While all senses have been optimized through evolution to provide unique information for survival, they must simultaneously be general-purpose systems that can respond to an ever-changing environment (Handel Reference Handel2006). For this reason, it seems inevitable that peripheral receptors must respond with differential excitations to varying stimuli (although there may be a small set of receptors that are narrowly tuned, Wilson Reference Wilson2007; Xu et al. Reference Xu, Atkinson, Jones and Smith2005). That, in turn, results in an across-fiber firing distribution representing the entire sensory continuum. Although across-fiber patterning appears to provide a functional mechanism for all senses, I argue that the properties of the sensory environment determine the number and distribution of the peripheral receptors and the subsequent computations that integrate the excitations of those diverse receptors. The issue of economy of coding can be understood only in terms of the goal of detecting objects in the environment.

1. Number of receptors

Why are there only three cones? The spectra of natural light and the reflectance of most materials are continuous and smooth (Judd et al. Reference Judd, MacAdam and Wyszecki1964; Wandell Reference Wandell1995). Typical excitations can be reproduced using different amounts of three independent functions: one function to represent the overall illumination level, one function to represent the blue-yellow contrast, and one function to represent the red-green contrast (Ruderman et al. Reference Ruderman, Cronin and Chiao1998). That both the illumination and reflectance functions can be represented by a small number of independent functions suggests that only a small number of receptors would be necessary to recover the illumination-independent color. However, the excitations of the three cones overlap to such a degree that there is a high correlation among the firing rates. The subsequent opponent color processing acts to decorrelate the responses of the receptors and yield independent contrast functions.

Why are there so many sound receptors? Sounds range from nonharmonic noise (wind) resulting from continuous frequency excitation to harmonic timbres produced by vibrating objects. For vibrating objects, the source excitation (e.g., vocal fold vibration) occurs at discrete non-overlapping (typically harmonic) frequencies, but the sound body resonances (the filter) occur at multiple overlapping frequencies based on the shape and material. Normally, the resonances create multiple excitation peaks at widely spread frequencies separated by regions of low amplitude. Neither the source spectra nor the filter spectra can be modeled by a small number of independent linear functions. To distinguish among different timbres (i.e., different sound objects) therefore requires many receptors (roughly 3,000 receptors/ear), necessarily tuned to narrow overlapping frequency bands to pick up the resonance peaks. (If important sounds were nonharmonic, then fewer receptors would be necessary).

What this means is that the organization of the visual and auditory periphery seems tuned to the properties of the environment. Economy of coding is not about the number of receptors; it is about the picking up of the redundancies that signify objects (Barlow Reference Barlow2001). Is this tuning also true for taste? Is there a reason for the small number of taste receptors? Furthermore, since there is no continuous physical attribute underlying the various tastes that would support excitation and inhibition among adjoining receptors (as found for seeing and hearing), would a cross-fiber patterning explanation require that all the peripheral taste receptors interact with each other? The across-fiber excitatory and inhibitory interactions found for vision and audition generate cells at higher levels with unique properties such as motion detection. Is this also true for taste, or is there just a simple population code? Without a physical continuum would it be possible to construct taste metamers analogous to color metamers, given the small number of different taste receptors?

2. Distribution of receptors

Young (Reference Young1802) argued that it would be physically impossible to have enough unique color detectors at each retinal point to account for color perception. Although the cones in the fovea underlie color perception, they are also responsible for the fine-grain shape and motion perception that creates figure-ground organization. I would argue that it is this latter function that is critical to survival and that color constancy enhances our ability to segment objects in spite of changes in shape resulting from motion, occlusion, or other variation. From this perspective, the color detectors occur throughout the fovea because they augment figure-ground organization. If only the background was colored but not the objects, then labeled-line color receptors could be localized at discrete points on the retina.

Here again, economy is about maximizing information. The perceptual goal is to construct a stable visual world in spite of the constant flux. If objects and backgrounds were not usually colored differently and if more independent functions were necessary for color perception, then a different receptor organization would have emerged. I feel that an economy argument for an across-fiber explanation of taste must go beyond the number of receptors. If taste perception is not about identifying foods, then I wonder if analogies to other senses are appropriate.

3. Perceptual constancy

Although we believe that color is an inherent property of objects, in reality perceived color is based on our inferences about the illumination and surrounding colors. It is a second-order calculation based on the relative ratios of absorptions in different parts of the field. Color matching under different illuminations requires imagining what the standard would look like under the second illumination, not matching cone absorptions. In this target article, Erickson presents results illustrating that the description of tastants is similarly context dependent. Are such judgments possible with a population code or must they also be based on higher-level excitation ratios?

In sum, senses convey specific survival information, as well as respond to an evolving environment. But, as I have argued earlier, it is the particulars of the environment that determines the organization and distribution of the receptors and the subsequent transformations at later neural regions. The neutral excitation at the eye (a photograph) or ear is transformed into a set of parallel intensity contrasts at multiple resolutions that segment sensory experience into objects and properties regardless of scale (De Valois & De Valois Reference De Valois and De Valois1988; Lewicki Reference Lewicki2002). I would expect similar sorts of transformations for taste, and deriving those transformations is necessary to create an adequate theory of taste perception.