| |
Background
The Primate Retina
Photoreceptors
In humans, the cone photoreceptors provide the second requirement of color vision: three cone types of differing spectral sensitivity. Rod photoreceptors provide night vision and may interact with cone signals, but it is the cones that provide the basis of normal human color vision. There is now broad agreement on the spectral sensitivities of the long-wavelength-sensitive (L), middle-wavelength-sensitive (M), and short-wavelength-sensitive (S) cones. Figure 58.1A shows a representative set of spectral sensitivities expressed on an energy basis. The S cone has peak sensitivity near 435 nm. The M and L cones peak at midspectrum at 534 nm and 555 nm, respectively. The equal energy spectral sensitivities of the M and L cones are highly correlated, r = 0.95. For images encountered in everyday life, the correlation of M- and L-cone responses is 0.99 (Zaidi, 1997). Burns and Lamb (Chapter 16) give a more detailed review of the primate photoreceptors.
Figure 58.1..
A, Relative cone spectral sensitivity functions (Smith and Pokorny, 1975) plotted as a function of wavelength. B, The cone spectral sensitivity functions plotted relative to luminance; the vertical scaling of the S-cone function is arbitrary.
For chromatic discrimination, we are concerned with the relative responses of the three types of cones under conditions where light stimuli are presented at equal luminance. Figure 58.1B shows the cone spectral sensitivities expressed in terms of relative excitation for lights equated in luminance. The concept of luminance has its roots in the early-eighteenth-century scientific and industrial need for measurement and specification of the visual effectiveness of lights. The CIE 1924 standard luminous efficiency function and its successors were originally proposed for this purpose, but today they also represent the spectral sensitivity of one of the postreceptoral pathways (Lennie et al., 1993).
Based on the receptoral spectral sensitivities alone, one would expect discrimination to be best at the wavelengths where the slopes of the sensitivity functions are steepest. This is qualitatively correct; wavelength discrimination is best in the spectral regions near 480 and 580 nm. However, changes in stimulus conditions cause chromatic discrimination to vary in complex ways that require incorporation of a representation of postreceptoral processing in current models.
Retinal pathways
Major advances in the understanding of retinal anatomy and electrophysiology in the past two decades (Dacey, 2000; Lee, 1996) have revealed three major pathways that convey photoreceptor signals to the brain. The common designations for these pathways arise from the layers of the lateral geniculate nucleus they traverse; the parvocellular (PC), koniocellular (KC), and magnocellular (MC) pathways. The circuitry for the three pathways consists of groups of cells which feed signals forward from the photoreceptor to the lateral geniculate nucleus, via bipolar and ganglion cells, with output to the visual cortex.
The PC pathway mediates spectral opponency of M and L cones and is responsive to luminance changes. PC ganglion cells show the classical center-surround receptive field organization, with a small center and a larger surround that signals with opposite polarity. Four subgroups of PC cells show characteristic response patterns (Derrington et al., 1984; Lee et al., 1994), reflecting the center activity of the typical center-surround retinal ganglion cell. Like classically defined center-surround cells (Kuffler, 1952), ON-center cells respond to an increase and OFF-center cells to a decrease in luminance contrast on their centers (Derrington and Lennie, 1982). The cell types are further distinguished by their spectral properties. The L-ON-center and the M-OFF-center cells respond to an increase in L- versus M-cone contrast (reddish-appearing lights), and the M-ON-center and the L-OFF-center cell respond to an increase in M- versus L-cone contrast (greenish-appearing lights).
Another class of spectral opponency is shown by +S − (L + M) cells, which combine inputs from S, M, and L cones (Dacey and Lee, 1994). These ganglion cells have spatially coextensive centers and surrounds. The cells project to the KC pathway (Hendry and Reid, 2000; Martin et al., 1997). The parallel −S + (L + M) pathway is recognized in physiological studies (Valberg et al., 1986) but its detailed anatomy has not yet been described.
