John O`Dea University of Tokyo

Representational Primitives in Perception
Towards a general theory of perceptual constancy
John O’Dea
University of Tokyo
Although I believe that the table is of the same colour all over, the parts that reflect the light look much brighter than the other parts...Here we
have the beginning of one of the distinctions that cause most trouble in Philosophy - the distinction between appearance and reality.
-- Russell, Problems of Philosophy (1912)
1. What is Perceptual Constancy?
“The perception of objects as constant in size, shape, color,
and other properties despite changes in their retinal image.
Berstein et al, Psychology (8th edition, 2008)
2. A Theory of Color Constancy
4. Towards a General Theory
Red lines in Figure 3 describe the range of possible perceptual experience given
the different stimuli (in this case, the corner of a uniformly coloured wall),
supposing that surface colour and illumination are inseparably experienced and
co-determining. The image is normally seen either as uniformly illuminated or as
uniformly coloured. It is possible to switch between these experiences in a topdown way (as for ambiguous images), though normally the visual system
automatically chooses, perhaps in Bayesian fashion, the most likely
interpretation. The capacity for top-down, deliberate interpretive switching creates
the false impression that both experiences are simultaneously present.
3. The Refrigerator Light Illusion
Fig. 1. Size constancy:
objects in the distance are
not usually perceived as
smaller.
Fig. 2. Shape constancy:
orientation does not usually
affect the shape things look
to be.
It is possible to see an ʻapparentʼ difference in surface colour in a context of
varied lighting. This is standardly assumed to be a continuous part of visual
experience, albeit not normally noticed. I claim that it is actually an unstable,
transitory experience involving a Gestalt-like switch. The veridical experience and
the false experience (which Russell calls the “appearance”) cannot occur
simultaneously, but the ability to switch at will creates the illusion otherwise.
Fig. 4. The apparent size of an object depends on how far it
appears to be. Size and distance are also co-primitive - to
see the size of an object at all is to see it at some distance or
other.
Fig. 3. Colour constancy: the white wall looks different in different lighting conditions.
A New Characterisation
The standard definition is indistinguishable
from mere successful perception. Rather,
perceptual constancy is best characterised
as:
A pair of properties;
One intrinsic and one extrinsic; which
are
Representationally and computationally
linked (ʼco-primitiveʼ); and
Fig. 5 The apparent shape of an object depends on its
apparent orientation, and conversely. Shape and orientation
are co-primitive in my sense. To see a shape at all is to see it
at an orientation.
Amenable to some degree of
conscious, top-down manipulation.
The Problem of Constancy:
Experiential Difference
On the one hand, perceptual experiences are paradigm
cases of intentionality; different perceptual experiences
plausibly have, thereby, different accuracy conditions.
However, due to perceptual constancy, one and the same
object can appear different without appearing to be
different.
Plausibly, a white experience is simply an experience that
represents an object as being white. The problem of
phenomenal variation arises because objects can look to
be white, and and yet that whiteness looks different in
different lighting (see Fig. 1).
The Myth of the Simple Colour Experience
Seeing the Light (and Dark)
Colour is a property not only of objects but also of the
illumination.
The amount (and quality) of illumination is represented
explicitly in perceptual experience
Several theorists recently argue persuasively for the
importance of this in accounting for the apparent remainder
(see Matthen, Hilbert, Mausfeld).
But how are illumination
and surface colour related
in perceptual experience?
x
x
A common approach, both in philosophy and psychology.
The experiential variation is attributed to sense data, a visual
field, a non-representational “qualitative” difference, et.
Example: Peacocke, Sense and Content (1983)
Imagine that you are in a room looking at a corner
formed by two of its walls. The walls are covered with
paper of a uniform hue, brightness and saturation. But
one wall is more brightly lit that the other. In these
circumstances, your experience can represent both
walls as being the same colour: it does not look to you
as if one of the walls is painted with brighter paint
than the other. Yet it is equally an aspect of your visual
experience itself that the region of the visual field in
which one wall is presented is brighter than that in
which the other is presented.
Hilbertʼs “Simple Solution”
Hilbert suggests that illumination might be regarded as a
dimension of colour, alongside or partly replacing hue,
saturation and brightness (HSB).
Some reasons to reject this approach:
Illumination is explicitly visible, but not HSB. Perceptual
matching is not possible for HSB separately (illustration the patches below (left) are composed of only two hues, but
this is not visibly obvious. In contrast, the two colours
composing the right figure are visibly apparent)
This approach is not generalizable. Orientation is not
plausibly a dimension of shape; distance is not plausibly a
dimension of size.
Some reasons to reject this approach:
Incompatible with a powerful approach to perceptual
experience: the idea that the experiential qualities of a
perceptual experience are its intentional contents.
There is no convincing account of the nature of the
postulated mental remainder.
There is a better alternative (see above).
A perceptually represented property is primitive iff it is explicitly
represented in perception and it is not the case that there is some
other property in virtue of the representation of which it has that
particular content.
Neither surface colour nor illumination are “primitive” in this sense.
No surface ever looks simply coloured - the experience will also
specify some illumination. Surface colour and illumination are only
jointly primitive in the above sense.
Hence, trivially, an un-illuminated object is not visible. This is not
only a practical truism; rather, it is a deep phenomenological fact,
the neglect of which is, I claim, the root cause of our failure to
understand constancy.
What is the apparent
experiential remainder?
A merely mental remainder?
A (stipulative) definition of “primitive”:
References
Matten, Mohan (2010) “How Things Look” in Perceiving the World; OUP.
Mausfeld, Rainer (2003) “The Dual Coding of Colour” in Colour Perception: Mind and
the Physical World; OUP.
Hilbert, David (2005) “Color constancy and the complexity of color” in Philosophical
Topics, 33.
Kelly, Sean (2008) “Content and Constancy: phenomenology, psychology, and the
content of perception” in PPR LXXVI (3).
Objection: It is completely obvious that
we see shadows and shading. How
could this be so misleading?
Ambiguity as a Feature of Constancy
Differences in the amount of
reflected light can mean either
differences in the light absorbed by
the surface, or differences in the light
received by the surface (or both).
Just as the conscious (or ʻtop-downʼ)
manipulation of perceptual
experience is possible in the case of
ambiguous scenes, it is also possible
the case of non-ambiguous scenes,
such as the white wall, where one interpretation is clearly more
plausible/
To see a scene such as the the wall in Fig.3 as receiving uniform
illumination (for example, by supposing it to be flat) is to see a
difference in surface colour (see Kelly 2008 for a similar proposal).
Because the “alternative”
way of seeing can be
experienced whenever
one tries to do so, it gives
the impression of being a
continual part of the visual
experience of the table.
This is a version of the
refrigerator light illusion
Acknowledgments: The paper on which this poster is based was presented at the Philosophy departments of Monash University, the University of Singapore and the University of Tokyo, and was greatly improved by feedback from participants, especially Jakob Howhy and Ben Blumson.
Photographs: Figure 2 is © 2009 sinkdd (Flickr username), used under a Creative Commons Attribution-Noncommercial license; the same license applies to the Face/Vase (© solsken (Flickr username)) and the Refrigerator Light (© Zen Sutherland).