Navigazione – Mappa del sito

HomeNumeri43ontologia dei coloriThe colours of things and the met...

ontologia dei colori

The colours of things and the metaphysics of colour samples

Olli Lagerspetz
p. 161-179


The essay explores the distinction between two main types of “colour grammar”, or ways of relating to colours: as samples and as features of a three-dimensional, materially heterogeneous and variably lit environment. (Realistic painting may be described as an interface where the two sets of colour concepts meet and conflict.) This duality gives rise to characteristic ambiguities, relating to the concepts of the same colour, real colour, and colour constancy. Hence we must give up the idea of specifying the one relation that always holds between “colour” and “physical properties”.

Torna su

Testo integrale


1Colour judgments occur in at least two important types of context, each involving colour discourses or “grammars of colour” of its own kind. The aim of the present essay is to argue that the clarification of their relations should be of central importance for the philosophy of colour.

2Colours are judged in or by means of colour samples – to name an obvious example, the chips of a colour chart, which should ideally be viewed in standard light. Colour samples may be matched and analysed both optically and in terms of their physical and chemical properties. The sample represents a certain colour, thus what is true of perceived colour relations between particular samples will be true of the relevant colours in general. We will not only be able to say, “this sample looks darker than this one”, but also, “this pigment in general is darker than this one” and, “this colour is darker than this one”. For instance, since saturated yellow is lighter than saturated red, then also a saturated yellow sample will be perceived as lighter than a saturated red sample in normal light. The study of samples involves the attempt to control and manipulate the viewing conditions – for instance, by taking the samples out of doors to study them in daylight, or by applying well-defined normal conditions in a laboratory setting.

  • 1  Hacker 1987: 108.
  • 2  Carruthers 2000: 138.
  • 3  Marie McGinn addresses a closely related question – the relation between the grammars of our abstr (...)

3Colour samples have a special role in philosophical debate, because that debate frequently involves attempts to define colours in terms of samples viewed in normal conditions by normal observers. For instance, Peter Hacker, who defends the existence of colours as objective properties of physical objects, states that the meanings of colour words are explained and defined by reference to samples that give us «the standard of correctness […] for the application of the term explained»1. Thus “red” is whatever matches the red sample in adequate light. Peter Carruthers, whose views otherwise are very different, gives us exactly the same description of how the concept red is to be explained2. The natural reason for such congruence is that the sample represents what appears to be a clear case of the property or concept that the authors think the discussion should be about, namely, colour – red, yellow, and so on. But we are left with a question that is, most of the time, not addressed: of how the study of samples relates to our ability to recognise colours outside the specific activity of matching samples3.

4Colours also enter our lives as the colours of objects in a three-dimensional environment, interacting with differences of illumination, shadows and highlights, and they are assessed with the general character of the relevant object in mind. There the relation between the colour of the object and its appearance to the viewer is less clear-cut. Differences between these two types of discourse introduce characteristically interrelated ambiguities in central colour concepts, to be discussed below – for instance, “the same” vs. “different colour”, “lighter” vs. “darker”, “real” vs. “apparent colour”, and “normal viewing conditions”.

5Experimental colour research deals both with samples and with environmental impacts on colour perception. Colour scientists have been conscious for a long time of many ways in which data from research on isolated samples are at variance with colour perception in a natural environment.

  • 4  Lee 2005: 50-51.

6In philosophical literature, the ambiguities of colour are in part glossed over because of the pre-eminence of the colour picture as the model for visual perception. Realistic pictures are treated as models of what meets the eye, or sometimes, of inner visual scenes4.

7My argument is that the picture constitutes an interface between two ways of talking about colour. Colours in the picture can either be discussed in terms of the colour properties of the surface, or in terms of the colours of the objects that make up the scene depicted. For each part of a painted canvas, a colour value can be assigned and its physical properties can be discovered. In other words, it can be treated as a colour sample. But it also represents some element of a visible environment. This may lead to the false impression that the samples discourse might serve as a blueprint for an eventual complete account of the relations between colour descriptions and physical descriptions, for all cases of normal visual perception.

  • 5  Thanks to Yrsa Neuman for helping me focus my paper more definitely on this theme. For discussions (...)
  • 6  This paper follows the general practice of defining “physical properties” as properties that have (...)

8Such ideas are rooted in an implicit or explicit approach that may be called a metaphysics of colour samples5. On that conception, visual perception is the perception of a scene, somehow ultimately consisting of coloured patches, each of which can be treated as a sample of a specific colour. Thus each part of the scene has a determinate colour value. Its colour value might be judged roughly as if we were to cut out a piece of painted canvas and study it as a colour sample. Since physical properties – for instance, reflectance properties – of colour samples an be determined, the description of the visual scene as a collection of samples would constitute a first step of an all-embracing account of the relation between colours and physical6 properties.

9My argument below suggests that there is no such one relation. We will not be able to isolate a unitary class of “colour phenomena” to be brought into law-like connections with “physical phenomena”. Colour descriptions involve several grammars that are, in characteristic ways, linked to other features of our lives in a three-dimensional and materially heterogeneous environment. Any one account, when successful, applies to the behaviour of colours in a specific context. There is room for accounts of colour in terms of physics – for instance, in connection with research on colour reproduction – but philosophers should be wary of generalising from those cases to all situations where colour concepts are employed.

10The main text of this essay will fall into three parts. There is a description of the samples discourse, including general comments on the empirical research on samples. Secondly, there is a discussion of colour judgments about objects in three-dimensional environments. Finally, there is a discussion of the role of the colour picture.

1. The Study of Samples

1.1. The Colour Sample

  • 7  In other words, the argument may be made in the context of either subtractive or additive colour p (...)