The third major pathway from retina to cortex is the nonopponent MC pathway, which sums inputs of M and L cones in either ON- or OFF-center receptive fields. M and L cones contribute to both centers and surrounds (De Valois et al., 1966; Lee et al., 1989; Wiesel and Hubel, 1966), which in most neurons have similar spectral sensitivities. It is not clear that S cones make any contribution to these receptive fields; if they do, it must be very small (Derrington et al., 1984; Kaiser et al., 1990). This pathway appears to form the physiological substrate for the psychophysical concept of luminance (Lee et al., 1988). Kaplan (Chapter 30) gives a more detailed review of the retinal postreceptoral pathways.
Color Matching
When a photoreceptor absorbs light quanta, subsequent processing stages receive signals proportional to the number of quanta producing isomerizations. All quanta, once absorbed in a given photoreceptor, have identical effects, regardless of wavelength. This property, termed univariance (Rushton, 1972), forms the basis for modern colorimetry and for the systems for specifying chromaticity. Many different spectral distributions of light can evoke identical rates of quantal absorption in each of the three types of cones. Let us create an example, starting from the cone photopigment spectra expressed in terms of relative excitation for lights equated in luminance (Fig. 58.1B). For wavelengths above about 545 nm, S-cone sensitivity is sufficiently low that significant quantal absorptions occur in only the L and M cones. A stimulus field containing 580 nm light can appear identical to another field containing a mixture of two lights, such as 550 nm and 650 nm lights. An identity of quantal absorptions in the two fields can be achieved because the 550 nm light produces a ratio of M- to L-cone absorptions that is higher than that of the 580 nm light, and the 650 nm light produces a ratio of M- to L-cone absorptions that is lower than that of the 580 nm light. A unique mixture ratio of 550 and 650 nm lights results in identical rates of quantal absorption in each of the cone types for the two fields. This is an example of a color match in which a test light (the 580 nm field) appears identical to a mixture of two primary lights (550 nm and 650 nm). Lights that appear identical but are composed of different spectral composition are called metamers.
This logic is the basis of color matching. Any arbitrary light exciting all three cone types can be matched using three primary lights. The term trichromacy is derived from this ability. Most humans make very similar matches and are called normal trichromats. When spectral lights are used for color matching, the test light and one primary light are mixed to match a mixture of the other two primaries.
The importance of the color match lies in the identity of appearance of the two mixture fields, not the actual appearance of the fields. The results of color mixture experiments reflect conditions where quantal catches in the three cone classes are equated. This is an important conclusion because it means that the color matching experiment is both very powerful and very limited. Neither the relative number of the three cone types nor the light level is important. When identity of appearance is attained at the quantal level, subsequent neural processing cannot undo the match. The color matching data are based on the spectral sensitivity of the photoreceptors, but that is all they tell us. Color matching obeys the properties of a linear system. Linearity holds for retinal locations that are cone-dominated (the fovea), and for light levels above cone threshold but below light levels that cause significant cone photopigment bleaching.
Chromaticity diagrams
The linearity property allows the results of color mixture experiments to be expressed in a variety of formats. A discussion of colorimetric transformation methods are beyond the scope of this chapter but may be found elsewhere (Brainard, 1995; Smith and Pokorny, 2003; Wyszecki and Stiles, 1982). One of the most useful formats is the chromaticity diagram, a two-dimensional representation of the results of a color mixture experiment. It is obtained by calculating the proportion of each primary in a match and then representing the proportions in a rectangular diagram formed by plotting two of the primaries.
