?
- Simplify the process of perception or the
outcome?
- Quantification How to Measure Simplicity?
- Work of Attneave, Garner, Leeuwenberg, Mumford
- How to test the claim?
- Do our perceptions minimize complexity?
- Is this plausible given evolution?
September, 2001
Pomerantz, BISCA, Bolzano, Italy
I -106
107Alternative to Prägnanz Likelihood Principle
- From Helmholtz, 1910
- Definition Sensory elements will be organized
into the most probable objects or event (distal
stimulus) in the environment consistent with the
sensory data (the proximal stimulus)
September, 2001
Pomerantz, BISCA, Bolzano, Italy
I -107
108Major Advocates of Likelihood Principle
- Helmholtz
- Hebb
- Hochberg
- Gregory
- Brunswik
- Rock
September, 2001
Pomerantz, BISCA, Bolzano, Italy
I -108
109Key Issues for Likelihood Principle
- Major Idea We organize our percepts in the way
that is most likely to be correct
- Evolutionarily plausible
- Question how to determine whats most likely
(Brunwick, Geisler, Palmer)?
- How to test?
September, 2001
Pomerantz, BISCA, Bolzano, Italy
I -109
110Challenges to Prägnanz
- Kanizsa many devastating counterexamples
September, 2001
Pomerantz, BISCA, Bolzano, Italy
I -110
111Challenges to Prägnanz
- Attneave, Rock the search for symmetry
- If symmetry is so important, why is our search
for it so brief and unsuccessful?
September, 2001
Pomerantz, BISCA, Bolzano, Italy
I -111
112Challenges to Likelihood
September, 2001
Pomerantz, BISCA, Bolzano, Italy
I -112
113Prägnanz or Likelihood?
The evidence favoring the prägnanz principle is
somewhat thin the Gestalt explanations of
Gestalt phenomena are often inadequate, vague, or
simply wrong. The law of symmetry, which can be
regarded as the keystone of prägnanz , lacks the
kind of clean and robust demonstration one should
expect of so preeminent a principle The
Helmholtzean approach, with a few exceptions, has
provided a more promising framework. Most
Gestalt laws can be regarded as indicators of
what distal stimulus is most likely to have given
rise to the proximal stimulus.
Pomerantz Kubovy, 1986
September, 2001
Pomerantz, BISCA, Bolzano, Italy
I -113
114Could a Prägnanz Bias Be Useful?
- Attneave, 1982 Likelihood Principle can
accommodate Prägnanz
- Simplicity is a diagnostic property of stimuli
- If a stimulus can be organized simply, that is
probably the correct organization
- (Cf. Symmetry, parallelism as a non-accidental
property).
September, 2001
Pomerantz, BISCA, Bolzano, Italy
I -114
115Attneaves Point
- A possible distal source that contains certain
regularities is more probable than one that does
not, or one that contains them to a lesser
degree. - His idea was subsequently echoed by Rock,
Pomerantz Kubovy, and Palmer
- Palmer Simplicity acts as a surrogate for
likelihood, since simple organizations are likely
to be correct.
September, 2001
Pomerantz, BISCA, Bolzano, Italy
I -115
116Mach, 1906
- The visual sense acts therefore in conformity
with the principle of economy, and, at the same
time, in conformity with the principle of
probability.
September, 2001
Pomerantz, BISCA, Bolzano, Italy
I -116
117Where the Field Is Headed
- Away from lists of principles
- Toward an analysis of specific mechanism, studied
both behaviorally and with techniques from
cognitive neuroscience
- Toward general models integrating processes
across levels (Biederman, Palmer)
September, 2001
Pomerantz, BISCA, Bolzano, Italy
I -117