11By colour sample, I will mean either a coloured light or a surface of unitary colour, which is supposed to be used by looking at it in uniform light. Most of my discussion will concern the properties of coloured surfaces, but a very similar argument could be formulated by using examples of coloured light7.

12Our interest in colour samples is accentuated in our handling of objects that we specifically choose because of their colour. They are objects we buy or use because of the fact that they are deliberately made to have a certain colour. Thus the study of samples is particularly relevant in our relation to such things as fabrics, wallpapers, pigments, and printed matter – and, in the case of samples of coloured light (instances of additive colour mixing), TV and computer images. Experimental research on samples is generally concerned with colour reproduction. For obvious reasons, colour reproduction plays an important role in electronic, graphic, textile, and other industries. The task is, typically, to match two samples in an optimal way, striking a balance between degree of visual equivalence and production costs.

  • 8  For an overview, see Rydefalk and Wedin 1997. Here, and generally in my discussion of the technica (...)

13The experimental research discussed in this section addresses the question how perceived colour is a function of micro-level properties of the sample8. The experiments have addressed such questions as which of two surfaces is perceived as lighter, more saturated, bluer etc., in standard viewing conditions, and how physical properties of the one might be changed in order to make it look the same as the other.

14The important thing for the experimental research is that both the perceived colour and the reflectance properties of the surface should be determined in an unequivocal way. In given illumination, a normal human observer will perceive this colour chip as, say, dark blue, with specifiable values for the three visual variables hue, saturation, and brightness. The spectral composition of the light reflected from it can also be measured exactly. When the spectral composition of the incident light is known, this gives us a specification of the absorption-reflection pattern of the surface. With the two sets of values thus established, it will be possible to look for law-like relations between the percept (perceived colour) and the stimulus (physical qualities of the surface).

  • 9  Rydefalk and Wedin 1997: 3.5.2 (p. 3-7) and 3.1 (p. 3-2).

15Experimental data on relations between spectral power distribution and perceived colour are the basis of standardised representations such as the CIE 1931 Standard Colorimetric Observer. It sums up correlations that make it possible, assuming normal human trichromatic vision, to predict the perceived colour of a surface sample from the spectral composition of the radiation it reflects (and similarly, to predict the colour of a light from the spectral composition of that light)9.

  • 10  This is known as metamerism. See e.g., Rydefalk and Wedin 1997: 3.7 (pp. 3-19 to 3-20). Any colour (...)

16The general empirical result is that, while the same spectral power distribution of radiation always gives the same visible colour if the conditions are kept exactly the same, the reverse is not true: we cannot predict the spectral power distribution of radiation from the colour that we see. The same visible colour may be achieved by mixing different pigments or illuminants, giving rise to different spectral power distributions10.

17The existence of fixed relations between physical properties of the sample and its appearance should not be treated as a radically modern discovery, even if the details are now much better known than previously. It is no more than natural to expect similar materials to have similar colours. We know there are recipes for pigments, including very ancient ones. Some colours even get their names from the substances from which their pigments are traditionally derived.

1.2. Real Colour and the Role of Standardisation

  • 11  Rydefalk and Wedin 1997: 3.1 (p. 3-2).

18However, the experimental results are of local character11. They presuppose standard conditions. I would argue that the significance of these discoveries does not lie in their being first, imperfect approximations of some future comprehensive theory relating “colour” to “physical properties”. They represent solutions to specific problems, important in their own right and capable of development without committing the inquiry to any ontological position about colour.

19The standard conditions applied here are obviously not the usual conditions for human colour vision. The research design for the study of samples is not even supposed to approximate the conditions that obtain in most natural settings – even less, to constitute a future model for all possible cases of colour perception.

  • 12  Rydefalk and Wedin 1997: 3.1 (p. 3-2).
  • 13  The CIE 1964 standard uses a 10o field instead, giving slightly different results. See Rydefalk an (...)

20The lawlike relations recorded in CIE 1931 Standard Colorimetric Observer issue from the following test design and from others comparable with it12. The surface sample should be something like a chip in a colour chart: an opaque, mat, flat surface of unitary colour. It should occupy two degrees of the visual field, framed by an area of standard grey13. It should be viewed in indirect light with the composition known as D65, which approximates daylight in parts of Europe (natural daylight has variations depending on both geographical area, season, and time of day). The need to control the illuminant is obviously due to the fact that reflected light is the part of the incident light that the surface does not absorb. Its spectral qualities thus depend on two factors: the light absorbing properties of the surface and the qualities of the incident light. Measurements of reflected light can inform us about the light absorbing properties of the surface only if measurements of the incident light are also available.

21On the whole, the point of the research design is to establish regularities that hold when the overall role of the sample in the general visual environment is controlled. The effect is that certain questions that emerge in more complex environments will not need to be asked. Philosophically, the important thing here is to see the ways in which the colour concepts used are determined by the research design; how certain distinctions that we make in other contexts are not applicable because of it. In particular, this is true of the distinction between real and apparent colour.

22In everyday situations, we recognise the risk that we may be mistaken about the real colour of a sample. This difficulty takes various forms and requires different solutions, some of which will be discussed later in this paper. In some cases, for instance, we must clean a surface in order to appreciate its real colour.

23The specific problems about “real colour” typically addressed in the study of samples are related to the effects of illumination and the colours of surrounding surfaces. The look of a surface depends in part on the quality of incident light. Moreover, it is a function of our adjustment to the general colouring of the environment.

24In this context, trying to see the real colour of a sample involves checking the sample in good light (which typically means indirect sunlight or incandescent light that approximates sunlight) and not placing it next to other strongly coloured objects. Thus the meaning of “real colour” here involves reference to “good light”. This is implicit in philosophical definitions of colour in terms of normal conditions. For instance, the colour red is often defined as the surface property that appears red to normally sighted human observers in normal conditions. This implies, conversely, that the recognition of normal or good light presupposes the ability to distinguish between real and apparent colours of objects. Good light is, in this context, light that lets the real colours show.

