Commentary on Alex Byrne and David R. Hilbert.

Abstract: 58 words
Main Text: 998 words
References: 111 words
Total Text: 1230 words

Ecological considerations support color physicalism

James J. Clark
Centre for Intelligent Machines
McGill University
3480 University Street
Montréal, Québec, H3A2A7
Canada

1-514-398-2654
clark@cim.mcgill.ca
http://www.cim.mcgill.ca/~clark

Abstract

We argue that any theory of color physicalism must include consideration of ecological interactions. Ecological and sensorimotor contingencies resulting from relative surface motion and observer motion give rise to measurable effects on the spectrum of light reflecting from surfaces. These contingencies define invariant manifolds in a sensory-spatial space that is the physical underpinning of all subjective color experiences.

 

The arguments for physicalism provided in the target article can be strengthened by considering the ecological aspects of surface reflectance. Humans develop and live in a complex visual environment, and this complexity should not be ignored in developing theories of color perception. There are a number of physical processes governing the spectra of light reflecting from surfaces that become important in non-trivial environments. These physical processes result in significant asymmetries that rule out many of the philosophical arguments against a physicalist view of color (Myin 2001).

For example, the problems caused by metamers vanish when surface inter-reflections (Gilchrist and Ramachandran 1992) are examined. Consider a V-shaped concavity formed from a folded surface with a given reflectance spectrum, illuminated by a diffuse light source. The spectrum of the light reflected from the surface will be a complicated non-linear function of the surface reflectance spectrum (Langer 1999). Suppose we take a second concavity with the same shape as the first, with a surface reflectance spectrum that is different but metameric to the first. Because of the non-linear effects of inter-reflection, the spectrum of the light reflected from the second concavity will, in general, be different than that reflected from the first. Thus, even though two different planar surface patches observed in isolation may appear to have the same color, when the same surfaces are observed in non trivial environmental arrangements (e.g. folded into concavities), the differences between them become apparent. As mentioned in the target article, it is clear that the human visual system takes the visual neighborhood into account when perceiving constancy of color across illuminant spectra changes. Bloj et al (1999) have shown that humans can likewise discount the effect of inter-reflection on perceived color when information as to relative surface orientation is available.

We contend that the human visual system identifies invariances in the sensory input corresponding to particular qualities such as color. These invariances are revealed through active experience in a complex environment. Experience allows the visual system to form statistical models and make predictions of the sensory effect of different motor acts and different environmental contexts on a given phenomenal quantity. In mathematical terms one can consider the invariances inherent in image formation as defining manifolds in an abstract space defined by sensory and environmental degrees of freedom. It is the identification of a particular sensory input with a specific sub-manifold that corresponds with the associated mental contents. The sub-manifold is defined by physical laws, and therefore the associated mental contents are dictated by physical reality. The brain, in this view, gathers evidence for a particular sub-manifold corresponding to a stable physical quantity. Many neurobiologists are beginning to believe that the brain is particularly designed to be able to learn and recognize such invariant manifolds (Seung and Lee 2000).

The apparent problems of the phenomenal structure of color can be handled by the manifold approach, as the particular invariant sub-manifold that the brain actually makes explicit depends on the specific circuitry of the brain. The phenomenal contents themselves are dictated by the physical reality, but the ones that are actually experienced depend on the precise wiring of the brain. For example, consider the problem of color similarity. We will approach this by first examining the sub-manifold (in the space of all possible sensory inputs and ecological configurations) corresponding to a surface with a particular reflectance characteristic. This sub-manifold is determined by purely physical processes and represents invariant properties of the light reflecting from the surface (e.g. what happens to the reflected light when another surface is brought close to it). The sensory apparatus and neural processes in the brain concerned with the perception of hue respond, however, only to a two-dimensional subset of this high-dimension color sub-manifold. They will still respond in a way that reflects the physical laws of sensorimotor and ecological contingency, but will be insensitive to many details. One can think of the brain as responding to a projected version of the high-dimension surface spectrum manifold. The specifics of the coordinate system used to represent the two-dimensional subset depend on the precise details of the neural implementation and will vary from person to person. Unique hues correspond to extreme values of these coordinates (i.e. red-green for one coordinate axis, and blue-yellow for the other). Binary hues correspond to linear combinations of the unique hues. Now, this choice of coordinates is arbitrary, as one could conceive of a brain that would use a different coordinate system to represent color. Thus the subjective experience would be different, much as the subjective experience of a color-blind person in viewing green grass is different than that of a normally sighted person. But this does not mean that the subjective experiences do not have a physical source. On the contrary, both the color-blind and normally sighted observers are merely perceiving different aspects of the same physical structure, the sensorimotor/ecological sub-manifold.

Our view of physicalism is very much in keeping with the recent proposals of O'Regan and Noë (2001). In their theory, color is determined by physical laws which describe how the sensory input is affected by motor acts of the observer, what they refer to as "sensorimotor contingencies". They add that color is also determined by the ways in which the sensory input generated by a colored patch depends on how the patch affects, and is affected by, its environment We refer to these ecological laws as the "ecological contingencies" of color. The role of the brain in our view is to extract, through neural circuits that are hardwired or developed through experience, the sensorimotor and ecological contingencies associated with a given phenomenon. The fact that the brain may only be able to extract a subset of all the contingencies does not eliminate the inherent connection to the physical world.

In summary, we suggest that the consideration of ecological and sensorimotor contingencies associated with the reflectance of light from surfaces leads to a natural elaboration of the physicalist theory of color being expressed in the target paper.

References

Bloj, M.G., Kersten, D. and Hurlbert, A.C. (1999) Perception of three-dimensional shape influences colour perception through mutual illumination. Nature 402(23/30):877-879.

Gilchrist, A.L. and Ramachandran, V.S. (1992) Red rooms in white light look different than white rooms in red light. Investigative Ophthalmology and Visual Science S 1992;4.

Langer, M.S. (1999) A model of how interreflections can affect color appearance. Technical Report No. 70, Max-Planck Institut for Biological Cybernetics, Tuebingen, Germany.

Myin, E. (2001) Color and the duplication assumption. Synthese, 129(1):61-77

O'Regan, J.K. and Noë, A. (2001) A sensorimotor account of vision and visual consciousness. Behavioral and Brain Sciences 24(5).

Seung H.S. and Lee D.D. (2000) The manifold ways of perception. Science 290:2268

 

Acknowledgements

Thanks go to Erik Myin for his comments on a draft of this commentary.