See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/332989419

What is the ‘‘Opposite’’ of ‘‘Blue’’? The Language of Color Wheels

Article  in  Journal of Perceptual Imaging · January 2019

DOI: 10.2352/J.Percept.Imaging.2019.2.1.010401

CITATION READS
1 2,877

1 author:

Neil Anthony Dodgson

Victoria University of Wellington
208 PUBLICATIONS   4,957 CITATIONS   

SEE PROFILE

Some of the authors of this publication are also working on these related projects:

animation from photographs View project

Stylised Rendering View project

All content following this page was uploaded by Neil Anthony Dodgson on 07 May 2020.

The user has requested enhancement of the downloaded file.

JOURNAL OF PERCEPTUAL IMAGING, VOL. 2, NO. 1, JAN–JUN 2019 1

What is the ‘opposite’ of ‘blue’?

the language of colour wheels
Neil A. Dodgson

pre-print of a paper in Journal of Perceptual Imaging 2(1), Jan-Jun 2019

©2019 Society for Imaging Science & Technology
https://doi.org/10.2352/J.Percept.Imaging.2019.2.1.010401

Abstract—Colour wheels are a tool for ordering and understanding yellow

hue. Different colour wheels differ in the spacing of the colours around primary
the wheel. The opponent colour theory, Munsell’s colour system, the
standard printer’s primaries, the artist’s primaries, and Newton’s rainbow
all present different variations of the colour wheel. I show that some of
this variation is owing to imprecise use of language, based on Berlin
green orange
and Kay’s theory of basic colour names. I also show that the artist’s
secondary secondary
colour wheel is an outlier that does not match well to the technical colour
wheels because its principal colours are so strongly connected to the
basic colour names.

1 I NTRODUCTION
blue red

C OLOUR wheels provide a way to describe the ordering

of hue and, in some cases, to aid understanding colour
mixing. The artist’s colour wheel (Figure 1), epitomised
primary primary

by Itten [Itt70], is used extremely widely in teaching. Its

purple
primary colours are red, yellow and blue.This is the colour
secondary
wheel that students meet in primary school. In this wheel,
the opposite of blue is orange. When students meet more
Fig. 1. An Itten colour wheel with twelve hues. The three primaries, red,
advanced material in colour theory, they find apparent con- yellow, and blue, combine to make three secondaries, green, orange,
tradictions. The printer’s colour wheel has primaries cyan, and purple. Each primary combines with its neighbouring secondaries
magenta, and yellow, which the student might be taught to make six tertiary colours.
to understand as a refinement of blue, red, and yellow.
But curiously for the student the colour labelled “blue” (a) (b) (c)
in the printer’s colour wheel is opposite to yellow, not to
orange. In my own early introduction to colour, I found the
art books’ insistence that orange was the opposite of blue
conflicted with my observation that, in many works of art
and design, yellow appeared to me to be the more apposite (d) (e) (f )
opposite. Further confusion comes to the student when they
meet the opponent colour theory, in which there are four
principal colours, with blue opposite yellow (Figure 3); and
Munsell’s colour system in which there are five principal
hues, with blue opposite yellow-red (Figure 4). The chal-
lenge for the educator is in explaining these differences. Fig. 2. The artist’s harmonious colour combinations: (a) complementary,
These differences can be downplayed in educational (b) split complementary, (c) triadic, (d) tetradic rectangle, (e) tetradic
square, (f) analagous. (b)–(e) are from Itten [Itt70, Figs. 54, 55].
material. For example, one text for art students states that
“Colour wheels must always have an even number of hues
and that number must be divisible by three. Any other com-
and five principal colours [Ble12, p.31] and which have a
bination would not be a true and accurate colour wheel” [Ble12,
well-defined notation for describing colours around the hue
p.66, emphasis mine]. This is a simplification by the author
wheel.
for the benefit of the students, as the author is well aware
One of the reasons to question the received wisdom
of the NCS and Munsell colour systems which have four
is that almost all art texts, inspired ultimately by Itten’s
School of Engineering and Computer Science, Victoria University of seminal work [Itt70], use angles on the colour wheel to
Wellington, Wellington, New Zealand e-mail: neil.dodgson@vuw.ac.nz determine “harmonious” colour combinations (Figure 2). If
JOURNAL OF PERCEPTUAL IMAGING, VOL. 2, NO. 1, JAN–JUN 2019 2

5Y 10YR
10Y 5Y
5 GY R

GY

10
R
10
10G 5G

5R
10RP 5RP
5BG
G

10
0B

P
1
5P
5B 10P
5PB 10B B

Fig. 3. An NCS colour wheel with four principal hues [HST96] [NCS18] Fig. 4. A Munsell colour wheel with five principal hues and five inter-
[Swe89]. The four principals are red, yellow, green, and blue. The mediate hues [Mun76] [Mun18]. The five principal hues are red, yellow,
circle is divided into 400 units, 100 between each pair of principals. green, blue and purple. The intermediates are indicated by combinations
Hues between the principals are indicated by numeral between the two of the colour letters (e.g., YR=yellow-red). The wheel is further subdi-
principals’ initial letters. In this chart we see units every ten steps for vided into ten sections for each principal and intermediate, indicated by
each of the four quadrants. numerals. In this chart we see the 5 and 10 units for each of the ten
sections. The “5” unit is the prototypical version of each hue. The “10”
unit is a half-and-half mix of the hues either side. Image used under a
Creative Commons 3.0 license from WikiMedia author Thenoizz.
the colour wheel is not immutable, as the different colour
wheels suggest, then these harmonies rest on insecure foun-
I first give a summary of the history of colour spaces and
dations. This is by no means a new problem [Ber81, Ch.6B]
colour wheels (Section 2), then a history of colour naming
[Bri07]. Some of the difference between the different colour
and an outline of Berlin and Kay’s theory of basic colour
wheels can be explained from the principles underlying
terms (Section 3). I describe five of the most commonly
their constructions and the uses for which they are designed.
used colour wheels (Section 4). I demonstrate that impre-
There is difference in how you construct your colour space
cise use of colour names explains a substantial amount of
depending on whether you are mixing coloured lights,
the apparent inconsistencies between the different wheels
mixing coloured pigments, or dealing with human visual
(Section 5), allowing us to reconcile these differences. This
perception [Der91]. For example, Itten’s artist’s colour wheel
leads to the observation that the technical colour wheels
is based in subtractive colour mixing of pigments; opponent
are broadly consistent with one another, provided we are
colour theory is based in visual perception; and Munsell
precise about our specification of the principal colours in
was aiming to bring clarity to colour communication by
those spaces, but that the standard artist’s colour wheel
establishing an orderly system for accurately identifying
is substantially different from the technical colour wheels
all colours. All the colour spaces discussed in this paper
(Section 6), because its primary and secondary colours are
are ways of specifying or mixing colours so all can be
so strongly related to use of basic colour terms in English.
considered ways of dealing with pigment.
The contribution of this paper is to argue that our under-
standing of colour wheels is mediated by the terms we use 2 H ISTORY
to describe colours, in particular in the use of basic colour Colour has fascinated philosophers and artists since antiq-
terms [BK69]. This leads to some of the apparent differences uity but it is only in the last century that we have come to
between colour wheels in two ways. First, generally across understand the psychophysical and biological mechanisms
all colour wheels, we use the same basic colour name, such of colour vision, so early writers could be said to be working
as blue, to represent subtly different colours in different in the dark. Aristotle described seven principal colours
wheels (Section 5), which confuses the student. Second, and (white, yellow, red, violet, green, dark blue, black) which
specific to the artist’s colour wheel (Figure 1), while the he considered all to be mixes of white and black [Sha94], a
artist’s (RYB) and printer’s (CMY) colour wheels should misconception that started to be challenged in the fifteenth
both be identical, because they are both subtractive colour century [Alb66] but still held some sway until the eighteenth
mixing models, I argue that the differences between them century. The discovery that red, yellow and blue are the
are largely owing to the artist’s colour wheel being actively artist’s primaries was made in the early seventeenth century.
driven by basic colour terms in a way that puts it at odds Shapiro cites Parkhurst and Gage as reporting that four
with the optimal physical colour mixing embodied in the scholars independently discovered the artist’s tri-chromatic
printer’s colour wheel (Section 6). primaries. All four scholars were conversant with both art
JOURNAL OF PERCEPTUAL IMAGING, VOL. 2, NO. 1, JAN–JUN 2019 3

