The origin of the idea that rods may give rise to chromatic sensations may be traced back to the hypothesis of Ebbinghaus (1893) that the photopigment of rods (rhodopsin) was the yellow-blue see-substance postulated by Hering (1878), and the suggestion of König (1894) that the rod receptors represented the primary ‘blue’ receptor system of photopic vision. Substantial evidence in favour of the idea was provided by von Kries and Nagel (1896) who found that twilight vision contained a tint of blue.

In order to identify more precisely this blue colour quality of rods, von Kries (1896) attempted to find the spectral light that did not change its hue when test intensity was reduced from photopic to scotopic levels, i.e. when the rod component increased. He used Nagel (a deutranope) as his subject and instructed him first to make a colour match at a photopic intensity level between a homogeneous spectral test light from the short-wave region of the spectrum (and hence with a relatively high potential scotopic value) and a mixture of spectrum red (670 nm) and violet (435 nm) lights (and hence with a relatively low potential scotopic value). Thereafter, the intensity of the test and comparison fields was reduced in the same proportion from the photopic level until the homogeneous light strongly activated the rod component. His results showed that the spectral light that did not change in colour quality with intensity reduction was situated between 480–485 nm, i.e. in the green-blue part of the spectrum.

Perhaps the most profound change in our conception of the sensitivity regulation mechanisms, however, came when it was realized that the rod and cone activities were not mutually independent, but that rod activity could influence cone sensitivity and vice versa. Anatomical, electrophysiological and psychophysical studies all contributed to this change in view.

Well-founded histological evidence in favour of the rod-cone interaction hypothesis was provided by Polyak's investigations of primates, summarized in Polyak (1941/1948). He discovered that, with the exception of the midget cone system, rods and cones were connected to the same pathways in the retina. Furthermore, he showed that the retina of primates was an extremely complex structure composed of a variety of different types of neural networks strongly indicating that the retina, and not just the brain, was involved in analyzing and synthesizing visual information.

More direct evidence in favour of rod-cone interaction was provided by Granit's ERG measurements. These results were interpreted to mean that rods and cones competed for ‘a final common path’ when excited simultaneously, and that the more active receptor types tended to exclude the other (see Granit, 1938, pp. 65–66).

Later on, Gouras and Link (1966) came to a similar conclusion. They investigated rod-cone interaction in Rhesus monkeys by analyzing the discharge pattern of ganglion cells connected to both rod and cone receptors. Their results showed that impulses from rods or cones arriving first at the ganglion cell would tend to block impulses from the other receptor type when these arrived shortly thereafter.

ROD-CONE INTERACTIONS UNDER MESOPIC CONDITIONS IN A CHROMATICALLY NEUTRAL STATE OF ADAPTATION

In the period from Schultze (1866) to Lie (1963) it had been generally agreed that rods and cones interacted in a kind of colour-mixture process in mesopic vision. Rods and cones were assumed to contribute an achromatic and a chromatic component, respectively. The most advanced attempt to further characterize the rod-cone interaction under mesopic conditions had been made by Granit (1938, 1947) and Lie (1963). They both suggested that rod and cone activities antagonized each other at the retinal level in that the most sensitive receptor system tended to suppress the other. The rise of the specific-hue threshold obtained when rods intruded during long-term dark adaptation was compelling evidence in favour of this antagonistic interaction. There could be little doubt that rod signals completely suppressed the chromatic cone signals within the intensity interval between the cone-plateau level and the specific-hue threshold.

Yet, in apparent opposition to the psychophysical data and the hypothesis of rod-cone antagonism proposed by Lie (1963), the histological and electrophysiological evidence obtained in the late 1980s and early 1990s indicated that rods and cones added their responses of the same polarity both at the receptor and bipolar cell levels (Daw et al., 1990; Schneeweis & Schnapf, 1995).

Fortunately, this apparent conflict between the psychophysical and the electrophysiological data could easily be resolved.

PHENOMENOLOGICAL ANALYSIS MAY REVEAL UNDERLYING MATERIAL PROCESSES

The almost complete lack of knowledge about colour processing in the visual pathway is quite understandable. How could it be possible to obtain such information without microelectrodes or other advanced instruments at hand to monitor the processing?

An ingenious way out of this apparently insurmountable difficulty was offered by Hering (1878). He held that there were actually two quite different routes to understanding the processes underlying colour vision: a direct physiological approach and an indirect psychological approach. The psychological approach was based on the presumption that information about material processes underlying colour vision may be obtained by analyzing the phenomenological characteristics of colour sensations. Actually, Hering accepted the psychophysical maxim of Mach (1865, p. 320) that made three basic assumptions:

1. Every mental process is unalterably correlated with an underlying material process.

2. Similar and different mental processes are, respectively, correlated with similar and different material processes.

3. Every detail in the mental process corresponds to a detail in the material process.

As may be seen, the maxim of Mach is a specification of Spinoza's principle of psychophysical parallelism. It may also be noted that the maxim is akin to Leibniz's presumption that there is a pre-established conformity between mind and body (e.g. Boring, 1957, pp. 165–168).