25These complications will not arise in the research design just described, because both illumination and surroundings are standardised. The illumination used is, by definition, good light, because it is in that light that the samples are to be matched. Similarly, no problem arises about the disturbing effect of the surroundings, since the standard grey around the colour chip is, by definition, “a neutral environment”. In this way, the standard conditions are designed to make the distinction between real colour and perceived colour disappear.

26The role of D65, or standard light, is ambiguous in this context. In strictly scientific terms, D65 represents a conventional standard among several possibilities. It is, in this respect, comparable with the Standard Metre as a measure of length. In physical theory, no such thing exists as the real colour of a sample, or the appearance that the sample ought to have if we were too see it correctly. Standard light helps us discern many reflectance differences between surfaces, but it also makes others disappear. (For this reason, one should choose a makeup that matches the evening dress in the likely viewing conditions – not daylight or D65.) On the other hand, it is not a coincidence that D65 is the most frequently used standard. It approximates the light we would use in most (though not all) everyday occasions where we want to study a sample – of paint, fabric, or wall paper – in “good light”.

27Thus experimental designs for studies of the reflectance properties of surfaces are constructed with specific practical interests in mind – interests whose intelligibility presupposes the essentially nonphysical notions of real colour and good light.

28The philosophical interest of this lies in the fact that results obtained in research on samples are sometimes falsely taken by philosophers to hold out the promise of establishing correspondences between specific colours as such and specific physical properties. “Colour”, on this philosophical conception, is understood as being independent of the context in which it is seen. But the sample is not free of context in this sense. Research on samples represents an attempt as far as possible to control the conditions in which samples are viewed and compared – not magically to make the context go away.

2. Colours in Three-Dimensional Environments

2.1. Aspects of Colour Constancy

29The concept of real colour is connected with the set of phenomena generally known as colour constancy. Our judgments of the colours of objects do not slavishly follow from the spectral properties of the light that meets our eyes.

30In a natural environment, the illumination is not fixed. Not only is there a difference between indoors and outdoors illumination, but also between daylight at different times or even moments of the day, and between illuminated and shaded parts of the visual environment. If our perception of the colours of objects were a direct function of the light that they reflect we would perceive them as constantly changing in colour. If colour vision is described as a process through which optical patterns are projected to the retina and then stimulate the rest of the visual system, questions will arise about how these changing patterns give us information about unchanging features of objects.

31Colour constancy is usually described as an adjusting mechanism by means of which “we”, or “our visual system” is able somehow to “discount” the changes of illumination typical of most visual environments.

32Three kinds of phenomena are at work here.

331) The perceived visual environment remains to some extent constant because of physiological adjustments. This involves the adjustment of the eye to lighting conditions. This always takes some time, hence we may initially have difficulties moving from outdoors to indoors and vice versa. More dramatically, afterimages are unexpected side effects of the physiological adjustment of the retina. This is a fairly uncontroversial class of phenomena that might be called physiological colour constancy.

  • 14  Lopes 1999. Lopes distinguishes between colour constancy and colour recognition. In his terminolog (...)

342) Secondly, when the visual scene as a whole has perceptibly changed, we are still able to recognise colours in the new situation. For instance, Dominic McIver Lopes prefers to speak here of colour recognition, not colour constancy, as we are conscious both of a general change and, simultaneously, of the permanence of colours in the changed environment14.

353) Thirdly, when there are local changes, we distinguish between the real colours of objects and various effects of highlights, shadows and other modifications of lighting. This is an aspect of our general familiarity with three-dimensional visual environments.

36Purely physiological adjustments should not be confused with the two other phenomena that should perhaps not be called adjustments at all – neither physiological nor even mental.

37One way to approach colour constancy is by trying to imagine an individual for whom it is lacking. For that individual, colours would be attributes of the patterns of light transmitted optically to his or her visual system. It would be correct to say that, for that person, “red” would be the same as the appearance of a red sample in standard light. Red objects in shadow would not count as properly red, while white objects placed in red light would do so. One consequence of this would be that the sameness of colour could not play any decisive role in that person’s identification of objects. Colours would, on the whole, play a much more marginal role in his or her negotiation with the material environment.

38The individual we are trying to imagine would not necessarily be physiologically deviant. His or her visual system might possess the normal adjusting mechanisms described above as physiological colour constancy. What we are imagining is someone for whom colour concepts have a different meaning. In part, the answer to how colour constancy is “possible” is admittedly a matter of identifying physiological adjusting mechanisms. But it also requires understanding the behaviour of our colour concepts; that is, understanding what we think of as constant when we think of the colours of objects as being constant.

2.2. Colour Recognition

  • 15  Lopes 1999: 426.

39Our recognition of the constant colours of objects does not require perceptual constancy. It is possible for us, at the same time, both to recognise the original colours in new lighting conditions and to see that they look different because of the light. As an analogy, Lopes points out that our recognition of the permanence of physical objects does not require that objects should have the same appearance regardless of distance and perspective. On the contrary, it is important that we should be aware both of their permanence and of the ways in which their visible shapes vary when seen from different angles15. The concept of looking the same has ambiguities that may cause some verbal confusion here. The visible shapes of objects change as we move in relation to them. But in another sense the objects do not look changed at all. On the contrary, we see their visible shapes changing exactly because the objects do not change. If their visible shapes were not to change as we move, this would indicate that the objects are moving as well. Similarly, if the quality of the light from a surface did not change with the illumination, that would mean the surface has either changed colour or is fluorescent.

