Are Colours Visually Complex? Pär Sundström 1. Introduction Squarehood is a visually complex property in the following sense. To be square is to have certain parts or aspects — four lines of equal length connected at right angles — that are visually accessible and none of which is identical with squarehood. And to see something as being square is to see it as having these parts or aspects. It is often supposed that colours are not thus visually complex. For example, I think we can take Locke to express this view when he men- tions colours among the simple Ideas; which being each in itself uncompounded, contains in it nothing but one uniform Appearance, or Conception in the mind, and is not distinguishable into different Ideas. (Locke 1975 [1689]: sect. 2.2.1) Similarly, Hume, discussing blue, green, scarlet and “particular sounds, and tastes and smells”, says that “their very nature … excludes all com- position” (Hume 1978 [1739]: 637).1 Perhaps this view of colour is ultimately correct. However, I am not sure it is correct. I think we should take seriously the hypothesis that colours — all colours — are visually complex in the above explained sense. This paper tries to explain why I think we should take this ser- iously. Section 2 presents a case that almost all shades of colour are visually complex. I will not fully articulate the case but I hope to say enough to convey that it is strong. Section 3 presents a more tentative case that the remaining colours are visually complex as well. 2. Almost All Shades of Colour Are Visually Complex There is, I believe, a strong case to be made that almost all shades of colour are visually complex. Consider a contemporary phenomeno- 1 See Mizrahi (2009) for another, recent endorsement of this kind of view. 628 logical colour model, like the Natural Color System, NCS.1 This system, or model, takes there to be six “elementary attributes”,2 blackness, whiteness, redness, yellowness, greenness and blueness, and it takes each shade of colour to be composed in a quantifiable way by 1–4 of these attributes. For example, the greyish-orange shade 3020-Y50R is composed by blackness, whiteness, redness and yellowness. The initial 30 in the notation says that this shade is composed by 30% blackness. The ensuing 20 says that its proportion of chromatic attributes (redness, yellowness, greenness or blueness) to achromatic attributes (blackness or whiteness) is 20:80. The Y50R says that its chromatic component is composed by equal proportions of yellowness and redness. One can derive from this that the shade in question is composed by the following shades in the following proportions: 10% redness, 10% yellowness, 30% blackness, and 50% whiteness.3 To judge this kind of model fairly it is important to keep it clearly apart from other models of colour, and two in particular. First, there are additive colour models that serve to systematise which perceived colours will be projected on a screen by various mixtures of monochromatic light. Additive colour models tell us, among other things, that we can project an image of yellow on a screen by blending, in the right propor- tions, monochromatic light that by itself would project an image of green on a screen and monochromatic light that by itself would project an image of red. Second, there are subtractive colour models that serve to systematise what colours result from mixtures of various pigments. Subtractive colour models tell us, among other things, that a colour printer will print green if it mixes certain proportions of yellow and cyan. It is clear that additive and subtractive colour models are designed for different purposes, make different and supplementary claims and that they systematise colours in nonequivalent ways without being in conflict with one another. It should also be clear that each of them differs in the 1 For expositions see Hård and Sivik (1981), Hård and Svedmyr (1995) and Hård et al. (1996). This kind of model is usually traced to Hering (1964 [1920]), but traces can be discerned earlier in history; see Hård and Svedmyr (1995: 39–40) and Pridmore (2006). 2 Hård et al. (1996: 189). 3 See Hård and Sivik (1981), Hård and Svedmyr (1995: chapter 2 and pages 138–9) and Hård et al. (1996: part 1). 