Presupposing Mach's maxim to be valid, it would be a straightforward undertaking to obtain information about material processes underlying colour sensation, once an unbiased and comprehensive phenomenological analysis of colour vision was worked out.

Although scientific theories and laws of nature may never be proved in any definite way by induction, as Hume had made clear, Popper (1969, 1975, 1994) pointed out that they may be falsified on the basis of empirical evidence and purely deductive reasoning, i.e. the modus tollens deduction form of classical logic (see Popper, 1975, pp. 75–77). Hence, he assumed that scientific theories could be tested by attempts to refute them, and by selecting the most successful theories that withstood severe falsification tests, Popper believed that science could progress ever closer to, but never reach, the ultimate truth. The key to scientific development, therefore, according to Popper, was not the collection of observational statements, but the emergence of competing, falsifiable theories. Only by searching for falsification of theories and, on this basis, selecting the most successful ones, could science hope to learn and advance.

Popper's view of scientific progress may be summarized as follows: progress starts when a theory is refuted or falsified. This will create a problem for the relevant scientific community. In order to solve the problem other falsifiable theories are proposed. These new theories are then criticized and tested. As a result one or a few of the theories will prove more successful than the others, i.e. they may withstand severe falsification tests, may have greater empirical information or content, may be logically consistent, may have greater explanatory and predictive power, and may be more simple. Eventually, however, even the most successful theory will be falsified.

In this developmental period, profound new knowledge about the anatomical and neurophysiological properties of the retina emerged as a result of advances in sophisticated instrumentation and research techniques. This knowledge greatly influenced the development of the duplicity theory: it provided an insight into the rod and cone processes, and also created and paved the way for new ideas of rod and cone functioning. Outstanding contributions to the development were provided by Polyak, Hartline, Kuffler and Granit. Using the Golgi impregnation method, Polyak's investigation of the primate retina helped to elucidate the extremely complex structure of the many types of retinal cells and the character of their connections. Hartline, Kuffler and Granit, using microelectrodes for registration of the action potential from individual nerve fibres in response to illumination, increased our knowledge about the relationship between light stimuli and nerve impulses in the retina.

THE DUPLICITY THEORY OF MAX SCHULTZE

The first section of Helmholtz's classical ‘Handbuch der Physiologischen Optik’ was published in 1856, the second (containing the Young-Helmholtz trichromatic colour theory) in 1860, the third partly at the beginning and partly towards the close of 1866, while the complete work was published in 1867. The Young-Helmholtz colour theory was, therefore, known to Max Schultze (professor in anatomy and the director of the Anatomic Institute in Bonn, Germany) when he published his important work in 1866. Indeed, he accepted its major assumption that three different cone fibres conveyed independent, qualitatively different colour-related processes.

Yet, in addition to the cone receptors and their functions, he introduced the much more numerous rod receptors presuming that they were responsible for night vision, and that they mediated achromatic sensation only. This theory of Schultze may be seen as the third major paradigm shift in vision research. It entailed a profound new insight into visual processing and generated basic questions about differences and similarities as well as about possible interactions between the information processing of the two different receptor systems.

EVIDENCE IN FAVOUR OF THE THEORY

Schultze's hypothesis that the retina contained two basically different kinds of receptor was based on extensive, comparative histological studies. Although he was aware of the fact that the structure of a receptor type in different species, and even in the same retina, may differ markedly, he found rods and cones in general to differ both with respect to structure and occurrence.

ALL PRINCIPLE HUES MAY BE OBSERVED IN SCOTOPIC VISION

A test of the prediction was carried out by Stabell and Stabell (1965). The successive phases of the experiment were as follows: (1) Dark adaptation for 30 min. (2) Pre-stimulation extrafoveally for 30 s using one of several colour filters in front of the eye at an intensity 1 log unit above the specific-hue threshold measured for the filter used. (3) Dark adaptation for 30 s. (4) Test stimulation at scotopic intensity levels in the pre-stimulated area using a green monochromatic test light.

Pre-stimulation with a red colour filter produced a blue colour upon the scotopic test stimulation, a yellow pre-stimulation filter produced blue or blue-violet, a green filter, blue-violet or violet, while blue and blue-green filters did not produce any chromatic effect upon scotopic test stimulation.

The reason for the failure to produce red, yellow and green scotopic contrast colours soon became apparent in a follow-up study where it was found that by increasing the level of chromatic adaptation (i.e. increasing the time of pre-stimulation, reducing the time interval between pre- and test stimulation, and increasing the size of the pre- and test-fields) scotopic red, orange and green colours could be produced. Thus, pre-stimulation with a green-blue colour produced a red colour upon test stimulation at scotopic intensities, a blue colour produced orange, and a purple colour produced a blue-green colour.