  • 16  Westphal 1991: 26-28 and note 28 p. 27. This effect is also utilised in paper industry, when blue (...)

40When coloured light is selectively directed at a white spot, we may be fooled into thinking that the spot is coloured. But a general coloured illumination does not have this effect. For instance, as Jonathan Westphal points out, blue illumination in a room does not make white objects seem blue. Even under a blue illuminant, white objects will appear less blue than blue ones would do in the same conditions. In fact, blue illumination often enhances whiteness since it cancels out the yellowing typical of many white objects16.

  • 17  See Westphal 1991: 101-108.

41What we see here is a degree of constancy in colour relations within the visual scene throughout the changes of illumination. The light reflected from an object displays a regular pattern in how it deviates from the colour of the general illuminant. A coloured object darkens the incident light by absorbing parts of it17. Absorbed parts of the spectrum correspond to the complementary of the colour of the object. Thus, for instance, blue objects absorb a disproportionate amount of orange. To see the colour of an object is, from this point of view, to see the under-representation of its complementary. To express this schematically, “looks blue” means “looks bluer than …” or perhaps, rather, “looks less orange than …”. And the general illumination gives us the point of comparison.

42The constancy that we recognise in the colours of objects is not that of their absolute colour values but that of constant relative values. Perhaps analogously, our judgment of the size of an object is not based on the absolute size of the area it covers in our visual field, but on its size relative to other objects we see.

2.3. Colour Judgment

  • 18  Baxandall 1995: 66-67.

43Colour judgment is an aspect of our general ability to distinguish between constant and transient elements of a visual scene. For instance, we can tell the difference between a shadow and a stain. When we read a book in daylight, there is a strong luminance contrast between the parts that lie in light and the parts that lie in shadow; white paper in shadow reflects less light than a lit black object next to it18. In one sense of the expression, the “perceived colours” of the different areas of the page are different. We might even say that the shadowed area looks darker, but we would not expect anyone to be fooled into believing that its real colour is darker.

44It would be possible for someone to dye the page – or, for instance, part of a wall – in such a way that we think we are seeing a shadow on it. This is the principle of trompe l’oeil paintings. We recognise our mistake if we are allowed to move in the setting or manipulate the object. Trompe l’oeil paintings do not deceive us unless we are stuck in one place. Our ability to make these judgments demonstrates that we do not, in a natural context, judge the colours of objects by matching them against a timeless model where a given spectral power distribution always corresponds to the same colour. We judge them in a context of illumination and shadows.

  • 19  Arguably, this fact has an evolutionary explanation. The ability to discern objects has mostly a m (...)

45Colour judgments of the kind that dominate in everyday life presuppose implicitly the concept of an individual object, in other words, an understanding of parts of the visual environment as relatively constant and detachable “things”19. Colour variations that move (to an extent) independently of the object on which they are seen, are variations of lighting conditions. Shadows and highlights indicate the shape and position of the object, but the object has its own colour, which can be determined independently of any particular lighting.

  • 20  Wittgenstein 1978: I, §§ 48-50.
  • 21  Wittgenstein 1978: I, § 56.

46Things have colours, but considered as optical patterns, the colours of things and the colours of shadows and highlights are indistinguishable. The same stimulus (the same quality of reflected light) is produced both by a grey object in good light, a white object in shadow, and a highlight on a polished dark object20. Are they the same colour or different colours? Ludwig Wittgenstein addresses cases of this kind in his Remarks on Colour, where he notes «the indeterminateness of our concept of sameness of colour»21.

  • 22  Hyman 2007: 674-675.

47The ambiguity can be resolved, for some purposes at least, with the introduction of the concepts of surface colour and aperture colour22. Surface colours are ascribed to objects regardless of lighting conditions. A uniformly white wall has the same surface colour all over, regardless of shadows and highligts. An aperture colour appears when some part of the visual environment is viewed through an opening, or aperture. For instance, we may look at different parts of a white wall through a tube of grey cardboard, which makes the object unrecognisable and blocks out any cue about the ambient light. The part of the wall that lies in shadow will have the same aperture colour as a grey wall that lies in good light. The two areas have different surface colours (white vs. grey) but the same aperture colour (grey). This may also be described by saying that the aperture enables us to see colours of objects as if we were looking at colour samples.

2.4. The Real Colours of Objects

48In the previous sections, the sameness of colour has mostly been discussed as an optical complication, caused by the fact that our visual environment is three-dimensional and variously lit. Those narrowly optical considerations already indicate that the concepts of the same colour and different colours are dependent on the concept of real colour. Moreover, the concept of real colour has a close relation with the concept of an object or a thing.

  • 23  Lee 2005: 60.

49Outside the specific context of pictorial representation (to be discussed below), our comparisons of the colours of objects are typically made between objects that are similar in important respects. We compare two painted surfaces, two liquids, two flames, the eyes or complexions of two persons. The important thing is that their colours have a similar “mode of appearance”. Thus we may compare the colour of a glass of beer with a transparent piece of amber. Comparisons are much less clear-cut between dissimilar objects. Already the juxtaposition of mat and glossy surfaces will create difficulties23.

  • 24  Cf. Wittgenstein 1978: I, § 58.

50To borrow an example from Wittgenstein, it is not immediately clear whether we can find a paint that is the exact match of somebody’s eyes24. A skilful artist might paint a good likeness of an eye. But in such cases, there is visual likeness between two objects – the model and the depiction – not between two colours. A mat surface of unitary hue would not look the same as an eye. Light brown paint may be said to match light brown eyes, but there is some arbitrariness in calling any hue the exact match.

51There is also a difficulty here because the ideal conditions in which real colours are seen will be understood differently for different objects. It is not obvious that bright sunlight is the illumination in which the real colour of a person’s eyes is most easily appreciated.