and the natural sciences, giving them access to the under- colours: red, yellow, green and blue, in addition to black
standings needed to make this discovery. Shapiro asserts and white [Har06]. Hering formalised this into the opponent
that it “. . . was the most important discovery in colour theory of colour vision [Der91] [HJ57]. Hering’s theory was
before Newton’s own theory” [Sha94, p.624]. further developed by Hård, Sivik and Tonnquist in their
In the late seventeenth century, Newton conducted ex- creation of the Natural Colour System (NCS) [HST96]. There
tensive investigations into the nature of colour, discovering is evidence that the four opponent principal colours are
that white light split into an infinite range of colours: the vi- physiologically determined [Gou91] [Har05]. Hering’s the-
sual spectrum. This discovery was at odds with the widely- ory was not widely embraced at the time because there was
held belief that white was “pure” and could not be split no understanding of how responses to two different colours
and also at odds with the three primary colours discovered of light could interact to create a colour-opponent signal. We
earlier that century, discrepancies that caused him much know today that the neurons in the retina process the out-
trouble to attempt to reconcile and which led to substantial puts of the light-sensitive cones to produce three channels
challenges in his work being accepted. Neverthless, in his of data to the brain: a high resolution luminance channel,
writings before Opticks, whenever he listed his principal a lower-resolution red-green channel and an even lower-
colours of the spectrum, he always added some phrase resolution blue-yellow channel [Ber81, p.16] [HJ57]. The
such as “with their innumerable intermediate gradations” opponent colour channels explain well several features of
to indicate that there were countless discernible colours, but human vision, including the way in which colour blindness
in Opticks he omits to say this in all but one place, possibly manifests and the complementary afterimages caused after
in an attempt to placate his critics [Sha94, p.619]. Newton’s fixating on a coloured field. Consistent with this theory is
early work described five principal colours: red, yellow, that you cannot perceive a colour as having simultaneously
green, blue and purple, but he later added orange and components from either end of an axis, so a yellowish-green
indigo, leading to English’s current seven-colour rainbow and a bluish-green both make sense, but a human can never
(see longer discussion in the Appendix). perceive a colour that is “reddish-green”, such a mixture
At the start of the nineteenth century, Goethe launched being a nonsense.
a challenge on Newton’s purely physical approach, tackling Over far more than a century, philosophers, scientists
colour instead as a perceptual phenomenon. To a technically and artists have grappled with ways to represent and under-
trained modern, some of Goethe’s arguments can seem mis- stand colour, leading to many systems of colour representa-
guided when compared with Newton’s empiricism. But tion. Basic introductions can be found in computer graphics
Newton was, in his own way, blinkered: fitting the data to and design texts [FvDFH90, Ch.13] [JMF94] [SAM09, Ch.20–
suit his hypothesis rather than the other way round [Pla06] 22] [Sto03], with more detailed explanations in specialist
[Rib85]. The challenge Newton faced was that his evidence texts [Ber81] [Ble12] [Bri07] [KB96], and a full history of
was inconsistent, because he was assuming that mixing colour spaces in Kuehni and Schwartz’s 2008 book [KS08].
lights (additive colour mixing) and mixing pigments (sub- A colour space is a three-dimensional representation
tractive colour mixing) should produce consistent results. of colour. We can restrict ourselves to three dimensions
It was only in 1852 that Helmholtz deduced that different because the human visual system has three types of receptor
rules apply to the mixing of pigments and of lights [Sha94]. for colour vision. All of the colour spaces are mathematical
In additive colour mixing, different coloured lights are used, transformations of one another. Hunter gives a detailed
each with its own spectrum. The mix of the lights is a history of nineteen colour spaces developed in the attempt
spectrum that is the weighted sum of the spectra of the to create a perceptually uniform space, starting with the CIE
individual lights, weighted by the intensity of the lights. 1931 colour space and Munsell’s original system, through
Additive colour is used in display devices and the usual to the CIELUV and CIELAB systems of 1976 [Hun75,
primary colours used are a red, a green, and a blue. This is Ch.8]. Derefeldt gives the background of the most impor-
in contrast to subtractive colour mixing, in which coloured tant colour appearance systems, including Munsell, NCS,
pigments, inks, paints or dyes are mixed together. Each CIELAB and CIELUV. She gives their basic attributes, and
pigment absorbs some part of the spectrum of light. The mix the principles for scaling and notation of the variables. In
of pigments absorbs a weighted sum of the absorptions of particular, she makes a comparison of the hue spacing of
the individual pigments. Subtractive colour mixing is used the different spaces [Der91]. Note that there is considerable
in painting, printing and dyeing [Ber81]. evidence that colour vision is non-Euclidean, so any colour
The chemist, Chevreul, dyemaster at the Gobelin work in space is not going to be a metric space, perceptually [Ber81,
Paris, published De la Loi du Contraste Simultané des Couleurs p.64]. For example, the CIELAB system has a cube-root
et de l’Assortiment des Objets Coloris in 1839 [Itt70]. This, and relationship with the signals that are received by the cones
other emerging colour theories, had substantial influence in the human eye. This is to better match the perceptual
on artists in the nineteenth and early twentieth centuries. response of the human visual system but means that linear
Itten says that “Delacroix. . . is the founder of the tendency, mixes in the CIELAB system do not necessarily match mixes
among modern artists, to construct works upon logical, ob- of pigments.
jective colour principles, so achieving a heightened degree A colour wheel is a representation of one dimension
of order and truth.” [Fry06, p.418]. The Impressionists and of a colour space: hue. Colour wheels have been used
Post-Impressionists, in particular, used theories of colour for centuries. The earliest known drawing of a colour
contrast and optical colour mixing. wheel dates from 1611 [PF82], a century before Newton’s
Several early commentators on colour, including da Opticks [New04].
Vinci, noted that there appear to be four fundamental A colour wheel or, more accurately, a hue wheel, is a
JOURNAL OF PERCEPTUAL IMAGING, VOL. 2, NO. 1, JAN–JUN 2019 4

