Light, neurons, and color

In the interest of clarity, the analysis of the chain of events from stimulus to perception requires a rigid distinction between coding and perception. At the earliest level, the visual system is confronted with images on the retina. Coding refers to the manner in which the information in the image is sampled, integrated, and transmitted in the visual system. For example, we can speak about the coding of light by photoreceptors. Subsequent transformations of the photoreceptors' signals also represent coding (or recoding) of the information in the retinal image. When discussing the relation between neural events and the stimulus, a fundamental question is how stimulus events are encoded by the responses of neurons.

In contrast, the term color perception ought to be reserved for what we actually experience: the hue, saturation, and brightness of color. In discussing the appearance of color, the term representation can be used to indicate how our experience of color is organized. This representation must have a correlate with neural activity at some level(s) of the visual system, so the coding should be related to the representation of color. The historically broad usage of the term color to refer to physical or physiological properties and events, rather than to perceived phenomena, continues to be a source of confusion that haunts the field. By restricting coding to the relation between neural activity and light, and perception to color appearance, many confusions in the literature are avoided, including ones implicit in misleading terms such as red light, red cones, red genes, and red/green retinal pathway. Red is only a percept.

In general, it is simplistic to expect the perception of color to be associated with the activity of single cells or even a single visual area. Early in the visual system, coding is highly local. At the ganglion cell and other precortical levels, activity is associated with the signals from small populations of photoreceptors and integrated over small regions of the visual field (Dacey, 1996). Progressing through the visual system, receptive fields grow to encompass larger and larger portions of the visual field (Zeki, 1993). This implies that at later stages of visual processing, the responses of cells are influenced by events occurring at locations quite distant from the centers of the receptive fields.

These physiological observations have a parallel in psychophysics. Metameric color matching (see below) can be explained by local retinal coding in the visual system (Brindley, 1960; Schnapf et al., 1987). The colors that we attribute to lights and surfaces, however, are highly dependent on the spatial, temporal, and chromatic properties occurring elsewhere in the visual field (Spillmann and Werner, 1996). In fact, when color is considered under natural conditions, it seems to be an intrinsic property of objects, and not of the light in retinal images. Our visual system, having evolved in an environment characterized by certain regularities, is conjectured to take advantage of such regularities, to resolve an intrinsically ambiguous retinal stimulus into an invariant representation of the properties of objects (such as their spectral reflectances) (Shepard, 1994). The neural processes mediating color perception, therefore, are intertwined with those involved in object perception. The operation of such a complex process is likely to involve several stages of processing, as well as multilevel interactions, in order to integrate and extract information from locations throughout the visual scene.