The most commonly encountered chromaticity diagram is the CIE 1931 XYZ diagram and, for visual psychophysics, the more accurate Judd (1951b) modified XYZ diagram (Fig. 58.2A). The values for the spectral wavelengths plot on a horseshoe-shaped curve; broadband spectral distributions plot within the confines defined by the spectrum locus and the line joining the short and long wavelengths of the spectrum, the line of extraspectral purples. The chromaticity of the equal energy spectrum (EES) for the Judd (1951b) modified XYZ diagram plots at x = 0.334, y = 0.336. The line joining the EES chromaticity and a spectral wavelength represents a purity dimension, having excitation purity ranging from 0 at EES to 1 at the spectrum locus. The two solid lines drawn through the diagram are important in that they represent the chromaticities under neutral adaptation that theoretically reflect activity in only one of the postreceptoral PC or KC pathways while silencing the other. The pairs of dashed lines flanking the solid lines represent stimulus conditions in which the mechanism silenced for the stimuli represented by the solid lines produce steady but nonzero excitation on the other spectral opponent axis.
Figure 58.2..
Chromaticity diagrams summarizing the 2 degree color matching data. A, The (x,y) diagram based on Judd's (1951b) revision of the 1931 CIE observer. The spectrum locus is the horseshoe shape with wavelengths noted. EES is the equal energy spectrum. The solid line and dashed lines converging on coordinate (x = 1, y = 0) is a line in which the S-cone excitation is constant (L/M vary). The solid line and dashed lines converging on coordinate (.1748, 0) is a line in which the L- and M-cone excitations are constant (S varies). The intersections of the dashed lines are shown as chromaticities A, B, C, and D. B, The (l,s) diagram of Boynton and MacLeod, representing cone excitations in an equiluminant diagram. C, The (l,s) diagram of panel B magnified to show the position of the EES. The horizontal lines represent chromaticities with constant S-cone excitation. The vertical lines represent chromaticities with constant L/M excitation. The coordinates A, B, C, and D are shown for comparison with panel A.
A physiological diagram
The CIE and Judd chromaticity diagrams have the disadvantage that they do not represent cone sensitivities directly. A more intuitive chromaticity space to present discrimination data was developed by MacLeod and Boynton (1979). This space is based on the photoreceptor sensitivities of Figure 58.1B, L/(L + M), S/(L + M). The MacLeod-Boynton chromaticity diagram represents a constant luminance plane. Since luminance is represented by a sum of L- and M-cone sensitivities, the proportions of L and M trade off on the horizontal axis. Figures 58.2B and 58.2C shows the MacLeod-Boynton chromaticity diagram. In this space, the ordinate and abscissa coordinates are proportional to L- and S-cone excitations for equiluminous colors. The abscissa is expressed in l-chromaticity (l), which is the proportion of L-cone excitation contributed by the total retinal illuminance (L/L + M). Since the luminous efficiency function is modeled as a linear sum of L- and M-cone contributions, a change in l-chromaticity is associated with a complementary change in m-chromaticity. Thus, a step “redward” is simultaneously an increment in L and a decrement in M. The ordinate value is in s-chromaticity (S/(L + M)), which is the proportion of S-cone excitation contributed by the total retinal illuminance. Therefore, the stimuli falling on the horizontal lines (L/M axis) show changes in M- and L-cone stimulation but constant S-cone stimulation. The stimuli falling along the vertical lines show changes in S-cone stimulation alone. Figure 58.2C shows the MacLeod-Boynton diagram with the solid and dashed lines through the diagram representing the same chromaticities as the lines in Figure 58.2A. Responses associated with the two postreceptoral spectral processing channels are now represented orthogonally, and it is straightforward to envision experimental designs that evaluate the discriminative capacities and possible interactions of the two major retinal pathways signaling spectral information.
Boynton and Kambe (1980) introduced an alternative normalization of the MacLeod-Boynton space with a scheme they called cone trolands. The L and M normalization was retained; the S normalization was changed so that for an illumination metameric to the EES, S was set equivalent to (L + M) and called S tds (trolands). This normalization allows presentation of data either in (l, s) chromaticity units or in (L, S) cone troland units. Cone troland units are particularly useful in evaluating physiologically based models of chromatic detection and discrimination (Smith and Pokorny, 1996).