circle that passes through all of the spectral colours and then brain [Har05]. Hardin notes that the four principal colours
through the purples to join the two ends of the spectrum (yellow, red, blue, green) “. . . prove to be both necessary and
(Figure 1). Hue is explicitly one of the three dimensions in sufficient for an English speaker to describe any spectral
some colour systems, including NCS (Figure 3) and Munsell stimulus” [Har98]. The other basic colour categories may
(Figure 4), and is implicit in others, where hue is a function be more culturally determined. Children are able to match
of two or three of the principal dimensions of the space. and discriminate colours long before they have consistently
For example, in the case of CIELAB, h◦ = tan−1 (b∗ /a∗ ). codified the boundaries in colour space of the basic colour
When considering a colour wheel, the hues always appear terms, so providing evidence that the boundaries are a
in the same order around the wheel but they differ in which social construct [ATF86]. In any case, in order to commu-
hues appear opposite each other and in the relative angular nicate clearly between members of a language group, the
separation of pairs of hues. learnt categories must be at least partly a social construct,
A student may make an assumption that a “true” colour reinforced by parents, kindergartens and primary schools
wheel exists and that the different colour wheels essentially because all members of the language group broadly agree
stretch or contract sections of the “true” wheel to fit their on them.
predilections, as if the colours were painted on a rubber Berlin and Kay identified that the number of basic colour
bicycle wheel and we nailed certain hues to certain points terms range between two (representing light and dark
on the rim. The stretching and contracting is epitomised colours) and twelve, depending on the language. In English
in the differences in the angles red–yellow and green–blue, there are eleven basic colour terms: red, orange, yellow,
shown in Table 1. When a colour wheel is used as a mech- green, blue, purple, pink, brown, black, grey and white.
anism to describe hue, then such stretching or contracting is As an example of the irreducibility of these basic terms,
fair: the wheel is not purporting to show precise physical consider how difficult it is to convince a child that brown
relationships. However, when a colour wheel is used to is really “dark orange” or that pink is “light red” [Har98,
describe relationships or mixes between distant hues, such p.210]. You may teach a particular child or student to make
as in defining the “opposite” of a hue or “harmonious finer distinctions, as between “cyan”, “azure”, “indigo” and
colour combinations” (Figure 2), then this stretching and “turquoise”, but there is a cultural push towards teaching
contracting becomes questionable. and agreeing on the eleven basic colour terms [KB96, Ch.11],
and there is demonstrated effect of these basic categories on
the ability to perform colour discrimination [WWF+ 07]. The
3 BASIC COLOUR TERMS maximum number of basic colour terms in any langague
Berlin and Kay proposed the theory that there are basic appears to be twelve. Russian, and a few other languages,
colour terms in all languages [BK69]. These are the terms distinguish light blue (Russian goluboy) from dark blue (Rus-
that you teach small children and which produce categories sian siniy) [Par05]. This paper considers the case of English
of colour that are irreducible, that is, all other colour terms though most other European languages use the same eleven
are considered, by most speakers of the language, to be categories, which is important to our discussion because
variations on these basic colour terms. Itten, in particular, was working in German.
In antiquity, classical scholars certainly privileged certain Rather than conducting new perceptual experiments,
colours above others. In the distant past, the fundamental we are able to make use of results from three previous
colours appear to have been severely limited. Berlin and studies [BK69] [RDD00] [RH72], which used colour chips
Kay quote Geiger as suggesting that “Democritus and the evenly chosen from Munsell’s colour space.
Pythagoreans [fifth century BC] assumed four fundamen- Ignoring the monochrome black, grey and white, there
tal colours, black, white, red and yellow” [BK69, p.136]. are eight basic colour terms in English. Roberson et
Elsewhere, Geiger comments that Aristotle [fourth century al. [RDD00] experimented with an array of 160 coloured
BC] “in his ‘Meteorology’ calls [the rainbow] tri-coloured, chips, evenly spaced within the Munsell colour system,
viz., red, yellow, and green” [Gei80, p.57]. By the fifteenth asking English speaking subjects to categorise each chip into
century, things had developed a little further. Alberti cites one of the eight colour categories.
three fundamental colours: red, green, blue, combined with Figure 5 shows the mean colour chosen by subjects for
grey [Alb66, Book I, paragraph 9] while da Vinci lists what each colour chip. In addition, each colour region contains a
we now call the colour opponent set of principal colours: small cross that marks the “best-example choice” for each
red, yellow, green, blue [Har06]. In the seventeenth century, of the eight colours, as described by Rosch [RH72]. Notice
Boyle listed the standard artist’s primaries: red, yellow and the difference in sizes of the different colour terms: orange
blue [Har06], but added green and purple when actually (5.5 cells), yellow (6.5 cells) and brown (9 cells) each take
conducting his experiments on colour [Boy64, p.187]. In up only a small part of the colour space compared with
the early eighteenth century, Newton started with these green (52.5 cells) and blue (36 cells). While I acknowledge
five principal colours: red, yellow, green, blue and purple, that Munsell’s colour space is non-uniform and is somewhat
then added orange and indigo (see longer discussion in the compressed in the yellow-red area and expanded in the
Appendix). blue-green area, that cannot explain the full magnitude of
There is a question of nature versus nurture: how much this difference. Over 50% of the chart is categorised as one
the colour categories are inherent in our psychophysiology of two terms blue and green; by contrast, red, orange and
and how much they are cultural constructs. There is good yellow between them take up just 14% of the chart (see
evidence that black, white, yellow, red, blue, and green are also Hardin’s comments on the relatively small sizes of the
strongly tied to the perceptual mechanisms in the human “warm” colours’ regions compared with the relatively large
JOURNAL OF PERCEPTUAL IMAGING, VOL. 2, NO. 1, JAN–JUN 2019 5

5R 10R 5YR 10YR 5Y 10Y 5GY 10GY 5G 10G 5BG 10BG 5B 10B 5PB 10PB 5P 10P 5RP 10RP

9 pink yellow
pink
8

6
orange
5 green blue
4 brown purple
3 red red
2

Fig. 5. Roberson, Davies and Davidoff’s diagram of the eight basic colours in English (redrawn from [RDD00, Fig.1a]). The colour space is that of
the Munsell colour system, which has five principal colours, red (R), yellow (Y), green (G), blue (B), purple (P), and their various combinations along
the horizontal axis, with brightness on the vertical axis (2=dark, 9=light). See Figure 4 for an explanation of the notation. The experiments used a
160 chip Munsell array and the array shows, for each of the 160 cells, the mean colour chosen by English speakers for each colour chip. Some
cells lie on the boundary, in which case the boundary passes through the centre of the cell. The small crosses mark the “best-example choices”
for each of the eight colours, as described by Rosch [RH72]. The colour of each area matches that best-example choice, within the limits of the
available gamut. The “best-example choices” are taken directly from Rosch’s 1972 paper [RH72]; the locations of several of these “best examples”
are placed incorrectly in Roberson et al’s 2000 paper [RDD00, Fig.1a].

5R 10R 5YR 10YR 5Y 10Y 5GY 10GY 5G 10G 5BG 10BG 5B 10B 5PB 10PB 5P 10P 5RP 10RP

9 yellow
8

7 pink
6 orange
5 green blue
4

3 red purple
brown
2

Fig. 6. Berlin and Kay’s diagram of the eight basic colours in English (redrawn from [BK69, Appendix I, p.119]). As in Figure 5, the colour space
is that of the Munsell colour system. See Figure 4 for an explanation of the notation. Berlin and Kay used a 320 chip Munsell array. They asked
participants to determine, for each basic colour term, x, (1) all those colour chips which they would, under any conditions, call x, and (2) the best,
most-typical examples of x. The small crosses mark the locations of the “best, most-typical example” for each colour. The colour of each area
matches that best most-typical example, within the limits of the available gamut. The white areas represent colour chips that were not given an
unequivocal colour name.
JOURNAL OF PERCEPTUAL IMAGING, VOL. 2, NO. 1, JAN–JUN 2019 6