52Not only illumination is important. For instance, in order to judge the real colour of fur we should turn it in our hands and look at it at different angles. The real colour of paint is the colour of dry paint. Sometimes the real colour of a thing is what is on the surface and sometimes what is under the surface. The latter is true, for instance, when a dyed surface is contrasted with the underlying substance. Considerations of this kind are important in our daily dealings with colours. The close relation between the nature of a given object and the ways in which its colour is determined and discussed is what, to an extent, accounts for the importance of colour judgments for us.

53On the other hand, considerations of this kind are likely to be sidetracked in philosophical debate, as they are seen as merely local complications of an otherwise unitary concept of colour. Colour is, instead, treated as the common optical feature that may be abstracted from the different individual cases. The reason is that the philosophical interest in those debates has a focus on the contribution of colours to vision (that is, on their role in creating the optical patterns deemed to be the proximal object of vision). This is a legitimate and important interest, also pursued in the present essay. But it may risk perpetuating the idea that colour recognition is only a matter of vision optically understood, instead of depending on a complex interaction with our other ways of understanding material features of things. It is this concentration on “pure” vision that contributes to the philosophical understanding of colour as something curiously detached from (other) physical qualities. And it contributes to the idea that this elusive relation is best studied by attending to colour samples – that is, objects that have no other important features except the fact that they are examples of a certain optical appearance.

  • 25  Lee 2005: 63.
  • 26  Lee 2005: 52.

54This kind of abstraction is important in one practical context, namely pictorial representation. It may be argued that the existence of a tradition of naturalistic picture making has played a guiding role in the development of a generalised conception of colour: the conception of colours as purely optical qualities, detached from other material features of objects25. As Alan Lee puts it, naturalistic pictures «accomplish a kind of reduction of the heterogeneity of visually discriminable qualities, reducing them to colours properly so-called»26.

3. The Colour Picture

3.1. The Realistic Picture as an Interface

55Philosophical interest in colour pictures is, at least to some extent, due to the idea that the study of pictures might give us a more manageable way to approach visual perception. A realistic painting or colour photo is treated as an approximation of what would be placed before a person’s eyes when he or she looks at coloured objects in three-dimensional space. The question “what colours do we see?” is approached by way of the question, “what pigments would we use in a realistic depiction of what we see?”.

56In a colour photo or a realistic painting in the traditional shading style known as chiaroscuro, shaded parts of an object are indicated as darkening in the corresponding part of the picture. Thus “what we see”, our “visual impression”, includes a darkened area in the place occupied by shadow in the visual scene.

  • 27  Wittgenstein 1978: III, § 79.
  • 28  Cf. Wittgenstein 1978: III, §§ 77-78.

57“Unconventional colours” – or surface features not captured in classifications of primary colours and their mixtures – involve a special challenge for the artist. This is true, for instance, of amber and gold. As Wittgenstein points out, the painter of The man with a golden helmet did not use gold paint to represent gold27. The artist (Rembrandt or, according to present knowledge, one of his students) created the effect of gold by means of an arrangement of yellow, brown, red, white, and black pigments. This highlights an ambiguity of the expression “this colour”. It means either “the colour of this part of the canvas” (the colour of this pigment – yellow, black, brown etc.) or “the colour of the object depicted on the canvas” (gold)28.

58In this sense, the realistic painting or photograph represents an interface between two types of colour discourse. We can speak either of the colours of samples – which could be studied in different kinds of light and from different angles – or of the colours of objects in a given illumination. The palette, with pigments neatly arranged on it, contains colour samples. By moving the pigment from the palette to the canvas, the artist represents real-life colours of objects. Black pigment can become the representation of some part of a golden surface. The colour of “this spot” may be described as either black or gold.

  • 29  Hyman 2007: 675.

59A different painting style, such as Egyptian art before the Ptolemaic period, would reproduce the real surface colours of each object depicted, representing white objects as white all over and grey objects as grey all over, and golden objects with unitary areas of gold leaf or yellow29. It takes a special effort not to do this in painting. We must ignore the identifiable objects before us and just try to see a visual pattern. One way to do this is by looking at individual parts of the scene through an aperture and trying to reproduce the aperture colour on the canvas. John Hyman argues that realistic painting involves matching aperture colour with pigment. As a result, the aperture colours seen on the canvas should be the same as the aperture colours in the scene:

  • 30  Hyman 2007: 675.

[W]hen shading is used, the aperture colours of depicted objects are the same as the aperture colours of the parts of the painting that depict them, as long as the aperture is small enough to prevent us from seeing the ambient light in the depicted scene, and the painting itself is uniformly and adequately lit30.

  • 31  Hyman 2007: 677.
  • 32  Newall 2006, Lee 2005: 61-62.
  • 33  This is not to imply that also the views described below should be attributed to Hyman.

60In principle, on Hyman’s view, the artist can reproduce a three-dimensional scene by matching, point by point, the aperture colour of its each individual part with pigment applied to the corresponding area of the canvas31. Visual correspondence between colour samples is thus a measure of realism: the sample obtained by isolating a given part of the visual scene is matched with a sample of pigment. There is some dispute here, to which I will return32. The important thing now is that this is the theoretical assumption in philosophy when colour images are treated as the paradigm of visual perception33.

  • 34  Westphal 1991: 147.

61Since each spot of the canvas may be given a determinate colour value (in standard illumination), this invites the idea that an inner visual scene could be similarly treated as an arrangement of coloured spots. Westphal refers to this idea as “the mosaic conception” of visual perception34. A complete philosophical analysis of our visual experience would achieve the same result on a conceptual level as the realistic painter supposedly does on the canvas: reducing a landscape to an arrangement of colour-patches. It would dispose of the concept of physical objects – such as golden helmets – and merely specify the individual colour value of each colour-patch.