CONTRIBUTION OF J. J. MCCANN AND J. L. BENTON

Soon after the discovery that rod signals may initiate all kinds of hue, a number of research workers contributed to the further exploration of this phenomenon.

Firstly, McCann and Benton (1969) convincingly demonstrated that rods had the ability to interact with the long-wave cones (L-cones) and thereby produce a multicoloured image. This was illustrated by first illuminating a multicoloured paper with a 656 nm monochromatic light at an intensity level just above the colour threshold in order to activate only the L-cone mechanism. Thereafter, they superimposed a monochromatic light of 546 or 450 nm that activated only the rod receptor system. Adding the scotopic light dramatically changed the colour of the display; red, orange, yellow, blue-green, brown, grey and black could be seen in the display.

A more sophisticated and detailed study of this rod-cone interaction colour effect was later reported by McKee, McCann and Benton (1977). To produce the multicoloured display, a transparent photographic picture was taken both through a red and a green filter and then combined. The display was then illuminated with a red 656 nm monochromatic light at an intensity just above the colour threshold and by one of ten monochromatic lights selected from the 420–600 nm region of the spectrum.

As can be seen, the theories of light and dark adaptation proposed by Rushton and Barlow disagreed both with regard to the site of the gain-determining mechanisms and the question of whether light and dark adaptation were equivalent.

In the 1965 version, Rushton held that the two processes were controlled in an AGC pool located centrally to the photoreceptors, while in the 1972 version he suggested that dark adaptation mainly occurred in the receptors and light adaptation in horizontal cells. In both versions he held that light and dark adaptation were quite different processes.

The noise theory of Barlow, on the other hand, presumed that light- and dark-adaptation processes were equivalent and located mainly in the receptors – that light adaptation mainly resulted from statistical fluctuation of photons of the background light, while dark adaptation was controlled by photon-like events, possibly originating from photoproducts as a stream of events fluctuating randomly like the photons.

During the last two decades the presumption that sensitivity regulation is basically controlled by photochemical processes located in the outer segment of the receptors has been replaced by much more complex theoretical conceptions. A recent review paper on gain controls in the retina by Dunn and Rieke (2006) gives an excellent illustration of this development. The authors argue that there are multiple retinal gain controls (i.e. adaptation mechanisms) that adjust sensitivity to different aspects of the light stimulus such as changes in mean intensity, variability about the mean (i.e. contrast variability) and spectral composition, and that the gain controls have diverse temporal and spatial properties, serve different functional roles and are located at different sites in the retina. Indeed, an additional dimension of complexity is introduced in that the gain controls are assumed to interact with other computations carried out in the retinal pathways.

In support of their suggestion that gain controls of the cone system may operate at different sites in the retina, they present evidence that both small and large steps in mean intensity and contrast may alter the gain adaptation level of ganglion cells, while only large steps change the gain of the receptors. Strong support in favour of the suggestion that gain controls may operate at different sites had previously been provided by Ahn and MacLeod (1993) and Yeh et al. (1996).

On the discoveries of Boll (1877) and Kühne (1877, 1878, 1879) that rhodopsin in rods is engaged in reversible cycles of bleaching and regeneration, Parinaud (1885) had suggested that changes in visual sensitivity were due to variation in the amount of rhodopsin. This view had a great impact. Thus, for a long period it became generally accepted that the alteration of visual sensitivity in light and dark adaptation reflected changes in the concentration of the visual pigments and hence their capacity to absorb light.

RESULTS OF TANSLEY

Tansley (1931) appears to be the first to measure quantitatively the change in rhodopsin concentration during dark adaptation. She light adapted albino rats almost completely and then measured the quantity of rhodopsin extracted after varying times (from 2.5 to 1140 min) in the dark. The results obtained could be explained both by bimolecular and monomolecular reactions, although the monomolecular reaction was found to fit slightly better. In accordance with Parinaud's (1885) assumption, she obtained a striking similarity between the regeneration curve of rhodopsin of the albino rat and the dark-adaptation curve measured in humans. Hence, she suggested that the sensitivity during dark adaptation was proportional to the amount of rhodopsin present in the retina.

RESULTS OF GRANIT

This simple photochemical theory of dark adaptation, however, eventually met with serious difficulties. Thus, evidence put forward by Granit et al. (1938, 1939) strongly suggested that the amount of rhodopsin played only a minor role in sensitivity regulation during light and dark adaptation.

The finding that stimulation of rods alone may give rise to qualitatively different colour sensations came as a surprise, since it challenged the fundamental Principle of Univariance. This principle follows from Helmholtz's (1896) specific fibre-energy doctrine and implies that a given receptor or nerve fibre does not discriminate between variation in intensity and wavelength of a test light and hence mediates only one sensory quality. Accordingly, when only the rod receptor system is stimulated, variation in wavelength can be simulated by variation in intensity – they both produce variation in brightness only.

