Perceptual ambiguities in the uncertain mind

advertisement
Perceptual ambiguities in the
uncertain mind
Johan Wagemans
Laboratory of Experimental Psychology
University of Leuven
CNCC Final Conference
Edinburgh, October 1-4, 2009
Uncertainty & ambiguity
• the visual system has to deal with incomplete
and noisy information but it has
evolved/learned to come to quick perceptual
decisions
– visually guided actions
– conscious percepts
• uncertainty is avoided but ambiguity remains
• different types of uncertainty & ambiguity
• underdetermination = crucial
Goals of this talk
• to explore the “boundaries” (possibilities and
limitations) of the subconscious mind in its attempt to
create certainties from inherently noisy and ambiguous
sources of information
• to bring order in the chaos of all kinds of perceptual
ambiguities by a phenomenological analysis of the
similarities and differences
• note: visual cases only
• note: focus on some of our on-going research projects
• note: I will probably raise more questions than I will
provide answers
Case 1
20% S
30% M
15% S
25% M
Case 1
• a few line segments, sometimes insufficient to
define a complete contour
• what is needed:
–
–
–
–
grouping
filling-in (contour completion)
(contour-based) shape description
interpretation
• what is experienced = transition between
unstructured, meaningless set of elements to
coherent, meaningful figure
Case 1 – additional observations
• we can examine necessary and sufficient
sources of information
• we can examine levels of uncertainty
• we can examine thresholds for identification
– from continuous stimulus change
– to abrupt change in percept
• we can investigate external and internal
factors influencing this
Case 2
Case 2
• a few line segments, embedded in a noisy background
• stimulus change is not additional contour fragments
but proper local alignment
• what is needed:
– grouping of target elements, segregation from background
elements
– filling-in (contour completion)
– (contour-based) shape description
– interpretation
• what is experienced = transition between unstructured,
meaningless set of elements to coherent, meaningful
figure
Case 2 – additional observations
• same, plus more:
– interplay between grouping and segregation
– same local changes, different global effects
– nice tool to investigate different mechanisms and
representations in the visual system’s hierarchy
Case 3
Case 3
• a few oriented elements, embedded in a noisy
background consisting of similar elements
• difficulty depends on local orientations and spatial
relations
• what is needed:
– grouping of target elements, segregation from background
elements
– filling-in (contour completion)
– (contour-based) shape description
– interpretation
• what is experienced = transition between unstructured,
meaningless set of elements to coherent, meaningful
figure
Case 3 – additional observations
• same, plus more:
– interplay between contour, surface, and
background grouping
– nice tool to investigate temporal dynamics of
different mechanisms and representations in the
visual system’s hierarchy
Case 4
c
a
b
d
Bravais (1848): Taxonomy based on crystallography
Kubovy & Wagemans (1995): Parameterization and experimentation
The space of lattices
90°
γ
60°
1
|b|
|a|
The space of lattices
90°
rectangular
γ
60°
1
|b|
|a|
The space of lattices
90°
square
γ
60°
1
|b|
|a|
The space of lattices
90°
γ
hexagonal
60°
1
|b|
|a|
The space of lattices
90°
γ
centered
rectangular
60°
1
|b|
|a|
The space of lattices
90°
γ
oblique
60°
1
|b|
|a|
Lattices vary in
shape
&
ambiguity
Rectangular
90°
Square
γ
60°
1
|b|
|a|
Hexagonal
Case 4
input is regular, not noisy
all the information is there
still, several organizations are possible
perceptual organization is influenced by the input but not
deterministically (noise in the system, rather than in the
input)
• what is experienced = transition between different
organizations (switching)
• different levels of ambiguity
•
•
•
•
– in some cases, one state is much more likely than others
– in other cases, two or three states are equally likely
• note: no intermediate or mixed percepts
Case 5
Case 5
• well-defined shape, only its perceived
orientation switches (tri-stability)
• perceived orientation depends on other cues
(gravitational framework, axis- and basealignment)
• switching occurs for the group as a whole
Case 6
Case 6
• insufficient information
• black and white patches, appearing random in shape
and distribution
• what is needed:
– grouping of target elements, segregation from background
elements
– filling-in (contour and surface)
– (contour- and surface-based) shape description
– interpretation
• what is experienced = transition between unstructured,