  • 35  Wittgenstein 1978: I, §§ 60-61; III, § 53. For discussions of Remarks on Colour I, § 60-61, see Le (...)
  • 36  Wittgenstein 1978: I, § 61.

62Wittgenstein has this idea in mind when he imagines that we actually cut up a painting in pieces and use the pieces as parts of a pointillist jigsaw puzzle. The position of each fragment in the puzzle will decide whether it will be seen as, say, part of a grey surface, a white object in shadow, or a highlight35. If this is used as a model for an analysis of what we see, the concepts of shadow and highlight will be disposed of. We get a description of the individual pieces as “the colours of places in our visual field, which are independent of any spatial or physical interpretation”36.

  • 37  Wittgenstein 1978: III, § 108; III, § 58; III, §§ 68-69; III, § 262.
  • 38  The effect of adjacent colours for colour perception is also a central area of experimental resear (...)

63Wittgenstein’s objection to this is that the individual fragment will also be seen in some illumination and in some surroundings37. Should we take the piece of fabric to the window and study it there? Should we place it on a neutral background to avoid interference from other parts of the picture38? Thus there is no such thing as determining the colour value of an individual colour-patch on its own, regardless of the circumstances. The mosaic conception fails.

64In the case of the actual coloured piece of fabric, this objection may not be appreciated because the piece of fabric still has “a determinate colour value” in another sense. It has a given pigment which can, of course, be studied. But that would not be the case with the colours of places in the visual field or in an inner image. A piece of an inner image cannot be detached and taken somewhere else to study it in better light.

  • 39  Something like this idea appears natural if “the visual field” is construed as a field, i.e., as a (...)

65Thus the analogy between “the visual scene that we see” and a realistic picture creates the false impression that individual parts of our visual field might be detached from their contexts and analysed as colour samples39.

3.2. Realistic Colouring in Pictures

66Some difficulty here is due to our tendency to think of depiction – at least, of realistic depiction – as a matter of copying the specific colour values of the environment in another medium. In his discussion of the role of colour constancy to which I already referred, Lopes questions this idea.

  • 40  See also Lee 2005: 61-62. On the other hand, this also means that photos taken on a cloudy day wil (...)

67Lopes’ specific discussion concerns the representation of colours in realistic painting. In certain ways, our recognition of colours in a picture works analogously with colour recognition under modified lighting conditions. For various technical reasons, a painting typically cannot reproduce the full colour range of outdoor scenes. Pictures are usually hung indoors and will not be as brightly lit as the original scenes. Moreover, pigments – especially in watercolour – mostly have a narrower range of available hues than those represented in the visual environment. Pictures can rarely reproduce both the tone (brightness), hue, and saturation of the original. Instead, they tend to reproduce relations between the original colours on a narrowed-down scale40.

  • 41  Lopes 2005: 421-422. Cf. Newall 2006, Hyman 2007.

68Different colouring styles opt for different ways of striking a balance. For instance, in chiaroscuro, the colours of surfaces are toned up with white to represent lighter parts of the scene. Thus saturation is sacrificed for brightness. In Impressionism, shadowed parts of surfaces are sometimes represented with toned-down mixtures of the surface’s complementary, thus preserving saturation but sacrificing brightness contrasts41. Within each colouring style, we are still able to see the result as realistic. We judge the colouring, not by making one-to-one comparisons between individual parts of the canvas and the corresponding real objects, but by comparing colour relations between real-life objects with relations that obtain within the general colour situation of the picture.

  • 42  Also see Newall 2006.

69Consequently, the realism of realistic colouring is not dependent on the requirement that the reflected light from each spot on the canvas should match the reflected light from each spot of the objects depicted42. This contrasts with Hyman’s views that I have cited. It seems, then, that Hyman describes a possible case of some realistic depiction, perhaps even an ideal implicit in much of the realistic painting tradition, but not a requirement that must or indeed can mostly be met.

3.3. Colour and Pigment

70An interesting empirical study by Brou et al. illustrates the role of pigment for our understanding of “the real colour”. This is an unintended side effect of their main argument, the aim of which is to demonstrate the role of colour constancy in the perception of colours in objects. Their general conclusion, which is in line with the argument of the present paper, is that

  • 43  Brou et al. 1986: 80.

the perceived colors of things in the visual world do not depend slavishly on the light from each object, sensed independently of all other things in the world, but on a comparison of the lights from an object and its surroundings43.

  • 44  Brou et al. 1986: 80.
  • 45  Brou et al. 1986 say “the perceived colors of things” depend on comparisons. But does their point (...)

71That point is skilfully illustrated with the help of various “colour ‘illusions’”. Usually, we call something a colour illusion if it makes us see (or think we see) a colour that is not there. For instance, afterimages can be colour illusions in this sense. However, by calling the effects of their experimental arrangements ‘illusions’ (consistently in inverted commas), the authors do not wish to suggest that the effects are somehow misleading. They are just unexpected, given certain received notions about colour. We attribute one colour to a surface although in some sense we “ought” to see another44. As the authors point out, colour constancy counts as illusion in this sense insofar as colour concepts are falsely taken to apply to spectral power distributions of light regardless of the surroundings45.

  • 46  Brou et al. 1986: passim.

72Brou et al. also present an effect that does count as illusion in a more straightforward sense. They consider a figure where patches of grey pigment are applied near the four corners of a non-figurative picture46. The same pigment is used for each of the hexagonal patches. However, we perceive their colour values quite differently depending on the colouring of the areas immediately surrounding them. For the viewer, it is almost impossible to see that the same pigment is used in the one hexagon surrounded by orange and in the three others surrounded, respectively, by blue, green, and purple. (We may find it hard to believe this even after we have been told the facts.)