sizes of the “cool” colours’ regions [Har05]). Describing a which is designed to be a reasonably uniform space, per-
colour as “red”, “orange” or “yellow” will always give a ceptually (see Section 2), and in the Natural Colour System
colour close to the “best-example choice”, that is, the colour (NCS) [HST96] [Ber81, p.39] [Fai05, p.99] [NCS18] [Swe89]
will be close to what an average person would imagine it to (Figure 3), which is designed for specifying colour in a
be. By contrast, describing a colour as “green” or “blue” can similar manner to Munsell’s colour system but, in the case
give a colour that is a significant distance from the “best- of NCS, using the observer’s in-built understanding of what
example choice”. Hardin discuses the consistency of such is meant by a “pure” red, yellow, green, blue, black and
studies, noting that, across the many studies,“No matter white [NCS18]. Derefeldt notes that “The development of
how many basic color terms languages might have, their foci the NCS began by psychometric testing of Hering’s con-
[“best-example choices”] tend to cluster reliably in relatively ceptual framework having observers estimate qualitative
narrow regions of the [Munsell] array, whereas boundaries colour attributes by assuming that observers could imagine
are drawn unreliably, with low consistency and consensus six elementary colours by means of verbal definitions only.
for any language.” [Har98, p.208]. As evidence that there These imaginary colours, which constitute cognitive, natural
is consistency between observers, consider that the NCS reference points, were used as references in absolute judg-
colour system is predicated on there being good agreement ments without any physical representation of the references.
between observers on what Berlin and Kay call the “best, . . . The definitions of the six elementary colours. . . follow
most-typical example” of the four principal colours red, Hering’s definitions of primary colours closely.” [Der91,
yellow, green and blue [HST96]. p.234].
Berlin and Kay undertook a different experiment [BK69], Both CIELAB and NCS are used for specifying colour
using a 320 Munsell chip array, in which, for each basic rather than purporting to represent colour mixing. Because
colour term, they asked English speaking subjects to select CIELAB is non-linear, a linear mix in CIELAB space will not
all those chips that they would, under any conditions, cate- necessarily create the same colour as mixing matching pig-
gorise as being of that colour. Figure 6 shows the regions in ments. Because NCS is entirely perceptual, there is, again, no
which they got an unequivocal response from their subjects. guarantee that a linear mix of two NCS colours will match
One important result, for our investigation, from Berlin and the mix of two pigments. The principal colours, RYGB, are
Kay’s work is that both cyan and indigo were unequivocally four of Berlin and Kay basic colour terms.
described as “blue” by their subjects.
Note that pink and brown do not appear on the standard 4.2 Munsell colour system, RYGBP, Figure 7(b)
colour wheel. Pink is a light variant of red. Brown is a dark
The five principal colours in this space are those used by
variant of orange. The basic English colour terms along the
Boyle in his seventeenth century colour experiments [Boy64,
visual spectrum are thus red, orange, yellow, green, blue
p.187] and are the original five colours of the rainbow spec-
and purple. Of these, orange is a relatively recent addition
ified by Newton. Munsell formalised this in the early twen-
to the basic colour terms in English. Red (Old English réod),
tieth century, using the “. . . guiding principal of equality of
yellow (geolu), green (grene) and blue (blaw) are all ancient
visual spacing” [Ber81, p.36]. The colour space was designed
colour terms. Purple was brought into English, from the
to aid in colour specification, originally for schoolchildren.
Latin, in the ninth century. Orange, by contrast, was adopted
The specific colours of the five principal hues were deter-
only in the early sixteenth century. Its first attested use as a
mined visually. Munsell’s colour order system was exten-
colour name was in 1512. Prior to this it had been known
sively reworked (“renotated”) in the 1940s by the Optical
as yellow-red (Old English geoluréod). It is unclear when
Society of America. This was a painstaking process of mea-
orange became a basic colour term in English, but it is a
suring the discriminability of the colors and adjusting the
possibility that Newton’s description of the spectrum was
spacing of the colours to optimize them for use in science
an influence. Similarly, in German, gelb, rot, blau, and grün
and industry [Mun76] [Mun18]. It should be made clear that
are ancient terms with words for orange and purple being
Munsell’s space is not a uniform colour space: small steps
more recent [Jon13].
should be roughly equal visually but large steps cannot be
compared. Berns implies that having five principal hues
4 T HE COLOUR WHEELS leads to greater visual equality between neighbouring hues
than a system based on the four principal unique hues of
Figure 7 illustrates the principal colours of five colour red, green, yellow, and blue [Ber81], though Kaiser and
wheels in common use. It is immediately obvious that they Boynton suggest that there is evidence that the Munsell
do not map linearly to one another. The colour that is dia- principal hues are not necessarily spaced evenly perceptu-
metrically opposite to blue ranges from yellow (Fig. 7(a),(c)) ally, in particular that P and PB are too far apart [KB96,
through orange (Fig. 7(b),(d)) to a red-orange (Fig. 7(e), p.494]. All five principal hues are Berlin and Kay basic
but see also the Appendix). I briefly describe each of the colour terms.
five colour spaces, including the purposes for which it was
designed and the principal colours it uses.
4.3 Printer’s colour wheel, RYGCBM, Figure 7(c)
This is the most pragmatic of the colour wheels, relat-
4.1 Opponent colour spaces, RYGB, Figure 7(a) ing directly to how printing works. The colour space is
The opponent colour spaces are based on the perceptual op- explicitly designed for colour mixing. The three principal
ponent colours of Hering. The opponent principal colours, colours, cyan, magenta and yellow, are the primaries of
RYGB, are used in the CIELAB colour space [Ber81, p.67], subtractive colour mixing, as used in printing. Each primary
JOURNAL OF PERCEPTUAL IMAGING, VOL. 2, NO. 1, JAN–JUN 2019 7

(a) Y (b) (c) Y (d) Y (e)

Y Y
G R G O
R O G
G
G R R
C M B R B
B B P B P I V

Fig. 7. Representations of the principal colours of five colour wheels, as they might be constructed by a student. All colours are represented by
their initial letter: Red, Orange, Yellow, Green, Cyan, Blue, Indigo, Purple, Violet, Magenta. All colour names are taken at face value. Left to right:
(a) colour opponent theory, with four principal colours, RYGB; (b) Munsell’s colour system with five principal colours, RYGBP; (c) the printer’s colour
wheel with the three subtractive primaries for colour printing CMY and their three secondaries RGB, RYGCBM; (d) the artist’s colour wheel with
the three painting primaries RYB and the three secondaries OGP, ROYGBP; (e) Newton’s colour wheel, with the seven colours of the rainbow,
ROYGBIV, notice that Newton’s colours are not evenly spaced round the wheel (see Appendix).