629 same ways from the kind of phenomenological colour model that the Natural Color System represents. The latter kind of model serves to sys- tematise how colour visually appear. It tells us about the visual charac- teristics of 3020-Y50R and other shades. But it makes no claim about which colour pigment mixtures or monochromatic light mixtures will produce these shades, and it is not in conflict with additive or subtractive models even when these systematise colours in ways that do not map onto the phenomenological one. For example, the claim that greenness is a phenomenologically “elementary attribute” is compatible with the claim that one can produce green by certain mixtures of pigments.1 To my eyes and mind, the Natural Color System is natural. It seems to me that 3020-Y50R is visually a mixture of blackness, whiteness, red- ness and yellowness. I gather that I share this sense with many others who are familiar with the model. For example, Hård and Sivik report that, “people without any previous knowledge of colour assessment, other than with common color names, understand and rapidly acquire the NCS method of describing colours — less than 15 minutes is generally required” (Hård and Sivik 1981: 137). Moreover, there is reportedly a high degree of agreement between different subjects’ specific assess- ments about how shades of colour are phenomenologically composed.2 The system is also widely adopted by professionals in, e.g., architecture, design and painting (see Hård and Svedmyr 1995: chapter 3 — and many local paint stores). This makes for a strong case, I think, that almost all shades of colour are visually complex in the present sense. For example, 3020-Y50R has multiple component parts or aspects — blackness, whiteness, redness and yellowness — that are visually accessible and none of which is identical with that shade, and to see something as 3020-Y50R is to see it as composed by these attributes in the relevant proportions. Please note: 1 For a bit more on these three types of model and their differences, see Sundström (2008: sect. 4.5) and Byrne and Hilbert (2008: sect. 2). 2 Interestingly, the agreement reportedly holds even between (a) subjects who make the estimates with the aid of samples of the elementary colours (the shades of colour that are composed by exactly one of the elementary colour attributes) and (b) sub- jects who make the estimates without such samples, drawing only on their own “in- ner” understanding of the elementary colours. For details, see Hård and Svedmyr (1995: 67–9) and Hård et al. (1996: 185–7). 630 I claim that the case is strong, not that it is conclusive. We may well be quite suggestible when it comes to perceptual and introspective reports.1 And, as far as I can tell, the test subjects that have been involved in the research and development behind the Natural Color System have had the system suggested to them. Their reports — and mine — may well be tainted by these background suggestions. Nonetheless it seems to me that these reports — in particular, the agreement between them and the related widespread use of the model among professionals in architecture, design and painting — provides strong support for the claim that almost all colours are visually complex.2 1 th The “imageless thought controversy” of the early 20 century comes to mind; for an overview see Thomas (2011: sect. 3.2). 2 As I announced, I do not here fully articulate the case that almost all colours are visually complex. To do so one should address at least two counter-proposals. The first is that, while the Natural Colour System provides a natural ordering of shades of colour, colours do not have this ordering because they are composed by ele- mentary colour attributes in different proportions. One can perhaps trace this kind of proposal to Hume. Hume claims that, “Blue and green are different simple ideas, but are more resembling than blue and scarlet: tho’ their perfect simplicity excludes all possibility of separation or distinction. ‘Tis the same with particular sounds, and tastes and smells. These admit of infinite resemblances upon the general appearance and comparison, without having any common circumstance the same” (Hume 1978 [1739]: 637). One may perhaps take this to suggest that, e.g., shades of orange bear some natural resemblance to one another but that this resemblance is not rooted in their “common circumstances” redness and yellowness. The second counter- proposal to what I have said and that a full defence should take into account is that the Natural Colour System does not even provide a natural ordering of colours. For example, Mizrahi, defending a “conventionalist approach to colour categorization” claims that “the fact that orange is steadily said to be both reddish and yellowish [is] not rooted in the phenomenology of colour experience” (Mizrahi 2009: sect. 4); quoted from the online version, which has the word “in” in place of the above bracketed “is”). Similarly, Saunders and van Brakel claim that the categorisation of chromatic colour in terms of four primitive hues is “rhetorical” (Saunders and van Brakel 1997: 173), by which I believe they mean that it is not grounded in the appearance of colours but has some other origin. See also Allen (2011: sect. 4). Note that the argument of the present paper does not obviously stand or fall with the claim that all shades of colour are composed by 1-4 of the elementary attributes of the Natural Colour System.