  • 47  Cf. Brou et al. 1986: 87. The authors present a physiological explanation that, insofar as I can s (...)

73Brou et al. point out that this illusion is also an effect of colour constancy. One way (mine) to describe the connection is, perhaps, as follows47. We tend to judge the colour of each hexagon in terms of how it appears in relation to its immediate surroundings. Looking at the picture as a whole, we tend to construe the large coloured areas occupying its different corners as coloured shadings or films laid over entire surfaces, from which the hexagons stand out. Hence, each hexagon appears to be tinted with the complementary of the colour that surrounds it.

74As viewers, we tend to look at the figure as if it was a depiction of objects of some kind. We instinctively take the task to be one of determining their colours, not one of comparing pigments. But, as Brou et al. understand the task they set for the viewer, the colours of the hexagons should be identified with their pigments on the printed page, not with the colours of any objects they might be seen as depicting. Thus there is a conflict between our natural way of understanding colour relations within the picture and the implicit task that we are expected to perform. We, as it were, import patterns of our perception of objects into our understanding of the nonfigurative image.

75Suppose the task really were to judge colours in a pictorial representation of objects. The result would be the same, but this time it would not count as illusion. The colour of an object in the picture would not need to be the same as the colour of the pigment that represents it. The same pigment may reasonably “stand in for” objects of different colour if the illumination of the depicted scene varies across the picture. Now we would be expected to see that the four hexagonal objects are not of the same colour. Our ability to see this belongs to our ordinary exercise of colour judgment.

76Thus colour constancy as a whole is not an illusion, with or without inverted commas, even if it may give rise to illusions. The illusion is, in the present case, created by the ambiguity between “colour” as the colour of an object and “colour” as sample or pigment. “This is the same colour as this” may either mean, “these two objects are the same colour” or “these two samples have the same pigment”.

4. Conclusion

77The effect of this discussion is to question the idea of giving an account of “the” relation between colour and physical properties. That is, the idea of specifying the one relation (of identity, response dependence, supervenience, elimination, independence, or whatever) whereby what is said about “the one thing” corresponds or fails to correspond to what is said about “the other”. In particular, I have concentrated on the differences and points of contact between colours as samples and colours as features of a three-dimensional material environment.

  • 48  In this passage, I am developing a formulation by David Cockburn, in a commentary to an earlier pa (...)

78Our understanding of colour is to a degree conditioned by the fact that there are pigments; that is, by the existence of painting and dyeing. In a painted scene, the same colour, i.e., the same pigment, may represent objects that we would not correctly call, nor correctly see as, being the same colour in a real-life situation. Yet what is on the painted surface came from the same pot – it could be shown to have the same chemical analysis. Our understanding of colour is, also, perhaps, conditioned – though less directly in our normal life – by the fact that we can measure spectral power distributions of light. Similar complications arise there. A key point of dispute is: What role should such things play in the philosophical understanding of “the real colours” of objects48?

  • 49  An earlier version of this paper was presented at the Department of Philosophy research seminar at (...)

79The study of colour samples shows that, for instance, certain reflectance properties of surfaces can be identified with certain colours in standardised conditions. One might even say that, for some purposes in that context, colours are reflectance properties of surfaces. But the standardisation itself presupposes the essentially nonphysical concepts of real colour and adequate studying conditions. Thus the physical description is subordinate to a more fundamental understanding of the role of colours in a three-dimensional, materially heterogeneous, and variably lit environment. The study of colour samples does not substitute for that understanding but it has a particular place within it49.

Torna su


Baxandall, M.

– 1995, Shadows and Enlightenment, New Haven-London, Yale University Press

Brou, P. et al.

– 1986, The Colors of Things, “Scientific American”, 255: 80-87

Carruthers, P.

– 2000, Phenomenal Consciousness: A Naturalistic Theory, Cambridge, Cambridge University Press

Hacker, P.M.S.

– 1987, Appearance and Reality, Oxford, Blackwell

Hyman, J.

– 2007, Depicting Colours: Reply to Newall, “Philosophical Quarterly”, 57: 674-678

Lagerspetz, O.

– 2009, Wittgenstein on Colour Geometry and Colour Perception, in C. Gefwert and O. Lagerspetz (eds.), Wittgenstein and Philosophical Psychology: Essays in Honour of Lars Hertzberg, Uppsala, “Uppsala Philosophical Studies”, 55: 223-235

Lee, A.

– 2005, Colour and Pictorial Representation, “British Journal of Aesthetics”, 45: 49-63

Lopes, D.M.M.

– 1999, Pictorial Color: Aesthetics and Cognitive Science, “Philosophical Psychology”, 12: 415-428

McGinn, M.

– 1991, Wittgenstein’s Remarks on Colour, “Philosophy”, 56: 435-453

Newall, M.

– 2006, Pictures, Colour and Resemblance, “Philosophical Quarterly”, 56: 588-595

Rydefalk, S. and Wedin, M.

– 1997, Literature Review on the Colour Gamut in the Printing Process, Stockholm, STFI

Westphal, J.

– 1991, Colour: A Philosophical Introduction, Second Edition, Oxford, Blackwell

Wittgenstein, L.

– 1978, Remarks on Colour, Berkeley-Los Angeles, University of California Press

Torna su


1  Hacker 1987: 108.

2  Carruthers 2000: 138.

3  Marie McGinn addresses a closely related question – the relation between the grammars of our abstract “colour geometry” and of our everyday colour-language. McGinn 1991.

4  Lee 2005: 50-51.

5  Thanks to Yrsa Neuman for helping me focus my paper more definitely on this theme. For discussions of the theme, see McGinn 1991: 445-447, Westphal 1991: 147-152.