is physically realised as a pigment that absorbs certain is used in the same sense as in the printer’s colour wheel,
wavelengths of light. When two primaries are mixed or Section 4.3), mixing to make the three secondaries: orange,
superimposed, the mixture absorbs the wavelengths that green and purple. All six of the primary and secondary
are absorbed by each pigment in proportion to the amounts colours are Berlin and Kay basic colour terms (see Section 6
of each primary mixed. The specific primaries chosen are for a discussion of the implications of this).
pigments that, when mixed, allow production of a large
gamut of colours. Good choices for primaries, that produce 4.5 Newton’s rainbow, ROYGBIV, Figure 7(e)
close to the largest gamut achievable with three pigments,
are broadband yellow, broadband cyan, and broadband ma- This is the prototypical early colour wheel, from when
genta (see Section 6). Mixing each pair of primaries produces evidence was beginning to be gathered about how coloured
the three secondaries, which are called red, green and blue, light and colour mixing worked. It is misguided in several
although these turn out to be rather imprecise descriptions respects (see the Appendix). Newton originally described
(see Section 5.6). While four of these six colours are Berlin the rainbow as having five colours, the same five that Mun-
and Kay basic colour terms, it is important to our discussion sell used two centuries later, but Newton quickly adopted
that cyan and magenta are not. As a consequence, cyan two extra colours (orange and indigo) to make the seven
and magenta are relatively precise terms, and each is well- colour rainbow that is taught in all English-language pri-
localised in colour space compared to say, blue or green. mary schools. His colour wheel is not evenly spaced and
his use of the terms “blue” and “indigo” do not match
their modern uses but this colour wheel has gained almost
4.4 Artist’s colour wheel, ROYGBP, Figure 7(d) unstoppable traction in English education about colour, to
This is the wheel that Itten exemplified (Figure 1) [Itt70]. the confusion of many students. Newton’s rainbow has the
This colour wheel has been known for over two centuries, same colours as the artist’s colour wheel plus indigo.
but Itten’s work in the 1950s and ’60s pushed it to preem-
inence. Prior to Itten, other colour wheels had been used 5 R ECONCILING THE DIFFERENT COLOUR WHEELS
in art teaching. For example, an opponent-colour system Consider the structure of the various colour wheels as a
designed by Wilhelm Ostwald was used in British art educa- student would view them. Figure 7(a)–(d) shows the result
tion between the two world wars, in which the colour wheel if you place the principal colours evenly spaced around
had four principal colours, though Ostwald used a bluish- the wheel, as they are in all diagrams in the student’s text
green opposite red [Bri07, Sec.7.3] rather than the pure green books (e.g., Figures 1, 3, 4). Table 1 tabulates the angle a
used by NCS and Hering. Itten, by contrast, designed his student would measure between red and yellow and that
colour wheel on the foundations that there must be three between green and blue. The red-yellow angle varies from
primaries and that diametrically-opposite colours must mix 60◦ (Figure 7(c)) to 120◦ (Figure 7(d)), while the green-blue
to grey [Itt70, p.21]. Briggs comments on how pervasive angle varies from 120◦ (Figure 7(c)) to 60◦ (Figure 7(d)). In a
Itten’s influence has become: “Itten’s book [The Art of Colour non-metric space, these angles are, at best, approximate, but
(1961)] has been so influential that it defines the limits of a student will still worry about why the angular distances
artistic colour theory for the majority of sources on the around the wheels differ so markedly, especially if they
internet today. . . As a result of its half century of ascendancy, have been trained to build the harmonious colour combi-
many artists today assume that traditional colour theory nations of Figure 2, which explicitly require consideration
has dominated art education continuously since its origins, of angle. They will also be concerned to understand why
and assume modern colour theory is a very recent intru- diametrically-opposite colour pairs differ between colour
sion” [Bri07, Sec.11.3]. Itten himself developed the concepts spaces.
of harmonious combinations of colour (Figure 2), which
are specified by precise angular relationships around the
colour wheel. The artist’s colour wheel is used to help artists 5.1 What is meant by “opposite”?
understand colour relationships and colour mixing. Red, Let us return to the question “what is the opposite of
yellow and blue are the primary colours (in which primary blue?”. The discussion above has implicitly assumed that
JOURNAL OF PERCEPTUAL IMAGING, VOL. 2, NO. 1, JAN–JUN 2019 8

TABLE 1 yellow
primary
The first four colour wheels from Figure 7, showing the differences in
angle between the pairs red-yellow and blue-green, and giving the
opposite colours to red and blue.
green red
secondary secondary
red- blue- opposite opposite
Colour wheel yellow green of red of blue
Opponent (RYGB) 90◦ 90◦ green yellow
Munsell (RYGBP) 72◦ 72◦ blue-green yellow-red
Printer’s (RYGCBM) 60◦ 120◦ cyan yellow
Artist’s (ROYGBP) 120◦ 60◦ green orange
cyan magenta
primary primary
the “opposite” of a given hue is the hue that is on the
opposite side of a diameter through the centre of a colour blue
wheel. There are at least three other useful definitions of secondary
opposite [Fry06] [Har06].
Fig. 8. Mixing real primary inks to produce the three secondaries.
• Additive complementaries: two coloured lights that, Inner ring: primary colour toners mixed on a colour laser printer [Fuji
when mixed, give white. Xerox FX ApeosPort-IV C3375 v3018.103 PS]. Outer ring: primary fluid
• Subtractive complementaries: two pigments that, when acrylic paints mixed with a paintbrush [paints from Golden Artist Col-
ors, Inc: Primary Cyan (pigments PW6/PB15:4/Titanium White/Phthalo
mixed together, produce a grey. In theory, oppo- Blue(GS)), Primary Magenta (pigment PV19/Quinacridone), Primary
sites on the artist’s and printer’s colour wheels (Fig- Yellow (pigments PY3/PY73/PW6/Hansa Yellow Light/Hansa Yellow
ure 7(c) and (d)) should do this. Medium/Titanium White)].
• Perceptual complementaries: a colour’s opposite is the
colour perceived as an afterimage after fixating on
the first colour for a significant period of time. The answer to “what is the opposite of blue?” depends
both on what you mean by “opposite” and on what you
As Harkness shows [Har06], these each give slightly differ- mean by “blue”. Some of the differences in the “opposite
ent opposites for any given colour. For example, fixating on of blue” column in Table 1 are owing to differences in the
Itten’s red and then looking away will give a blue-green meaning of “opposite” and some are explained by the word
sensation rather than to the green of the artist’s colour “blue” refering to different hues in the different cases.
wheel [Bri07, Sec.11.3]. So the word “opposite” needs to
be defined carefully in order to give a clear answer to
our question. This means that we should not expect the 5.3 Imprecision in colour naming
Munsell or opponent-colour wheels to have the same colour
More generally than blue, we find that the colour names
diametrically opposite blue as do the printer’s or artist’s
are imprecise in several cases in our various colour wheels,
colour wheels, because the Munsell and opponent-colour
where the actual principal colour used in the colour system
colour spaces were not designed using criteria by which op-
does not match the “best most-typical” example of that
posite colours necessarily represent complementary colours.
colour name. If we consider the colours by how they actu-
Indeed, these colour spaces are non-linear spaces and there-
ally appear, rather than by their basic colour names, we find
fore attempts to use them for accurate colour mixing will
that it is possible to reconcile a great deal of the apparent
fail.
differences between the wheels.
However, we would expect the printer’s and artist’s
Consider the colour wheels in light of the linguistic
colour wheels to have the same diagonally-opposed colours,
ambiguity inherent in the colour names and in terms of the
because they are both constructed by the same principle
true appearance of each colour. In Figure 7, we assumed,
of subtractive complementarity. We find that they do not:
as a student might, that each colour represented the “best,
the diagonal opposite of blue is yellow in the printer’s
most-typical example” of that colour, marked by the crosses
colour wheel and orange in the artist’s. As the definition
in Figure 5. Figure 9 redraws those diagrams to reflect the
of “opposite” is the same in these two wheels, we must
actual colour that is represented by each of the general
consider the definition of “blue”.
colour terms. We consider each wheel in turn.

5.2 What is meant by “blue”?

5.4 Opponent systems (CIELAB and NCS), RYGB
When asked to imagine a blue, the average person will
choose a colour close to the “best, most-typical example” In the standard CIELAB system (Figure 9(a)), yellow and
at 10B/4.5. But when asked if a particular colour is “blue”, blue match their best, most-typical examples, but red and
the answer is “yes” for a range from cyan through to indigo green do not. CIELAB red is a purplish-red; CIELAB green
(Figure 6). “Blue” can refer to any spectral colour from about is a cyanish-green [Har06].
490nm (a greenish-blue, cyan) to 450nm (a purplish-blue, By contrast, in the Natural Colour System (NCS), the
indigo). What we mean by “blue” changes the answers to four principal colours all do match their best, most-typical
questions about that colour. As Itten says, “unless our color examples (Figure 9(f)) because NCS is defined explicitly in
names correspond to precise ideas, no useful discussion of terms of the colours that an observer would consider to be
color is possible” [Itt70, p.30]. “pure” red, yellow, green and blue [HST96, p.181] [NCS18].
JOURNAL OF PERCEPTUAL IMAGING, VOL. 2, NO. 1, JAN–JUN 2019 9

