Upload
vohanh
View
216
Download
2
Embed Size (px)
Citation preview
Phenomenological Psychophysics & GroupingCurved vs. Collinear Organizations
Feldman’s “Eureka!” to the Rescue?
Proximity, Curvature, and Feldman’s Eureka!
Michael Kubovy and Lars Strother
Sunday, November 10, 2004, 9:40–9:5545th Annual Meeting of the Psychonomic Society
Hyatt Regency, Minneapolis
Kubovy & Strother Proximity, Curvature, & Eureka!
Phenomenological Psychophysics & GroupingCurved vs. Collinear Organizations
Feldman’s “Eureka!” to the Rescue?
Outline
1 Phenomenological Psychophysics & the Strength of GroupingGrouping by proximityGrouping by similarity
2 Curved vs. Collinear OrganizationsStandard early-vision resultsCurves are more complexDot-sampled grids
3 Feldman’s “Eureka!” to the Rescue?
Kubovy & Strother Proximity, Curvature, & Eureka!
Phenomenological Psychophysics & GroupingCurved vs. Collinear Organizations
Feldman’s “Eureka!” to the Rescue?
Outline
1 Phenomenological Psychophysics & the Strength of GroupingGrouping by proximityGrouping by similarity
2 Curved vs. Collinear OrganizationsStandard early-vision resultsCurves are more complexDot-sampled grids
3 Feldman’s “Eureka!” to the Rescue?
Kubovy & Strother Proximity, Curvature, & Eureka!
Phenomenological Psychophysics & GroupingCurved vs. Collinear Organizations
Feldman’s “Eureka!” to the Rescue?
Outline
1 Phenomenological Psychophysics & the Strength of GroupingGrouping by proximityGrouping by similarity
2 Curved vs. Collinear OrganizationsStandard early-vision resultsCurves are more complexDot-sampled grids
3 Feldman’s “Eureka!” to the Rescue?
Kubovy & Strother Proximity, Curvature, & Eureka!
Phenomenological Psychophysics & GroupingCurved vs. Collinear Organizations
Feldman’s “Eureka!” to the Rescue?
Two take-home questions
1 Particular: Are collinear organizations more salient thancurvilinear ones?
2 General: Can middle-level vision be reduced to early vision?
Kubovy & Strother Proximity, Curvature, & Eureka!
Phenomenological Psychophysics & GroupingCurved vs. Collinear Organizations
Feldman’s “Eureka!” to the Rescue?
Two take-home questions
1 Particular: Are collinear organizations more salient thancurvilinear ones?
2 General: Can middle-level vision be reduced to early vision?
Kubovy & Strother Proximity, Curvature, & Eureka!
Phenomenological Psychophysics & GroupingCurved vs. Collinear Organizations
Feldman’s “Eureka!” to the Rescue?
The difference between early vs. middle-level vision
Middle-level vision is a.k.a. perceptual organization
Like respiration, it is bottom-up most of time but top-downsome of the timeOne goal of research in early vision: to account for perceptualorganizationDifferences in method
Early: studies responses at threshold (by and large)Middle-level: studies responses to easily-perceived stimuli
Kubovy & Strother Proximity, Curvature, & Eureka!
Phenomenological Psychophysics & GroupingCurved vs. Collinear Organizations
Feldman’s “Eureka!” to the Rescue?
The difference between early vs. middle-level vision
Middle-level vision is a.k.a. perceptual organizationLike respiration, it is bottom-up most of time but top-downsome of the time
One goal of research in early vision: to account for perceptualorganizationDifferences in method
Early: studies responses at threshold (by and large)Middle-level: studies responses to easily-perceived stimuli
Kubovy & Strother Proximity, Curvature, & Eureka!
Phenomenological Psychophysics & GroupingCurved vs. Collinear Organizations
Feldman’s “Eureka!” to the Rescue?
The difference between early vs. middle-level vision
Middle-level vision is a.k.a. perceptual organizationLike respiration, it is bottom-up most of time but top-downsome of the timeOne goal of research in early vision: to account for perceptualorganization
Differences in method
Early: studies responses at threshold (by and large)Middle-level: studies responses to easily-perceived stimuli
Kubovy & Strother Proximity, Curvature, & Eureka!
Phenomenological Psychophysics & GroupingCurved vs. Collinear Organizations
Feldman’s “Eureka!” to the Rescue?
The difference between early vs. middle-level vision
Middle-level vision is a.k.a. perceptual organizationLike respiration, it is bottom-up most of time but top-downsome of the timeOne goal of research in early vision: to account for perceptualorganizationDifferences in method
Early: studies responses at threshold (by and large)Middle-level: studies responses to easily-perceived stimuli
Kubovy & Strother Proximity, Curvature, & Eureka!
Phenomenological Psychophysics & GroupingCurved vs. Collinear Organizations
Feldman’s “Eureka!” to the Rescue?
The difference between early vs. middle-level vision
Middle-level vision is a.k.a. perceptual organizationLike respiration, it is bottom-up most of time but top-downsome of the timeOne goal of research in early vision: to account for perceptualorganizationDifferences in method
Early: studies responses at threshold (by and large)
Middle-level: studies responses to easily-perceived stimuli
Kubovy & Strother Proximity, Curvature, & Eureka!
Phenomenological Psychophysics & GroupingCurved vs. Collinear Organizations
Feldman’s “Eureka!” to the Rescue?
The difference between early vs. middle-level vision
Middle-level vision is a.k.a. perceptual organizationLike respiration, it is bottom-up most of time but top-downsome of the timeOne goal of research in early vision: to account for perceptualorganizationDifferences in method
Early: studies responses at threshold (by and large)Middle-level: studies responses to easily-perceived stimuli
Kubovy & Strother Proximity, Curvature, & Eureka!