Sensitivity Regulation
The range of light levels encountered in the natural world is over a trillionfold, from a dark night in a forest to the snowfields of the Rocky Mountains. No single neuron can accommodate this range. A neuron has a threshold when its response is above the dark noise, increases with light level, and reaches a maximum response that is about 200 to 300 times the threshold response. At the maximum, the neuron is said to have saturated, since it cannot generate a differential response with a further increase in light level. Different organisms have adopted different strategies to cope with the range of light levels, including restricting the natural habitat and using pupil variation to control light admitted into the eye. One important strategy is the development of two receptor systems, a low-threshold system for lower levels (rods) and a high-threshold system for higher levels (cones). Even with these strategies, there remains a need for neural mechanisms to regulate sensitivity to the ambient illumination in order to avoid saturation. This behavior is sometimes called adaptation, although we prefer to restrict ourselves to the term sensitivity regulation. Sensitivity regulation means that the neuron resets its response range to the ambient level in order to match its 300-fold response range to that level. When light levels are high enough to bleach photopigment, receptors that have maintained response sensitivity are protected from saturation. Bleaching reduces the number of photopigment molecules available for absorption. It thus acts as a self-regulating sensitivity mechanism.
The classical psychophysical method used to study sensitivity regulation is the increment threshold (König and Brodhun, 1889). In an increment threshold experiment, a test stimulus is presented on a background of fixed luminance. The test stimulus radiance is adjusted to determine the detection threshold. The measurement is repeated at a series of background radiance levels, and a threshold versus radiance (TVR) function plotting the logarithm of threshold versus the logarithm of the background may be created.
Sensitivity regulation in rods
It is instructive to look at the TVR function for a single receptoral pathway. We can do this for rods. Aguilar and Stiles (1954) isolated the rod system using a long-wavelength background to suppress the cone photoreceptors and a middle-wavelength test spot to stimulate the rod photoreceptors. Figure 58.3 shows the mean of the thresholds from the four Aguilar-Stiles observers. The threshold is relatively constant at low scotopic levels (scotopic is the term used to describe light levels for rod-mediated vision), starts to increase near −4 log scotopic tds, and then rises linearly between background illuminances of −3.0 and 2.2 log scotopic tds. Above 2.2 log scotopic tds, the rod system begins to saturate, and the amplitude of the thresholds increases rapidly.
Figure 58.3..
Increment thresholds for the rod system. (Data from Aguilar and Stiles, 1954.)
Below 2.2 log scotopic tds, the Aguilar-Stiles increment threshold value, ΔI, can be described for any scotopic illuminance, I, by an equation of the form
ΔI=k(
I
n
+
I
0
)
(1)
I0 is the so-called dark light. The exponent n and the constant k are defined for the asymptotic portion of the curve when I is much greater than I0 but before saturation begins: n is the limiting slope of the TVR function plotted on log-log coordinates and, for n = 1.0, k is the Weber fraction, the fractional increment above the background light required for detection.
The TVR function is often characterized as composed of four segments. Starting from darkness, as background luminance is increased, the threshold does not differ from the absolute threshold for a dark background. This is called the linear region of the TVR function, and sensitivity in this section is hypothesized to be limited by neural (internal) noise (Barlow, 1958), the dark light. This neural noise is internal to the retina, and examples include thermal isomerizations of photopigment, spontaneous opening of photoreceptor membrane channels, and spontaneous neurotransmitter release. The second part of the TVR curve represents the transition between the linear region and the region where sensitivity regulation occurs. The linearly rising portion of the function is sometimes called the Weber law region; ideally, it has a slope (n, in equation 1) of 1.0. This means that the sensitivity regulation is exactly compensating for the steady ambient illumination. Although the psychophysically observed Weber behavior with a slope of 1.0 is often attributed to adaptation in the rod receptors themselves, this cannot be the case since shallower slopes, of about 0.7, are found for some stimulus conditions different from those of Aguilar and Stiles (Shapiro et al., 1996; Sharpe et al., 1989). The rightmost portion of the TVR curve shows rod saturation at high background luminance. Sensitivity falls so drastically that no amount of increment light reaches detection threshold.