Y Y colour space that the average observer would call the “best
(a) (b) most-typical example” of those colour names.
Y G Y
G R R R
5.7 Artist’s, ROYGBP
cG
cG pR Following Itten [Itt70], the artist’s principal colours match
their “best, most-typical” examples, by definition, because,
cB P
B in the absence of any other concept, the student will use
their internal linguistic definition of the colour to ensure
B B that their red is neither shading towards yellow nor shading
Y towards blue, and likewise for all the other colours (see
Y
Section 6).
(c) (d)
G Y G Y
oR R O 5.8 Discussion
G G
Consider the six revised wheels of Figure 9. Here, I have
C RP B indicated the actual hue of each of the principal colours and
R indicated an approximate location for the “best most-typical
pB B P R
examples” of yellow, red, blue and green. In particular,
B
compare the approximate locations of those four colours
between the six wheels and compare the lengths of the
Y Y arcs between red and yellow and between green and blue,
(e) (f ) though recall that CIELAB, Munsell, Newton and NCS are
G Y Y not metric spaces and so the angles subtended by these arcs
O give only an approximate indication of the actual distance
G R R
R G R between two colours.
G We find that CIELAB (Figure 9(a)), Munsell (Figure 9(b)),
C
B rP and the printer’s colour wheel (Figure 9(c)) are now all
B
very similar. Our linguistic adjustments push all three of
these colour spaces close enough to one another that we
B B can see that they are describing much the same thing.
Newton’s ROYGCBV (Figure 9(e) and Appendix) is a little
Fig. 9. The colour wheels of Figure 7(a)–(d) and Figure 11(b) redrawn
to reflect the actual colour represented by each of the principal colour distorted, with a larger red-yellow angle than those three
names. An uppercase letter represents the same colour as in Figure 7, a colour wheels, emphasising the role of orange, but this was
lowercase letter indicates a colour modifier, with “-ish” added to the end an early attempt at a colour system so we can accept it as a
of the colour name; for example, oR is “orangish-red”. Inside each wheel rough approximation. We include it because of its continu-
are labels for the actual colours represented by each coloured disc.
Outside each wheel are the approximate locations of the best, most- ing influence on children’s education about colour. The NCS
typical examples of Red, Yellow, Green, Blue. The thick arc indicates opponent colour system (Figure 9(f)) is distorted further, but
the green-blue angle. The thick dotted arc indicates the red-yellow angle. it is designed for colour specification not for colour mixing
The six colour wheels are: (a) CIELAB, one variant of RYGB; (b) Mun-
sell’s RYGBP; (c) the printer’s colour wheel, RYGCBM; (d) the artist’s and it is known to be perceptually uneven: there are more
colour wheel, ROYGBP; (e) Newton’s ROYGCBV where we have used visually distinct hues between red and blue than between
the correct angles from Figure 10 and applied the colour corrections yellow and green [Fai05, p.100]. The outlier is the artist’s
discussed in the Appendix; (f) NCS, a second variant of RYGB in which colour wheel (Figure 9(d)) where the red-yellow section is
the colour names match their best, most-typical examples of that colour.
clearly expanded and the blue-green section compressed
compared with all the other wheels.
5.5 Munsell, RYGBP
Both the principal green and the principal blue have a 6 W HY IS THE ARTIST ’ S WHEEL DIFFERENT ?
cyanish cast to them (Figure 9(b)), so true green and true
In theory, the artist’s RYB and the printer’s CMY colour
blue are closer to yellow and purple respectively.
wheels should be identical. Both purport to have primaries
that cannot be made by mixing other colours. Both purport
5.6 Printer’s, RYGCBM to be able to create all hues from the three primaries.
To assess the printer’s colour wheel (Figure 9(c)), consider Both purport to have diametrically-opposite colours that
the colours in Figure 8. This shows two different exam- mix to make grey. However, the CMY colour wheel is
ple sets of CMY inks mixed to make two sets of RGB demonstrably the correct way to do this, given that these are
secondaries. Note the consistency between the two sets of the colours used in the vast majority of commercial colour
secondaries. Here “green” is close to the “best, most-typical” printing processes. The underlying theory is that each of
green, but “red” is an orangish-red, rather than the “best, the primaries theoretically absorbs exactly one third of the
most-typical example” of a red, and “blue” is far removed visual spectrum. Berns [Ber81, Ch.6], for example, suggests
from “best most-typical”, being a deep purplish-blue: an splitting the spectrum into thirds at 500 nm and 600 nm.
indigo. So, although these secondaries are informally called A theoretical cyan ink absorbs all red and orange light.
“red” and “blue”, they are not sitting at the positions in A theoretical magenta ink absorbs all yellow and green
JOURNAL OF PERCEPTUAL IMAGING, VOL. 2, NO. 1, JAN–JUN 2019 10

light. A theoretical yellow ink absorbs all blue and violet either colour on either side [Itt70, p.29]. The former is
light. Combinations of these three primaries can produce achieved correctly by the printer’s colour wheel; the latter
any hue. In practice, the spectra of the three inks are not pushes the primary and secondary colours to their “best,
perfect squares [Ber81, pp.154–5] [Kip01, Figs.1.4-20,-22], so most-typical locations” in perceived colour space. The fact
the range of colours achievable is not as broad as would that the printer’s colour wheel and the artist’s colour wheel
be possible with perfect theoretical primaries, but we can are substantially different demonstrates that these two aims
manufacture inks of sufficient quality to satisfy the vast cannot both be satisfied in the same colour wheel1 .
majority of our printing needs. The artist’s RYB is thus an approximation to CMY, yet the
The dramatic difference between the theoretically-correct artist’s colour wheel remains by far the most popular colour
CMY colour wheel and the artist’s RYB colour wheel can wheel, outside the technical sphere. This, in spite of the fact
be explained by considering the mechanism by which the that the printer’s wheel is technically superior for mixing
artist’s colours are chosen. In Itten’s seminal writing on the the widest possible range of colours. We must ask why it is
colour wheel [Itt70], he writes: “. . . a person with normal that a colour wheel that appears technically inferior should
vision can identify a red that is neither bluish, nor yellowish; be so tenaciously held. I hypothesise that the artist’s colour
a yellow that is neither greenish, nor reddish; and a blue that wheel’s success is owing to its use of the six basic colour
is neither greenish, nor reddish. . . The primary colors must terms that correspond to spectral colours: if you are teaching
be defined with the greatest possible accuracy.” There is no colour theory to children, you will gravitate towards using
freedom here to allow red to be magenta, because magenta the colour names with which they are most familiar.
is a red that is distinctly bluish, nor is there freedom to One of the challenges in teaching technical printing is
allow blue to be cyan, because cyan is a blue that is dis- to explain to the student the special role of magenta and
tinctly greenish. I hypothesise that Itten is placing his three cyan, and to describe what they are in terms of the basic
primaries at or near the “best, most-typical locations” in the colour terms (“reddish-pink” and “greenish-blue” respec-
Berlin-Kay sense. Itten then mixes his secondaries, which are tively). For example, Gleeson, in her text on the illustration
all also Berlin-Kay basic colour terms. So while Itten says of picture books, identifies magenta with red and cyan with
that his hues are “. . . evenly spaced with complementary blue [Gle03, p.53], while Cianciolo, writing on the same
colours diametrically opposite each other, ” his even spacing topic, twice mentions the four process colours, naming them
is in a linguistic sense rather than a physical one. as red, yellow, blue and black [Cia76, pp.61,88]. This is not
Note also how Itten defines his red, yellow and blue. necessarily a misunderstanding on the author’s part but
Red is “neither bluish, nor yellowish”, defined relative to a need to explain the technical concepts (“magenta” and
the other two primaries. But yellow and blue are “neither “cyan”) in language that is accessible to the general reader
greenish, nor reddish” (emphasis mine), so Itten’s three pri- (“red” and “blue”).
maries are defined relative to the four principal colours of As a framework for teaching colour, RYB does admit
opponent-colour theory [Bri07, Sec.11.3]. The NCS colour the possibility of using colours other than the “best most-
system defines its principal colours in exactly the same typical” and artists over the centuries have used a range
way [HST96, p.181], but uses this defining mechanism to of different reds, yellows, and blues as their primaries.
create four principal colours rather than the three primary However magenta is outside the red zone in the Berlin
colours of the artist’s colour wheel (Figure 9(f) cf. 9(d)). and Kay diagrams. I hypothesise that the untrained ob-
By having red and yellow as primaries in the artist’s server has a challenge with accepting magenta as a primary,
colour wheel, orange becomes a natural secondary and, because it is not intuitively satisfying. Though magenta
because it is a basic colour term, it is possible to mix is the correct colour for printing, it does not fall at one
orange so that, to the artist’s eye, it is neither too reddish of the optimal points in Berlin and Kay’s diagram, being
or too yellowish, thereby occupying its “best most-typical” somewhere between red, pink, and purple. Red is much
position in colour space. With blue in the third of the more satisfying, being one of the key colours in Hering’s
primary positions, it is obvious from the English Berlin theory of perception. Yellow, by contrast, is both a primary
and Kay chart (Figure 5) that the other two secondaries and an optimal point in linguistic colour space, so we have
are going to be green and purple, if we wish them to also no trouble accepting it. Cyan is a blue but it is not the most
be basic colour terms. As with orange, it is possible to typical blue.
mix these, as Itten says we should, so that they are well- There is a further gloss on the use of RYB. Despite it
balanced and not leaning towards the colour on either side, being taught to children as a way of “mixing colours”, it
which I hypothesise places them at the “best most-typical” would be extremely unusual for a professional artist to have
positions. just three colours on their palette. Rather, the artist’s colour
The substantial difference between the artist’s colour wheel is used as a framework within which to understand
wheel and the other colour wheels would not be a problem colour relationships. This is because it is not possible to
if the artist’s colour wheel, as designed by Itten, were not so achieve all colours by mixing just a red, yellow and blue;
pervasive in education about colour. and because having a pure hue allows for consistency of
The challenge with creating the artist’s colour wheel colour not achievable in repeated mixings. For example,
is that Itten had two aims that cannot be satisfied simul- Matisse used a palette of 17 colours, van Gogh 9 colours, and
taneously: he wants his diametrically-opposed colours to
be perfect subtractive complementaries [Itt70, p.20] and he 1. It is possible to create narrowband pigments for opposing colours
in Itten’s scheme, so that the opposing colours mix to grey, but broad-
wants his primary and secondary colours to be mixed “very band pigments for the primaries do not allow coverage of as large a
carefully” so that, perceptually, they do not “lean towards” gamut as CMY.
JOURNAL OF PERCEPTUAL IMAGING, VOL. 2, NO. 1, JAN–JUN 2019 11