Phenomenological Psychophysics & GroupingCurved vs. Collinear Organizations
Feldman’s “Eureka!” to the Rescue?
Grouping by proximityGrouping by similarity
How we study middle-level vision
Use ambiguous stimuli: dot lattices
Near equilibriumDisplace from equilibriumRecord how the stimulus is perceivedNo correct responsePhenomenological psychophysics
Kubovy & Strother Proximity, Curvature, & Eureka!
Phenomenological Psychophysics & GroupingCurved vs. Collinear Organizations
Feldman’s “Eureka!” to the Rescue?
Grouping by proximityGrouping by similarity
How we study middle-level vision
Use ambiguous stimuli: dot latticesNear equilibrium
Displace from equilibriumRecord how the stimulus is perceivedNo correct responsePhenomenological psychophysics
Kubovy & Strother Proximity, Curvature, & Eureka!
Phenomenological Psychophysics & GroupingCurved vs. Collinear Organizations
Feldman’s “Eureka!” to the Rescue?
Grouping by proximityGrouping by similarity
How we study middle-level vision
Use ambiguous stimuli: dot latticesNear equilibriumDisplace from equilibrium
Record how the stimulus is perceivedNo correct responsePhenomenological psychophysics
Kubovy & Strother Proximity, Curvature, & Eureka!
Phenomenological Psychophysics & GroupingCurved vs. Collinear Organizations
Feldman’s “Eureka!” to the Rescue?
Grouping by proximityGrouping by similarity
How we study middle-level vision
Use ambiguous stimuli: dot latticesNear equilibriumDisplace from equilibriumRecord how the stimulus is perceived
No correct responsePhenomenological psychophysics
Kubovy & Strother Proximity, Curvature, & Eureka!
Phenomenological Psychophysics & GroupingCurved vs. Collinear Organizations
Feldman’s “Eureka!” to the Rescue?
Grouping by proximityGrouping by similarity
How we study middle-level vision
Use ambiguous stimuli: dot latticesNear equilibriumDisplace from equilibriumRecord how the stimulus is perceivedNo correct response
Phenomenological psychophysics
Kubovy & Strother Proximity, Curvature, & Eureka!
Phenomenological Psychophysics & GroupingCurved vs. Collinear Organizations
Feldman’s “Eureka!” to the Rescue?
Grouping by proximityGrouping by similarity
How we study middle-level vision
Use ambiguous stimuli: dot latticesNear equilibriumDisplace from equilibriumRecord how the stimulus is perceivedNo correct responsePhenomenological psychophysics
Kubovy & Strother Proximity, Curvature, & Eureka!
Phenomenological Psychophysics & GroupingCurved vs. Collinear Organizations
Feldman’s “Eureka!” to the Rescue?
Grouping by proximityGrouping by similarity
Dot Lattices
Hexagonal Lattice: three-way equilibrium
Kubovy & Strother Proximity, Curvature, & Eureka!
Phenomenological Psychophysics & GroupingCurved vs. Collinear Organizations
Feldman’s “Eureka!” to the Rescue?
Grouping by proximityGrouping by similarity
Dot Lattices
Hexagonal Lattice: three-way equilibrium
Kubovy & Strother Proximity, Curvature, & Eureka!
Phenomenological Psychophysics & GroupingCurved vs. Collinear Organizations
Feldman’s “Eureka!” to the Rescue?
Grouping by proximityGrouping by similarity
Dot Lattices
Square Lattice: two-way equilibrium
Kubovy & Strother Proximity, Curvature, & Eureka!
Phenomenological Psychophysics & GroupingCurved vs. Collinear Organizations
Feldman’s “Eureka!” to the Rescue?
Grouping by proximityGrouping by similarity
Dot Lattices
Square Lattice: two-way equilibrium
Kubovy & Strother Proximity, Curvature, & Eureka!
Phenomenological Psychophysics & GroupingCurved vs. Collinear Organizations
Feldman’s “Eureka!” to the Rescue?
Grouping by proximityGrouping by similarity
Dot Lattices
Rectangular Lattice: ambiguous, but not at equilibrium
Kubovy & Strother Proximity, Curvature, & Eureka!
Phenomenological Psychophysics & GroupingCurved vs. Collinear Organizations
Feldman’s “Eureka!” to the Rescue?
Grouping by proximityGrouping by similarity
Dot Lattices
Rectangular Lattice: ambiguous, but not at equilibrium
Kubovy & Strother Proximity, Curvature, & Eureka!
Phenomenological Psychophysics & GroupingCurved vs. Collinear Organizations
Feldman’s “Eureka!” to the Rescue?
Grouping by proximityGrouping by similarity
Dot Lattices
They vary continuously
Kubovy & Strother Proximity, Curvature, & Eureka!
Phenomenological Psychophysics & GroupingCurved vs. Collinear Organizations
Feldman’s “Eureka!” to the Rescue?
Grouping by proximityGrouping by similarity
Dot Lattices
The dots are at the intersections of a regular grid
Kubovy & Strother Proximity, Curvature, & Eureka!
Phenomenological Psychophysics & GroupingCurved vs. Collinear Organizations
Feldman’s “Eureka!” to the Rescue?
Grouping by proximityGrouping by similarity
Dot Lattices
Unit: the basic parallelogram
a → short sideb → long sidec → short diagonald → long diagonal
Kubovy & Strother Proximity, Curvature, & Eureka!
Phenomenological Psychophysics & GroupingCurved vs. Collinear Organizations
Feldman’s “Eureka!” to the Rescue?
Grouping by proximityGrouping by similarity
Dot Lattices
Unit: the basic parallelogram
a → short sideb → long sidec → short diagonald → long diagonal
Kubovy & Strother Proximity, Curvature, & Eureka!
