Skip to main content
Log in

Seeing with color: Psychophysics and the function of color vision

Synthese Aims and scope Submit manuscript

Cite this article


What is the function of color vision? In this paper, I focus on perceptual phenomena studied in psychophysics and argue that the best explanation for these phenomena is that the color visual system is a perceptual enhancement system. I first introduce two rival conceptions of the function of color vision: that color vision aims to detect or track the fine-grained colors of distal objects and scenes (Seeing Color) and that it aims to help organisms discriminate, detect, track and/or recognize ecologically important objects, properties, and relations more directly (Seeing with Color). I then discuss two kinds of systematic perceptual phenomena investigated by psychophysicists: approximate color constancy and color induction. I argue from the premise that Seeing with Color better accommodates and explains these phenomena to the conclusion that it is the conception that an empirically-guided philosopher of color ought to adopt.

This is a preview of subscription content, log in via an institution to check access.

Access this article

Price excludes VAT (USA)
Tax calculation will be finalised during checkout.

Instant access to the full article PDF.

Institutional subscriptions


  1. I am inspired by M. Chirimuuta’s language here. Chirimuuta writes that “color vision doesn’t help us see the colors of things; it helps us see things” (2015, p. 86) and that “color perception should be examined in terms of what seeing with color allows one to do” (2017, p. 164).

  2. I am not suggesting that the relevant sciences will directly answer the function question; empirical data requires interpretation, which is a task in which both scientists and philosophers can participate. For discussion, see Wright, 2010; Chirimuuta, 2017.

  3. Hilbert argues that we need to use human color vision as a paradigm case to “[break] into the cycle” to explain what color vision amounts to (1992, p. 364). For criticism, see Thompson, 1995a, pp. 7–10.

  4. This is essentially what Cohen (2015) suggests in his review of Chirimuuta (2015). It is true that, in neuroscience, there are two relevant hypotheses concerning the cortical mechanisms of color vision: (i) that chromatic information is processed in isolation of other visual attributes, and (ii) that chromatic information is processed jointly with those attributes. The debate largely boils down to the question of how a cortical ‘color cell’ is to be defined. Conway et al. (2002) adopt a strict criterion: only cells that respond exclusively to colored stimuli are color cells. Many others, however, maintain that any cortical cell that subtracts cone inputs deserves the label (e.g., Gegenfurtner, 2003).

  5. Functional analysis is often seen as a means for ascribing content to the color visual system (Hatfield, 1992, p. 492). Philosophers might engage in “Wrightian” etiological analysis (Wright, 1973) and seek to determine the selected function(s) of color vision, or they might adopt a Cummins-style approach (Cummins, 1975) and seek to uncover the specific role(s) color vision plays in our perceptual and cognitive economy. In both of these approaches, the function of color vision is conceived of as the aim of the system (think of a heart pumping blood), distinct from mere side effects and other accidents (think of a heart making heart sounds). Answering the function question generally relies on IBE-type reasoning: one articulates the competing conceptions of the function of color vision and determines which one best accommodates and explains the available evidence.

  6. We might need other considerations, including a priori ones, to arbitrate between these alternatives, but the space of plausible options has already been substantially narrowed. This is the benefit of the empirically-guided approach.

  7. Saying that the function of color vision is to help perceiving organisms see faster and better in general allows for the existence of multiple sub-functions, all of which enhance perception in some relevant way.

  8. That said, it would also be possible to advocate a hybrid conception and maintain that one of the functions of color vision is to track stable fine-grained color properties and another is to enhance perception in a more general sort of way. I discuss this option in Sect. 6.2.

  9. Light sources can also differ in other ways. For example, direct sunlight has a much higher intensity than standard artificial lights.

  10. For a helpful review, see Foster, 2011. For discussion of this approach, see Hatfield, 2009, pp. 184-5, 193.

  11. The color constancy index is defined by CI = 1-b/a, where “a” is the distance between the test patch and an ideal match and “b” is the deviation of the observer’s judgment from the ideal match, when plotted into some appropriate color space (see Arend et al., 1991, p. 665; Foster, 2011, Fig. 4).

  12. See e.g., Smithson, 2005, p. 1329; Foster, 2011.

  13. As Arend and Reeves write: “Color constancy was weak for our hue matches (direct sensory representation), although two of the three observers could, if required, approximate the latter type of color constancy (the paper matches)” (1986, p. 1749). This suggests that observers are generally capable of separating the two types of judgments. It is an interesting question whether the “paper match” condition measures a genuinely visual experience (a second perceptual mode in addition to the hue and saturation mode) or a cognitive judgment instead, as Hatfield (2009) suggests. If the latter is true, then there is reason to be suspicious of the use of paper match data as evidence in philosophizing about color constancy. That said, the debate surrounding this issue is intricate and I will not rehearse it here. In any case, even the constancy of perceptual judgments about the way things are is only approximate (e.g., Foster, 2011, Table 1). When it comes to the other type of constancy, i.e., the constancy of how things phenomenally appear, adaptation to the illuminant likely plays an important role. For example, Arend (1993) reports substantially higher constancy indices in maximum-adaptation conditions than in brief-adaptation conditions. But, again, even in the maximum-adaptation conditions, constancy is still far from perfect (CI ≈ 0.7).