6  This paper follows the general practice of defining “physical properties” as properties that have an explanatory role in physical theories. In this sense, colours do not count as physical properties even if they obviously are familiar properties of middle sized material objects.

7  In other words, the argument may be made in the context of either subtractive or additive colour processes.

8  For an overview, see Rydefalk and Wedin 1997. Here, and generally in my discussion of the technical aspects of colour reproduction, I have profited from discussions with Associate Professor Jan-Erik Nordström.

9  Rydefalk and Wedin 1997: 3.5.2 (p. 3-7) and 3.1 (p. 3-2).

10  This is known as metamerism. See e.g., Rydefalk and Wedin 1997: 3.7 (pp. 3-19 to 3-20). Any colour can be visually matched with a suitable combination of any three pigments or light sources (with certain restrictions). Metamerism is enormously important for the graphic and electronic industries, as this helps to minimise the number of pigments or light sources and hence production costs. – Two samples with different spectral properties are metamers of each other under a given illuminant, when the two samples cannot be visually distinguished under that illuminant.

11  Rydefalk and Wedin 1997: 3.1 (p. 3-2).

12  Rydefalk and Wedin 1997: 3.1 (p. 3-2).

13  The CIE 1964 standard uses a 10o field instead, giving slightly different results. See Rydefalk and Wedin 1997: 3.5.2 (p. 3-9).

14  Lopes 1999. Lopes distinguishes between colour constancy and colour recognition. In his terminology, the items (2) and (3) here would count as instances of colour recognition.

15  Lopes 1999: 426.

16  Westphal 1991: 26-28 and note 28 p. 27. This effect is also utilised in paper industry, when blue pigment is added for extra whiteness.

17  See Westphal 1991: 101-108.

18  Baxandall 1995: 66-67.

19  Arguably, this fact has an evolutionary explanation. The ability to discern objects has mostly a much higher evolutionary survival value than any capacity for determining absolute spectral values of light.

20  Wittgenstein 1978: I, §§ 48-50.

21  Wittgenstein 1978: I, § 56.

22  Hyman 2007: 674-675.

23  Lee 2005: 60.

24  Cf. Wittgenstein 1978: I, § 58.

25  Lee 2005: 63.

26  Lee 2005: 52.

27  Wittgenstein 1978: III, § 79.

28  Cf. Wittgenstein 1978: III, §§ 77-78.

29  Hyman 2007: 675.

30  Hyman 2007: 675.

31  Hyman 2007: 677.

32  Newall 2006, Lee 2005: 61-62.

33  This is not to imply that also the views described below should be attributed to Hyman.

34  Westphal 1991: 147.

35  Wittgenstein 1978: I, §§ 60-61; III, § 53. For discussions of Remarks on Colour I, § 60-61, see Lee 2005: 56, McGinn 1991: 445-446, Westphal 1991: 147-151.

36  Wittgenstein 1978: I, § 61.

37  Wittgenstein 1978: III, § 108; III, § 58; III, §§ 68-69; III, § 262.

38  The effect of adjacent colours for colour perception is also a central area of experimental research.

39  Something like this idea appears natural if “the visual field” is construed as a field, i.e., as a flat, coloured area at which I am somehow looking, and not as an expression for describing what objects, persons and other elements of my surroundings I am able to see without moving. The analogy turns awkward, if not earlier, at least when we try to incorporate three-dimensionality and movement.

40  See also Lee 2005: 61-62. On the other hand, this also means that photos taken on a cloudy day will generally reproduce colour nuances more faithfully than those taken in bright sunshine. – Jan-Erik Nordström, communication with the author.

41  Lopes 2005: 421-422. Cf. Newall 2006, Hyman 2007.

42  Also see Newall 2006.

43  Brou et al. 1986: 80.

44  Brou et al. 1986: 80.

45  Brou et al. 1986 say “the perceived colors of things” depend on comparisons. But does their point concern the mechanisms of perception or is it about the meaning of our colour concepts? Do the authors mean we perceive colours in ways that falsely suggest their independence of reflected light, or do they say that colours are to a degree independent of it? Brou et al. do not recognise the idea of there being “real colours” of objects to start with, hence the question does really make sense for them.

46  Brou et al. 1986: passim.

47  Cf. Brou et al. 1986: 87. The authors present a physiological explanation that, insofar as I can see, is not in conflict with the account I am giving here, rather demonstrating its physiological background.

48  In this passage, I am developing a formulation by David Cockburn, in a commentary to an earlier paper by me.

49  An earlier version of this paper was presented at the Department of Philosophy research seminar at Åbo Academy. I have also profited from discussions around another paper by me on Wittgenstein and Lichtenberg (Lagerspetz 2009). I have received helpful comments to issues discussed in these two papers from David Cockburn, Lars Hertzberg, Jan-Erik Nordström, Göran Torrkulla, Yrsa Neuman, Antony Fredriksson, and others.

Torna su

Per citare questo articolo

Notizia bibliografica

Olli Lagerspetz, «The colours of things and the metaphysics of colour samples»Rivista di estetica, 43 | 2010, 161-179.

Notizia bibliografica digitale

Olli Lagerspetz, «The colours of things and the metaphysics of colour samples»Rivista di estetica [Online], 43 | 2010, online dal 30 novembre 2015, consultato il 18 septembre 2021. URL:; DOI:

Torna su

Diritti d’autore

Licenza Creative Commons
Rivista di Estetica è distribuita con Licenza Creative Commons Attribuzione - Non commerciale - Non opere derivate 4.0 Internazionale.

Torna su
  • Logo Rosenberg & Sellier
  • DOAJ - Directory of Open Access Journals
  • OpenEdition Journals
Cerca su OpenEdition Search

Sarai reindirizzato su OpenEdition Search