Phenomenological Psychophysics & GroupingCurved vs. Collinear Organizations
Feldman’s “Eureka!” to the Rescue?
Grouping by proximityGrouping by similarity
Dot Lattices
Unit: the basic parallelogram
a
a → short side
b → long sidec → short diagonald → long diagonal
Kubovy & Strother Proximity, Curvature, & Eureka!
Phenomenological Psychophysics & GroupingCurved vs. Collinear Organizations
Feldman’s “Eureka!” to the Rescue?
Grouping by proximityGrouping by similarity
Dot Lattices
Unit: the basic parallelogram
a
b
a → short sideb → long side
c → short diagonald → long diagonal
Kubovy & Strother Proximity, Curvature, & Eureka!
Phenomenological Psychophysics & GroupingCurved vs. Collinear Organizations
Feldman’s “Eureka!” to the Rescue?
Grouping by proximityGrouping by similarity
Dot Lattices
Unit: the basic parallelogram
a
bc
a → short sideb → long sidec → short diagonal
d → long diagonal
Kubovy & Strother Proximity, Curvature, & Eureka!
Phenomenological Psychophysics & GroupingCurved vs. Collinear Organizations
Feldman’s “Eureka!” to the Rescue?
Grouping by proximityGrouping by similarity
Dot Lattices
Unit: the basic parallelogram
a
bc
d
a → short sideb → long sidec → short diagonald → long diagonal
Kubovy & Strother Proximity, Curvature, & Eureka!
Phenomenological Psychophysics & GroupingCurved vs. Collinear Organizations
Feldman’s “Eureka!” to the Rescue?
Grouping by proximityGrouping by similarity
What a trial might look like
Task: which organization did you see?
Four possible responses: a,b,c,d . No correct response
Kubovy & Strother Proximity, Curvature, & Eureka!
Phenomenological Psychophysics & GroupingCurved vs. Collinear Organizations
Feldman’s “Eureka!” to the Rescue?
Grouping by proximityGrouping by similarity
What a trial might look like
Task: which organization did you see?
Four possible responses: a,b,c,d .
No correct response
Kubovy & Strother Proximity, Curvature, & Eureka!
Phenomenological Psychophysics & GroupingCurved vs. Collinear Organizations
Feldman’s “Eureka!” to the Rescue?
Grouping by proximityGrouping by similarity
What a trial might look like
Task: which organization did you see?
Four possible responses: a,b,c,d . No correct response
Kubovy & Strother Proximity, Curvature, & Eureka!
Phenomenological Psychophysics & GroupingCurved vs. Collinear Organizations
Feldman’s “Eureka!” to the Rescue?
Grouping by proximityGrouping by similarity
The data
Variables
four IVs: the lengths |a|, |b|, |c|, |d|four DVs: the probabilities p(a), p(b), p(c), p(d)
p(a) + p(b) + p(c) + p(d) = 1 ⇒ three dfSo, new variables:
three IVs: the relative distances |b|/|a|, |c|/|a|, |d|/|a|three DVs: the odds p(b)/p(a), p(c)/p(a), p(d)/p(a)
We plot the log-oddslog[p(b)/p(a)], log[p(c)/p(a)], log[p(d)/p(a)]
Kubovy & Strother Proximity, Curvature, & Eureka!
Phenomenological Psychophysics & GroupingCurved vs. Collinear Organizations
Feldman’s “Eureka!” to the Rescue?
Grouping by proximityGrouping by similarity
The data
Variablesfour IVs: the lengths |a|, |b|, |c|, |d|
four DVs: the probabilities p(a), p(b), p(c), p(d)
p(a) + p(b) + p(c) + p(d) = 1 ⇒ three dfSo, new variables:
three IVs: the relative distances |b|/|a|, |c|/|a|, |d|/|a|three DVs: the odds p(b)/p(a), p(c)/p(a), p(d)/p(a)
We plot the log-oddslog[p(b)/p(a)], log[p(c)/p(a)], log[p(d)/p(a)]
Kubovy & Strother Proximity, Curvature, & Eureka!
Phenomenological Psychophysics & GroupingCurved vs. Collinear Organizations
Feldman’s “Eureka!” to the Rescue?
Grouping by proximityGrouping by similarity
The data
Variablesfour IVs: the lengths |a|, |b|, |c|, |d|four DVs: the probabilities p(a), p(b), p(c), p(d)
p(a) + p(b) + p(c) + p(d) = 1 ⇒ three dfSo, new variables:
three IVs: the relative distances |b|/|a|, |c|/|a|, |d|/|a|three DVs: the odds p(b)/p(a), p(c)/p(a), p(d)/p(a)
We plot the log-oddslog[p(b)/p(a)], log[p(c)/p(a)], log[p(d)/p(a)]
Kubovy & Strother Proximity, Curvature, & Eureka!
Phenomenological Psychophysics & GroupingCurved vs. Collinear Organizations
Feldman’s “Eureka!” to the Rescue?
Grouping by proximityGrouping by similarity
The data
Variablesfour IVs: the lengths |a|, |b|, |c|, |d|four DVs: the probabilities p(a), p(b), p(c), p(d)
p(a) + p(b) + p(c) + p(d) = 1 ⇒ three df
So, new variables:
three IVs: the relative distances |b|/|a|, |c|/|a|, |d|/|a|three DVs: the odds p(b)/p(a), p(c)/p(a), p(d)/p(a)
We plot the log-oddslog[p(b)/p(a)], log[p(c)/p(a)], log[p(d)/p(a)]
Kubovy & Strother Proximity, Curvature, & Eureka!
Phenomenological Psychophysics & GroupingCurved vs. Collinear Organizations
Feldman’s “Eureka!” to the Rescue?