Fryer shows an example of his own work with a palette of

14 colours [Fry06, p.419]. And, while the colour harmonies
of Figure 2 are widely taught, any professional artist or
designer will use their own judgment of harmony rather
than slavishly depend on this basic framework. Indeed, Itten
himself says that different students find different combina-
tions harmonious so that there cannot be a general principle
that appeals to all [Itt70, p.23].

7 C ONCLUSION
Some of the apparent difference between the colour wheels
can be explained linguistically. The most obvious example
is that we recognise that “blue” has a broad spectrum
of meanings and that the “best, most-typical” blue is an
imprecise approximation to the true colour represented by
the word. In the printer’s colour space, “blue” is an indigo, Fig. 10. Newton’s colour wheel (adapted from [New04, Book I, Part II,
in Munsell’s colour space it is tending to cyan, in the NCS Plate III, Fig. 11]). Notice that orange and indigo have segments of
and artist’s colour spaces it is sitting at the “best most- only 30◦ compared to the 60◦ allocated to the five other colours. The
typical” position, and in the traditional rainbow, ROYGBIV, uppercase letters A − G are intended to correspond to the notes of
the musical scale, with orange and indigo corresponding to semi-tone
it transpires that “blue” was originally used by Newton to intervals. The lowercase letters p − x are at the centres of the seven
mean “cyan”. But all of these are described informally by the colour arcs. Notice the blue (t) is directly opposite the boundary (E )
single term “blue”. Likewise “red” and ”green” are used in between red and orange.
some colour spaces to refer to colours that are not the “best
most-typical” example of the colour.
What can educators conclude from this? Using and indigo (which was a recently discovered, imported
commonly-understood terms, like “opposite” or “blue”, to dyestuff). The addition of these two colours appears to have
also have a specific technical meaning leads to problems, been driven by his desire to get the spectrum to agree with
unless one is careful to define those terms to have precise the notes of the musical scale [New04, p.114]. Goethe cri-
meaning. When educating students about colour, we need to tiques Newton for adding orange and indigo and criticises
be careful to be precise in what we mean when we use terms his musical analogy as an attempt to impose on the colours
like “blue”. In the printing industry, we already have this a mathematical order they do not in fact have [Rib85].
precision when talking about the CMY space, because cyan To get from a linear spectrum to a circular colour wheel
and magenta are not basic colour terms, so our students “Newton also notes that purples could be created by com-
understand them to have precise meanings, and yellow is a bining light from the two ends of the spectrum. . . ” [Wal02,
precise term in common usage, because it occupies such a p.193] so allowing us to join up the two ends into a cir-
small part of the overall colour space (Figure 5). But terms cle [Sha94, p.620]. Looking at Newton’s own drawings of
like red, green, blue and purple all have imprecise meaning his colour wheel (Figure 10), we see an oddity: in order to
in English and we must be careful to ensure that we are match the tones and semi-tones of a musical scale, Newton
defining them appropriately. gives the new colours, orange and indigo, only half as much
There is a remaining challenge, which is that the artist’s space on the wheel as the original five colours. If we take his
colour wheel is at such odds with all of the other colour colour wheel at face value, we see that the opposite of blue is
wheels and yet is the first colour system that most people the boundary between red and orange. This is significantly
will meet. I hypothesise that one reason for its tenacity is different from any of the modern understandings, where the
that it is a convenient approximation that allows educators opposite of blue lies between orange and yellow.
to use six of the basic English colour terms in explaining With regard to his use of the colour indigo, “a careful
how colour works. reading of Newton’s work indicates that the color that he
called indigo, we would normally call blue; his blue is then
what we would name blue-green or cyan” [Wal02, p.193].
A PPENDIX Finlay points out that, in the eighteenth century, indigo
N EWTON ’ S COLOUR WHEEL referred to a much wider range of colours than it does
Newton’s seven-colour rainbow is pervasive in English- today [Fin07, p.340], generated by different concentrations
language education but it is based on shaky foundations. of indigo dye. Taking into account both this information
Newton performed some of the earliest scientific work on about the meanings of words and also the non-uniform
understanding colour. He had access to some of the earliest spacing of colours means that a naı̈ve version of Newton’s
optical components that were of good quality and demon- colour wheel (Figure 11(a)) is incorrect and what he meant is
strated that a prism split white sunlight into a spectrum much better represented by Figure 11(b), where we replace
of colours. In Newton’s earliest work on this, he names Newton’s “blue” by “cyan” and his “indigo” by our modern
five colours of the spectrum: red, yellow, green, blue and “blue”. We now find that the opposite of blue is orange-
violet [Sha94]. In his later work he augments this to seven yellow.
colours adding orange (a relatively new word in English) How much easier would our explanation of the rainbow
JOURNAL OF PERCEPTUAL IMAGING, VOL. 2, NO. 1, JAN–JUN 2019 12