Indeed, the Principle of Univariance had been directly demonstrated by Graham and Hartline (1935). By analyzing the nerve impulses arising in the retina of the Limulus (horseshoe crab), where each photoreceptor is linked with a separate nerve fibre, they found that the variation in the response of the single fibres with wavelength could be simulated by suitably adjustment of the incident light energy. Thus, when the intensity was suitably adjusted, any test wavelength could be made to evoke the same frequency of impulses from a given receptor cell. Hence, it appeared that single retinal receptors alone had no power to discriminate between wavelength and intensity.

How, then, could the Principle of Univariance be reconciled with the new discovery that test stimulation of rods may give rise to all the principle hues of the spectrum? An answer to this question became apparent when it was discovered that the scotopic hues were due to rod-cone interactions.

ROOTS OF THE DUPLICITY THEORY OF VISION: ANCIENT GREEKS

The duplicity theory is the most basic and comprehensive theory within vision research. Yet, there has been no attempt to describe its developmental history. In the present work, therefore, our aim has been to throw some light on this dark area in the history of science. As will be seen, the duplicity theory is not an old, static, antiquated theory dating back to Schultze's (1866) original formulation of the theory, as is generally held, but is a living body that expands and deepens as new knowledge of the rod and cone systems is obtained.

The beginning of the scientific study of vision may be traced back to the Ancient Greeks. However, due to an almost complete lack of knowledge about optics and sensory information processing at that time, the Greeks made two serious mistakes in their functional interpretation of the visual system. Thus, they generally held that (1) the crystalline lens of the eye was the most important organ of vision, being the actual sense organ, and (2) visual perception depended in a fundamental way on some sort of ‘rays’ that emanated from the lens toward the objects of the environment.

Both assumptions were accepted and adhered to in one form or other by many of the leading Ancient Greek philosophers and research workers. The most important among them, because of his strong and long-lasting influence on science in Western Europe, was Galen – also named Galenos (about AD 130–200).

It is apparent that neither of these models fit well with our description of the development of the duplicity theory. As regards Popper's model, none of the classical theories of Newton, Young, Schultze, Kühne and Hering was triggered by an attempt to falsify or refute current hypotheses or theories. Newton's starting point was the rectangular form of the prismatic solar spectrum he observed when he looked at its beautiful colours; Young based his theory on the old and well-known fact that three pigments were sufficient to produce every object colour; Schultze reflected on the fact that nocturnal and diurnal animals tended to have rod- and cone-dominated retinas, respectively, and that colours were absent in night vision; Kühne's theory was instigated by the great discovery of Boll that rhodopsin bleached in light and regenerated in the dark; and Hering was spurred on by Mach's psychophysical maxim. Thus, there is little evidence of long, fruitful falsification periods ending with falsification of the most successful hypothesis and the development of a new, better one triggered by this last falsification.

It is evident that Maxwell (1855, 1860) provided conclusive evidence in support of Young's trichromatic colour theory by demonstrating that three standard spectral lights were sufficient to produce all spectral colours, but his experiments represented a confirmation of an already existing theory and not a falsification. It should also be noted that Young's theory was immediately accepted by the relevant scientific community without much debate following the discoveries of Maxwell.

HECHT'S PHOTOCHEMICAL THEORY

The finding of Loeser (1904) that cones also had the ability to increase their sensitivity during early dark adaptation was confirmed by Hecht (1921/1922). He found that cones could increase their sensitivity markedly even during the first few seconds after bleaching.

More importantly, however, Hecht developed a photochemical theory for dark and light adaptation of rods and cones that had a strong influence on a whole generation of research workers. Certainly, he has played a central role in the developmental history of the duplicity theory. In a series of papers he provided an array of evidence supporting his photochemical theory (see Hecht, 1919/1920a, b, c, 1921/1922). In its essence and in its most simple version, the theory runs as follows: light acts on a photosensitive substance S and decomposes it into two precursors called P and A. The sensitivity of the eye, then, depends on the concentration of these precursors, not on the quantity of the photosensitive substance S. Thus, the model states that the amount of fresh precursors necessary for a threshold response is always a constant fraction of the amount of the precursors already present in the system. Hence, dark adaptation was thought to depend on the regular decrease in the concentration of the residual precursors present in the sensory system. This decrease was assumed to proceed according to the dynamics of a bimolecular reaction, to be independent of light stimulation, and, in accord with Kühne's (1879) ‘Optochemische’ hypothesis, to result in a reformation of the photosensitive substance S. Thus, the model states that the amount of fresh precursors necessary for a threshold response is always a constant fraction of the amount of the precursors already present in the system.