Grouping by proximityGrouping by similarity
The data
Variablesfour IVs: the lengths |a|, |b|, |c|, |d|four DVs: the probabilities p(a), p(b), p(c), p(d)
p(a) + p(b) + p(c) + p(d) = 1 ⇒ three dfSo, new variables:
three IVs: the relative distances |b|/|a|, |c|/|a|, |d|/|a|three DVs: the odds p(b)/p(a), p(c)/p(a), p(d)/p(a)
We plot the log-oddslog[p(b)/p(a)], log[p(c)/p(a)], log[p(d)/p(a)]
Kubovy & Strother Proximity, Curvature, & Eureka!
Phenomenological Psychophysics & GroupingCurved vs. Collinear Organizations
Feldman’s “Eureka!” to the Rescue?
Grouping by proximityGrouping by similarity
The data
Variablesfour IVs: the lengths |a|, |b|, |c|, |d|four DVs: the probabilities p(a), p(b), p(c), p(d)
p(a) + p(b) + p(c) + p(d) = 1 ⇒ three dfSo, new variables:
three IVs: the relative distances |b|/|a|, |c|/|a|, |d|/|a|
three DVs: the odds p(b)/p(a), p(c)/p(a), p(d)/p(a)
We plot the log-oddslog[p(b)/p(a)], log[p(c)/p(a)], log[p(d)/p(a)]
Kubovy & Strother Proximity, Curvature, & Eureka!
Phenomenological Psychophysics & GroupingCurved vs. Collinear Organizations
Feldman’s “Eureka!” to the Rescue?
Grouping by proximityGrouping by similarity
The data
Variablesfour IVs: the lengths |a|, |b|, |c|, |d|four DVs: the probabilities p(a), p(b), p(c), p(d)
p(a) + p(b) + p(c) + p(d) = 1 ⇒ three dfSo, new variables:
three IVs: the relative distances |b|/|a|, |c|/|a|, |d|/|a|three DVs: the odds p(b)/p(a), p(c)/p(a), p(d)/p(a)
We plot the log-oddslog[p(b)/p(a)], log[p(c)/p(a)], log[p(d)/p(a)]
Kubovy & Strother Proximity, Curvature, & Eureka!
Phenomenological Psychophysics & GroupingCurved vs. Collinear Organizations
Feldman’s “Eureka!” to the Rescue?
Grouping by proximityGrouping by similarity
The data
Variablesfour IVs: the lengths |a|, |b|, |c|, |d|four DVs: the probabilities p(a), p(b), p(c), p(d)
p(a) + p(b) + p(c) + p(d) = 1 ⇒ three dfSo, new variables:
three IVs: the relative distances |b|/|a|, |c|/|a|, |d|/|a|three DVs: the odds p(b)/p(a), p(c)/p(a), p(d)/p(a)
We plot the log-oddslog[p(b)/p(a)], log[p(c)/p(a)], log[p(d)/p(a)]
Kubovy & Strother Proximity, Curvature, & Eureka!
Phenomenological Psychophysics & GroupingCurved vs. Collinear Organizations
Feldman’s “Eureka!” to the Rescue?
Grouping by proximityGrouping by similarity
The data: proximity
Kubovy, Holcombe, & Wagemans (1998)
basic parallelogram
a
b
|b||a|
1 1.1 1.2 1.3
log p(b)p(a)
0c
|c||a|
log p(c)p(a)
attraction function
Kubovy & Strother Proximity, Curvature, & Eureka!
Phenomenological Psychophysics & GroupingCurved vs. Collinear Organizations
Feldman’s “Eureka!” to the Rescue?
Grouping by proximityGrouping by similarity
The data: proximity
Kubovy, Holcombe, & Wagemans (1998)
basic parallelograma
b
|b||a|
1 1.1 1.2 1.3
log p(b)p(a)
0c
|c||a|
log p(c)p(a)
attraction function
Kubovy & Strother Proximity, Curvature, & Eureka!
Phenomenological Psychophysics & GroupingCurved vs. Collinear Organizations
Feldman’s “Eureka!” to the Rescue?
Grouping by proximityGrouping by similarity
The data: proximity
Kubovy, Holcombe, & Wagemans (1998)
basic parallelograma
b
|b||a|
1 1.1 1.2 1.3
log p(b)p(a)
0c
|c||a|
log p(c)p(a)
attraction function
Kubovy & Strother Proximity, Curvature, & Eureka!
Phenomenological Psychophysics & GroupingCurved vs. Collinear Organizations
Feldman’s “Eureka!” to the Rescue?
Grouping by proximityGrouping by similarity
The data: proximity
Kubovy, Holcombe, & Wagemans (1998)
basic parallelograma
b
|b||a|
1 1.1 1.2 1.3
log p(b)p(a)
0c
|c||a|
log p(c)p(a)
attraction function
Kubovy & Strother Proximity, Curvature, & Eureka!
Phenomenological Psychophysics & GroupingCurved vs. Collinear Organizations
Feldman’s “Eureka!” to the Rescue?
Grouping by proximityGrouping by similarity
The data: proximity
Kubovy, Holcombe, & Wagemans (1998)
basic parallelograma
b
|b||a|
1 1.1 1.2 1.3
log p(b)p(a)
0c
|c||a|
log p(c)p(a)
attraction function
Kubovy & Strother Proximity, Curvature, & Eureka!
Phenomenological Psychophysics & GroupingCurved vs. Collinear Organizations
Feldman’s “Eureka!” to the Rescue?
Grouping by proximityGrouping by similarity
The data: proximity
Kubovy, Holcombe, & Wagemans (1998)
basic parallelograma
b
|b||a|
1 1.1 1.2 1.3
log p(b)p(a)
0c
|c||a|
log p(c)p(a)
attraction function
Kubovy & Strother Proximity, Curvature, & Eureka!
Phenomenological Psychophysics & GroupingCurved vs. Collinear Organizations
Feldman’s “Eureka!” to the Rescue?
Grouping by proximityGrouping by similarity
The data: proximity
Kubovy, Holcombe, & Wagemans (1998)
basic parallelograma
b
|b||a|
1 1.1 1.2 1.3
log p(b)p(a)
0
c
|c||a|
log p(c)p(a)
attraction function
Kubovy & Strother Proximity, Curvature, & Eureka!
Phenomenological Psychophysics & GroupingCurved vs. Collinear Organizations
Feldman’s “Eureka!” to the Rescue?
Grouping by proximityGrouping by similarity
The data: proximity
Kubovy, Holcombe, & Wagemans (1998)
basic parallelograma
b
|b||a|
1 1.1 1.2 1.3
log p(b)p(a)
0
c
|c||a|
log p(c)p(a)
attraction function
Kubovy & Strother Proximity, Curvature, & Eureka!
Phenomenological Psychophysics & GroupingCurved vs. Collinear Organizations
Feldman’s “Eureka!” to the Rescue?
Grouping by proximityGrouping by similarity
The data: proximity
Kubovy, Holcombe, & Wagemans (1998)
basic parallelograma
b
|b||a|
1 1.1 1.2 1.3
log p(b)p(a)
0
c
|c||a|
log p(c)p(a)
attraction function
Kubovy & Strother Proximity, Curvature, & Eureka!
Phenomenological Psychophysics & GroupingCurved vs. Collinear Organizations
Feldman’s “Eureka!” to the Rescue?
Grouping by proximityGrouping by similarity
The data: proximity
Kubovy, Holcombe, & Wagemans (1998)
basic parallelograma
b
|b||a|
1 1.1 1.2 1.3
log p(b)p(a)
0
c
|c||a|
log p(c)p(a)
attraction function
Kubovy & Strother Proximity, Curvature, & Eureka!
Phenomenological Psychophysics & GroupingCurved vs. Collinear Organizations
Feldman’s “Eureka!” to the Rescue?
Grouping by proximityGrouping by similarity
The data: proximity
Kubovy, Holcombe, & Wagemans (1998)
basic parallelograma
b
|b||a|
1 1.1 1.2 1.3
log p(b)p(a)
0c
|c||a|
log p(c)p(a)
attraction function
Kubovy & Strother Proximity, Curvature, & Eureka!
Phenomenological Psychophysics & GroupingCurved vs. Collinear Organizations
Feldman’s “Eureka!” to the Rescue?
Grouping by proximityGrouping by similarity
The data: proximity
Kubovy, Holcombe, & Wagemans (1998)
basic parallelograma
b
|b||a|
1 1.1 1.2 1.3
log p(b)p(a)
0c
|c||a|
log p(c)p(a)
attraction function
Kubovy & Strother Proximity, Curvature, & Eureka!
Phenomenological Psychophysics & GroupingCurved vs. Collinear Organizations
Feldman’s “Eureka!” to the Rescue?
Grouping by proximityGrouping by similarity
The data: proximity
Kubovy, Holcombe, & Wagemans (1998)
basic parallelograma
b
|b||a|
1 1.1 1.2 1.3
log p(b)p(a)
0c
|c||a|
log p(c)p(a)
attraction function
Kubovy & Strother Proximity, Curvature, & Eureka!
Phenomenological Psychophysics & GroupingCurved vs. Collinear Organizations
Feldman’s “Eureka!” to the Rescue?
Grouping by proximityGrouping by similarity
The data: proximity
Kubovy, Holcombe, & Wagemans (1998)
basic parallelograma
b
|b||a|
1 1.1 1.2 1.3
log p(b)p(a)
0c
|c||a|
log p(c)p(a)
attraction function
Kubovy & Strother Proximity, Curvature, & Eureka!
Phenomenological Psychophysics & GroupingCurved vs. Collinear Organizations
Feldman’s “Eureka!” to the Rescue?
Grouping by proximityGrouping by similarity
The data: proximity
Kubovy, Holcombe, & Wagemans (1998)
basic parallelograma
b
|b||a|
1 1.1 1.2 1.3
log p(b)p(a)
0c
|c||a|
log p(c)p(a)
attraction function
Kubovy & Strother Proximity, Curvature, & Eureka!
Phenomenological Psychophysics & GroupingCurved vs. Collinear Organizations
Feldman’s “Eureka!” to the Rescue?
Grouping by proximityGrouping by similarity
The data: proximity
Kubovy, Holcombe, & Wagemans (1998)
basic parallelograma
b
|b||a|
1 1.1 1.2 1.3
log p(b)p(a)
0c
|c||a|
log p(c)p(a)
attraction function
Kubovy & Strother Proximity, Curvature, & Eureka!
Phenomenological Psychophysics & GroupingCurved vs. Collinear Organizations
Feldman’s “Eureka!” to the Rescue?
Grouping by proximityGrouping by similarity
The data: similarity
van den Berg & Kubovy (in prep.): Similarity along b
basic parallelograma
b
|b||a|1 1.1 1.2 1.3
log p(b)p(a)
0
Kubovy & Strother Proximity, Curvature, & Eureka!
Phenomenological Psychophysics & GroupingCurved vs. Collinear Organizations
Feldman’s “Eureka!” to the Rescue?
Grouping by proximityGrouping by similarity
The data: similarity
van den Berg & Kubovy (in prep.): Similarity along b
basic parallelograma
b
|b||a|1 1.1 1.2 1.3
log p(b)p(a)
0
Kubovy & Strother Proximity, Curvature, & Eureka!
Phenomenological Psychophysics & GroupingCurved vs. Collinear Organizations
Feldman’s “Eureka!” to the Rescue?