(a) R (b) R [Gle03] Libby Gleeson. Making Picture Books. Scholastic Press,
Gosford, NSW, Australia, 2003.
V O O
V [Gou91] P. Gouras. Cortical mechanisms of colour vision. In
P. Gouras, editor, The Perception of Colour, pages 179–197.
Y Macmillan, Basingstoke, 1991.
I Y B [Har98] Clyde L Hardin. Basic color terms and basic color cat-
egories. In Werner G. K. Backhaus, Werner Backhaus,
B G C G Reinhold Kliegl, and John Simon Werner, editors, Color
Vision: Perspectives from Different Disciplines, chapter 11,
pages 207–217. Walter de Gruyter, 1998.
Fig. 11. Two representations of Newton’s colour wheel. (a) The wheel as [Har05] Clyde L Hardin. Explaining basic color categories. Cross-
it would be generated from a naı̈ve literal reading of the colour names Cultural Research, 39(1):72–87, 2005.
and from equal spacing around the wheel. (b) The wheel when we take [Har06] Nick Harkness. The colour wheels of art, perception, sci-
into account that Newton’s “indigo” is a modern blue, his “blue” is cyan ence and physiology. Optics & Laser Technology, 38(4):219–
and his spacing around the wheel is non-uniform (Figure 10). 229, 2006.
[HJ57] Leo M Hurvich and Dorothea Jameson. An opponent-
process theory of color vision. Psychological Review, 64(6,
be if Newton had chosen to stick with his original five pt. 1):384–404, 1957.
[HST96] Anders Hård, Lars Sivik, and Gunnar Tonnquist. NCS,
colours, or had chosen to introduce cyan as his seventh Natural Color System—from concept to research and ap-
colour instead of indigo? We may even have been on the plications. Part I. Color Research & Application, 21(3):180–
way to having a twelfth basic colour term adopted into 205, 1996.
English, as does Russian. Russian does have seven basic [Hun75] Richard S. Hunter. The Measurement of Appearance. Wiley,
1975.
colour terms in its rainbow [Par05]. As it is, indigo is a
[Itt70] Johannes Itten. The Elements of Color. Van Nostrand
constant source of confusion in teaching colour in English: Reinhold, New York, 1970.
children are taught that indigo is a fundamental colour in [JMF94] Richard Jackson, Lindsay MacDonald, and Ken Freeman.
the rainbow, but most people do not distinguish it as a Computer Generated Color: A Practical Guide to Presentation
and Display. Wiley, 1994.
separate colour. The artist’s colour wheel discards it with-
[Jon13] William Jervis Jones. German Colour Terms: a study in their
out hesitation: the seven colour ROYGBIV becomes the six historical evolution from earliest times to the present. John
colour ROYGBP. Benjamins Publishing, 2013.
[KB96] Peter K Kaiser and Robert M Boynton. Human Color Vision.
Optical Society of America, 1996.
ACKNOWLEDGEMENTS [Kip01] Helmut Kipphan, editor. Handbook of Print Media: Technolo-
gies and Production Methods. Springer, 2001.
Thanks to the anonymous reviewers for their careful read- [KS08] Rolf G Kuehni and Andreas Schwarz. Color ordered: a
ing of the manuscript and for their substantial input that survey of color systems from antiquity to the present. Oxford
considerably improved the quality of this final version. University Press, 2008.
[Mun76] Munsell Color. Munsell Book of Color. Munsell Color
Company, Baltimore, 1976.
R EFERENCES [Mun18] Munsell Color. Munsell Color website, 2018.
http://www.munsell.com/, visited on 2018–12–24.
[Alb66] Leon Battista Alberti. On Painting. Yale University Press, [NCS18] NCS Colour AB. NCS Colour website, 2018.
second edition, 1966. A translation, with introduction http://www.ncscolour.com/, visited on 2018–12–24.
and notes, by J. R. Spencer of the Latin De Pictura, first [New04] Isaac Newton. Opticks: or, a Treatise of the Reflexions, Refrac-
published 1435. tions, Inflexions and Colours of Light. Royal Society, 1704.
[ATF86] Gail Rex Andrick and Helen Tager-Flusberg. The acquisi- [Par05] Galina V Paramei. Singing the Russian blues: An argument
tion of colour terms. Journal of Child Language, 13(1):119– for culturally basic color terms. Cross-cultural research,
134, 1986. 39(1):10–38, 2005.
[Ber81] Roy S. Berns. Billmeyer and Saltzman’s Principles of Color [PF82] Charles Parkhurst and Robert L Feller. Who invented the
Technology. Wiley, third edition, 1981. color wheel? Color Research & Application, 7(3):217–230,
[BK69] Brent Berlin and Paul Kay. Basic Color Terms: Their Univer- 1982.
sality and Evolution. Univ of California Press, 1969.
[Pla06] Jim Platts. Newton, Goethe and the process of perception:
[Ble12] Steven Bleicher. Contemporary Color: Theory and Use. Cen-
an approach to design. Optics & Laser Technology, 38(4):205–
gage Learning, second edition, 2012.
209, 2006.
[Boy64] Robert Boyle. Experiments and Considerations Touching
Colours. Henry Herringman, 1664. [RDD00] Debi Roberson, Ian Davies, and Jules Davidoff. Color
[Bri07] David J. C. Briggs. The Dimensions of Colour: modern categories are not universal: replications and new evidence
colour theory for traditional and digital painting me- from a stone-age culture. Journal of Experimental Psychology:
dia, 2007. http://www.huevaluechroma.com/, visited on General, 129(3):369–398, 2000.
2018–12–24. [RH72] Eleanor Rosch Heider. Universals in color naming and
[Cia76] Patricia J Cianciolo. Illustrations in Children’s Books. Wm. memory. Journal of Experimental Psychology, 93(1):10–20,
C. Brown Company Publishers, second edition, 1976. 1972.
[Der91] G. Derefeldt. Colour appearance systems. In P. Gouras, [Rib85] Neil M Ribe. Goethe’s critique of Newton: A reconsider-
editor, The Perception of Colour, pages 218–261. Macmillan, ation. Studies in History and Philosophy of Science Part A,
Basingstoke, 1991. 16(4):315–335, 1985.
[Fai05] Mark D Fairchild. Color Appearance Models. Wiley, 2005. [SAM09] Peter Shirley, Michael Ashikhmin, and Steve Marschner.
[Fin07] Victoria Finlay. Color: A Natural History of the Palette. Fundamentals of Computer Graphics. AK Peters, third edi-
Random House, 2007. tion, 2009.
[Fry06] Michael J. Fryer. Complementarity. Optics & Laser Technol- [Sha94] Alan E Shapiro. Artists’ colors and Newton’s colors. Isis,
ogy, 38(4):417–430, 2006. 85(4):600–630, 1994.
[FvDFH90] James D. Foley, Andries van Dam, Steven K. Feiner, and [Sto03] Maureen Stone. A Field Guide to Digital Color. AK Peters,
John F. Hughes. Computer Graphics: Principles and Practice. 2003.
Addison-Wesley, second edition, 1990. [Swe89] Swedish Standard. Swedish Standard SS 01 91 02 Colour
[Gei80] Lazarus Geiger. Contributions to the History of the Develop- Atlas. Swedish Standards Institution (SIS), Stockholm,
ment of the Human Race. Trübner & Company, 1880. second edition, 1989.
JOURNAL OF PERCEPTUAL IMAGING, VOL. 2, NO. 1, JAN–JUN 2019 13

[Wal02] Gary Waldman. Introduction to Light: The Physics of Light,

Vision, and Color. Dover, 2002.
[WWF+ 07] Jonathan Winawer, Nathan Witthoft, Michael C. Frank,
Lisa Wu, Alex R. Wade, and Lera Boroditsky. Russian
blues reveal effects of language on color discrimination.
Proceedings of the National Academy of Sciences, 104(19):7780–
7785, 2007.

View publication stats