meaningless set of elements to coherent, meaningful
figure
Case 6 – additional observations
• once you have interpreted it in one way, it is
impossible to switch back to the uninterpreted
state
• sometimes the visual system exaggerates and
imposes templates from visual memory on the
input image that are inappropriate (i.e., do
not correspond to the true source of the
image): pareidolia
Case 7
Case 7
• unclear, misleading information
• black and white patches with clear edges, appearing
random in shape and distribution
• source of difficulty:
– edges between black and white are not always proper
transitions between surfaces (spurious edges due to
thresholding of shading)
• what is needed:
–
–
–
–
deconstruction of low-level units
reconstruction of surface patches
shape-description
interpretation
• what is experienced = transition between unstructured,
meaningless set of elements to coherent, meaningful figure
Case 8
Case 8
• clearly-defined black and white figures
• two possible interpretations
• in some cases, one interpretation is very salient (defined by
strong cues) and the second may be hidden and must be
searched for or even pointed out
• in this case, the second has partly real and partly illusory
contours
• in other cases, both interpretations can be more or less
equally strong, in which case switching takes place regularly
• the contour is owned by the figure, hence switching
• border-ownership plays a crucial role (cf. model by Naoki
Kogo, yesterday’s poster)
• figure-ground ambiguities
Case 8 – additional observations
Case 9
Case 9
• clearly, richly defined shapes, and yet ambiguous
• two different interpretations of the same input
• switching does not imply changing borderownership and yet they cannot be perceived
simultaneously
• in the case of Boring’s old/young woman
(wife/mother-in-law), it does imply a different 3D surface-interpretation of the 2-D edges
Case 9 – additional observations
• context effects
• top-down effects
Case 10
A
100% Duck
0% Rabbit
B
80% D
20% R
C
60% D
40% R
D
50% D
50% R
E
40% D
60% R
F
20% D
80% R
G
0% Duck
100% Rabbit
Verstijnen & Wagemans, 2004
RABBIT
CAKE
DUCK
TENT
Duck-Church
100
80
60
40
20
0
0%100%
20%80%
30%70%
40%60%
50%50%
level of morphing
60%40%
70%30%
80%20%
100%0%
Cat-Butterfly
100
80
60
40
20
0
0%100%
20%80%
30%70%
40%60%
50%50%
level of morphing
60%40%
70%30%
80%20%
100%0%
Case 10
• some sequences give steep transitions (categorical
perception), others give smooth transitions (cf. Mijke
Hartendorp et al., Utrecht)
• sometimes the intermediate cases are perceptually
disambiguated, sometimes they are not (nothing
interpretable is seen, or many different things may be
seen)
• individual differences, both in conceptualization
(starting with the middle stimulus going to one end)
and in reconceptualization (starting at one end of the
continuum and going to the other end) (cf. autism
research by Hollie Burnett & Tjeerd Jellema, Hull)
• additional differences w.r.t. animate – inanimate
Case 10 – additional observations
• sequential order plays a role: hysteresis effect
Case 10 – additional observations
• sometimes all morphs are equally
interpretable, e.g., within-category morphs
Case 10 – additional observations
• sometimes the morphs are seen as mixtures,
i.e., novel, hybrid objects with features of
both source objects
Case 11
Case 11
• two different 3-D interpretations of the same
2-D input
• clearly a case of underdetermination
• also no two interpretations at the same time,
so switching occurs
• but this time not between two figure-ground
organizations but between two 3-D
perspectives (reversible figures)
Case 11 – additional observations
• mixtures are possible
• impossible figures (the visual system can
tolerate global inconsistencies)
• applications in advertisements
• applications in art (e.g., Escher, Dali)
Conclusions
• a wide variety of perceptual ambiguities:
– noisy vs. clear (but ambiguous) input
– switching from unclear to resolved state (reversing is
impossible)
• strongly affected by visual memories (templates)
– switching between two or more clear states (reversals
are quite common)
• figure-ground
• morphed shapes
• reversible figures
– intermediate states (ambiguous, hybrid, impossible
combinations)
Conclusions
• a wonderful tool to study perception
– subjective nature
– unconscious, subconscious, conscious
processes/states
– perceptual and neural mechanisms (interplay
between different mechanisms, temporal dynamics)
– external and internal factors influencing these
– still many unresolved issues and questions
– excellent playground for interdisciplinary
collaborations
Thank you! Questions?
• johan.wagemans@psy.kuleuven.be
Download