Grouping by proximityGrouping by similarity
The data: similarity
van den Berg & Kubovy (in prep.): Similarity along a
basic parallelograma
b
|b||a|1 1.1 1.2 1.3
log p(b)p(a)
0
Kubovy & Strother Proximity, Curvature, & Eureka!
Phenomenological Psychophysics & GroupingCurved vs. Collinear Organizations
Feldman’s “Eureka!” to the Rescue?
Grouping by proximityGrouping by similarity
The data: similarity
van den Berg & Kubovy (in prep.): Similarity along a
basic parallelograma
b
|b||a|1 1.1 1.2 1.3
log p(b)p(a)
0
Kubovy & Strother Proximity, Curvature, & Eureka!
Phenomenological Psychophysics & GroupingCurved vs. Collinear Organizations
Feldman’s “Eureka!” to the Rescue?
Standard early-vision resultsCurves are more complexDot-sampled grids
It’s harder to detect curves than straight lines
Pizlo, Salach-Golyska, & Rosenfeld (1997): stimulus
Kubovy & Strother Proximity, Curvature, & Eureka!
Phenomenological Psychophysics & GroupingCurved vs. Collinear Organizations
Feldman’s “Eureka!” to the Rescue?
Standard early-vision resultsCurves are more complexDot-sampled grids
It’s harder to detect curves than straight lines
Pizlo, Salach-Golyska, & Rosenfeld (1997): stimulus
Kubovy & Strother Proximity, Curvature, & Eureka!
Phenomenological Psychophysics & GroupingCurved vs. Collinear Organizations
Feldman’s “Eureka!” to the Rescue?
Standard early-vision resultsCurves are more complexDot-sampled grids
It’s harder to detect curves than straight lines
Pizlo, Salach-Golyska, & Rosenfeld (1997): data
4
3
2
1
0none low high
d’
curvature
Kubovy & Strother Proximity, Curvature, & Eureka!
Phenomenological Psychophysics & GroupingCurved vs. Collinear Organizations
Feldman’s “Eureka!” to the Rescue?
Standard early-vision resultsCurves are more complexDot-sampled grids
Feldman & Singh (2004), Psychological Review
Curve sampled every ∆s; α = ∆φ = angle between successivetangents φ
Kubovy & Strother Proximity, Curvature, & Eureka!
Phenomenological Psychophysics & GroupingCurved vs. Collinear Organizations
Feldman’s “Eureka!” to the Rescue?
Standard early-vision resultsCurves are more complexDot-sampled grids
Complexity of a curve
The expected change in tangent direction α is distributed as a vonMises distribution centered on 0 (straight).
Kubovy & Strother Proximity, Curvature, & Eureka!
Phenomenological Psychophysics & GroupingCurved vs. Collinear Organizations
Feldman’s “Eureka!” to the Rescue?
Standard early-vision resultsCurves are more complexDot-sampled grids
At a point where the angle is α, the surprisal is:
µ(α) = − log[p(α)]
If for p(α) we substitute the value given by the von Misesdistribution, we get:
µ(α) = − log A− b cos (α)
Kubovy & Strother Proximity, Curvature, & Eureka!
Phenomenological Psychophysics & GroupingCurved vs. Collinear Organizations
Feldman’s “Eureka!” to the Rescue?
Standard early-vision resultsCurves are more complexDot-sampled grids
Complexity of a curve
Contour information (surprisal) on the boundary of a shape
Kubovy & Strother Proximity, Curvature, & Eureka!
Phenomenological Psychophysics & GroupingCurved vs. Collinear Organizations
Feldman’s “Eureka!” to the Rescue?
Standard early-vision resultsCurves are more complexDot-sampled grids
Dot-sampled grids
The stimulus
Kubovy & Strother Proximity, Curvature, & Eureka!
Phenomenological Psychophysics & GroupingCurved vs. Collinear Organizations
Feldman’s “Eureka!” to the Rescue?
Standard early-vision resultsCurves are more complexDot-sampled grids
Dot-sampled grids
The grid
Kubovy & Strother Proximity, Curvature, & Eureka!
Phenomenological Psychophysics & GroupingCurved vs. Collinear Organizations
Feldman’s “Eureka!” to the Rescue?
Standard early-vision resultsCurves are more complexDot-sampled grids
Possible organizations
The grid
Kubovy & Strother Proximity, Curvature, & Eureka!
Phenomenological Psychophysics & GroupingCurved vs. Collinear Organizations
Feldman’s “Eureka!” to the Rescue?
Standard early-vision resultsCurves are more complexDot-sampled grids
The response screen
The response screen
Kubovy & Strother Proximity, Curvature, & Eureka!
Phenomenological Psychophysics & GroupingCurved vs. Collinear Organizations
Feldman’s “Eureka!” to the Rescue?
Standard early-vision resultsCurves are more complexDot-sampled grids
Prediction
Strother & Kubovy (in prep.)
nomenclature
σκ
|κ||σ|
1.2−1 1.1−1 1 1.1 1.2
log p(κ)p(σ)
0
attraction functionwe expect collinearityto group more strongly
Kubovy & Strother Proximity, Curvature, & Eureka!
Phenomenological Psychophysics & GroupingCurved vs. Collinear Organizations
Feldman’s “Eureka!” to the Rescue?
Standard early-vision resultsCurves are more complexDot-sampled grids
Prediction
Strother & Kubovy (in prep.)
nomenclature
σκ
|κ||σ|
1.2−1 1.1−1 1 1.1 1.2
log p(κ)p(σ)
0
attraction functionwe expect collinearityto group more strongly
Kubovy & Strother Proximity, Curvature, & Eureka!
Phenomenological Psychophysics & GroupingCurved vs. Collinear Organizations
Feldman’s “Eureka!” to the Rescue?
Standard early-vision resultsCurves are more complexDot-sampled grids
Prediction
Strother & Kubovy (in prep.)
nomenclature
σκ
|κ||σ|
1.2−1 1.1−1 1 1.1 1.2
log p(κ)p(σ)
0
attraction function
we expect collinearityto group more strongly
Kubovy & Strother Proximity, Curvature, & Eureka!
Phenomenological Psychophysics & GroupingCurved vs. Collinear Organizations
Feldman’s “Eureka!” to the Rescue?
Standard early-vision resultsCurves are more complexDot-sampled grids
Prediction
Strother & Kubovy (in prep.)
nomenclature
σκ
|κ||σ|
1.2−1 1.1−1 1 1.1 1.2
log p(κ)p(σ)
0
attraction functionwe expect collinearityto group more strongly
Kubovy & Strother Proximity, Curvature, & Eureka!
Phenomenological Psychophysics & GroupingCurved vs. Collinear Organizations
Feldman’s “Eureka!” to the Rescue?
Standard early-vision resultsCurves are more complexDot-sampled grids
Prediction
Strother & Kubovy (in prep.)
nomenclature
σκ
|κ||σ|
1.2−1 1.1−1 1 1.1 1.2
log p(κ)p(σ)
0
attraction function
we expect collinearityto group more strongly
Kubovy & Strother Proximity, Curvature, & Eureka!
Phenomenological Psychophysics & GroupingCurved vs. Collinear Organizations
Feldman’s “Eureka!” to the Rescue?
Standard early-vision resultsCurves are more complexDot-sampled grids
The Experiment
Method10 observers50 trials/conditionfactorsaspect ratio: |κ|/|σ| = 1.3−1, 1.2−1, 1.1−1, 1, 1.1, 1.2, 1.3
curvature: none, moderate, highdensity: high, moderate, low
duration: 200ms, 300 msaperture: circular, square
Kubovy & Strother Proximity, Curvature, & Eureka!
Phenomenological Psychophysics & GroupingCurved vs. Collinear Organizations
Feldman’s “Eureka!” to the Rescue?
Standard early-vision resultsCurves are more complexDot-sampled grids
The Experiment
The Data
nomenclature
σκ
|κ||σ|
1
log p(κ)p(σ)
0
no curvature
moderate curvature
high curvature
Kubovy & Strother Proximity, Curvature, & Eureka!
Phenomenological Psychophysics & GroupingCurved vs. Collinear Organizations
Feldman’s “Eureka!” to the Rescue?
Standard early-vision resultsCurves are more complexDot-sampled grids
The Experiment
The Data
nomenclature
σκ
|κ||σ|
1
log p(κ)p(σ)
0
no curvature
moderate curvature
high curvature
Kubovy & Strother Proximity, Curvature, & Eureka!
Phenomenological Psychophysics & GroupingCurved vs. Collinear Organizations
Feldman’s “Eureka!” to the Rescue?
Standard early-vision resultsCurves are more complexDot-sampled grids
The Experiment
The Data
nomenclature
σκ
|κ||σ|
1
log p(κ)p(σ)
0
no curvature
moderate curvature
high curvature
Kubovy & Strother Proximity, Curvature, & Eureka!
Phenomenological Psychophysics & GroupingCurved vs. Collinear Organizations
Feldman’s “Eureka!” to the Rescue?
Standard early-vision resultsCurves are more complexDot-sampled grids
The Experiment
The Data
nomenclature
σκ
|κ||σ|
1
log p(κ)p(σ)
0
no curvature
moderate curvature
high curvature
Kubovy & Strother Proximity, Curvature, & Eureka!
Phenomenological Psychophysics & GroupingCurved vs. Collinear Organizations
Feldman’s “Eureka!” to the Rescue?
Standard early-vision resultsCurves are more complexDot-sampled grids
Prediction falsified!
More resultsWhen density is low, curvature is less salient and the effect isreduced.When duration is only 200 ms the effect is reduced.When aperture is square the effect is preserved, hence thesalience of curvilinearity is not a frame of reference effect
Kubovy & Strother Proximity, Curvature, & Eureka!
Phenomenological Psychophysics & GroupingCurved vs. Collinear Organizations
Feldman’s “Eureka!” to the Rescue?
Feldman’s “Eureka!”
Feldman (2004), CognitionQuantifies the principle of nonaccidentality
Given the properties of a set of objects, how unlikely wouldthe pattern be if they were generated at random?In statistics we answer this question by reference to a thedistribution under H0 (such as the t-distribution).The paper gives the answer in the realm of concept learning.
Kubovy & Strother Proximity, Curvature, & Eureka!
Phenomenological Psychophysics & GroupingCurved vs. Collinear Organizations
Feldman’s “Eureka!” to the Rescue?
Feldman’s “Eureka!”
Feldman (2004), CognitionQuantifies the principle of nonaccidentalityGiven the properties of a set of objects, how unlikely wouldthe pattern be if they were generated at random?
In statistics we answer this question by reference to a thedistribution under H0 (such as the t-distribution).The paper gives the answer in the realm of concept learning.
Kubovy & Strother Proximity, Curvature, & Eureka!
Phenomenological Psychophysics & GroupingCurved vs. Collinear Organizations
Feldman’s “Eureka!” to the Rescue?
Feldman’s “Eureka!”
Feldman (2004), CognitionQuantifies the principle of nonaccidentalityGiven the properties of a set of objects, how unlikely wouldthe pattern be if they were generated at random?In statistics we answer this question by reference to a thedistribution under H0 (such as the t-distribution).
The paper gives the answer in the realm of concept learning.
Kubovy & Strother Proximity, Curvature, & Eureka!
Phenomenological Psychophysics & GroupingCurved vs. Collinear Organizations
Feldman’s “Eureka!” to the Rescue?
Feldman’s “Eureka!”