  14. Foster (2011, Table 1) provides a convenient overview of results from experimental testing of constancy levels tabulated against experimental method, stimulus configuration, illuminants, judgment condition, experimental apparatus, etc. The reported constancy indices vary from 0.11 to 0.92.

  15. Changes in the perceived saturation and brightness of the target can also be induced. ‘Saturation’ refers to the vividness of perceived color. A single hue comes in different degrees of perceived vividness, ranging from completely desaturated grey to fully saturated pure color.

  16. The effect increases with the chromatic difference between the target and surround up to a certain point (e.g., for one particular surround chromaticity (0°, +L−M) the maximum effect was achieved at a distance of around 45° in color space) after which it decreases again (Klauke and Wachtler, 2015, p. 3).

  17. This is an issue Fechner may have missed because his targets were grey (achromatic). Note that Fechner’s complementarity hypothesis and the direction hypothesis make identical predictions in many familiar test cases (e.g., when the target is grey or when its color is complementary to that of the surround), as Ekroll and Faul (2012, p. 109) point out.

  18. The first type of effect is known as the “Bezold effect,” after von Bezold (1876). Akiyoshi Kitaoka’s illustrations, which involve more complex spatial patterns, can be found here: See also Shevell and Kingdom, 2008, Fig. 2.

  19. The most prominent theories of the evolution of primate trichromacy link the red-green dimension in primate color vision to the feeding strategies of our ancestors: the “frugivory hypothesis” appeals to an improved ability to discriminate ripe fruit against dappled background foliage (e.g., Mollon, 1989) and the “folivory hypothesis” appeals to an improved detection of nutritious young leaves that are often reddish in the tropics (e.g., Dominy and Lucas, 2001).

  20. E.g., Maloney and Wandell, 1986; Matthen, 1988; Poggio, 1990; Hilbert, 1992; Tye, 2000; and Byrne and Hilbert, 2003. Most of the philosophers who champion this view take (surface) color to be either identical to, supervenient on, or otherwise straightforwardly dependent on surface spectral reflectance.

  21. The advocates of the comparative approach would add that there might also be important interspecies differences in the degree of constancy. It might even be that the color vision of some animals displays no constancy at all. Hilbert, in his blatant anthropocentrism, would argue that such animals lack color vision altogether (see 1992, pp. 363–4).

  22. For empirical evidence, see Bramão et al., 2011.

  23. Chirimuuta makes a similar case against Michael Tye and other color “realists” when she writes that painting deviations from perfect constancy as inefficiencies of the reflectance recovery process “no longer supports the idea that in virtue of having properly working constancy mechanisms we are thereby gaining perceptual access to the fine grained, physical color of things” (2015, p. 56).

  24. For discussion, see Mausfeld, 2003.

  25. Here is my reading of how the illuminant estimate might be retained as part of the visual representation of the scene in Hilbert’s view: when we view scenes, our visual systems compute both illuminant colors and surface colors (supposedly the illuminant color is estimated first, and this estimate is then used to solve for the surface spectral reflectances of objects). The estimates are then somehow combined to produce a conscious visual experience of “how an object is illuminated” (2005, p. 151). Hilbert suggests, somewhat mysteriously, that “we need to not assume that any precise representation of the way an object is illuminated is visually represented” (Hilbert, 2005), but the conscious visual experience of how an object is illuminated is supposed to account for the deviations from constancy at the level of chromatic appearance.

  26. If the “stability” that Hilbert appeals to is supposed to be the constancy of perceptual judgments about the way things are, recall that this kind of constancy has also been shown to be only approximate.

  27. This is in line with Arend’s (1993) observation that maximum adaptation to the illuminant facilitates constancy.

  28. Consider the task of identifying the ripest strawberry in good light. Here our color vision plausibly guides our decision-making, and a proponent of Seeing Color maintains that it does so by tracking fine-grained color. But why think that the color visual system needs to track fine-grained color to help track ripeness? When it comes to strawberries, we know that redness of appearance correlates with ripeness. Provided that our chromatic experiences preserve this correlation (this requires some stability but not perfect), our color visual system can facilitate the successful completion of the identification task. The same state of ripeness need not (and often doesn’t) correspond to the same chromatic appearance in different perceptual conditions. Because perfect constancy is neither needed nor observed, the fact that we can complete such tasks with relative ease gives no reason to conclude that the color visual system aims to track fine-grained color. The best explanation for the role that color vision plays in these tasks is simply that one of the goals of the system is to help us track ecologically important properties in our environment.