Sensitivity regulation in M and L cones
Sensitivity regulation in cones differs from that in rods. Saturation does not occur in the steady state for stimuli activating the M and L cones. At high background light levels, >10,000 tds, a significant amount of photopigment is depleted due to bleaching. Following the viewing of a bright light for a period of time, the loss of quantum catch ability reaches equilibrium and protects the cones from saturation.
Sensitivity regulation for stimulation of the M and L cones depends on the postreceptoral pathway in which measurements are made. In the MC pathway, the TVR function moves from threshold and attains Weber's law with a slope of unity. For foveal stimuli, threshold is near 0.1 td and Weber's law is achieved at 1 td (Hood and Finkelstein, 1986). In the PC pathway, the TVR function moves from threshold to a limiting slope near 0.6 to 0.7 until bleaching levels (>10,000 td) are reached. For foveal stimuli, threshold is near 1 td and the limited sensitivity regulation is achieved by 3 to 8 td. This pathway difference is evident both in psychophysical studies (Swanson et al., 1987) and in physiological studies (Lee et al., 1990).
Now, Weber's law requires n = 1.0 in equation (1). The finding that Weber's law occurs only in the MC pathway implies that sensitivity regulation is hierarchical. Some regulation is common to both pathways and may be in the cones themselves. Intracellular recordings from primate horizontal cells, one synapse removed from the receptors, show slopes of 0.6 to 0.7 (Smith et al., 2001) and independent sensitivity regulation in each cone type (Lee et al., 1999). The MC and PC pathways are differentiated at the first synapse, the cone-bipolar synapse. The additional regulation of the MC pathway thus occurs in the bipolar or ganglion cell complexes.
The linear slopes of 0.7 or 1.0 are sometimes referred to as multiplicative regulation (Hood and Finkelstein, 1986). The sensitivity regulation mechanism scales the response both to the background and to the test stimulus. The neural mechanisms that underlie multiplicative regulation are not delineated. In the photoreceptors, sensitivity regulation probably involves the complex interactions of the outer segment photocurrent. Additionally time-dependent mechanisms, in which temporal resolution is traded for sensitivity, probably play a role both in the photoreceptor and in the MC-pathway mechanism.
There is another form of regulation in which the mechanism scales only the background (Hood and Finkelstein, 1986). This is called subtractive regulation. Subtractive regulation by itself is usually not considered an effective mechanism of light adaptation since it provides only a single scaling of the background light level. As the sole method of sensitivity regulation, it will delay but not protect the postreceptoral neurons from saturation. However, in combination with other mechanisms (e.g., following partial multiplicative regulation), it can be effective. Subtractive feedback is a possible neural substrate for subtractive sensitivity regulation. There is evidence that the PC pathway shows subtractive feedback (Krauskopf and Gegenfurtner, 1992; Smith et al., 2000). Subtractive feedback as a second form of hierarchical regulation in the PC pathway is attractive. It was stated that the L- and M-photopigment sensitivities are highly correlated. As a result, at a given luminance level the range of differential spectral stimulation is limited. There is only 0.3 log unit differential in M-cone to L-cone stimulation between 480 nm and 700 nm. This small range is easily handled by a subtractive feedback mechanism in place following the spectral opponent receptive field.
Sensitivity regulation in S cones
Sensitivity regulation in S cones differs considerably from that in L and M cones. In some ways, S-cone regulation is more similar to that in rods in that the S cones do show saturation (Mollon and Polden, 1977). They do not obey Weber's law. They may have some multiplicative regulation but with a limiting slope of 0.6 to 0.7. The S-cone system does show subtractive regulation. This has been termed second-site regulation in the literature (Pugh and Mollon, 1979). However, the range of S-cone stimulation relative to an adapting white stimulus is high, 60 to 1. Thus, the subtractive mechanism does not protect the S-cone system from saturation. There is still no final understanding of sensitivity regulation in the S-cone system.
| |