Feldman (2004), CognitionQuantifies the principle of nonaccidentalityGiven the properties of a set of objects, how unlikely wouldthe pattern be if they were generated at random?In statistics we answer this question by reference to a thedistribution under H0 (such as the t-distribution).The paper gives the answer in the realm of concept learning.
Kubovy & Strother Proximity, Curvature, & Eureka!
Phenomenological Psychophysics & GroupingCurved vs. Collinear Organizations
Feldman’s “Eureka!” to the Rescue?
Feldman’s “Eureka!”
Application to our resultFeldman & Singh: each curve has a certain probability, thelikelihood under the smooth hypothesis.
Curvilinearity is less likely than collinearityWhat happens when you have a second curve?When a complex, low-probability event is precisely repeated,this is stronger evidence of a relationship then when a simple,high-probability even is repeated.The significance level (probability under H0) is lower, H0 canbe rejected more readily, and the alternative hypothesis of“pattern” can be accepted more readily.
Kubovy & Strother Proximity, Curvature, & Eureka!
Phenomenological Psychophysics & GroupingCurved vs. Collinear Organizations
Feldman’s “Eureka!” to the Rescue?
Feldman’s “Eureka!”
Application to our resultFeldman & Singh: each curve has a certain probability, thelikelihood under the smooth hypothesis.Curvilinearity is less likely than collinearity
What happens when you have a second curve?When a complex, low-probability event is precisely repeated,this is stronger evidence of a relationship then when a simple,high-probability even is repeated.The significance level (probability under H0) is lower, H0 canbe rejected more readily, and the alternative hypothesis of“pattern” can be accepted more readily.
Kubovy & Strother Proximity, Curvature, & Eureka!
Phenomenological Psychophysics & GroupingCurved vs. Collinear Organizations
Feldman’s “Eureka!” to the Rescue?
Feldman’s “Eureka!”
Application to our resultFeldman & Singh: each curve has a certain probability, thelikelihood under the smooth hypothesis.Curvilinearity is less likely than collinearityWhat happens when you have a second curve?
When a complex, low-probability event is precisely repeated,this is stronger evidence of a relationship then when a simple,high-probability even is repeated.The significance level (probability under H0) is lower, H0 canbe rejected more readily, and the alternative hypothesis of“pattern” can be accepted more readily.
Kubovy & Strother Proximity, Curvature, & Eureka!
Phenomenological Psychophysics & GroupingCurved vs. Collinear Organizations
Feldman’s “Eureka!” to the Rescue?
Feldman’s “Eureka!”
Application to our resultFeldman & Singh: each curve has a certain probability, thelikelihood under the smooth hypothesis.Curvilinearity is less likely than collinearityWhat happens when you have a second curve?When a complex, low-probability event is precisely repeated,this is stronger evidence of a relationship then when a simple,high-probability even is repeated.
The significance level (probability under H0) is lower, H0 canbe rejected more readily, and the alternative hypothesis of“pattern” can be accepted more readily.
Kubovy & Strother Proximity, Curvature, & Eureka!
Phenomenological Psychophysics & GroupingCurved vs. Collinear Organizations
Feldman’s “Eureka!” to the Rescue?
Feldman’s “Eureka!”
Application to our resultFeldman & Singh: each curve has a certain probability, thelikelihood under the smooth hypothesis.Curvilinearity is less likely than collinearityWhat happens when you have a second curve?When a complex, low-probability event is precisely repeated,this is stronger evidence of a relationship then when a simple,high-probability even is repeated.The significance level (probability under H0) is lower, H0 canbe rejected more readily, and the alternative hypothesis of“pattern” can be accepted more readily.
Kubovy & Strother Proximity, Curvature, & Eureka!
Phenomenological Psychophysics & GroupingCurved vs. Collinear Organizations
Feldman’s “Eureka!” to the Rescue?
The take-home answers
1 Particular: Collinear organizations moresalient than curvilinear ones? No.
2 General: Middle-level vision reducible toearly vision? #1 suggests that it cannot.
1 Particular: No.2 General: No.
Kubovy & Strother Proximity, Curvature, & Eureka!
Phenomenological Psychophysics & GroupingCurved vs. Collinear Organizations
Feldman’s “Eureka!” to the Rescue?
The take-home answers
1 Particular: Collinear organizations moresalient than curvilinear ones? No.
2 General: Middle-level vision reducible toearly vision? #1 suggests that it cannot.
1 Particular: No.2 General: No.
Kubovy & Strother Proximity, Curvature, & Eureka!
Phenomenological Psychophysics & GroupingCurved vs. Collinear Organizations
Feldman’s “Eureka!” to the Rescue?
The take-home answers
1 Particular: Collinear organizations moresalient than curvilinear ones? No.
2 General: Middle-level vision reducible toearly vision? #1 suggests that it cannot.
1 Particular: No.2 General: No.
Kubovy & Strother Proximity, Curvature, & Eureka!
Phenomenological Psychophysics & GroupingCurved vs. Collinear Organizations
Feldman’s “Eureka!” to the Rescue?
The take-home answers
1 Particular: Collinear organizations moresalient than curvilinear ones? No.
2 General: Middle-level vision reducible toearly vision? #1 suggests that it cannot.
1 Particular: No.2 General: No.
Kubovy & Strother Proximity, Curvature, & Eureka!
Phenomenological Psychophysics & GroupingCurved vs. Collinear Organizations
Feldman’s “Eureka!” to the Rescue?
ThanksDaniel Dunbar for programmingNIE and NIDCD for grantingThe developers of LATEX and BEAMER for making possiblePDF presentations created with free software.
Download this presentation from MK’s home page:http://www.people.virginia.edu/~mk9y/.
Kubovy & Strother Proximity, Curvature, & Eureka!