  29. Sometimes this is made explicit. For example, Hilbert begins with the thesis that the function of the visual system, on a general level, is to “extract information about the properties of distal objects from the structured light in the environment” (1992, p. 360). Then, honing in on color vision, he goes on to claim that “the relevant function that provides the criterion for possession of color vision is a function defined in terms of the visual acquisition of information about some distal property” (Hilbert, 1992, p. 362). But there are obvious issues here. First, the claim about the function of color vision need not follow from the claim about the function of vision. Vision as a whole might be in the business of extracting information about the distal world, but not all the visual qualities we experience need to correspond to such extracted information; some might be how the extracted properties are experienced. Second, the claim about the overall function of the visual system can be—and has been—challenged. For example, Hatfield (2009) argues that the function is to guide action via perceiver-relative phenomenal structures. See also Purves et al., 2015.

  30. Some researchers have suggested that there is a tight connection between color induction and perceptual grouping. Xian and Shevell note that “several perceptual properties, such as depth, form, and brightness, which affect chromatic induction, also play a role in perceptual grouping” (2004, p. 383). King (2001) suggests that assimilation is tied to the perception of one “whole” and simultaneous contrast to the perception of two separate “wholes.”

  31. For my purposes here, it is important to show that color induction sometimes aids perceptual grouping and results from bottom-up processing, though this is consistent with the idea that there could also be top-down influences on assimilation (see e.g., Fuchs, 1923).

  32. Klauke and Wachtler note that in natural environments objects are generally illuminated by (yellow) sunlight or (blue) skylight. Because illumination within natural scenes tends to be uneven, the authors suggest that induction effects brought on by “shifting the gaze between differently illuminated areas” might help achieve color constancy (2015, p. 8).

  33. The researchers themselves take this to indicate that the contrast mechanisms operate at low levels of visual processing, before the image segmentation mechanisms based on computing relative motion or depth (Hurlbert and Wolf, 2004, p. 154).

  34. The sub-functions falling under Seeing with Color might not always be simultaneously fulfilled either. A particular color experience could help the perceiver with scene segmentation but not with property identification, for example. But such “conflicts” are easier to deal with than the conflicts between Seeing Color and Seeing with Color. For example, a color experience could be said to be correct when some pre-specified threshold of usefulness is reached, e.g., when the color visual system contributes to some/most of its functions.

  35. Assume that, under the hybrid conception, a color experience is correct when it accurately represents fine-grained color and/or contributes to its enhancement function. Because there is so much variation at the level of fine-grained color experience, the first function is rarely fulfilled. The correctness of color experiences then depends mainly on the fulfilment of the second function. But if the first function is rarely fulfilled, then why consider it a function of color vision in the first place? I have argued that Seeing with Color can account for approximate color constancy (i.e., the empirical phenomenon motivating Seeing Color) and so, for simplicity reasons alone, it seems that we are justified in adopting Seeing with Color.

  36. Tye’s modification is motivated by a desire to specify the colors of objects in the face of widespread interpersonal variation in color perception. Tye assumes that interpersonal variation doesn’t extend to the perception of coarse-grained hues. Cohen et al. (2006) argue that he is wrong.

  37. Some other philosophers (e.g., Gert, 2017) advocate this view.


Download references


The research for this paper was supported by the Kone Foundation. I am grateful to Gary Hatfield for commenting on multiple drafts. I also thank Zab Johnson, Lisa Miracchi Titus, Quayshawn Spencer, members of the MIRA group at the University of Pennsylvania, conference participants in Milan, Almería and Warsaw, and two anonymous reviewers for their helpful comments, questions, and suggestions.

Author information

Authors and Affiliations


Corresponding author

Correspondence to Tiina Carita Rosenqvist.

Ethics declarations

conflicts of Interest

There are no conflicts of interest.

Additional information

Publisher’s Note

Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

Rights and permissions

Springer Nature or its licensor (e.g. a society or other partner) holds exclusive rights to this article under a publishing agreement with the author(s) or other rightsholder(s); author self-archiving of the accepted manuscript version of this article is solely governed by the terms of such publishing agreement and applicable law.

Reprints and Permissions

About this article

Check for updates. Verify currency and authenticity via CrossMark

Cite this article

Rosenqvist, T.C. Seeing with color: Psychophysics and the function of color vision. Synthese 202, 20 (2023).

Download citation

  • Received:

  • Accepted:

  • Published:

  • DOI: